RAS AGREEMENT BETWEEN THE MINISTRY OF LABOR, PENSION SYSTEM, FAMILY AND SOCIAL POLICY
 (MLPSFSP) AND THE INTERNATIONAL BANK FOR RECONSTRUCTION AND DEVELOPMENT (WORLD
                                       BANK)




DELIVERABLE – Report on the "Analysis of the Long-Run Pension
Adequacy and Sustainability in the Croatian Pension System, with the
          focus on Policy Options for its Improvement"

                                       Final Version



   Under the “Analysis and policy proposals for the improvement of the
  Long-Run Adequacy and Sustainability of the Croatian Pension System
                            (RAS, P178457)”




EUROPE AND CENTRAL ASIA REGION
HECSP – SOCIAL PROTECTION & LABOR




DATE July 16, 2024
DISCLAIMER

© 2024 The World Bank
1818 H Street NW, Washington DC 20433
Telephone: 202-473-1000; Internet: www.worldbank.org

Some rights reserved


This work is a product of the staff of the International Bank for Reconstruction and Development/ The
World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily
reflect the views of the Executive Directors of The World Bank or the governments they represent. The
World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors,
denominations, and other information shown on any map in this work do not imply any judgment on the
part of The World Bank concerning the legal status of any territory or the endorsement or acceptance
of such boundaries.

Rights and Permissions

The material in this work is subject to copyright. Because the World Bank encourages dissemination of
its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long
as full attribution to this work is given.

Please cite the work as follows: “World Bank. 2024. Report on the Pension Adequacy in Croatia © World Bank.”
All queries on rights and licenses, including subsidiary rights, should be addressed to World Bank
Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-
2625; e-mail: pubrights@worldbank.org.




                                                     2
Table of Acronyms and Abbreviations

Acronym/Abbreviation        Definition
AIF                         Alternative investment fund
APD                         Accumulated Public Debt
AR                          Aging Report
AUM                         Assets under management
APV                         Actual point value
EU AWG                      European Union’s Aging Working Group
EU SILC                     European Union Statistics on Income and Living Conditions
ETF                         Exchange-traded fund
GTC                         Gross Transition Cost
                            Croatian Pension Insurance Institute (Hrvatski zavod za
HZMO
                            mirovinsko osiguranje)
IFRS                        International Financial Reporting Standards
MLPSFSP                     Ministry of Labor, Pension System, Family and Social Policy
NTC                         Net Transition Cost
OMF                         Mandatory pension fund
OR                          Personal account in the second pillar
P1                          Pension provided only by the first PAYG pension pillar
P2                          Combined pension from first PAYG and second pillar
PAR                         Pension Adequacy Report
PAYG                        Pay-as-You-Go
PF                          Initial factor
                            Pension insurance company (Mirovinsko osiguravajuće
PIC
                            društvo)
PROMIS                      Pension Reform Options Microsimulation Model
RAS                         Reimbursable advisory services
REGOS                       Central registry of affiliates
TRR                         Theoretical replacement rate
                            Undertaking for Collective Investment in Transferable
UCITS
                            Securities
VP                          Value points
ZOMO                        Law on pension insurance




                                      3
                             A brief overview of the document



This Report is intended to provide analytical support to the Ministry of Labor, Pension System,
Family and Social Policy in designing pension reform policies with the objective to improve
pension adequacy but also preserve the fiscal and social sustainability of the system. It has
been created as a supporting document in the process of implementation of the National
Recovery and Resilience Program 2021-2026 of the Republic of Croatia which envisages the
preparation of amendments to the Pension Insurance Law in the fourth quarter of 2025.
The Report gives a short overview of recent legal changes and pension policies in Croatia and
discusses current pension adequacy in Croatia and its determinants. Key pension system
outcomes related to adequacy and fiscal balance are compared with the trends in pension
benefit provisions in the EU. Key system challenges in all three pension pillars are identified
(see overview below).
Based on the individual beneficiary career model the key policy issues are discussed and a
number of policy measures recommended (see overview below). Main recommended policies
are then simulated and analyzed with the microsimulation model and tested in policy
packages. The policy package optimizing the improvement in pension adequacy and fiscal
sustainability is identified and recommended.
The key message of this Report is that Croatian pension system does not require a
fundamental reform such as the abolishment of the second pension pillar or reintroduction
of the mandatory second pillar pension (abolishment of the pension choice). The current
three-pillar system can, with appropriate adjustment and upgrade of its parameters, continue
providing adequate social protection to Croatian population in the old age better than only
the first or only the second pension pillar. The policy package recommended in this Report
includes a set of measures in all three pillars to set them in balance and on a long-run
sustainable trajectory.
According to the analyzes conducted in this Report, simultaneous implementation of these
measures, especially the strongest and mutually balancing ones would i) reverse the trend of
eroding pension adequacy, ii) set the Croatian pension system on a sustainable fiscal path,
and iii) assure intergenerational sustainability by keeping the total pension contribution rate
unchanged.




                                              4
Croatian pension system – challenges and potential solutions

   Main challenges / features                          Main reform proposals


  •short service period                              •increase of early retirement age with
  •low average exit age                               shortening the gap between early and
  •low support ratio (contributors /                  statutory retirement age
   pensioners)                                       •increase of penalty for early retirement
  •abundant early retirement options                  pension
  •choice option between PAYG and                    •increase of retirement age and
   combined (1st and 2nd pillar) pension              insurance period for old-age pension
   at the time of retirement                          for long-term insured
  •significant amount of non-earnings                •increase of statutory retirement age
   related transfers included in pension             •adjustment of valorization / indexation
   benefit calculation affecting the                  rotational formula with new weights
   choice between PAYG and combined                   (85:15)
   pension                                           •increase of disability and survivor
  •less generous valorization and more                pensions by extending added period
   generous indexation than in most EU               •increase of pension supplement (from
   countries                                          the current 4%-27% to 7%-30%)
  •high coverage by minimum pensions                 •increase of minimum pension (from
  •lower contributions from other                     103% to 106% of APV per year of
   income (due to either lower                        insurance)
   contribution rate or lower                        •equalization of contribution rate for
   contributions base)                                other income with the general
  •contained pension system deficit                   contribution rate
  •low coverage and level of voluntary               •application of pension-type-choice-
   pension savings                                    neutral transfer scheme
                                                     •promotion of active choice of
                                                      investment allocation and
                                                      implementation of IAS
  •low adequacy of pensions
                                                     •increase of the 2nd pillar contribution
  •high old-age at-risk-of-poverty
                                                      rate (when the social and fiscal
                                                      conditions are appropriate and in
                                                      balance)
                                                     •further liberalization (simplification) of
                                                      investment limits bor both mandatory
                                                      and voluntary pillar
                                                     •application of the non-linear cascading
                                                      incentive scheme for better targeting
                                                      low-income workers for voluntary
                                                      savings
                                                     •increase of existing cash withdrawal
                                                      option from 30% to 50% for voluntary
                                                      pillar, but not earlier than statutory
                                                      retirement age




                                               5
Contents
1.     About the RAS task........................................................................................................................ 12
2.     Croatian pension system: current setting and overview of past changes .................................... 13
3.     Pension adequacy in Croatia – definition, performance, and determinants................................ 21
     3.1      Definition of pension adequacy ............................................................................................ 21
     3.2      Pension adequacy in EU policy documents .......................................................................... 22
     3.3      Pension adequacy in Croatia ................................................................................................. 24
       3.3.1         Poverty risk in old age and retirement in Croatia ......................................................... 24
       3.3.2         Income replacement of Croatian pensions ................................................................... 26
       3.3.3         Gender pension gap ...................................................................................................... 30
       3.3.4         Minimum pensions ....................................................................................................... 32
       3.3.5         Disability pension .......................................................................................................... 34
       3.3.6         Survivors’ pensions ....................................................................................................... 35
       3.3.7         Pension adequacy of multipillar pensions .................................................................... 36
     3.4      Key determinants of pension adequacy in international comparisons ................................ 40
       3.4.1         Statutory retirement age and early retirement options ............................................... 41
       3.4.2         Contribution period ...................................................................................................... 45
       3.4.3         Accrual rate ................................................................................................................... 48
       3.4.4         Support ratio ................................................................................................................. 49
       3.4.5         Pension indexation........................................................................................................ 50
       3.4.6         Supports for low-income pensioners ............................................................................ 52
       3.4.7         Options to combine public pensions with income from work ...................................... 53
4.     Pension projections for hypothetical careers ............................................................................... 55
     4.1      Projection settings and assumptions .................................................................................... 55
     4.2      TRR Baseline: no-policy change scenario .............................................................................. 56
     4.3      Impact of investment policy and wage profile on future pensions ...................................... 60
     4.4      Impact of wage growth and investment returns on choice of pension................................ 61
     4.5      Choice between P1 and P2 and the impact on the pension adequacy and fiscal balance ... 63
     4.6      Cash withdrawal and the choice between P1 and P2 ........................................................... 65
5.     Fiscal sustainability of the PAYG system ....................................................................................... 67
     5.1      Pension expenditure, contribution revenue and deficit of the PAYG system ...................... 67
     5.2      PAYG deficit and transition cost of multipillar reform.......................................................... 71
       5.2.1         Pension Institute’s (HZMO) accounting approach to identification of transition cost . 71
       5.2.2         Economic identification of the transition cost and fiscal effects of the reform ........... 72
6.     Discussion on key policy issues ..................................................................................................... 77
     6.1      Multipillar pension system vs. PAYG-only: is there a need for systemic change again? ...... 77

                                                                          6
     6.2       Indexation formula: effects of weight rotation and no-negative adjustment rules ............. 79
     6.3       A proposal for pension-type-choice-neutral social transfers paid as pension benefits ....... 83
     6.4       Second pillar pension payments and pension insurance companies ................................... 85
     6.5       Investment policy in mandatory pension funds ................................................................... 87
     6.6       Contribution of voluntary pensions to pension adequacy ................................................... 93
     6.7       Autoenrollment..................................................................................................................... 96
7.      Policy options to improve pension adequacy and fiscal sustainability....................................... 106
     7.1       The proposed policy measures – initial list ......................................................................... 106
        7.1.1         Measures for improvement of pension adequacy and/or fiscal sustainability in the
        PAYG          106
        7.1.2         Measures for improvement of pension adequacy in PAYG only ................................ 107
        7.1.3         Measures for improvement of the contribution to benefit ratio ............................... 108
        7.1.4         Measures for improvement of second pillar and/or combined pensions .................. 109
        7.1.5         Measures for improvement of voluntary pension savings ......................................... 109
     7.2       The impact of selected policy measures estimated by the TRR model .............................. 110
        7.2.1         Change in indexation/valorization formula ................................................................ 110
        7.2.2         Increase in minimum age for early retirement ........................................................... 113
        7.2.3         Increase in statutory retirement age .......................................................................... 113
        7.2.4         Increase in pension supplement and minimum pension ............................................ 115
        7.2.5         Pension-type-choice-neutral lump-sum social transfers paid as pension benefits .... 116
        7.2.6         Increase in the second pillar contribution rate ........................................................... 118
8.      Baseline (Status Quo) Projection of Croatian Pension System Using PROMIS MODEL .............. 120
9.      Simulation of policy measures .................................................................................................... 135
     9.1       Valorization and indexation policy...................................................................................... 136
     9.2       First Pillar reform framework.............................................................................................. 141
     9.3       Second pillar reform framework ......................................................................................... 144
     9.4       Later retirement and shorter early retirement window ..................................................... 147
     9.5 Policy packages and its adequacy and fiscal outcomes ............................................................ 153
10. Conclusion ................................................................................................................................... 162
Annexes ............................................................................................................................................... 164
     Annex 1: Pension Savings, Credit Ratings and Economic Development ......................................... 164
     Annex 2: Alternative determination of the basic pension factor ................................................... 166
     Annex 3: How many first pillar pensions in Croatia are earnings-related? .................................... 168
     Annex 4: Net pension benefits and net replacement rates projected by the TRR model .............. 170
     Annex 5: Description of the Croatia PROMIS microsimulation model ........................................... 175
References .......................................................................................................................................... 179

                                                                            7
List of Figures
Figure 2.1 Projected pension expenditures for Croatia in consecutive Aging Reports .......................... 19
Figure 2.2 Projected gross benefit ratio in public pension scheme in Croatia in consecutive Aging
Reports .................................................................................................................................................. 20
Figure 3.1 At risk of poverty or social exclusion (AROPE) among people aged 65+, 2015 and 2022
(percent)................................................................................................................................................ 25
Figure 3.2 At risk of poverty or social exclusion (AROPE) among people aged 18-64 and retired persons
aged 65+, Croatia and EU-27, 2015-2022 (percent) .............................................................................. 26
Figure 3.3 At risk of poverty or social exclusion (AROPE) among people aged 18-64 and retired persons
aged 65+, 2022 (percent) ...................................................................................................................... 26
Figure 3.4 Aggregate replacement ratio of the population aged 65-74 to population aged 50-59, 2010
and 2022 (percent) ................................................................................................................................ 27
Figure 3.5 Net benefit ratio, 2017-2023 ................................................................................................ 28
Figure 3.6 Gross and net theoretical replacement rates for wage earners with 50%, 100% and 200% of
the average wage .................................................................................................................................. 30
Figure 3.7 Gender pension gap – Croatia and EU in 2022 ..................................................................... 31
Figure 3.8 Female pension benefit in % of male pension benefit – first pillar pensions ....................... 31
Figure 3.9 At-risk-of-poverty rate of older people by sex – Croatia and EU in 2022 ............................. 32
Figure 3.10 Net replacement rate: Full PAYG pension for retirement in January 2024......................... 34
Figure 3.11 Personal accounts closed due to retirement with the combined pension as a proportion of
all accounts closed for retirement......................................................................................................... 37
Figure 3.12 Number of new contracts on second pillar pension by savings transferred to PIC, 2023 .. 37
Figure 3.13 Distribution of second pillar pensions paid in December 2023 .......................................... 38
Figure 3.14 Basic and overall PAYG old-age pensions in December 2023 (monthly, in EUR) ................ 39
Figure 3.15 Insurance period of pensioners with basic and overall PAYG old-age pensions (without
international service), as of December 2023......................................................................................... 39
Figure 3.16 Annual decrement for early retirement in Selected countries, 2020 or later .................... 42
Figure 3.17 Annual pension bonus for late retirement in selected countries, 2020 or later ................. 44
Figure 3.18 Average exit age from labor market, Croatia and the EU, 2022 ......................................... 44
Figure 3.19 Average age of new old-age pensioners (including early retirement, excluding
international service)............................................................................................................................. 45
Figure 3.20 Average contributory period of new earning-related public pensions in the EU, 2022...... 46
Figure 3.21 Average insurance period of new old-age pensioners (including early retirement,
excluding international service) ............................................................................................................ 46
Figure 3.22 Expected years in retirement in OECD and selected non-OECD countries in 2020 ............ 47
Figure 3.23 Average accrual rate in EU in 2022 ..................................................................................... 48
Figure 3.24 Support ratio in EU and Croatia, 2022 ................................................................................ 49
Figure 3.25 Net (theoretical) replacement rates, base case and 10 years after retiring, average earner,
2069, %.................................................................................................................................................. 51
Figure 4.1 Gross TRR for a male average wage earner in cases of old-age and early retirement
(M65/40 and M60/35) .......................................................................................................................... 56
Figure 4.2 P2/P1 for three generations of new old-age pensioners and different wage levels (M65/40)
 .............................................................................................................................................................. 58
Figure 4.3 Theoretical gross replacement rates for female average wage earner with flat wage profile
having two children in cases of old-age and early retirement (F65/40 and F60/35) ............................. 59


                                                                               8
Figure 4.4 Gross TRR for old-age pension of average wage earner (M 65/40) depending on wage
profile and investment behavior ........................................................................................................... 60
Figure 4.5 Gross TRR depending on real wage growth and real rate of pension fund returns ((M65/40,
average wage earner)............................................................................................................................ 61
Figure 4.6 Simulation of the P2/P1 ratio depending on real wage growth and real rate of pension fund
returns (M65/40, average wage earner) ............................................................................................... 62
Figure 5.1 Pension expenditure, contribution revenue and PAYG deficit in % of GDP .......................... 67
Figure 5.2 Pension expenditure, contributions, and balance of the public scheme (% of GDP) ............ 69
Figure 5.3 Budget transfers for covering PAYG deficit in % of GDP 2012-2022..................................... 72
Figure 5.4 Multipillar pension reform net fiscal cost (MPRNC) since inception of the reform in % of
GDP, with (MPRNC2) and without (MPRNC1) cost of accumulated portion of public debt .................. 75
Figure 6.1 Hypothetical evolution of the pension point value under current indexation rules applied to
quarterly, half-yearly and yearly adjustment frequency ....................................................................... 81
Figure 6.2 Pension point indexation, CPI inflation and wage growth (January 2019 =100) ................... 82
Figure 7.1 The impact of indexation rules on the gross TRR ............................................................... 111
Figure 7.2 Gross TRR with higher minimum age for early retirement (women, retirement with the first
conditions for early retirement) .......................................................................................................... 113
Figure 7.3 Gross TRR assuming rising statutory retirement age by one month each year .................. 114
Figure 7.4 Gross TRR after increase of the pension supplement to 30% and the minimum pension by
3% (M65/40, flat wage profile)............................................................................................................ 116
Figure 7.5 Gross TRR for alternative calculation of social transfers (W 60/35, two children) ............. 117
Figure 7.6 Combined-to-full PAYG pension ratio (P2/P1) for alternative calculation of social transfers,
2024 and 2034 .................................................................................................................................... 118
Figure 8.1 Croatia – Population Projection (2023-2065) ..................................................................... 120
Figure 8.2 Croatia – Demographic Structure Projection (2023-2065) ................................................. 121
Figure 8.3 Macroeconomic assumptions in Croatia PROMIS model 2023-2065 ................................. 122
Figure 8.4 Share of Contributors in Working Age Population - PROMIS model 2023-2065 ................ 122
Figure 8.5 Croatia – Pension Support Ratio, Baseline simulation (2023-2065) ................................... 123
Figure 8.6 Share of new old-age and disability pensioners without second pillar in total................... 124
Figure 8.7 Baseline replacement rates for new old age pensioners with and without benefit choice 125
Figure 8.8 Length of service of new old age pensioners by gender..................................................... 126
Figure 8.9 Baseline replacement rates for new old age pensioners who chose PAYG-only pensions and
combined pensions by gender ............................................................................................................ 126
Figure 8.10 Average baseline replacement rate for old-age, disability and survivor pension ............. 128
Figure 8.11 Share of new old-age pensioners choosing combined pension........................................ 129
Figure 8.12 Share of old age pension categories retiring at statutory retirement age choosing
combined pension ............................................................................................................................... 130
Figure 8.13 Combined pension beneficiaries – share in total pensioners 2023- 2065 ........................ 131
Figure 8.14 Baseline benefit ratio of combined pensioners vs. PAYG-only pensioners ....................... 132
Figure 8.15 Baseline PAYG expenditures and pension system deficit ................................................. 133
Figure 8.16 Second pillar payouts will remain a small share of future pension expenditures ............. 133
Figure 9.1 Simulation of 85%W+15%P valorization and indexation patterns – replacement rates..... 137
Figure 9.2 Simulation of 85%W+15%P valorization and indexation patterns – benefit ratios ............ 138
Figure 9.3 Simulation of 85%W+15%P valorization and indexation patterns – choice of P2, flow and
stock .................................................................................................................................................... 139
Figure 9.4 Simulation of 85%W+15%P valorization and indexation patterns – PAYG expenditures .... 140
Figure 9.5 Simulation of 85%W+15%P valorization and indexation patterns – PAYG gross and net
deficit .................................................................................................................................................. 140

                                                                              9
Figure 9.6 Simulation of higher supplement, minimum pension and choice-neutral transfer – average
replacement rates and benefit ratios .................................................................................................. 141
Figure 9.7 Simulation of higher supplement, minimum pension and choice-neutral transfer – choice of
combined pension of new pensioners and all pensioners ................................................................... 142
Figure 9.8 Simulation of higher supplement, minimum pension and choice-neutral transfer – PAYG
expenditures, gross and net deficit ..................................................................................................... 143
Figure 9.9 Simulation of higher second pillar contribution rate (7%) – choice of combined pension for
new pensioners and all pensioners ..................................................................................................... 145
Figure 9.10 Simulation of higher second pillar contribution rate (7%) – average replacement rates and
benefit ratios ....................................................................................................................................... 146
Figure 9.11 Simulation of higher second pillar contribution rate (7%) – total PAYG expenditures and
gross deficit ......................................................................................................................................... 146
Figure 9.12 Assumed increases in retirement age and early retirement age 2025-65 and increase in
life expectancy according to EU projections, RETAGE SLOW .............................................................. 149
Figure 9.13 Assumed increases in retirement age and early retirement age 2025-65, and increase in
life expectancy according to EU projections, RETAGE FAST ................................................................ 149
Figure 9.14 Retirement age increase, RETAGE SLOW and RETAGE FAST, support ratio ..................... 150
Figure 9.15 Retirement age increase, RETAGE SLOW and RETAGE FAST, length of service at retirement
 ............................................................................................................................................................ 151
Figure 9.16 Retirement age increase, RETAGE SLOW and RETAGE FAST, pension adequacy ............. 151
Figure 9.17 Retirement age increase, RETAGE SLOW and RETAGE FAST, expenditures and deficit .... 152
Figure 9.18 Retirement age increase, RETAGE SLOW and RETAGE FAST, choice of P2 ....................... 153
Figure 9.19 Policy packages SLOW, FAST – replacement rates and benefit ratios .............................. 154
Figure 9.20 Policy packages SLOW, FAST – Total PAYG expenditures ................................................. 156
Figure 9.21 Policy packages SLOW, FAST – GROSS PAYG deficit ......................................................... 158
Figure 9.22 Policy packages SLOW, FAST – NET PAYG deficit .............................................................. 159
Figure 9.23 Policy packages SLOW, FAST – share of old age pensioners choosing combined pension160
Figure 0.1 Sovereign ratings vs. total assets in private and funded pension plans in 2021 ................. 164
Figure 0.2 Sovereign ratings vs. GDP per capita at PPP in 2021 .......................................................... 165
Figure 0.3 Implicit contribution rate to finance total pension expenditure and implicit basic pension
factor ................................................................................................................................................... 166
Figure 0.4 Non-earnings related PAYG pensions (in % of total pensions) ........................................... 169
Figure 0.5 Net TRR for a male average wage earner in cases of old-age and early retirement (M65/40
and M60/35) ....................................................................................................................................... 170
Figure 0.6 P2/P1 in net terms for three generations of new old-age pensioners and different wage
levels (M65/40) ................................................................................................................................... 170
Figure 0.7 Theoretical net replacement rates for female average wage earner with flat wage profile
having two children in cases of old-age and early retirement (F65/40 and F60/35) ........................... 171
Figure 0.8 Net TRR for old-age pension of average wage earner (M 65/40) depending on wage profile
and investment behavior..................................................................................................................... 172
Figure 0.9 Net TRR depending on real wage growth and real rate of pension fund returns ((M65/40,
average wage earner).......................................................................................................................... 173
Figure 0.10 The impact of indexation rules on the net TRR ................................................................ 173



List of Tables
Table 3.1 Statutory retirement age and early retirement age in EU and Western Balkans ................... 41

                                                                              10
Table 3.2 Eligibility rules for Long Service Retirement in EU, 2023 ....................................................... 42
Table 3.3 Pension valorization and indexation rules in EU countries .................................................... 52
Table 3.4 Pension benefit of employed pensioners in EU 2023 ............................................................ 54
Table 4.1 Payouts from the Second Pillar .............................................................................................. 63
Table 6.1 Taxonomy of possible interactions between multipillar pension system and PAYG- in CEE EU
countries ............................................................................................................................................... 77
Table 6.2 MIREX* based net annualized returns for A, B and C portfolios ............................................ 88
Table 6.3 An Example of IAS .................................................................................................................. 89
Table 6.4 Illustrative calculation of auto-enrollment effects ............................................................... 103
Table 7.1 Gross replacement rates in case of increasing contribution rate for the second pillar to 7%
and decreasing contribution rate for the first pillar to 13% since January 1, 2025 (old-age pension with
40 years of service for average wage earner, men) ............................................................................ 119
Table 8.1 Comparison of macroeconomic assumptions PROMIS- AR24 ............................................. 123
Table 9.1 Individual Pension Policies and Policy Packages Considered and/or Simulated with Croatia
PROMIS Model .................................................................................................................................... 135
Table 9.2 Average annual differences to baseline by 5-year periods and decades in pension adequacy
– policy packages................................................................................................................................. 155
Table 9.3 Cumulative differences to baseline TOTAL PAYG EXPENDITURES – individual policies and
policy packages, % of GDP ................................................................................................................... 157
Table 9.4 Annual differences to baseline GROSS PAYG DEFICIT – individual policies and policy
packages, % of GDP ............................................................................................................................. 158
Table 9.5 Annual differences to baseline NET PAYG DEFICIT – individual policies and policy packages, %
of GDP ................................................................................................................................................. 160
Table 0.1 Determination of the net pension benefit for the full-PAYG pension and the combined
pension depending on the average value point (female average wage earner having two children in
case of early retirement F60/35), retirement at January 1, 2024 (in EUR) .......................................... 172
Table 0.2 Net replacement rates in case of increasing contribution rate for the second pillar to 7% and
decreasing contribution rate for the first pillar to 13% since January 1, 2025 (old-age pension with 40
years of service for average wage earner, men) ................................................................................. 174




                                                                            11
1. About the RAS task

The Ministry of Labor, Pension System, Family and Social Policy (MLPSFSP) has requested
assistance from the World Bank in preparing the Analysis of the long run adequacy and fiscal
sustainability of the Croatian pension system. The activity is envisaged in the National
Recovery and Resilience Plan (NRRP) 2021-26 and other strategic documents, supported by
the European Commission. The Bank’s support is being provided in the form of a
Reimbursable Advisory Services (RAS) signed on July 1, 2022.
Analysis of Croatian pension system was foreseen in the NRRP (C4.2) entitled "Increasing
pension adequacy through continued pension reform". The activity has commenced with the
establishment of the MLPSFSP's Working Group (WG) on the analysis of the situation of the
pension system in Q2 2022 tasked to assess the impacts of the Government’s 2019 pension
reform and identify options for future pension system improvements including raising the
pension adequacy. The WG is expected to yield conclusions and recommendations by the end
of 2024 and coordinate the public debate before the expected amendments to the Pension
Insurance Law in 4Q/2025.
The role of the World Bank is to facilitate the work of the Working Group by assessing Croatian
pension system's long run adequacy and fiscal sustainability, identifying policies to improve
it, and proposing policy recommendations. According to the Terms of Reference (ToR), the
activities of the World Bank's team envisaged: i) development of a microsimulation model of
Croatian pension system (CROATIA PROMIS) and ii) preparing the Report on "Analysis of the
Long-Run Pension Adequacy and Sustainability in the Croatian Pension System, with the focus
on Policy Options for its Improvement" to serve as a background document for MLPSFSP's
WG designing the policies to strengthen the Croatian pension system.
The World Bank’s engagement included i) gathering the agreed datasets from the client’s
pension administration agencies (HZMO and REGOS), ii) development of the microsimulation
model of Croatian pension system, iii) conducting a set of agreed policy scenario simulations,
and iv) drafting this Report. The list of policy scenarios and analyzed measures suggested by
the clients follow the policies announced in the NRRP stating pension adequacy as a primary
goal.
The Report is prepared by the World Bank team comprising the World Bank staff (Zoran
Anušić, Sergiy Biletskiy, Iva Tomić) and consultants covering all aspects of the pension system
(Ljiljana Marušić, Danijel Nestić, Velimir Šonje, Mitchell Wiener). The logistical preparation of
the Report was supported by Martina Vojković.




                                               12
2. Croatian pension system: current setting and overview of past
   changes

The current Croatian three-pillar pension system is a result of a comprehensive structural
reform enacted in 1998. Overarching objective of the reform was to bring the Croatian
pension system to a sustainable fiscal and social sustainability path, i.e., to reach a balance
between standard pension twin-peaks conflicting objectives of stable pension expenditures’
share in GDP providing rising pension adequacy despite ageing population.
The reform started with thorough changes within the public PAYG system in 1998. Pension
entitlement conditions were tightened, and a new earnings-related German style point
formula was introduced, complemented by elements of solidarity and social redistribution in
the form of service-related minimum and maximum pension. The PAYG first pillar covers the
risks of old age, disability (including work-related injury and diseases) and death of a family
income-earner. Employees in arduous and hazardous jobs are entitled to more favorable
pension entitlements, financed by employers’ contributions. The first pillar also provides
pensions determined by specific laws that stipulate more favorable retirement entitlements
for specific categories of population, where transfers from the state budget cover respective
pension expenditures.
The main objective of the 1998 Reform was to diversify sources of retirement income. The
mandatory fully funded defined contribution private pension scheme (second pillar) based on
individual pension accounts began functioning in 2002. Enrollment was obligatory for people
in employment at the age of 40 or less. Persons at 40 to 50 in 2002 could opt in on a voluntary
basis. All new entrants to the labor market since 2002 enroll the second pillar on a mandatory
basis. The second pillar design is substitutive. A part of first pillar contributions (5% of gross
wage) is diverted to the second pillar individual accounts, with total mandatory pension
contribution remaining unchanged at 20%. In exchange, the second pillar participants have
been entitled to a reduced first pillar benefit (basic pension) with added second pillar lifetime
benefit drawn from the individual account. The initial setting in 2002 was to start with a
quarter of contributions transferred to the second pillar and gradually raise the second pillar
rate to 10%. However, until today the second pillar contribution rate remained at 5%.
Voluntary third pillar with added government incentives for individuals and employers also
begun operations in 2002.
The second important objective of the multipillar reform was to link benefits tighter to
contributions. While the second pillar yields the proportionality by design, the first pillar’s
new point-formula was programmed to be accompanied by other rules strengthening
proportionality between benefits and contributions such as widening of the calculation period
to full career, equalizing the accrual rate for every year of the service period and by applying
the adequate penalties/bonuses in cases of early/late retirement.
Third, the bigger objective of the multipillar reform was intergenerational. The more the
current generations contribute to their own future pension benefits, the less expenditure
would need to be financed by future generations. Given the aging process, the substitutive
                                               13
second pillar was seen as one of the tools for more justful redistribution of expanding
intergenerational liabilities.
PAYG reform undertaken in 1998 resulted in consolidating the old system’s liabilities into a
new framework and stabilizing short and medium-term pension expenditures by slowing
down the inflow of new pensioners, but at the expense of pension adequacy. Shrinking
pension adequacy in terms of replacement rates (the ratio of first pension to last wage) and
benefit ratios (average pension benefits relative to average wage) resulted from the widening
of the calculation period from 10 best consecutive years to a full career and new below-wage
valorization and indexation formula. Curtailing the first pillar expenditures opened the fiscal
space for restoring and improving the long-run adequacy through a larger second pillar with
contributions set above 5%. However, the declining pension adequacy in the early phase of
the pension reform and emerging replacement rate differentials between subsequent
generations raised political concerns and triggered short-run compensatory policy reactions.
Policy reactions largely disregarded the principle of benefits to contributions proportionality
and the long-run objectives of the pension reform.
Most of the interventions over the past two decades altered the pension entitlement
criteria, parameters of the pension formula, pension valorization and indexation patterns,
as well as provisions about the combined two-pillar pension. Chronologically, they included:
-   Social insurance contributions have been levied also on income of persons engaged in
    temporary jobs, professional, and other freelance-type economic activities. The aim was
    to eliminate distortions in the labor market by equalizing taxation of all types of earnings.
    Until today, this initiative has not been completed. The pension contribution rate for
    salaried workers stands at 20% of the wage (15% for first and 5% for second pillar), while
    the contribution rate for “other income earners” (those working on intellectual property
    contracts or service contracts) stands at 10% of gross earnings (7.5% for first and 2.5% for
    second pillar). The pension formula is the same for both, implying that “other income
    earner” is likely to earn the same PAYG pension rights as a salaried worker with one-half
    of the pension contributions paid. Student contract implies only a rate of 5% for the first
    pillar. In addition, the contribution base for the self-employed, i.e., different professional
    and other freelance-type economic activities, depends on the type of activity, while the
    applied rate is the same as for salaried workers. For example, for small businesses and
    artisans it is 65% of the average wage in the previous year; for independent professional
    activities and athletes it is 110%; and for persons engaged in agriculture and forestry it is
    either 38% or 55%, depending on whether agriculture is performed as a registered family
    farm or a regular self-employment activity.
-   Pension supplements amounting up to 27% to post-1999 pensions were introduced in
    2007, aiming to restore the balance between pre-reform and post-reform pensioners by
    increasing the level of post-reform pensions. The imbalance stemmed mostly from
    widening the calculation period from 10 best consecutive years to a full career. The step
    of 3 service years (13 best, 16 best, to the whole career as of 2010) started in 2000 and
    led to erosion of the replacement rates by 1-2 percentage points per cohort. As a remedy,
    the Government decided to partly compensate new pensioners for the loss in

                                               14
    replacement rates by raising the point value / accrual rate for future PAYG beneficiaries.
    The measure was initially designated for beneficiaries insured only in the first PAYG pillar.
    Later, it was extended to all beneficiaries including those with combined pension from the
    first and the second pillar where supplement is applied only to the basic pension
    component from the first pillar.
-   After raising the statutory retirement age for old-age and early retirement, equalization
    of retirement age for women started in 2010, to be finalized in 2030. The minimum age
    for acquiring survivors’ pension has also been increased from 45 to 50 years, still lower
    than in most EU countries. Raising the retirement age aimed to adapt the pension system
    to population ageing and higher life expectancy, improve long-term sustainability of the
    pension system in parallel with the adequacy of pensions, and achieve gender equality.
-   Two additional types of pensions have been introduced in 2014: old-age pension for long-
    term insured persons that attained 60 years of age and 41 years of insurance period, and
    early retirement without application of pension decrement factors for persons being
    unemployed due to enterprise bankruptcy.1 The first option has been used extensively in
    recent years, while the second has been exercised by a small number of new pension
    beneficiaries. These pension benefits are not reduced in proportion to retirement before
    statutory retirement age, which is the case for workers with lower number of years of
    service. Moreover, after legislative changes in 2023, the benefit for long-service
    beneficiaries increases by 0.15% per month of deferred retirement after completing 60
    years of age, i.e., five years before reaching the statutory pension age. With such early
    retirement rule pension system further departed from proportionality principle of linking
    benefits to contributions paid. This is contrary to practice in most EU countries where
    decrement-free early retirement has been largely abolished or limited to exceptional long
    service cases.
-   Since 2014, the basic pension (the first pillar benefit paid to the two-pillar pension
    beneficiaries) became earnings-related except for its part falling under minimum and
    maximum pension regulation. Starting from 1999, basic pension was a combination of flat
    and earnings-related benefit, with low earners having higher accruals than the high
    earners. Since 2014, it is calculated by the same point formula as pensions realized in the
    first pillar only. However, basic pension is lower in comparison to full PAYG pension due
    to the basic pension factor of 0.75 which reflects the proportion between the first pillar
    contribution rate of 15%, paid by persons insured in both mandatory pillars, and the rate
    of 20%, paid by persons insured in the first pillar only. The basic pension factor is applied
    in calculation of basic pension for the period of payment of the second pillar contribution
    (i.e., since 2002). The basic pension beneficiaries are entitled to minimum pension, scaled
    down by the basic pension factor of 0.75 for the years of insurance completed since 2002.
-   Mandatory pension funds, initially designed as conservative bond funds with at least 70%
    of portfolio invested in government bonds, undergone several waves of liberalizations of
    investment rules with the most important change introduced in 2014 when lifecycle


1
  The insured person has the right to an early old-age pension due to the employer's bankruptcy when s/he
fulfills the regular pension entitlement conditions for an early old-age pension and spends a continuous period
of at least two years as a registered unemployed person following the bankruptcy.

                                                      15
    portfolios A (high risk), B (balanced risk) and C (low risk) were introduced. After initial
    period of allocation of new young entrants who behave passively by not choosing their
    mandatory pension fund to fund category B, since 2019 they are allocated to A. In parallel,
    several cycles of liberalization of investment limits were implemented in the second and
    the third pillar, which was largely associated with Croatia’s EU entry in 2013. In the last
    cycle of flexibilization of investment limits in 2024, pension insurance companies’ (PICs)
    which pay second pillar annuities were also included in the flexibilization, thus allowing
    more exposure to riskier assets and real estate.
-   The minimum pension per year of insurance period was set at 97% of the Actual pension
    (point) value (APV) in 1999. It was increased to 100% of APV in 2019, and to 103% of APV
    since January 2023. The pension supplement of 27% is not applied to minimum pension.
    For the time being, all persons having average career earnings below 81.1% (=1.03/1.27)
    of the average wage in Croatia end up as the minimum pension recipients.
-   Early retirement decrement has been altered on several occasions after having been set at
    0.3% per month in 1999. From January 2020 it stood at 0.2% per month, i.e., up to 12%
    lower benefit for up to five years of early retirement. The current decrement is lower than
    the actuarially neutral one and below the decrements in EU countries. It implies that
    pensioners retiring at older ages partly subsidize pensions of early-retirement pensioners.
    Actuarially low early retirement decrement and lax entitlement requirements for early
    retirement combined with other early retirement options have been resulting in growing
    incentive for retirement before reaching the statutory retirement age.
-   A bonus to deferred old-age retirement was introduced in 2010 in an amount of 0.15% per
    month. It was increased to 0.45% in January 2023 and applied up to 5 years after the
    statutory retirement age. Application of bonus on deferred old-age pension strengthens
    the relation between the contributions paid and cost of lifelong pensions, reflecting the
    shorter period of pension payment, thus lower cumulative expenditures for deferred old-
    age pensions compared to pensions realized at the younger ages. The factor of 0.45% per
    month is close to actuarially neutral level and is like bonuses applied in other EU countries,
    however the entitlement to the deferred retirement bonus is enabled only after minimum
    35 years of insurance period has been completed. In other EU countries there is no such
    restriction and bonus on old age pension depends on the retirement age only, in line with
    principles of actuarial mathematics.
-   Since 2014, pensions are indexed by a combination of growth of CPI and average gross
    wage according to rotational formula where weight of 70% is attached to the higher of
    these two rates; if the indexation rate is negative, pensions are not indexed. In the period
    1999-2013, PAYG pensions were indexed twice a year by a 50:50 combination of growth
    of average gross wage and CPI (the so-called Swiss formula). The current rotational
    formula results in indexation at rates that are higher than with the fixed proportion
    formula (see Section 6.2). Since 2019 this effect is augmented by the fact that indexation
    of pensions by rotational formula is performed twice a year with consecutively calculated
    semestral indexation rates.
-   Six months of insurance period per child was added for calculation of mother’s pensions
    as of 2019. This measure contributed to a lower gender pension gap.


                                               16
-   The pension supplement of 27% on the basic pension of multi-pillar pensioners was
    introduced in 2019. Missing supplement contributed to the majority of multipillar
    participants returning to the first pillar through the opt-out process described below. In
    2019 the supplement was extended to the basic pension of the multipillar participants,
    amounting up to 27% for the part of the first pillar pension based on the insurance period
    realized until 2001. However, the method of calculation implied adjustment of the basic
    pension by 20.25% (=0.75 x 27%) although the basic pension has been already reduced by
    25% due to share of contribution paid to the second pillar (5/20). Although the extended
    supplement to basic pension narrowed the pension benefit gap between the benefits of
    PAYG-only and multi-pillar beneficiaries, the discriminatory double-reduction of the basic
    pension supplement for the second pillar participants continued to distort the
    proportionality feature of the basic pension formula. This has been reverted as of 1
    January 2024, when the rate of supplement for basic pension was equalized to 27%
    multiplying the adjusted basic pension.
-   The 2019 reform extended the opt-out from the second pillar at retirement to all second
    pillar members. In 2011 such an option was given to voluntary members only (those
    between 40 and 50 years in 2002). It allowed the second pillar members to choose
    between a pension from the first pillar only in exchange for handing over the second pillar
    savings to the government budget. The choice of the better-of pension, based on the
    presentation of the first pensions paid upon retirement, was meant to be an insurance for
    members in the second pillar in the periods of low returns in mandatory pension funds
    and protection of low-income earners. However, comparison of the first pensions in the
    case of opt-back to the first pillar vs. combined pension may not provide appropriate
    information. It does not show the difference between the present values of two pension
    streams, which would be more appropriate given different indexation patterns. Since its
    introduction in 2019, the opt-out from combined pension based on the first pension
    motivated the majority of second pillar pension fund members to choose the first pillar
    PAYG-only pension with the state having absorbed their pension savings into the state
    budget.
-   2019 regulatory reform introduced an option to withdraw 15% in cash from the second
    pillar pension savings in the moment of retirement, applicable only to retirees with basic
    pension from the first pillar 15% higher than the minimum pension. Withdrawal option is
    not available for persons who opt-back to full payment of pension from the first pillar.
    Cash withdrawal option was extended to 20% in 2024 and cash withdrawal condition was
    relaxed to include second pillar members with basic pension higher than the minimum
    pension (no more excess 15% requirement).
-   Since 2014 old-age pensioners working up to half of their regular working time are eligible
    to retain pension benefits in full amount. In 2019 this option was extended to early
    retirement beneficiaries (including long service old-age pensioners); and in the year 2021
    also to survivor’s pensioners.2 In addition, part-time working pensioners were allowed to
    retain minimum pension instead of receiving lower pension based on their earnings. This


2
 Since January 2024 the eligibility to part-time work without reducing or suspending the disability pension are
extended to Homeland War Veterans.

                                                      17
    is different from the practice before 2019 when part-time workers were entitled only to
    pension based on their earnings while working instead of a more favorable minimum
    pension. Not surprisingly, new unconstrained rules for working pensioners have boosted
    their number from 5 thousand at the end of 2018 to almost 30 thousand in 2023. It also
    intensified early retirement. Although the possibility of cumulating pension with earnings
    prolongs the work activity of older people, it may also encourage workers to retire earlier.
    Having this in mind, many countries are reluctant to extend the possibility of cumulating
    pensions with earnings to all categories of pensioners and often apply some sort of ceiling
    on earnings and/or pensions of working pensioners.
-   As of 2023, survivors’ pensions have been increased by 10%, from 70% to 77% of the
    pension of deceased beneficiary for one survivor, up to 110% for four or more of survivors’
    pension beneficiaries and a new pension benefit is introduced in the form of a part of
    survivor’s pension that can be combined with the own old age, early retirement or
    disability pension, if beneficiary has at least 65 years of age. The “inheritable” part of
    survivor’s pension amounts to 27% of the survivors’ pension assigned to one person ,
    subject to a ceiling of 80 APVs on a sum of own old-age or disability pension plus the
    “inheritable” part of survivor’s pension. The part of survivor’s pension targets mostly the
    widows in older age brackets as they prevail among the older population. Enabling the
    individuals to “inherit” a share of spouse’s or partner’s pension in addition to own’s
    without own pension reduced further defies the proportionality principles of linking
    benefits to contributions. For the same contributions, a married person may get a higher
    benefit than a single. For comparison, the joint annuity in the second pillar insures against
    the risk of partner’s death by reducing the initial pension by actuarially determined
    amount.
-   The disconnect between the pension benefits and contributions paid is widening in 2024.
    The latest amendments to the Contributions Law, effective as of January 2024, reduce
    PAYG pension contributions for insured individuals with lowest salaries. Consequently,
    lower net wages increased proportionally while future pension benefits of such
    contributors stay the same. Effectively, the government endorsed further weakening of
    the link between pensions and individual contributions paid. The second pillar
    contribution is regularly paid since pension in the second pillar is tightly linked to the
    contributions paid.
-   The obligatory shift to the second pillar pension fund category C was abolished in 2024
    and transitions between funds category A and B were made more flexible subject to
    individual choice. Until 2024, second pillar pension accounts of members at the age of 60
    (5 years before statutory retirement age) were shifted to C portfolios by Law. It led to
    significant deterioration of the value of pension savings of persons who were about to
    retire due to rising interest rates and falling bond prices during bond conundrum in 2022-
    2023. It also contributed to greater preference for opt-back to the first pillar in the payout
    stage in 2022 and 2023.
-   In January 2024, the Law on Veterans of the Homeland War and Their Family Members
    was amended in line with the amendments to the Pension Insurance Act on the increase
    of survivors’ pensions and the payment of minimum pension to employed pensioners. Also,
    reduction in war veteran’s pensions by 10%, where they cannot fall below 3,500kn (EUR

                                               18
    464.53) that was introduced in 2010 was abolished partially in July 2023 and completely
    in January 2024.
Frequent interventions in the Croatian pension system since the 1998-2002 reform have
altered the pension system’s rules, incentives, and outlook, while raising the expected
pension costs future generations would have to pay. Figures 2.1 and 2.2 show projections
from the four latest Aging Reports (AR15, AR18, AR21 and AR24). The projected decline in
pension adequacy and lower costs in AR15 has been gradually reversed in recent projections
towards a more stable adequacy at a higher fiscal cost. With these policies, the first pillar has
been gradually squeezing out the second pillar. The time has come to reassess the pension
reform objectives from social, fiscal, and intergenerational points of view. This document
suggests how a higher adequacy of pensions could be attained without disturbing the fiscal
and social fiber of society.
Figure 2.1 Projected pension expenditures for Croatia in consecutive Aging Reports




Source: Ageing Reports 2015, 2018, 2021, and 2024.




                                                     19
Figure 2.2 Projected gross benefit ratio in public pension scheme in Croatia in consecutive Aging Reports




Source: Ageing Reports 2015, 2018, 2021, and 2024.




                                                     20
3. Pension adequacy in Croatia – definition, performance, and
   determinants

3.1 Definition of pension adequacy


Adequacy of retirement income or pension adequacy usually implies maintaining the
income of persons for the duration of their retirement and preventing old-age poverty.3
However, the definition is not straightforward. Although the adequacy of pensions is
intuitively, or at least approximately, a clear term that evokes positive association related to
one’s satisfaction with the amount of pension income, the attempt to precisely define it is
related to deep methodological problems.4 On the one hand, an adequate pension can
indicate protection against poverty in old age. On the other hand, adequate pension income
may entail largely enabling retention of earlier way of life. Different viewpoints reflect the
historical development of pension systems which were initially established as systems for
protection from poverty. With time, views gradually evolved towards systems of maintaining
standards of living in old age. Also, policymakers look at adequacy from an overall public
policy point of view, which is a different perspective compared to what individuals would take
to assess the adequacy of their own retirement income.5 Generally, pension adequacy is a
subjective category for each person, while socially desirable adequacy includes issues of
financial sustainability and intergenerational fairness.
Historically, the ILO Convention 102, Social Security Minimum Standards (1952), was
the first and remains one of the most frequently quoted references of adequate pension a
public pension system should provide. The Convention stipulates that for earnings-related
pensions the minimum benefits should be provided at least to those with earnings lower than
average earning levels, replacing at least 40% of previous earnings. The less frequently
referenced ILO Convention 128 on Invalidity, Old age and Survivors benefits sets the minimum
replacement rate at 45%. The ILO conventions are not explicit if the minimum is set in gross
or net terms. However, the contents of the conventions suggest the emphasis on effective
replacement, i.e., net replacement.
Following different definitions of pension adequacy, different indicators are used to assess
current and future adequacy of retirement income. According to OECD,6 there is no universal
understanding of pension adequacy, its measurement, and assessment of retirement
incomes. Still, indicators commonly used to assess retirement income adequacy are
retirement income or pension replacement rates.

The replacement rate indicates the extent to which the pension system enables a person to
preserve her standard of living when she goes from employment to retirement, i.e., it

3
   European Commission, Directorate-General for Employment, Social Affairs and Inclusion, 2021 Pension
Adequacy Report – Current and future income adequacy in old age in the EU.
4
  Nestić and Tomić (2012).
5
  OECD (2020).
6
  Ibid.

                                                 21
indicates the degree of replacement of the former main source of income (salary) with a
new main source of income (pension). It is calculated as ratio of pension to salary, with
numerous variants depending on the definition of pension and salary. For example, wages
and pensions can be measured before or after tax, so the gross (before tax) pension
replacement rate differs from the net (after tax) pension replacement rate. Similarly, the
replacement rate can also differ between the own or economy-wide, or the period
considered, etc. For individual pension replacement rates, usually the so-called theoretical or
typical pension, or hypothetical individual replacement rates are used.7 The theoretical case
for a hypothetical individual is considered for comparison of her retirement and pre-
retirement income. A prime example of the theoretical case is a worker that earns an average
wage during the entire working life and retires at statutory retirement age after 40 years of
service. This example is frequently used in this report.

Besides individual, there are also aggregate measures of retirement income that use
different economic aggregates to compare retirement and pre-retirement income.
Aggregate measures of retirement income, sometimes called quasi-replacement rates,8 rely
on economic aggregates to compare income in retirement to income before retirement. In
this case, the aggregate replacement rate compares the replacement rates of two
generations - one representing a generation before retirement, and one after retirement
(e.g., median individual gross pension income in early years of retirement (people aged 65-
74) to the median individual gross earnings of late career workers (people aged 50-59).

The replacement rate should be distinguished from the benefit ratio. Benefit ratio is a
measure of the pension system’s adequacy at a macro level because it compares average
pensions to average earnings in the economy. Finally, gross average replacement rate
compares the average first pension of all those who retire in the given year to the (economy-
wide) average wage of people at the point of retirement.


3.2 Pension adequacy in EU policy documents

There is a growing recognition of the need for pension adequacy in the EU Member States.
Income replacement capacity of pension benefits is expected to decrease relative to wages in
most EU Member States reflecting current policies and population ageing unless labor market
and pension systems’ reforms are implemented. The EU institutions encourage increases in
the statutory retirement age while advocating that restrictions of early retirement should be
gradual and known in advance so that people can plan for their professional life and
retirement.9 Additionally, retirement of workers with sufficiently long careers should be


7
  For example, the Social Protection Committee (SPC) and the European Commission (DG EMPL) have developed
a set of adequacy indicators based on the individual replacement rates of a hypothetical retiree with strictly
defined characteristics. Such indicators are called theoretical replacement rates. They assess the current state
of pensions and prepare projections for the next 40 years, with the standard assumption of unchanged pension
and tax regulations in the future and with selected assumptions about trends in the economy and the labor
market.
8
  OECD (2020).
9
  European Commission, Economic Policy Committee - Social Protection Committee, 2020, Joint Paper on
Pensions 2019.

                                                      22
addressed, ensuring fair pensions for workers entering the labor market early. The right to
work beyond the statutory retirement age and the flexible old age retirement pathways
should also be encouraged by combining part-time work and part-time retirement which
might help alleviate shrinking of working age population, build additional pension rights, and
ease the transition from work to retirement.

The EU Pension Adequacy Report, published every three years, provides an overview of the
current and future adequacy of old-age incomes in the EU Member States in three
dimensions: (i) poverty prevention, (ii) income maintenance, and (iii) retirement duration .
Principle 15 of the European Pillar of Social Rights10 promotes the right to old age income and
pensions. Every three years, the Ageing Report (AR) and the Pension Adequacy Report (PAR)
provide an in-depth analysis of the sustainability of ageing-related expenditure and the
adequacy of old-age income. Ensuring both the adequacy of pensions and financial
sustainability of pension systems over the long term are interdependent issues. The scope of
the two reports differs: the AR focuses on macrolevel projections of public pensions, while
PAR analyses pension adequacy for various population groups, covering both public and non-
public schemes. Important aspects of both reports are complementary and consistency as
they use the same assumptions for future demographic and economic developments.11

Additionally, Green paper on ageing (2021) aimed to launch a broad policy debate on ageing
to discuss options on how to anticipate and respond to the challenges and opportunities,
considering the UN 2030 Agenda for Sustainable Development and UN Decade for Healthy
Ageing. European Commission’s Green paper on ageing12 examines the impact of ageing and
policy responses to provide adequate, safe, and sustainable pensions. The document is based
on a life-cycle approach that analyses the universal impact of ageing, focusing on both
personal and wider societal implications. These include everything from lifelong learning and
healthy lifestyles, to how to fund adequate pensions. In the absence of further reforms, more
pensioners, and fewer people of working age, are likely to lead to higher contribution rates
and lower pensions relative to wages to ensure sustainability of public finances. Pension
systems could support longer working lives by adjusting retirement ages or benefits to reflect
higher life expectancy. Reforms should consider the redistribution and fairness of pension
systems and the phasing-out of preferential pension schemes.

The 2021 edition of the Pension Adequacy Report relies heavily on the European Pillar of
Social Rights Action Plan from 2021. The European Pillar of Social Rights sets out principles
that relate directly or indirectly to the areas impacted by ageing such as old age income and
pensions, long-term care, health care, inclusion of people with disabilities, social protection,
work-life balance and education, training, and life-long learning. The European Pillar of Social
Rights Action Plan,13 published in 2021, covers the implementation and materialization of the
principles and rights contained in the European Pillar of Social Rights. The plan set headline
targets to be achieved by 2030, including employment (at least 78% of the population aged

10
   https://ec.europa.eu/social/main.jsp?catId=1606&langId=en
11
    European Commission, Directorate-General for Employment, Social Affairs and Inclusion, 2021 Pension
Adequacy Report – Current and future income adequacy in old age in the EU.
12
   European Commission - Secretariat-General (European Commission), 2021, Green Paper on Ageing: Fostering
solidarity and responsibility between generations.
13
   https://ec.europa.eu/social/main.jsp?catId=1607&langId=en

                                                   23
20 to 64 should be in employment), skills, social protection (number of people at risk of
poverty or social exclusion should be reduced by at least 15 million), and others. To retain
more people in the 55-64 age group active in the labor market, useful tools might involve
postponing retirements, improving working conditions for older workers (i.e., through better
digital connectivity), or subsidizing companies hiring older workers and encouraging senior
entrepreneurship.

The Pension Adequacy Report 2021 (PAR 2021) analyzes the adequacy of current and future
pensions in the EU Member States by assessing how national pension systems help
maintain the income of men and women for the duration of their retirement and prevent
old-age poverty. PAR’s focus is on old age and survivors’ pensions in providing adequate old-
age income and the minimum income provisions for older people. The main conclusions from
the PAR 2021 are the following:
       • After a decade of improvement, no further progress has been made recently to
          reduce the risk of poverty or social exclusion for older people in the EU.
       • Maintaining adequate living standards throughout retirement remains a
          challenge, particularly for women.
       • Income inequalities among older people persist, though pension and tax policies
          can help mitigate them.
       • Future careers will need to be longer to maintain adequate pensions.
       • Pension systems evolve amid changing economy and labor market, and
          maintaining their adequacy may require reconsidering financing sources.

3.3 Pension adequacy in Croatia


      3.3.1 Poverty risk in old age and retirement in Croatia


While the old-age poverty trend has been similar to other EU countries, Croatia is among
the Member States with the highest at-risk-of poverty and social exclusion (AROPE) rates
among elderly population (Figure 3.1).14 In 2015, the AROPE rate in Croatia was 29.6 percent,
and it increased to 33.5 percent in 2022. There are five countries in the EU (Bulgaria, Estonia,
Latvia, Lithuania, and Romania) with higher AROPE rates among elderly population than
Croatia. The number of elderly persons who are at risk of poverty and social exclusion in
Croatia increased from 228,000 in 2015 to 283,000 in 2022, which represents an increase of
almost 25 percent. The poverty risk for 65+ population in many countries in the EU, including
the EU average, grew at a similar rate as in Croatia, which points to the fact that old age
poverty is a systemic issue that must be addressed at the EU level as it reflects growing
number of old-age citizens.


14
  The at-risk-of-poverty or social exclusion rate (AROPE rate) is an indicator that measures the proportion of the
population that is either at risk of poverty, or severely materially and socially deprived or living in a household
with a very low work intensity. The definition for each of these three statuses can be found on Eurostat web
page                                                                      https://ec.europa.eu/eurostat/statistics-
explained/index.php?title=Glossary:At_risk_of_poverty_or_social_exclusion_(AROPE)


                                                        24
Figure 3.1 At risk of poverty or social exclusion (AROPE) among people aged 65+, 2015 and 2022
(percent)




Note: Sorted by the AROPE rates in 2022.
Source: Eurostat (online data: ilc_peps01n)



The old-age poverty and social exclusion risk in Croatia has been high and rising, unlike for
the working-age population. This is equally true for the general 65+ population and retired
population aged 65+. In 2022, the AROPE rate for persons 65+ regardless of labor market
status was 33.5 percent, while for retired persons 65+ it was 32% and increasing. In contrast,
poverty, and social exclusion risk for the working age population (18-64 age group) is much
lower and has been falling in both the EU and Croatia (Figures 3.2 and 3.3). Diverging trends
are especially concerning in Croatia. The AROPE rate for the working-age population in Croatia
is among the lowest in the EU and it is falling faster than the average rate in the EU-27. The
same rate for retired elderly people aged 65+ is growing faster than in the EU by some 27
percentage points.15 That worrisome trend calls for decisive policy measures to reverse it.




15
  The same findings are related to standard poverty measures, such are income poverty rate (i.e. at-risk-of-
poverty (AROP) rate), relative median at risk of poverty gap, or severe material and social deprivation.

                                                    25
Figure 3.2 At risk of poverty or social exclusion (AROPE) among people aged 18-64 and retired persons
aged 65+, Croatia and EU-27, 2015-2022 (percent)

 33

 31

 29

 27

 25

 23

 21

 19

 17

 15
                       2015   2016         2017         2018          2019         2020             2021       2022

                                          HR Retired 65+                   EU Retired 65+
                                          HR Population 18-64              EU Population 18-64

Source: Eurostat (online data: ilc_peps02n)



Figure 3.3 At risk of poverty or social exclusion (AROPE) among people aged 18-64 and retired persons
aged 65+, 2022 (percent)

                  65
                                                                      EE

                  55
                                                                              LV
                                                                                               BG
                  45                                                  LT
    Retired 65+




                  35                                                                                          RO
                                                       HR
                                                       MT
                  25                                             IE
                                                       CY
                                            SI                       DE
                                     CZ                       HU PT EU
                                                           NL BE
                                                          PL                              ES
                  15                                        DK SE
                                                           FI
                                                           AT         FR                IT          EL
                                                       SK         LU
                   5
                        5        10               15             20                25                    30           35
                                                                18-64

Note: The AROPE rate is the share of the total population which is at risk of poverty or social exclusion. The reference (solid
orange) line represents a 45-degree line.
Source: Eurostat (online data: ilc_peps02n)


                  3.3.2 Income replacement of Croatian pensions


Adequacy of pensions in Croatia, measured by the replacement rate, is one of the lowest in
the EU, although it increased somewhat in the last decade. The near-retirement aggregate


                                                                       26
replacement rate (ARR)16, defined as the gross median pension income of the population aged
65–74 years relative to gross median individual earnings from work of the population aged
50–59 years, excluding other social benefits, indicates a considerable heterogeneity across
the EU-27. The highest replacement rate in 2022 was in Luxembourg (90 percent) and the
lowest of around 33 percent were in Lithuania and Ireland (Figure 3.4). On average, EU
pensions replaced almost 60 percent of the income received by people who are close to
retirement. In Croatia, the aggregate replacement rate has been substantially lower: in 2010
it was the lowest in the EU at 32 percent, which was 21 percentage points below the EU-27
average; from 2010 to 2022, the replacement rate in the EU-27 grew by 5 percentage points
and it increased slightly faster in Croatia lifting it from the EU bottom to the fourth-lowest
position in 2022.
Figure 3.4 Aggregate replacement ratio of the population aged 65-74 to population aged 50-59, 2010
and 2022 (percent)

 90
 80
 70
 60
 50
 40
 30
 20
 10
     0
                EU-27
                 Latvia



            Denmark




               Austria
              Finland

            Romania
             Hungary
               Ireland




               France




                  Italy
             Bulgaria
              Croatia
               Cyprus

             Slovenia




         Netherlands
            Germany




               Poland

               Greece

                 Spain
            Lithuania




              Estonia



             Belgium




                 Malta




             Slovakia




         Luxembourg
             Sweden
              Czechia




             Portugal




                                               2010     2022

Note: The aggregate replacement ratio is gross median individual pension income of the population aged 65–74 relative to
gross median individual earnings from work of the population aged 50–59, excluding other social benefits.
Source: Eurostat (online data: ilc_pnp3).



Net benefit ratio, defined as the average net pension benefit as proportion of the average
net wage has been on a downward trend for a longer period, but the reversal happened in
2023. This ratio is calculated for three different beneficiaries’ categories in the period 2017-
2023 (Figure 3.5). Between 2019 and 2021, the decline was quite intensive in all three
categories. The most comprehensive indicator that covers all pensions shows that the average
net pension was at 39.5% of the average net wage in December 2022 compared to 41.5% in
December 2019. Ten years ago, in December 2012, the net benefit ratio for all pensions was
43.4%. This benefit ratio which covers all public pensions in payment is the most often cited
indicator of the adequacy of the pension system in Croatia, however it should be used with

16
  The ARR represents the extent to which pensions in initial retirement years replace earnings obtained in late
years of working and contributing life.

                                                          27
caution as it is highly misleading in assessment of the social protection provided by the
pension system. Low net benefit ratio is partly due to inclusion of smaller pension benefits
based on incomplete careers which only partly happened in Croatia, while the other part of
pension related to work in other countries is not included in the calculation. Such benefits are
much lower than benefits in cases where entire career was spent in Croatia. In September
2023, the average old-age pension by general legislation without partial pensions due to
international service was 14% higher than the comparable average of all benefits, standing at
41.7% (Figure 3.5). If only pensions under general pension legislation and without those
approved to pensioners with international careers17 are considered, the net benefit ratio was
42.6% in 2019 and 41.1% in 2021. If focus is on old-age pensions only, the net benefit ratio
was 46.5% in 2019 and 44.6% in 2021 (Figure 3.5).
Figure 3.5 Net benefit ratio, 2017-2023
             48
             47
             46
             45
             44
      In %




             43
             42
             41
             40
             39
                  2017       2018          2019         2020         2021          2022         2023
                             Total pensions (December)
                             All pensions by general legislation - w/out int'l service*
                             Old-age pensions by general legislation - w/out int'l service*

Notes: Net benefit ratio is calculated as the average net pension benefit in percent of the average net wage. * Average
pension benefit and average wage are year averages.
Source: HZMO.



A part of the decline in the net benefit ratio in 2019-2021 is due to factors outside the
pension system. Firstly, in 2020, the Croatian Bureau of Statistics has changed the definition
of the average wage to reflect the full-time equivalent wage. That resulted in a one-off
increase in the average gross wage of around 2.6% and the average net wage of around 2%.
The methodological change that impacted the official average wage level was not reflected in
valorization/indexation of pensions as official wage growth was based on comparable wage
coverage and without methodological break. Only statistics changed – there was no actual
change of pensions in relation to wages. Secondly, in 2021 there was a significant change in
income taxation. Personal income tax rates were reduced from 24% to 20%, and 36% to 30%.
Lower income tax burden helped to increase net wages more than net pensions. In 2022, a
mild decline in the benefit ratio happened mostly due to valorization/indexation that was
below wage growth.


17
     If such career is reported to HZMO by countries with which Croatia has social protection agreements.

                                                            28
The increase in the net benefit ratio observed in 2023 is mainly due to policy measures. The
reform of survivors’ pensions led to their increase by 10% since the beginning of 2023. In
addition, the minimum pension increased by 3% in 2023. For pensions realized according to
the special legislation under more favorable conditions18, the reduction of 10% introduced
some 10 years ago has been gradually abolished leading to an increase in their benefits. The
additional health insurance contribution of 3% levied on pensions higher than the average
net wage was abolished in April 2023. Policy changes contributed to an additional increase in
the average pension on top of regular pension indexation, which helped increase the benefit
ratio in 2023.
Tax and pension policy measures in the last several years have helped to protect the relative
value of new pensions, especially for low-wage earners. The concept of the theoretical gross
and net replacement rate will help to isolate the impact of policy measures on pensions for
the low-wage, average-wage, and high-wage earners. The replacement rate is calculated for
hypothetical careers by applying tax and benefit rules valid in for a given year. Figure 3.6
shows the results. First, there is a considerable difference between the gross and net
replacement rate regardless of wage class (low, average, high). It confirms more beneficial
tax treatment of pensions compared to wages. One of the reasons is the absence of social
security payments from pensions.19 The other reason is halved tax obligation on pensions
compared to wages and other incomes from work. The evolution of the net replacement rates
over time suggests that policy measures compensate for a part of unfavorable trends
observed in gross rates. For average- and high-wage earners, the trend of mild reduction in
the net replacement rate stopped in 2022 and reversed in 2023. For low-wage earners (in our
example, 50% of the gross average wage), increase in the gross replacement rate in 2023 was
mainly due to an increase in the minimum pension point value by 3%. For high-wage earners
(200% of the gross average wage), the increase in the net replacement rate in 2023 was due
to abolishment of the additional health contribution of 3% of gross pensions for pensions
above the average net wage. The effects of policy measures can also be seen in other years:
in 2019 there was an increase in the minimum pension point value that resulted in higher
gross and net replacement rates for low-wage earners. Earlier decline in net replacement
rates for average- and high-wage earners in 2017 was caused by changes in income taxation
of wages: tax burden was reduced by increased personal allowance and reduced tax rates,
which increased net wages by a higher proportion than net pensions. However, the reform
had no influence on the replacement rates of low-wage earners because they didn’t pay any
income tax even before 2017.




18
   Pensions acquired/determined according to special legislation, like pensions for army, police, Homeland war
veterans, firefighters, people suffering from asbestos, ex-YU army and officials, and WWII veterans.
19
   Up to April 2023, health contribution of 3% was paid out of pensions above the average net wage, but that
was much lower rate than pension contributions of 20% paid out of the wage income.

                                                     29
Figure 3.6 Gross and net theoretical replacement rates for wage earners with 50%, 100% and 200% of
the average wage
 85.0%                                                                    65.0%

 80.0%                                                                    60.0%

 75.0%                                                                    55.0%

 70.0%                                                                    50.0%

 65.0%                                                                    45.0%

 60.0%                                                                    40.0%

 55.0%                                                                    35.0%
          2016   2017     2018   2019    2020   2021   2022   2023                2016   2017     2018 2019 2020 2021 2022            2023
                                                                                                Gross RR -100% of the average wage
                        Gross RR -50% of the average wage                                       Net RR - 100% of the average wage
                                                                                                Gross RR - 200% of the average wage
                        Net RR - 50% of the average wage                                        Net RR - 200% of the average wage

Note: The gross (net) theoretical replacement rate is the ratio between the first gross (net) old-age pension benefit (full PAYG
benefit paid in the second half of the year) and the last gross (net) wage (paid on average in the first half of the year) for a
man retired at the age of 65 years after 40 years of service. Net amounts are calculated by applying relevant tax rules for a
person without dependent household members living in a place with a surtax rate of 10%.
Source: WB Team.


         3.3.3 Gender pension gap

Gender differences in pensions are smaller in Croatia than in the EU on average. Gender
pension gap, defined as a percentage difference between average pension of men and
women aged over 65, was 26% in the EU in 2022, while in Croatia it was 25% (Figure 3.7).
There are 15 EU countries with a larger gap and 11 with a lower gap.

HZMO data point to an even smaller gender gap of around 10% for all pensions approved
by the general legislation in 2023, and to gradually declining trend in the last decade. As
shown in Figure 3.8, in 2023 women received pension benefit which was on average for all
pensions 89.4% of men’s benefit.20 In 2012, that proportion was lower, 84.1%, and since then
it has increased gradually. The trend is the same among the largest part of all pensions in
payment, for old-age pension benefits without pensioners with international service,
although the gender gap is somewhat larger than among overall pensions. The gender
pension gap is related to gender wage gap as the average wage is one of two key factors in
the pension benefit formula. In the last couple of years gender wage gap is hovering around
7% (women earn 93% that of men, on average) as reported by the Croatian Bureau of
Statistics. Such wage gap is relatively low by international standards. Some ten years ago, the
gap was around 10%. However, pension formula considers earnings over entire career,
meaning that lower wage gap needs time for transposing into lower gender pension gap. Also,
average pension gap is calculated for all pensions in payment, not only new ones, reflecting
wage disparities from decades ago.


20
   A higher gender gap shown previously in Eurostat’s international comparisons is mostly due to broader
coverage of pensions in their survey-based data. Eurostat’s data includes pensions under special regulations,
including war veteran pensions, as well as pensions for retired police and army personnel. Pension benefits
approved under special regulation are higher than those approved by general legislation and more often taken
by men. Therefore, their inclusion results in a higher gender pension gap.

                                                                     30
Gender pension gap reduced significantly in 2023 due to policy push by an increase in
minimum pensions and reform of survivors’ pensions. More women than men benefited
from these two policy measures.

Despite declining gender pension gap in Croatia, gender poverty differences remained
sizable and among the highest in the EU (Figure 3.9). The difference in at-risk-of-poverty rate
was about 10 pp (36.5% for females vs. 26.6% for males), while in the EU it was somewhat
less than 6pp (14.2% vs. 19.8%). Although a part of the poverty gap might be due to larger
proportion of women living alone in old age as they live longer on average than men, but a
part is for sure caused by lower current incomes including pensions.

Figure 3.7 Gender pension gap – Croatia and EU in 2022




Note: % difference between pension of men and women aged over 65.
Source: Eurostat (ilc_pnp13)



Figure 3.8 Female pension benefit in % of male pension benefit – first pillar pensions
  92.0%
  90.0%
  88.0%                                                                                             89.4%
  86.0%
  84.0%                                                       85.1%       85.4%        85.6%
                                               84.8%                                                   85.0%
          84.1%                   84.4%
  82.0%
                                                                                            82.1%
  80.0%                                                                        81.5%
                                                  80.1%           80.7%
  78.0%                                79.1%
  76.0%
  74.0%
  72.0%
             2012                    2018        2019           2020         2021         2022        2023

                         Total PAYG pensions           Total old-age pensions w/out int'l service



                                                         31
Source: HZMO.



Figure 3.9 At-risk-of-poverty rate of older people by sex – Croatia and EU in 2022
                           Males                                                         Females




Note: At risk of poverty rate (cut-off point: 60% of median equalized income after social transfers).
Source: Eurostat (ilc_li02)


       3.3.4 Minimum pensions


Minimum pension regulation in Croatia covers a sizable portion of existing retirees with
increasing importance among new pensioners. Simple calculation using the current rules for
application of minimum pension shows that new retirees with career-wide average wage
below 81.1% (= (1.03/1.27) x 100) of the national average wage will receive the minimum
pension. According to our estimates, around 40% of new old-age pensions in 2023 were
minimum pensions. There are significant gender differences: our estimate indicates that 34%
of new male old-age pension beneficiaries and 51% of new female pensioners will receive
minimum pension. In total, there were 340,001 beneficiaries of minimum pension in
September 2023 or 30% of all pensions according to general legislation, of which:
     a. 306,259 of minimum pensions are realized since January 1999 according to Law on
        pension insurance, representing 36% of pensions realized since 1999 according to
        general legislation,
     b. 33,742 of minimum pensions were realized until 1998, covering minimum pensions
        and means-tested supplements to pensions according to legislation in force up to
        December 1998.

Around 68% of minimum pension benefits realized since 1999 were taken by women. The
share of women among minimum pension recipients is augmented because more than 50
thousand beneficiaries, mostly men, have realized minimum pension according to the war
veteran’s legislation which stipulates higher minimum pension compared to minimum
pensions according to general legislation. In September 2023 there were 58,276 pensions


                                                             32
realized according to the Law on pension insurance but determined according to the Law on
Croatian Homeland War Defenders.

Average minimum pension according to the Law on pension insurance amounted to EUR
315.86 in September 2023, which is EUR 101.01 or 47.4% greater than the average pension
these beneficiaries would receive if it would be calculated by the general earnings- and
service-related pension benefit formula. The low average amount of minimum pension is
connected to short average service period. The average minimum pension is realized after 28
years of insurance period on average. Pension expenditures arising from the difference
between minimum pension actually paid and pension that would be paid if determined based
on general pension formula amounted to around EUR 560 million or 7.6% of total pension
expenditures in 2023.21

Minimum pension in Croatia is not means tested; it is proportional to the length of service
and applies the same bonus and penalty scheme in case of deferred/early retirement as the
general point formula for PAYG pensions. Such regulation may encourage underreporting of
wages since the amount of minimum pension for those having career-average wages below
81% of national-wide average wage does not depend on contributions paid. However, the
current rules may encourage low-wage earners to work longer.
Although minimum pension beneficiaries receive relatively low benefits in absolute
amounts, their replacement rates are high, reaching more than 100% in certain cases. This
is illustrated with theoretical net replacement rate, i.e., the first pension benefit expressed as
a ratio of the last wage earned for a predetermined service period and earnings profiles in
Figure 3.10. It presents several hypothetical cases regarding career length and age of
retirement. The comparison is shown for the full PAYG pension benefit according to the
current pension formula and system parameters as of January 2024. Male average wage
earner (wage ratio of 100%) can expect the net replacement rate between 42% and 48%
depending on the length of service and age at retirement. Low-wage earners can reach the
net replacement rate over 100% if they retire after a sufficiently long career. A hypothetical
career of 42 years and retirement at the age of 65 years and 2 months with wage ratio of 45%
(which corresponds to minimum wage over the entire career) will result in the replacement
rate slightly above 100%. Minimum wage in Croatia in 2023 was EUR 700 gross, or EUR 554
net, and the estimated pension benefit, if retired in January 2024, is EUR 556. In the case of a
longer career (M 67y/45y) and having the wage ratio of 40% (which is possible due to
lower/non-existing minimum wages in a more distant past), the net replacement rate is close
to 130%.




21
     Authors’ estimate.

                                               33
Figure 3.10 Net replacement rate: Full PAYG pension for retirement in January 2024

     130%
     120%
     110%
     100%
      90%
      80%
      70%
      60%
      50%
      40%
      30%




            320%
             40%
             50%
             60%
             70%
             80%
             90%
            100%
            110%
            120%
            130%
            140%
            150%
            160%
            170%
            180%
            190%
            200%
            210%
            220%
            230%
            240%
            250%
            260%
            270%
            280%
            290%
            300%
            310%
                                   Career-wide individual-to-average wage ratio
                          M 65y2m/42y service                         M 65y/35y service
                          M 67y/45y service                           M 60y/35y service

Source: WB Team.



         3.3.5 Disability pension

There are two types of disability pensions: a) total disability pension paid in 100% amount,
and b) partial disability pension, amounting 80% in case of unemployment, 50% if
beneficiary is employed, and 66.7% if disability of employed beneficiary is work-related.
Insurance period for calculation of disability pension consists of actual insurance period and
added period aiming to compensate for the part of working career lost due to occurrence of
disability. Added period is calculated as 2/3 of the period between the date of disability
recognition and the age of 55, plus ½ of the period from the age of 55 to 60. If disability is
work related, the minimum number of years for pension calculation is 40.
Expenditures for disability pensions of 1.4% GDP out of total pension expenditures of 9.1%
GDP in 202222 were one of the highest among the EU countries. The average expenditures
for disability pensions in EU was 0.8% GDP out of average total pension expenditures of 11.4%
GDP. Very high share of disability pensions in Croatia occurred due to high inflow of disability
pensioners in the past and expenditures for disability pensions related to the Homeland War
veterans’, police, and army. After reaching the statutory retirement age, disability pensioners
are translated to old-age beneficiaries and their pensions are counted as an old age pension
expenditure. The share of expenditures for disability pensions has been decreasing in recent
years. Similar trend is expected in the future as the annual inflow of new disability pensioners

22
     European Commission, The 2024 Ageing Report.

                                                     34
has decreased from the average of ten thousand up to 2012 to around three thousand in the
last ten years.
In September 2023, there were 94,799 disability pensions realized according to general
legislation, amounting to EUR 372.95 on average, and 78,818 old age pensions translated
from disability pension amounting on average EUR 389.02. Average disability pension of
EUR 372.95 is lagging the average survivor’s pension of EUR 428.72 (i.e., by 13%) and average
old-age and early retirement pension of EUR 539.88 (by 30.9%). Amounts of average old-age
and disability pension in 2023 include a part of survivor’s pension for those who are eligible
for such pension supplement. Amounts refer to pensions net of taxes, without pensions
realized according to international social agreements. The average age of new disability
pensioners retired in 2023 according to general pension legislation was 54 years and 8
months, average insurance period was 24 years and 3 months, while average disability
pension amounted to EUR 367.66, which is also lagging behind the average new survivor's
and old age pensions by 16.5% and 33.3% respectively.
Net benefit ratio for disability pensions approved by the general legislation for those
without international service was 30.8% in 2023, substantially below the same indicator for
old-age pensions (44.6%). As disability pensions are valorized/indexed by the same formula
as old-age pensions, the long-term declining trend in their net benefit ratio has been equally
present. However, policy measures taken in 2023 enabled a mild increase in this ratio
compared to 2022. Relatively low average amounts of disability pensions and their low net
benefit ratio compared to other types of pensions would become even more stringent when
these beneficiaries get older and disability becomes even heavier burden then in the younger
ages, especially for the poor ones. This situation is calling for some kind of legislative
intervention to increase disability pensions and bring them closer to the level of old age and
survivor’s pensions, thus improving the adequacy of disability pensions of todays and future
beneficiaries.


     3.3.6 Survivors’ pensions

According to amendments to the Law on pension insurance from January 2023, survivors’
pensions are increased by 10%, while surviving spouse ageing 65 or more got the right to
receive own early retirement/old-age/disability pension together with the part (27%) of
survivor’s pension if such combined pension is higher than survivor’s pension . That reform
increased significantly pension benefits of surviving partners. In September 2023, there were
197,047 survivor’s pensions realized according to general legislation, amounting EUR 428.72
on average, and 73,470 partial survivor’s pensions paid together with own old age/ disability
pension, amounting EUR 101.43 on average. In the last decade, number of survivor’s pensions
was steadily decreasing, which is expected to continue in the coming years due to increasing
employment rates of women resulting in growing share of women that will realize their own
old-age pension instead of survivor’s pension. On the other hand, as the right to receive own
pension and the part of survivors’ pension is a new institute introduced in 2023, it is expected
that this right will be used by around 110 thousand pensioners in coming years.

                                              35
Similar as for disability pensions, expenditures for survivors’ pensions of 1. 4% GDP in 2022
are quite high (EU average is 1.3% GDP). This is the result of lose entitlement conditions,
relatively high level of survivors’ pension relative to other types of pensions and payment of
the above-the-average special kind of war-related survivor’s pensions. These expenditures
have further increased in 2023 due to a 10% increase of pension factor for survivors’ pensions
and a possibility to receive part of survivor’s pensions (up to 27%) in addition to own old-age
or disability pension.


      3.3.7 Pension adequacy of multipillar pensions


Participants in the multipillar pension insurance scheme, when they retire, choose between
combined pension (basic pension from the PAYG plus second pillar pension annuity) and
full PAYG pension; this choice affects both pension adequacy and fiscal sustainability of the
PAYG system. For ease of expression, in more condensed discussions we call the full PAYG
pension P1, and the combined pension P2, but otherwise continue to use the usual terms.23
Motivation for opt-back to PAYG (P1) depends mostly on the relationship between P1 and P2.
However, other factors also affect the choice. Firstly, there are dynamic effects of different
indexation methods applied to the second pillar annuity component of P2 and P1 (including
the basic pension).24 Secondly, the option to withdraw a part of the second pillar savings at
retirement (20% of savings since the beginning of 2024, 15% earlier) plays a role, too. Time
preferences can motivate the choice of P2 for retirees who have P2<P1 and meet the cash
withdrawal condition. Thirdly, accumulation of savings in the second pillar mandatory pension
funds due to contribution payments and net investment returns critically determine the
relationship between P1 and P2. The endowment effect can also play a role in the choice of
the payout model.25 Finally, some retirees may have a safety preference towards the first
pillar – believing government more than private pension providers.
In 2023, around 15 thousand pension savings accounts in the second pillar were closed due
to retirement, and around EUR 280 million were transferred either to the state budget for
those who opt back to pension payment from the first pillar only, or to pension insurance
companies (PIC) for annuity payments. The number of account closures has been increasing
each year. It was around 8 thousand in 2018 and 2019 (Figure 3.11). Many personal account
closures so far have been for those who voluntarily participated in the second pillar (those
between 40 and 50 years of age in 2002) and for disability and survivors’ pensions. Out of
those who were mandatory members, the accounts are mostly closed for early retirement. A
higher inflow in retirement is expected in 2026 (women) and 2027, when the first generation
of mandatory members who were at the age of 40 in 2002 will reach statutory retirement
age.
Only 15.1% of retired mandatory pension funds members chose the combined pension in
2023, which is a lower proportion than in earlier years. The share of new pensioners who

23
   Croatian pension administration sometimes refers to it as models A and B, while we call it P1 and P2.
24
   Basic pension is indexed by the same formula as entire P1, see further in the text.
25
   Higher absolute amount of cash withdrawal may matter.

                                                      36
opt back for PAYG-only pension increased significantly in 2023 after bond market turmoil
reduced the value of pension savings in the second pillar. For comparison, the proportion of
accounts closed opting for the combined pension was 29% of total accounts closures in 2021.

Figure 3.11 Personal accounts closed due to retirement with the combined pension as a proportion of
all accounts closed for retirement

 50.0%                                                                                                               16,000
                                                                                                                     14,000
 40.0%
                                                                                                                     12,000

 30.0%                                                                                                               10,000
                                                                                                                     8,000
 20.0%                                                                                                               6,000
                                                                                                                     4,000
 10.0%
                                                                                                                     2,000
  0.0%                                                                                                               0
                                    2018           2020
                                                  2019            2021              2022                  2023
                                              Number of accounts (in % of total, left scale)
                                              Personal account balance (in % of total, left scale)
                                              Total number of accounts closed (right scale)
Source: Regos (Statistical Indicators, various issues).


The average account balance of members choosing the combined pension in 2023 was EUR
34,052, considerably higher than of members choosing the full PAYG pension with EUR
15,337. Around 20% of new contracts for payments of the second pillar pension are based on
savings of less than EUR 20 thousand. Close to 80% of annuitized accounts accumulated the
individual savings below EUR 40 thousand (Figure 3.12). This information goes against the
claims that the combined pension is attractive only to persons with extremely high savings.
The wage profile impacts the choice stronger than the amount of savings (more on the impact
of the wage profile on the choice available in Section 4).
Figure 3.12 Number of new contracts on second pillar pension by savings transferred to PIC, 2023

                              600
    Number of new contracts




                              500

                              400

                              300

                              200

                              100

                               0
                                     Up to   10.001 - 15.001 - 20.001 - 25.001 - 30.001 - 35.001 - 40.001 - 50.001
                                    10,000    15.000 20.000 25.000 30.000 35.000 40.000 50.000 and more
                                                         Pension savings received by PICs, in EUR

Sources: HRMOD and Raiffeisen MOD.



                                                                             37
Those who had the best years of earnings in the period of multi-pillar insurance are more
likely to choose the combined pension. Also, persons with disproportionally higher rights to
pension benefits from the first pillar compared to their contribution payments are less likely
to choose combined pensions. For example, this applies to minimum pension beneficiaries.
At the end of December 2023, there were 13,311 pensioners (1.1% of total) receiving the
annuity from the second pillar from two pension insurance companies (PICs) operating in
Croatia. In December 2023, the average annuity was EUR 97. According to PICs records,
around 83% of their pensioners took the opportunity to withdraw 15% of pension savings at
retirement, reducing thereby their annuities. Most pensions paid by PICs were between EUR
50 and 75 as shown in Figure 3.13. Around half of pensions in payment were below EUR 75,
and 85% were below EUR 150.
Figure 3.13 Distribution of second pillar pensions paid in December 2023

                            5000
                            4500
     Number of pensioners




                            4000
                            3500
                            3000
                            2500
                            2000
                            1500
                            1000
                             500
                               0




                                           Monthly pension, in EUR

Note: One-off withdrawal of 15% is not counted as monthly pension payment.
Sources: HRMOD and Raiffeisen MOD.

Beneficiaries of the combined pension receive substantially higher total pension benefits
than the average PAYG pensioner. Although data on total combined pensions are not
published, a combination of existing data from HZMO and PICs shed some light on the
difference in pensions between two groups of pensioners. The average basic old-age PAYG
pension was EUR 744 (EUR 779 for men and EUR 704 for women, Figure 3.14), while the
average overall PAYG pension was EUR 481 (EUR 500 for men and EUR 461 for women).26 It
turns out that the average combined pension in December 2023 was EUR 841 (EUR 744 basic
PAYG pension plus EUR 97 of annuity from the second pillar). If we focus on old-age pensions
without international service, then the combined pension is practically unchanged, EUR 846


26
  We present only the average for old-age pensions here because almost all of the combined pensions are old-
age pensions. HZMO in their statistical reports shows that substantial number of basic pensions are survivors’
pensions (2.241 pension at the end of 2022), whereas PICs pay annuities for only a dozen survivors‘ pensions. A
possible reason for that difference is the behavior of descendants who first decide to inherit the full amount of
pension savings for the deceased member, and then later opt for survivor's basic pension from the first pillar.
The number of basic old-age pension in HZMO registers is close to the total number of annuities paid by PICs.

                                                         38
but average PAYG pension increases to EUR 547. The average combined old-age pension is
more than 50% higher than the average old-age PAYG pension without international service.


Figure 3.14 Basic and overall PAYG old-age pensions in December 2023 (monthly, in EUR)

        1000

         800
                          779            785
         600                                                   704             709
 Euro




                                               595
         400                     500                                                   506
                                                                       461

         200

            0
                                     Male                                Female

        Basic PAYG pensions                          Overall (full and basic) PAYG pensions
        Basic pensions w/out int'l service           Overall (full and basic) PAYG pensions w/out int'l
Source: HZMO.

The average gender pension gap for the basic PAYG pension is lower than the gap among
all PAYG pensions. For pensions without taking into account international service, the
gender gap was 11% among basic pension and 18% for all pensions under general
legislation. There are two main reasons explaining the difference. The first is a higher average
wage of those who take the combined pension. Higher average personal account balance of
those who take combined pension supports such conclusion.27 It appears that women who
take combined pensions had wage histories more similar to those of men. Unfortunately,
there are no statistical data available by gender. The second is a much lower difference in
length of service between men and women for those who take combined pensions, compared
to the overall average for new old-age pensioners, which is then transposed to lower
difference in pensions. Those who take combined pensions have a higher length of service
than the average pensioner. The difference is striking: 6.8 years for women and 5.4 years for
men (Figure 3.15).
Figure 3.15 Insurance period of pensioners with basic and overall PAYG old-age pensions (without
international service), as of December 2023




27
     REGOS does not report account closures and account balances by gender.


                                                       39
             45.0
             40.0
             35.0          39.8
                                                                        38.2
             30.0                        34.4
                                                                                   31.4
             25.0
     Years




             20.0
             15.0
             10.0
              5.0
              0.0
                                  Male                                      Female

                     Basic PAYG pensions        Overall (full and basic) PAYG pensions

Source: HZMO.



The relatively low average annuity paid by the second pillar, EUR 97 in December 2023, is
frequently misinterpreted as a failure of the second pillar. At least two additional indicators
should be considered for making a more informed judgment on pensions paid by the second
pillar. Firstly, most pensions from the second pillar are determined after 15% of pension
savings are withdrawn. Secondly, mandatory second pillar pension contributions are small in
comparison with contributions paid to the first pillar. We can illustrate that with pension
calculations for a hypothetical case of a man, an average earner, with a flat wage profile
retiring in 2023 at the age of 65 after 40 years of uninterrupted service.28 20% of his gross
wage was paid in pension contributions to the first pillar up to 2001, and after that, 15% of
the gross wage was paid to the first pillar and 5% to the second pillar. Out of total
contributions paid up to retirement in 2023, only 13.4% was directed to the second pillar.
Based on that, the estimated second pillar annuity (without one-off withdrawal of 15%) would
be EUR 75. This amount may seem low, but it is 12.9% of the total combined pension, and
12.5% of the full PAYG pension, which is not far from the proportion of contributions paid to
the second pillar. With the expected recovery in returns of mandatory pension funds, the
annuity will increase in absolute and relative terms. A person with the same characteristics,
but retired in 2028, will pay 16.6% of total pension contributions to the second pillar and
receive an annuity of EUR 110 in constant 2023 prices. That is 17.6% of the combined pension
and 17.2% of the full PAYG pension. It means that, despite an annuity of only EUR 110, each
percentage of pension contributions paid into the second pillar is expected to result in a
higher unit pension benefit than if paid to the first pillar.



3.4 Key determinants of pension adequacy in international comparisons




28
  For retirement in 2023, we took all the available parameters of the pension system as of July 2023, as well as
historical rates of return of pension funds to calculate pension benefits. For future benefits, we prepare a set of
macroeconomic assumptions, as explained in Section 4.

                                                         40
       3.4.1 Statutory retirement age and early retirement options


Statutory retirement age in Croatia is similar to other EU countries, but early retirement
options are much more flexible. Statutory retirement age stands at 65 for men. The age for
women is to be raised to 65 until 2030 in annual incremental steps. In 2023 the retirement
age for women was 63 years and 3 months. Early retirement can be attained five years before
the statutory age with 60/58.25 years of age and 35/33.25 years of service for men/women.
It is among the lowest compared to other EU countries (Table 3.1). Pension reform in 2019
aimed at further increasing the length of working lives. However, under the pressure of trade
unions and obligation to call for the national referendum, the Government gave up main
elements of the reform related to rising retirement ages and decrement factors for early
retirement. The intended package included an accelerated increase in the statutory
retirement age to 67, accelerated equalization of the retirement age for men and women,
and higher decrement for early retirement.


Table 3.1 Statutory retirement age and early retirement age in EU and Western Balkans
 2019*-22
 HR (2023)        retirement age (early retirement)            2019*-22         retirement age (early retirement)
 country            Men                    Women              country            Men                     Women
     BE           65 (63)                  65 (63)                NL          66.6 (66.6)              66.6 (66.6)
     BG          64.4 (63.4)             61.8 (60.8)              AT            65 (60)                  60 (55)
     CZ           63.9 (60)              62.2 (59.2)              PL            65 (65)                  60 (60)
     DK           67 (63.5)               67 (63.5)               PT           66.6 (60)                66.6 (60)
     DE           65.9 (63)               65.9 (63)               RO            65 (60)                61.8 (56.8)
     EE          64.2 (59.2)             64.2 (59.2)              SI            65 (60)                  65 (60)
     IE           66 (66)                  66 (66)                SK          62.8 (60.8)                62 (60)
     EL           67 (62)                  67 (62)                 FI          64.5 (61)                64.5 (61)
     ES          66.2 (64.2)             66.2 (64.2)              SE            65 (62)                  65 (62)
     FR           67 (62)                  67 (62)                NO            67 (62)                  67 (62)
 HR (2023)        65 (60)                63.3 (58.3)
     IT           67 (64)                  67 (64)               MNE           66 (62.6)               62.3 (57.3)
     CY           65 (65)                 65.0 (65)              SRB           65 (59.6)                63.4 (59)
     LV          64.3 (62.3)             64.3 (62.3)             FBIH          65 (62.6)                65 (57.6)
     LT          64.3 (59.3)             63.7 (58.7)              RS            65 (60)                  65 (58)
     LU           65 (57)                  65 (57)              N.MAC           64 (64)                  62 (62)
     HU           65 (65)                  65 (65)               ALB            65 (62)                  62 (57)
     MT           63 (61)                  63 (61)               KOS              65                       65
Source: The 2024 Ageing Report: Underlying assumptions and projection methodologies for EU Member States; national
statistics for Croatia (2023) and Western Balkan (*2019) countries.



Long service early retirement has been abolished in most of the EU Member States, while
it has been restored in Croatia at favorable conditions . Table 3.2 shows that there are only
5 countries in the EU allowing decrement-free early retirement for long service contributors.

                                                         41
Croatia abolished it in 1998 and restored in 2014. In 2023 it was made even more attractive
due to a bonus scheme integrated into it. A retiree is eligible for a pension benefit bonus after
60 years of age and 41 years of service even if retirement is taken before the statutory
retirement age. In other countries, the age and/or service requirement for long service
retirement is higher than in Croatia (Table 3.2) because there is nothing intrinsic in the service
period of 41 years that justifies early retirement (unless it coincides with service in hazardous
and arduous occupation). Also, there is nothing intrinsic which would require special
decrement-free treatment in comparison with someone of the same age and fewer service
years who would get actuarially reduced pension. The long service early retirement window
should be reviewed for actuarial fairness and brought in line with general early retirement
principles.
Early retirement in Croatia is encouraged by low early retirement decrement. A first pillar
early retirement decrement close to actuarially neutral rate of 3.6% was set from the start of
pension reform in 1998 until 2014 when the Government decided to introduce the sliding
scale with reducing decrement factor. After the trade unions ’ initiative in 2019, the
decrement was unified at 0.2% per month which is the lowest among the selected EU, OECD,
and neighboring countries (Figure 3.16).
Table 3.2 Eligibility rules for Long Service Retirement in EU, 2023
                                              Long Service Early retirement
                                  age requirement                     service requirement
 Belgium                                  -                                    43
 Croatia                                 60                                    41
 Germany                                 65                                    45
 Greece                                  62                                    40
 Italy                                    -                              43(M), 42(W)
 Slovenia                                60                                    40
Source: MISSOC



Late retirement bonus in Croatia has been among the lowest in the EU in 2022, which is
ineffective because it does not encourage longer stay in the labor market (Figure 3.17). The
new first-pillar late retirement bonus of 5.4% per year (for a maximum of 5 years) stands close
to the actuarially neutral level. The bonus is, however, awarded only for a service longer than
35 years. A person at statutory retirement age with 30 years of service intending to work
more years would get no pension bonus and may decide to withdraw from the labor market
early.
Figure 3.16 Annual decrement for early retirement in Selected countries, 2020 or later




                                                    42
Note: In countries with a range of decrements, the average is used.
Source: MISSOC, Pensions at a Glance, World Bank ASPIRE.




                                                            43
Figure 3.17 Annual pension bonus for late retirement in selected countries, 2020 or later




Note: In countries with a range of bonuses, the average is used.
Source: MISSOC, Pensions at a Glance, World Bank ASPIRE.



Figure 3.18 Average exit age from labor market, Croatia and the EU, 2022




Source: EU Ageing Report 2024.



The average exit age from the labor market in Croatia is below the EU average. Due to
retirement parameters, primarily related to early retirement, overall exit age in Croatia was
almost a year lower than the EU average (Figure 3.18). Difference between men and women
– one year for women and 0.5 years for men – will be narrowing as the statutory and early
retirement ages for women converge, but it will apparently remain lower than the EU average
for some time. The main factors of low exit age from Croatian labor market remain to be the
                                                            44
widely open early retirement window of 5 years and low early retirement decrement or no
decrement at all for the long-term insured. Policy actions should be directed towards
reshaping two early retirement parameters.
Increase of the statutory retirement age is the policy that may effectively increase the
average age at retirement and service period. There was close correlation between statutory
and effective retirement age in the last 20 years (Figure 3.19). In the last 15 years, average
age of the old-age pensions among men stayed unchanged at slightly above 63 years. In this
period, statutory retirement age for old-age pension was intact, 65 years, while early
retirement options were broadened, as described above. If only the last 10 years are
considered, effective retirement age for men is on a mild downward trend, suggesting that
the attractiveness of early retirement is on the rise. Between 2012 and 2022 the statutory
retirement age for women increased by 2.5 years, while the effective age increased by 2.1
years, confirming that current retirement regulation does not provide enough incentives for
longer work (Figure 3.19).29
Figure 3.19 Average age of new old-age pensioners (including early retirement, excluding international
service)

             65.0
             64.0
             63.0                                         63.8
                                          63.2                            63.3
             62.0                                                                          63.4
                                                                                                  62.4
             61.0
     Years




                         61.0                                                    61.1
             60.0
                                                                 60.3
             59.0
                                                 59.1
             58.0
             57.0
             56.0
                                56.0
             55.0
                          2003              2008           2012             2017            2022

                    Age of new old-age pensioners - men    Age of new old-age pensioners - women
                    Legal retirement age - men             Legal retirement age - women

Note: Pensions for war veterans, police, and army forces are not considered.
Source: HZMO („Statističke informacije“, various issues).



             3.4.2 Contribution period

Average service period for contribution payments in Croatia is shorter than in most other
EU countries. In 2022, including disability, survivors’, and partial foreign contributors, average
was 32.2 years. Only two countries in the EU had a lower average contribution period (Figure
3.20).


29
  Data for 2023 are not shown as they are incomparable with data for previous due to changes in survivors’
pension legislation causing considerable benefit switch from a survivor pension to an old-age pensions, where
previous survivors’ pensions are counted as “new” old-age pensions.

                                                                 45
Figure 3.20 Average contributory period of new earning-related public pensions in the EU, 2022




Source: EU Ageing Report 2024.


Recent changes in the average contribution period for new pensioners in Croatia seem to
be related to changes in legal retirement age and minimum service period requirements.
Between 2003 and 2022, the average contribution period (more precisely: insurance period
counted according to pension legislation) for men increased by 3.8 years, while the legal
retirement age and the minimum service period requirement for early retirement increased
by 2.5 years (Figure 3.21). For women, increase in the average service period was 6.6 years,
while the legal requirements increased by 5.5 years.

Figure 3.21 Average insurance period of new old-age pensioners (including early retirement, excluding
international service)

             38
             36
                                    35.7                                              35.8
             34                                         34.6            34.7                 34.6
             32                                                                32.8
                  32.0                     31.8
             30                                                31.5
     Years




             28
                         28.0
             26
             24
             22
             20
                   2003               2008               2012             2017         2022

                                                  Men   Women

Note: Pensions for war veterans, police, and army forces are not considered.
Source: HZMO (“Statističke informacije”, various issues).



                                                               46
In the last five years the effective insurance period increased by less than the legal
retirement age and the minimum service period requirement for early retirement . Although
the average service period found in the official statistics increased faster than legal retirement
age, two other regulations lead to conclusion that the effective service period has not
changed more than the legal retirement age. For women, since 2019, the official insurance
period increases by six months for each child raised (this is also possible for men, but that is
a rare case). Unofficial estimates indicate that 0.9 years is added based on that regulation.30
This means that out of total increase in the average insurance period for women of 1.8 years
between 2017 and 2022 (Figure 3.21), 0.9 years was due to added insurance period for
parenting. Remaining 0.9 years was acquired in the period when the legal retirement age
increased by 1.25 years. Among men, one must consider the added insurance period for war
veterans. War veterans can retire according to general rules but have the right to added
insurance period related to the length of service during the Homeland War. As more war
veterans are approaching the retirement age and getting the added insurance period, the
average insurance period for men is increasing. We do not have data on the extent of using
the added insurance period for war veterans, but it is reasonable to assume that observed
increase of one year in the average insurance period among men between 2017 and 2022
(Figure 3.21) was in large part due to legally added period for veterans. Average insurance
period for men is almost the same in 2022 as it was in 2008. Supposedly, attractive early
retirement options prevented effective increase in average insurance (contribution) period.

Figure 3.22 Expected years in retirement in OECD and selected non-OECD countries in 2020




Source: OECD Pension at a Glance 2021 (data extracted on 30 May 2023 from OECD.Stat).

International comparison indicates that women in Croatia can expect one of the longest
periods spent in retirement, while men can expect their years in retirement to be around
the EU average. OECD estimates that women in Croatia retired in 2020 can expect 25.7 years
in retirement and men 18.6 years (Figure 3.22). The averages for the EU-27 are 24 and 19.4

30 According to the HZMO, up to the end of 2022 there were 58 thousand women with a recognized additional
insurance period for 119 thousand children, which is around two children per woman. At the same time, around
65 thousand women took old-age and disability pensions. Therefore, around 1.8 children per new female
pensioner brought around 0.9 additional insurance period in years since 2019.

                                                         47
years for women and men respectively. Expected years in retirement is estimated as life
expectancy measured at the age of effective exit from the labor market. The difference of 7.1
years between women and men in Croatia is the result of a higher effective age of retirement
for men and lower life expectancy than for women.

      3.4.3 Accrual rate

Effective net accrual rate (percentage of average net wage earned as pension for one year
of service) in Croatia is around the EU average. In 2022 it stood at 1.29 (including supplement
of 27% and international pension service), slightly above the accrual rate in 2019. The increase
is explained by new valorization and indexation pattern and the Covid-period real wage
developments. Figure 3.23 shows that the net accrual rate in Croatia is below the EU average.
For one year of work, Croatian insured individuals earn the pension benefit in the lower half
of the EU list, suggesting that the pension formula should not be blamed for relatively low
pension adequacy.
Figure 3.23 Average accrual rate in EU in 2022




Source: EU Ageing Report 2024.


A year of work on average wage in Croatia accrues a unit pension similar to other point-
formula countries but contributory periods are lower. As shown in Figure 3.23 the value of
one personal point in Estonia is close to Croatian and in Germany well below. All three
countries have a similar points-based PAYG pension system with proportional point formula
between the floor and the ceiling. Nevertheless, in both Estonia and Germany, the
replacement rates have been higher (Figure 3.4) than in Croatia due to longer service periods,
not due to the point value or the pension formula.




                                                 48
      3.4.4 Support ratio


Croatian support ratio (number of workers per retiree) is one of the lowest in the EU. One
of the factors affecting pension adequacy is the low support ratio (SR: Figure 3.24). It is a
problem for a PAYG system because the declining number of prime age contributors must
finance pensions for an increasingly large number of pensioners. Since the establishment of
the PAYG systems in Europe, when SRs stood above 5, ratio has been continuously declining
due to the aging process and post WWII reductions in the statutory retirement age. Figure
3.24 shows that majority of EU countries in 2022 had the SR between 1 and 2, with Croatian
1.31 (=1,648,887 / 1,225,246) among the lowest in the EU. Currently, positive effects of a
strong business cycle are dominating over the negative population aging trend. In 2023, the
SR in Croatia improved to 1.35, primarily due to strong economic growth and rising
employment rate. From the end of 2021 to 2023, the number of contributors increased by
4.9% despite a tight labor market related to population ageing and prior emigration. The
number of contributors increased by 6.7% compared to the end of 2019. Employment growth
is also related to growing employment of immigrant workers. Increase in the support ratio is
also related to stagnant number of pension beneficiaries: as of end 2023, their number was
only 0.6% higher than in end 2012, and 1.3% lower than in 2019. Gradual increase of the
statutory retirement age for women contributed to SR improvement. We expect the
economic cycle and ongoing adjustment of the retirement age for women would continue to
dominate over demography for several years during this decade. This assumption is reflected
in the simulations in Section 8 of this report. However, very soon, in the later part of this
decade, demographic fundamentals will probably take over again when growth of the
employment rate will begin to saturate. So, cyclically good times represent an opportunity to
undertake substantive labor market and pension system improvements. The reforms
undertaken in the EU have been attempting to keep the support ratio as high as possible by
slowing down the entry into retirement and increasing the employment rate, which is the gist
of recommendations presented in the second half of this document.31
Figure 3.24 Support ratio in EU and Croatia, 2022




31
  It should also be noted that historical reasons (Homeland War 1991-1995, emigration) caused low SR level in
Croatia: there are many pension beneficiaries living abroad. Ratio of number of people working in Croatia and
number of pension beneficiaries living in Croatia is around 1.65 according to 2021 Census, which is closer to EU
median value of SR. If number of pension beneficiaries is counted without beneficiaries at the basis of
international contracts, end 2023 SR ratio was 1.58.

                                                      49
Source: EU Ageing Report 2024.


      3.4.5 Pension indexation


Pension valorization applied in Croatia has been less generous, while pension indexation
has been more generous than in other EU countries. Croatia applies the same rule for
adjustments of new pensions and pensions in payment, meaning that the rate of valorization
is equal to the rate of indexation. Valorization and indexation32 of pensions in Croatia are
based on average wage growth and the CPI inflation according to rotational formula: 70%
weight is attached to the higher of two rates, 30% to the smaller.33 The
valorization/indexation practice varies substantially in the EU (Table 3.3): countries with social
contract requiring the same relative starting pension position of new retirees apply wage
valorization while countries that commit to protect the real pension apply the price (CPI)
indexation. Some countries implement pensioners' participation in economic growth; they
add the wage or GDP growth to price indexation. Croatian combination means that each new
generation of retirees starts relatively worse than the previous one, but then participates in
fruits of economic (wage) growth. It also means that pensioners with the same career and
earnings receive the same nominal benefit regardless of the year of retirement, unless other
parameters of the pension system have meanwhile changed. Such a social contract has been
deeply rooted in practice of pension policy making in Croatia. Several previous proposals for
delinking valorization from indexation rules have been swiftly rejected on the “nominal
equality” grounds. Although net replacement rates in Croatia are among the lowest in the
EU, meaning that the first pension are low, but pensions in payments are indexed more
favorably than in many other EU countries. Figure 3.25 shows net replacement rates for
those retired in 2058 compared to these rates 10 years after retirement for average wage
earner with career of 40 years and retired at standard (statutory) retirement age. There are
four EU countries where net replacement rate according to projections reported in the EU
Adequacy Report 2021 will be higher 10 years after retirement than at retirement. For all

32
   Valorization upgrades past earnings to current period wages prior to awarding first pension, while indexation
adjusts all pensions in payment.
33
   More on the effects of such valorization/indexation in Croatia, available in Section 6.2.

                                                      50
other countries, the net replacement rate will be somewhat lower. However, decrease in
Croatia will be among the mildest. Only Estonia’s and Ireland’s beneficiaries will lose less in
terms of net replacement rate after 10 years in retirement.

Figure 3.25 Net (theoretical) replacement rates, base case and 10 years after retiring, average earner,
2069, %




Notes: Ranked by the difference in Theoretical Replacement Rates (TRR); base case refers to a 40 years’ uninterrupted car eer
ending with the standard pensionable age; countries on the left have higher replacement rates 10 years after retirement;
those on the right have a lower TRR after 10 years.
Source: The Adequacy Report 2021 (p.90).




                                                            51
Table 3.3 Pension valorization and indexation rules in EU countries
            Valorization rule                           Indexation rule
 BE         Prices                                      Prices
 BG         Wages                                       50%p+50%w
 CZ         Wages                                       50%p+50%w
 DK         Not applicable (basic pension)              Wages
 DE         Wages                                       Wages with sustainability factor
 EE         Wages                                       20%p + 80% contribution collection
 IE         Not applicable                              No fixed rule, discretionary
 EL         Prices and wages or GDP                     50%p+50%GDP (Max 100%CPI)
 ES         Prices                                      Prices+0.50%
 FR         Prices                                      Prices
 HR         70%w+30%p or 30%w+70%p, whichever is        70%w+30%p or 30%w+70%p, whichever is more
            more favorable                              favorable
 IT         GDP growth rate                             Prices
 CY         Wages                                       Basic pension wages; pensions prices
 LV         Contribution wage bill                      Prices and wage bill varies from 50%w+50%p to
                                                        80%w+20%p, depending on service years
 LT         Wage bill                                   Wage bill
 LU         Prices and wages                            Prices and supplemental wage when fiscal space
 HU         Wages                                       Prices
 MT         Prices                                      70%w+30%p
 NL         Not applicable (national pension)           Wages
 AT         Wages                                       Prices or discretionary on degressive scale in
                                                        proportion to pension
 PL         NDC 1st: Wages, NDC 2nd: GDP                80%p + 20%w
 PT         Prices and wages                            Between p-0.75 percentage points. and 120%*GDP
                                                        growth, depending on the GDP growth and
                                                        individual pension amount
 RO         Prices and wages                            50%p+50%w
 SI         Wages                                       40%p+60%w
 SK         Wages                                       Prices
 FI         80%w + 20%p                                 80%p + 20%w
 SE         Wages                                       w-1.6pps (and automatic adjustment mechanism)
 NO         Prices and wages                            w-0.75%
Source: OECD Pensions at a Glance 2023 (p. 141); The 2024 Ageing Report: Underlying assumptions and projection
methodologies (p. 125); MISSOC; World Bank ASPIRE database.


         3.4.6 Supports for low-income pensioners


Countries employ different practices to maintain minimum living standards of their
pensioners, either directly through the general pension system rules or through social
policies financed from other sources of taxation from the state budget. There are four main
ways in which the OECD countries might provide retirement incomes to meet a minimum
standard of living at the old age34: (i) (non-contributory) residence-based basic benefits, (ii),
targeted benefits, (iii) contribution-based basic benefits, and (iv) minimum contributory
benefits, with overlaps between some of these tools in some countries. Also, benefits in some

34
     OECD Pensions at a Glance 2023.

                                                     52
countries are means-tested, while in others they are not. For example, all OECD countries
provide targeted benefits that are subject to means tests.35
Three quarters of the OECD countries have either a contribution-based basic or a minimum
contributory pension. Minimum contributory pensions are more widespread than
contribution-based basic pensions, and more commonly have only one monetary value
irrespective of the contribution period. Fewer than half of countries apply higher rates for
longer careers. On average, 19 years of contribution are required for eligibility to a minimum
contributory pension, with 29 years required on average for the full minimum pension. The
incidence of receiving a minimum contributory pension (% of population aged 65 and over) is
highly diverse across countries, from 44% in Portugal, 34% in France and 31% in Belgium, to
only around 5% in Poland and the Slovak Republic and 2% or lower in Hungary and Slovenia.
Minimum pensions vary between a low of about 5% of the average wage in Hungary and 11-
12% in Czechia and Latvia, to a high of about 35-36% in Belgium and Spain and even 63% in
Colombia where the minimum contributory pension is set at the level of minimum wage. In
Croatia, the incidence of people above 65 who receive minimum contributory pension benefit
in Croatia stood at 35% in 2023, with minimum pension benefit at 30.3% of the average wage
for 40 years of service and 11.4% for 15 (minimum) years of service. High incidence of
qualifying for minimum pension in Croatia is partly due to the minimum pension formula and
eligibility based on average point value regardless of the service period.36


      3.4.7 Options to combine public pensions with income from work


The possibility to cumulate income from work and full retirement income in Croatia has
been extended to almost all categories of retirees including old age, early, survivors, and
the partial disability pensioners. Table 3.4 shows that most EU member states apply
unlimited cumulation for regular old age retirees (those retiring at statutory retirement age)
but reduces or suspends the pension benefit for early old age retirees to deter early
retirement. Furthermore, eligibility to cumulate pension with income from work has been
dependent on the duration of contract and type of work only in Portugal and Hungary. In
Portugal, the argument to limit the cumulation to part time work was implemented to avoid
stimulating the earliest possible retirement. However, like in Croatia, part time work was
difficult to supervise as such arrangement led to excessive early retirement with nominally




35
   In Australia, Finland, Germany and the United States these are the only first-tier schemes in place; however,
Germany recently introduced a new supplemental pension to the points scheme, which will provide additional
contributory benefits to low earners with careers of at least 33 years (Pension at a Glance, 2023). On the other
hand, residence-based basic pensions do not reduce the targeted benefit in some countries (such as Canada,
Denmark and Iceland), while in others (Greece, the Netherlands and New Zealand) those eligible to the
residence-based basic pensions cannot receive targeted benefits on top.
36
   For example, a retiree with 30 years of service and below-average point value may qualify for the minimum
pension, while a person with 15 years of service and above-average point value would not regardless of a lower
benefit in absolute terms.

                                                      53
part time arrangements accompanied by shadow payouts for the remaining work
engagement.
Table 3.4 Pension benefit of employed pensioners in EU 2023
                        Early old-age pension                                              Regular old-age pension
  The full amount of                   Reduced pension                    The full amount                Reduced pension
     the pension                         (% reduction)                     of the pension                  (% reduction)
                          Belgium (up to 100% depending on                                     Belgium (up to 100% depending on
  Bulgaria                                                               Bulgaria
                          income)                                                              income)
                          Denmark (PAYG pension up to 64%                                      Denmark (PAYG pension up to 50%
  France                                                                 Germany
                          depending on income)                                                 depending on income)
                          Germany (reduction up to 100% in
  Netherlands                                                            Estonia               Greece (30%)
                          proportion to income)
                          Spain (reduction up to 100% in the case                              Spain (reduction of up to 50% in the
  Sweden                                                                 France
                          of full-time work)                                                   case of full-time work)
                          Cyprus (continued payment of
  Croatia (up to 4hrs     contributions)                                                       Iceland (up to 100% depending on
                                                                         Ireland
  per day worktime)       Finland (up to 100% depending on                                     income)
                          income)
                          Luxembourg (up to 100% depending on                                  Slovenia (reduction of 40% in the first
  Ireland                                                                Italy
                          income)                                                              three years; 80% afterwards)
                          Estonia (100%)                                 Cyprus
                          Greece (up to 100% for civil servants)         Lithuania
                          Czech Republic (100% above EUR170
                                                                         Luxembourg
                          per month)
                          Poland (up to 100% depending on
                                                                         Malta
                          income)
                          Latvia (100%)                                  Netherlands
                          Lithuania (100%)                               Austria
                          Hungary (up to 100% dependent on
                                                                         Poland
                          income and type of work)
                          Malta (100%)                                   Portugal
                          Austria (100% above EUR500 per month)          Romania
                          Romania (100%)                                 Finland
                          Slovenia (reduction of 40% in the first
                                                                         Slovak Republic
                          three years; 80% afterwards)
                          Iceland (up to 100% depending on
                                                                         Croatia (up to 4hrs
                          income)
                                                                         per day worktime)
                          Italy (up to 100% depending on income)
                          Slovak Republic (up to 100% depending
                          on income)
                                                                         Sweden
                          Portugal (reduction depends on duration
                          and type of work
                                                                         Czech Republic
                                                                         Hungary
                                                                         Latvia
Source: EU MISSOC Database (Social Protection Systems in Member States)




                                                                    54
4. Pension projections for hypothetical careers

4.1 Projection settings and assumptions

Findings and the discussion in this section are based on the results showing theoretical gross
replacement rates (TRR), i.e., expected gross replacement rates (first pension vs. last wage)
for selected individual careers under common assumptions on macroeconomic
developments and current pension system parameters and retirement rules. These are not
forecasts; calculations represent scenario analyses, where theoretical replacement rates
involving policy changes are compared to no-policy change scenario. Effects of policy changes
aim to signal the potential impacts of pension policy measures and inform the selection of
policy scenarios to be tested and discussed in Section 9 of this Report where microsimulation
results are presented. Background assumptions used in the illustrative calculations in this
Section reflect the state of the pension system as of January 2024 after legislative changes
enacted at the end of 2023. Background assumptions include the following:
    • retirement of a representative worker on January 1,
    • rates of return of mandatory pension funds up to December 31, 2023, are based on
        MIREX (obligatory pension funds’ - OPFs’ index) true values,
    • starting from 2024, the real rate of return for fund categories A (high risk), B (balanced)
        and C (low risk) are assumed to be 3.4%, 2.2% and 1.0%, respectively (annualized
        constant rates),
    • contributions to the second pillar (5% of gross wage) are paid at the end of the month
        for work done during the month,
    • annualized constant growth rate of the real average wage is set at 1.5%, while wage
        levels and wage growth rates until December 31, 2023, are taken from the official
        statistical sources (hence gross replacement rates),
    • a passive member of the second pillar follows regulatory default rules for changing
        pension fund risk categories A, B, C which were effective before 2024 and does nothing
        to change the category over entire career,
    • an active member is assumed to take the most active investment position allowed by
        the legislation before and after changes in 2024 which show effects of more flexible
        rules for portfolio allocation,
    • two types of wage evolution over the career are considered: a) flat wage profile (the
        individual-to-average wage ratio is constant over the entire career), and ii) rising wage
        profile (individual-to-average wage ratio starts with 65% of assumed career average
        ratio and linearly increases to 135% by the end of career),
    • annuity factor for the calculation of pension from the second pillar refers to a single
        life annuity without partial one-off cash withdrawal of savings at retirement; annuity
        factors are those valid as of February 15, 2024, applied by the pension insurance
        company which offered the highest annuity factor on the specified date,
    • career is uninterrupted until retirement, and the start of the career is adjusted by its
        assumed length,
    • projections are prepared for the next 40 years (up to 2064).


                                               55
4.2 TRR Baseline: no-policy change scenario


Gross theoretical replacement rates (TRR) are projected to decline gradually for both the
full PAYG pensions (P1) and the combined pensions (P2) due to pension point valorization
in the first pillar below the rate of wage growth.37 Figure 4.1 shows cases of old-age and early
retirement of male average wage earners. For retirement at the age of 65 after 40 years of
career, the gross TRR of full PAYG pensions is projected to decline from 38.4% in 2024 to
32.2% in 2064. In case of early retirement at the age of 60 years after 35 years of service, the
trend in gross TRR is the same, but at lower rates that starts at around 30% in 2024. The
projected decline is somewhat lower for combined pensions in both cases. The decline in both
P1 and P2 is caused by continuous 70%:30% valorization of pension point values with wage
growth and inflation applied in calculation of PAYG pensions. A word of caution is needed
here to remind that projections are based on the model with wage growth always above non-
negative inflation rate. In real-life situation, variability of wage growth and inflation can cause
that valorization takes values much closer to wage growth and therefore replacement rate
decline to be milder.38 But, even with declines in projected replacement rates, purchasing
power of future pension benefits is expected to rise for both P1 and P2 due to the expected
valorization of the first pillar pensions above inflation rate. Positive real rates of return of
pension funds will contribute to P2 real growth as well.
Figure 4.1 Gross TRR for a male average wage earner in cases of old-age and early retirement (M65/40
and M60/35)




37
   In this and other sections of the Report the gross replacement rates are reported and used for analysis. Annex
4 shows selected policy projections of net replacement rates. Results, trends and conclusions are the same as
with gross replacement rates.
38
    Section 6.2 confirms that the current valorization/indexation formula between 2019 and 2023 enabled
pension point adjustments that were relatively close to wage growth.

                                                       56
Note: Person is assumed to be passive pension fund member and experiences flat wage profile in the career.
Source: WB team projections.



In a no-policy change scenario, the projected P2 exceeds the projected P1 for old-age and
early retirement of a male average wage earner after 2029 (Figure 4.1). Before that, full
PAYG pension is expected to be somewhat higher. Poor performance of pension funds in 2022
and mandatory allocation to C fund at the age of 60 for men, which was effective until 2024,
results in lower P2 than P1 as funds of category C recorded worse results than A and B in
recent years. This is the reason why P2 for early retirement at the age of 60 did not suffer as
much as for those who reached statutory pension age. This finding holds under the
assumption that male fund members are passive and remain in fund B until the age of 60.
In the case of early retirement, the relationship between P1 and P2 depends on early
retirement decrements (penalty factors). These factors work differently in the first and the
second pillar. For the same amount of pension savings, those retired at the age of 60 will
receive 15.4% lower annuity than those retired at the age of 65.39 In the first pillar, pension
reduction for 5 years of retirement before statutory requirement age is only 12%. Therefore,
early retirement increases likelihood of choice of P1, other things being equal. Further
comparison of P1 and P2 for early retirement and retirement at statutory age shows that P2
is considerably higher than P1 for regular old-age retirement than for early retirement. For
example, in 2040, the P2/P1 ratio is estimated at 1.055 for M65/40 and 1.040 for M60/35
(early retirement).
The difference between the late retirement bonus factors applied in pension calculation in
the case of late retirement in the first pillar, and increases in annuity factors for comparable


39
     HRMOD annuity tables for single life annuity.

                                                          57
ages, also makes a difference regarding choice between P1 and P2. For late retirees, the
annuity factor for retirement at the age of 70 is 21.3% higher than at the age of 65,40 while
rapid bonification of late retirement in the first pillar will result in 27% higher pension benefit
5 years after the statutory retirement age (0.45% per month). Bonification of late retirement
is a logical policy to defer a pension in the PAYG. Some EU member states have even stronger
bonification scheme (e.g., Germany, Slovakia and Romania 0.5%, and Sweden 0.7% per
month, see Figure 3.17). The discussion here is about the implicit disincentives for the choice
of P2 created by an excessive bonification scheme only in the first pillar, motivating the new
retirees to opt-back to P1. We shall discuss policy options for elimination of this distortion
below.
For most pensioners, wage level itself is not the decisive factor that impacts attractiveness
of P2 relative to P1. Figure 4.2 shows P2/P1 ratio for wage levels from 50% to 330% of average
wage, and for generations of retirees in 2023, 2043 and 2063. P2/P1 is below 1 in 2023, except
for very high wage earners; however, for generations expected to retire in 2043 and 2063,
P2/P1 is mostly above 1, except for low wage earners. These findings contradict the common
opinion that the attractiveness of the combined pension depends on the wage level. Although
the wage level is important for minimum pension beneficiaries (those with career-wide wage
level below 81% of the national average wage), and for maximum pension beneficiaries (those
with career-wide wage level above approx. 300% of the national average), for most new
pensioners with benefits between minimum and maximum, and in cases of flat wage profile
and unchanged wage ratio over entire career, wage level per se is not the factor that impacts
attractiveness of P2 relative to P1. The same pattern is found for shorter careers and early
retirement (not shown here).
Figure 4.2 P2/P1 for three generations of new old-age pensioners and different wage levels (M65/40)

1.200

1.150

1.100

1.050

1.000

0.950

0.900
                         70%




                                                                                                                                                           260%
             50%
                   60%


                               80%
                                     90%
                                           100%
                                                  110%
                                                         120%
                                                                130%
                                                                       140%
                                                                              150%
                                                                                     160%
                                                                                            170%
                                                                                                   180%
                                                                                                          190%
                                                                                                                 200%
                                                                                                                        210%
                                                                                                                               220%
                                                                                                                                      230%
                                                                                                                                             240%
                                                                                                                                                    250%


                                                                                                                                                                  270%
                                                                                                                                                                         280%
                                                                                                                                                                                290%
                                                                                                                                                                                       300%
                                                                                                                                                                                              310%
                                                                                                                                                                                                     320%
                                                                                                                                                                                                            330%




                                                                  Career-wide individual-to-average wage ratio
                                                                                     2024                      2044                     2064
Note: Person is assumed to be passive pension fund member and experiences flat wage profile in the career.
Source: WB team projections.




40
     Ibid.

                                                                                                          58
The P2/P1 ratio also depends on the social policy element in PAYG pensions, such as the
rules for determining the minimum pension. Like in previous considerations about the late
retirement and old-age pensioners with long-term service, social transfer component of
pension biases the choice of the payout towards P1. Minimum pension provisions are applied
to basic pension by multiplication with ¾ factor, which implies lower effect of minimum
pension on multipillar pension. However, it can be argued that neither PAYG nor multipillar
insured persons “earned” minimum pension by payment of contributions.41 Therefore,
minimum pension beneficiaries will almost always choose P1 as shown in Figure 4.2. Since the
combined pensions are expected to increase faster than the PAYG pensions, a part of
minimum pension beneficiaries - those with somewhat higher wages and higher contributions
paid to the second pillar - are expected to find P2 more appealing than P1 in the long run.
Gender affects the pension choice in the transitional period until equalization of the
statutory retirement ages for men and women in 2030. For the same age and the same
career characteristics, women will receive higher first pillar pension benefits P1 up to 2030
(Figure 4.3, in comparison with Figure 4.1 above for men). Pension from the second pillar will
be the same as for men. Women’s combined pension for the 65/40 career will be relatively
lower than full PAYG pension in the entire transitional period. The added insurance period for
parenthood, which is almost always granted to mothers, will increase P1 relatively more than
P2 if applied. Overall, women will have somewhat less incentives to choose P2 compared to
men with comparable career in the next five years, and even longer in cases when women
are qualified for the added insurance period for mothers.
Figure 4.3 Theoretical gross replacement rates for female average wage earner with flat wage profile
having two children in cases of old-age and early retirement (F65/40 and F60/35)




Source: WB team projections.


41
  Minimum pension in Croatia is related to the insurance period, meaning that pension contributions must be
paid, but general pension formula determines pension benefit below the value of the minimum pension.

                                                    59
4.3 Impact of investment policy and wage profile on future pensions

Career wage profile and active/passive investment behavior can have a significant impact
on the second pillar pension benefit and on the choice between P1 and P2. Pension benefit
formula in the PAYG system takes into account the average career-wide relative wage
(individual-to-average wage) meaning that variation in the relative wage during career, i.e.
wage profile is not relevant. However, the personal account balance and therefore annuity
paid from the second pillar significantly depends on the wage profile in transition period when
the time spent in the second pillar is shorter than entire period of pension insurance. If in the
first part of the career, up to 2002, individual wage was well below the career-wide average,
and in the second part well above that average, that would not change PAYG pension if the
wage profile is flat. However, with a rising wage profile it would result in above-average
contributions paid to the second pillar later in the career. The resulting pension savings and
second pillar annuity will be higher due to wage evolution in the second part of the career.
Therefore, P2 for a rising wage profile will be higher than P2 for a flat wage profile. This effect
will gradually vanish as the length of career with pension insurance in the second pillar
converges to full career, i.e., as transitional generations retire. The effect on TRR can be seen
in Figure 4.4 by comparing a baseline P2 (passive fund member, flat wage profile) with a P2
for a passive fund member with rising wage profile. While baseline P2 is below P1 in 2024, in
case of rising wage profile P2 is above P1. Such a person is more likely to choose a combined
pension. Although the effect of wage profile is temporary, it may significantly influence
pension choices in the next 10 years.
Active investment choice also strongly affects gross replacement rate and P2/P1 ratio. For
an active member with a steeply rising wage profile, P2 exceeds P1 from the beginning of the
projection period.42 If flat wage profile is applied, P2>P1 condition is met a few years later,
in 2028. In case of active investment behavior, the difference between P2 and P1 becomes
striking and peaks 20 years from now. This is the result of longer time spent in portfolio A for
younger generations and avoiding allocation to fund C at the end of career. This result should
be taken with caution because of the assumption of no financial market volatility. Nevertheless,
more active investment choice can be beneficial for future combined pensions in the long run.43
Figure 4.4 Gross TRR for old-age pension of average wage earner (M 65/40) depending on wage profile
and investment behavior




42
   It appears that rising wage profile among a part of multipillar contributors is likely to be the main explanation
of actual choices of P2 made by new pensioners, around 21% in the first quarter of 2024, despite TRR calculations
based on flat wage profile suggesting that for all but maximum pension beneficiaries, P1 will be higher than P2
for retirement in 2024.
43
   Therefore, in Section 6.5 we propose a reform on mandatory pension funds investment policy in the form of
the Incremental Allocation System (IAS), which aims to provide a unifying framework for promoting more active
investment behavior.

                                                        60
Note: A passive member follows default ABC fund category distribution, while an active member chooses the most active
ABC fund category allowed (pre- and post-2024 legal changes).
Source: WB team projections.


4.4 Impact of wage growth and investment returns on choice of pension

Although it seems counterintuitive, higher real wage growth leads to lower replacement
rates for the first pillar pensions, which is due to valorization/indexation of pensions at the
rate below wage growth. Figure 4.5 shows how our assumption on optimistic (higher) real
wage growth results in lower replacement rates for full PAYG pension compared to situation
with pessimistic (lower) assumed wage growth. Due to valorization/indexation formula with
higher real wage growth, the difference between pension point value and the average wage
becomes larger affecting the replacement rate. That pattern also affects the basic PAYG
pension paid as a part of combined pension (P2). Therefore, P2 replacement rate also declines
faster with higher wage growth, other things being equal.
Impact of pension fund investment returns on replacement rate is straightforward: other
things being equal, the higher the net return, the higher the annuity from the second pillar
and the higher the P2/P1 ratio. Incentives for active choice of investment portfolio do not
imply fundamental change in pension funds’ returns. Long-term net returns depend primarily
on the quality of pension fund management, trends in financial markets and regulatory
investment limits discussed below.
Figure 4.5 Gross TRR depending on real wage growth and real rate of pension fund returns ((M65/40,
average wage earner)




                                                         61
     41.0%

     39.0%

     37.0%

     35.0%

     33.0%

     31.0%

     29.0%

     27.0%
                    2026




                                                2034
             2024



                           2028

                                  2030

                                         2032



                                                       2036

                                                              2038

                                                                     2040

                                                                            2042

                                                                                    2044

                                                                                           2046

                                                                                                  2048

                                                                                                         2050

                                                                                                                2052

                                                                                                                       2054

                                                                                                                              2056

                                                                                                                                     2058

                                                                                                                                            2060

                                                                                                                                                   2062

                                                                                                                                                          2064
                              Full PAYG pension - P1 - pessimistic real wage growth, baseline returns
                              Combined pension - P2 - pessimistic real wage growth & returns
                              Full PAYG pension - P1 - optimistic real wage growth, baseline returns
                              Combined pension - P2 - optimistic real wage growth & returns
Note: Baseline scenario assumes annualized real wage growth of 1.5% per annum and net real investment returns of 3.4%
in A, 2.2% in B and 1.0% in C. Pessimistic/optimistic scenarios assume +/- 1 pp in these rates.
Source: WB team projections.


A higher wage growth in combination with unchanged returns leads to a somewhat lower
P2/P1 compared to baseline scenario, but P2 is projected to rise faster than P1 in the future
in our TRR model (Figure 4.6). The reason for P2 to continue to grow faster is that a higher
wage growth means more contributions that are paid to the second pillar mandatory pension
funds, and therefore higher annuities from the second pillar. Therefore, annuities increase by
a higher percentage than the first pillar pensions which are subject to current
valorization/indexation at rates below the wage growth. Lower wage growth will result in a
higher P2/P1 ratio, other things being equal. More optimistic/pessimistic returns than in the
baseline scenario will lead to higher/lower P2/P1 ratio.44
Figure 4.6 Simulation of the P2/P1 ratio depending on real wage growth and real rate of pension fund
returns (M65/40, average wage earner)




44
   We did not show result for combinations of optimistic wage & optimistic returns and pessimistic wage &
pessimistic returns because results depend on the difference in rates of wage growths and returns. If the
difference is similar to baseline scenario, the P2/P1 ratio is similar. As the discussion on the long-term relation
between wage growth and returns is beyond the current ambition of this report, we did not report further
results.

                                                                                   62
120.0%


115.0%


110.0%


105.0%


100.0%


 95.0%
                                                   2036
         2024

                2026

                       2028

                              2030

                                     2032

                                            2034



                                                          2038

                                                                 2040

                                                                         2042

                                                                                 2044

                                                                                        2046

                                                                                               2048

                                                                                                       2050

                                                                                                              2052

                                                                                                                     2054

                                                                                                                              2056

                                                                                                                                     2058

                                                                                                                                            2060

                                                                                                                                                   2062

                                                                                                                                                          2064
                                               R - ratio P2/P1 (%) - baseline wage and returns
                                               R - ratio P2/P1 (%) - baseline returns, wage growth 0.5%
                                               R - ratio P2/P1 (%) - baseline returns, wage growth 2.5%
                                               R - ratio P2/P1 (%) - pessimistic returns, baseline wage

Note: Baseline scenario assumes annualized real wage growth of 1.5% per annum and net real investment returns of 3.4%
in A, 2.2% in B and 1.0% in C. Pessimistic/optimistic scenarios assume +/- 1 pp in these rates.
Source: WB team projections.


4.5 Choice between P1 and P2 and the impact on the pension adequacy and fiscal
    balance

The share of new pensioners choosing P2 is important for pension adequacy as higher P2/P1
improves it. Choice of P2 means that the pension benefit is higher than it would be without
multipillar insurance. The frequency of choice of P2 rapidly declined when the ratio of the
second pillar annuity to PAYG pension deteriorated in 2022 and early 2023, when higher
inflation and interest rates led to deterioration of the value of the second pillar savings in risk
category C. Out of total closed pension accounts of living members in the second pillar, share
of those who transferred their pension assets to PICs for annuity payouts decreased from 29%
in 2021 to 24% in 2022 and 15% in 2023 (Table 4.1).45 The declining trend was reversed in the
third quarter of 2023 after application of higher annuity factors by PICs. The share increased
further to 16.5% in the fourth quarter of 2023 and 21% in the first quarter of 2024.
Table 4.1 Payouts from the Second Pillar

                                                                                     2020              2021                 2022              2023        2024:Q1
 Number of PICs' annuity contracts,
 second pillar                                                     1                 4,419             7,813           11,097               13,288
 Change in number of PICs' new annuity
 contracts in the period                                           2                 2,906             3,394                3,284            2,191
 Closed pension accounts in the second
 pillar pension funds*                                           3=4+5           10,944               12,094           13,816               15,367               2,378


45
  There is hidden benefit of the second pillar, which is rarely mentioned, related to inheritance. In recent years
around 10% of closed accounts are accounts of members of the second pillar who passed away. Inheritance paid
out from those accounts amounted to EUR24 million in 2022.

                                                                                63
 Opt-back to the PAYG (number)                     4            7,974     8,588     10,527     13,047      1,879
 Transferred to PIC                                5            2,970     3,506      3,289      2,320        499
 Opt-backs as % of total closed pension
 accounts                                         6=4/3         72.9%     71.0%      76.2%      84.9%      79.0%
 Transferred to PIC as % of total closed
 pension accounts                                 7=5/3         27.1%     29.0%      23.8%      15.1%      21.0%
 Assets in opt-back (transferred to gvt.
 budget) in EUR million                            8             97.1     116.3      149.3      200.0       26.0
   Assets in opt-back as % of total
 transferred assets                                9            52.0%     51.6%      58.9%      71.7%      59.8%
 Assets transferred to PICs (for payment
 of second pillar annuities)                       10            89.7     109.3      104.1       79.0       17.5
    Assets transferred to PICs as % of
 total transferred assets                        11=9/5         48.0%     48.4%      41.1%      28.3%      40.2%
 Average assets in opt-back per
 beneficiary in EUR                              12=8/4    12,182.1     13,547.4   14,186.3   15,329.2   13,837.1
 Average assets transferred to PICs in
 EUR                                             13=10/5   30,191.1     31,174.3   31,637.2   34,051.7   35,070.1
 Ratio of average transferred assets (PICs
 vs. opt-back)                                  14=13/12          2.5        2.3        2.2        2.2        2.5
 *Without closures due to inheritance for
 people who passed away                                                              1,723      1,812        553
Source: REGOS Statistical Indicators, HANFA Monthly Report.




Choice between P1 and P2 has an impact on the shares of assets transferred to PICs and
PAYG scheme, which in turn has complex implications for the fiscal sustainability: in the
short run, the more pensioners choose P1, the higher PAYG scheme current revenues due
to transfer of pension assets to the state budget; in the long run, this effect may reverse. As
the average value on the individual pension account transferred to PICs is more than twice
higher than the average value of accounts transferred back to the PAYG, shares of assets
transferred are higher than the share of members. Asset shares were also declining from 48%
in 2021 to 41% in 2022 and 28% in 2023, but it reversed recently and reached 40% again in
the first quarter of 2024. Asset shares are important because fiscal effects of the multipillar
reform are proportional to assets. However, fiscal effects are complex: the difference
between P1 and basic component of P2 equals PAYG savings, which is present in case of
choice of P2, but asset transfer to the state budget in case of choice of P1 is recorded as fiscal
revenue. This problem is elaborated in detail in the section on fiscal transition cost.
Since the second half of 2023, the share of new beneficiaries choosing P2 has been
recovering due to: (i) new annuity factors and (ii) declining share of people who opt-back
as early retirees and/or enjoy disability or survivors’ benefits. In the coming years, old age
pensioners who accumulated second pillar pension assets for longer service periods will gain
a larger share in the population of new pensioners and determine the choice of payout. The
application of a new 27% supplement formula for basic pension will also contribute to a higher
share of P2s.




                                                           64
4.6 Cash withdrawal and the choice between P1 and P2

The opportunity to withdraw 20% of the second pillar accumulation at retirement may bias
individual decision towards P2 irrespective of the P2/P1 ratio. From 2024, the combined
pensioners fulfilling the legal condition are eligible to withdraw 20% from their second pillar
accumulation at retirement (15% was applied in 2019-2023). There are opposing views
related to cash withdrawal and regulatory conditions for access to this right. Paternalistic
argument is that (myopic) people have distorted (very high) subjective discount factors. If
they are (relatively) low-wage earners, withdrawal will jeopardize their long-term pension
adequacy and possibly lead to old-age poverty. Higher demand for old-age social protection
may eventually cancel the positive short-term fiscal sustainability effect associated with more
frequent choice of P2 due to cash withdrawals. On the other hand, freedom of choice and
equality arguments call for abandoning any regulatory condition for access to withdrawal
right and giving it to all. People may have good personal reasons for withdrawals: some enter
pension with (expensive) debt, and they can repay it after withdrawal; some know better
opportunities for investment, and others may have other good reasons such as inheritance,
medical treatments, terminal illness, even enjoying the first few years of retirement while
health is still good, which may be welfare enhancing. Arguments in favor of freedom of choice
get stronger as general living standards, employment rates and service years grow, and
threats of old-age poverty diminish for younger generations who participate in the second
pillar. Following recent regulatory change of lowering the threshold for access to cash
withdrawal right from having the basic PAYG pension of at least 115% to 100% of the
minimum pension, full abandonment of that threshold should be considered after
implementation of research to estimate the risk of choice of P2 by people with P1>P2 due to
severely distorted time preferences. Such people may be at risk of old-age poverty later in
their pension age.
Lump sum withdrawal from the individual account is not the only reason for imperfect
predictive power of P2/P1 for the choice of payout method: different indexation methods
can also explain why some people may prefer P1 despite P1<P2.46 With P2/P1 closely around
value of 1, and if beneficiaries expect to live long enough (and expect long-term growth of
real wages in the economy), opting-back to PAYG and choosing P1 may be a rational choice
despite P2>P1. Namely, PAYG pension benefits are indexed by the combination of wage
growth and inflation, while second pillar pensions paid out from PICs are indexed only with

46
  Lack of data and research made it impossible to decompose many opposing factors which bias choice in favor
of P2 or P1, but, overall, P2/P1 is a reasonably good predictor of choice. To better understand the impact of cash
withdrawal, we segmented the database on informative pension calculations for new pensioners eligible for
obtained from REGOS (for 2020) and found out the following: (i) 48.5% of all retirees from the second pillar had
the right to withdraw 15%, and (ii) 63.7% of members who decided to take multipillar pension had the right to
withdraw 15%. Shares are in line with significantly higher pension assets held by members who chose combined
multipillar pension. 81.4% of members with P2/P1 > 1 chose P2 and 93.5% of members with P2/P1 > 1.02 chose
P2. Furthermore, 79.3% of members with P1>P2 chose P1, however, share increases under condition P1/P2>1.02
up to 90.1%. Therefore, P2/P1 was roughly 90% confident predictor of choice outside +/- 2% interval around
P2/P1 = 1. For the remaining 10%, even a large difference between P2 and P1 outside +/-2% interval has no
predictive power.


                                                       65
inflation, meaning that P1 is highly likely to be indexed by a higher percentage than P2.
However, different indexation methods make the prediction of dynamic differences produced
by different indexation methods very hard, even for the best macroeconomists and financial
professionals. Also, different indexation dynamics have ambiguous fiscal implications: more
generous P1 indexation attracts more people to P1, which reduces potential fiscal
sustainability effect. Opting-back to P1 creates short-term fiscal gain for PAYG system due to
the transfer of pension savings from obligatory pension funds to the state budget, whereas
long-term fiscal burden grows due to the obligation to pay full P1 pensions instead of basic
component of the multipillar pension in line with indexation rules.




                                             66
5. Fiscal sustainability of the PAYG system

5.1 Pension expenditure, contribution revenue and deficit of the PAYG system


The average annual PAYG deficit exceeded 4% of GDP in Croatia in the 2012-2022 period.
Croatian PAYG system has high fiscal deficit as the average pension expenditure was 10.3% of
GDP and average contribution revenue was 5.9% of GDP in 2012-2022 period. A recent drop
in relative pension expenditure below 10% of GDP (Figure 5.1) was partly cyclical. It was driven
by rapid GDP recovery in the post-pandemic period (13.0% in 2021 and 7.0% in 2022) and
high inflation during pandemic after-shock.47 The upcoming cyclical normalization will lead to
higher pension expenditure ratio. It increased from 9.1% of GDP in 2022 to 9.4% in 2023 and
the deficit was widened from 3.5% to 4.0% of GDP in the same period (without transfer of
pension assets from the second pillar to the central government budget due to opt-backs,
which amounted at 0.26% of GDP in 2023).
Figure 5.1 Pension expenditure, contribution revenue and PAYG deficit in % of GDP

                 12

                 10

                 8

                 6
      % of GDP




                 4

                 2

                 0

                 -2

                 -4

                 -6
                      2012.   2013.   2014.    2015.     2016.      2017.    2018.   2019.      2020.   2021.   2022.

                                 Expenditure for pensions                Contribution revenue
                                 o/w transfer from the 2nd pillar        Deficit

Notes: Pension expenditure does not include administrative cost and transfers not related to pensions, which is in line with
EC’s Ageing Report methodology. Contribution revenue includes capitalized contributions – transfer of pension savings to
the budget for members in the second pillar who opt back to PAYG in the payout stage.
Source: Croatian Pension Insurance Institute (HZMO).

Transfer of capitalized contributions from the second pillar to the state budget improves
the PAYG fiscal balance in the short term but is likely to increase challenges for achieving
fiscal sustainability in the long term. Therefore, it is important to view transfers from the
second pillar in a different way from regular pension contributions of workers. As more new

47
     Sudden increase in the rate of inflation in 2021-2022 created a time lag related to pension indexation.

                                                                    67
retirees per year come from the multipillar pension insurance system in the coming years,
and more of them - in terms of number - may choose opting-back to the PAYG in the payout
stage, public revenue from the transfer of unused pension savings from the second pillar will
technically improve fiscal balances. However, in the long run, there will be higher expenditure
on PAYG pensions compared to a situation without opt-back, and therefore fiscal
sustainability of the PAYG system will be under pressure in the long run.48
Transfer of assets from the second to the first pillar requires careful monitoring of its effects
in the future when this revenue becomes more important. For the time being, the barely
visible gray line in Figure 5.1 above is related to the transfer of capitalized contributions. In
the national fiscal accounts, and in HZMO’s financial reports, which are widely used by
politicians and public administration, this transfer is recorded as current contribution
revenue. There is no accounting for a counterbalancing financial item which would measure
the increase in liabilities related to this transfer. The missing counterbalancing item is one of
the major flaws in national fiscal statistics and public sector financial reports.49 This
accounting principle is different from private pension insurance on a defined contribution
basis (unfunded or funded) where any change in the present value of the expected long-term
cash outflow (technical reserve) must be priced-in and covered. The problem will grow as the
share of the second pillar members among new pensioners increases. Policy decisions must
internalize the long-run negative fiscal impact on top of positive short-term fiscal impact of
asset transfers to the budget.
The decline in the contributions-to-GDP ratio in the PAYG system in 2021 and 2022 was
mainly due to a high GDP growth, while contribution revenues continued to grow in
absolute terms driven not only by wage growth but also by increase in employment.
Average annual employment growth measured by the rise in the number of pension insurance
contributors in 2021-2023 was 2.4%. Total number of contributors in 2023 was the highest in
the last three decades.
International comparison of contribution revenues of the public pension scheme shows that
Croatia collects less than most of the EU countries (Figure 5.2). Only Ireland, Bulgaria and
Sweden collect less contribution revenues in the first pillar in terms of GDP than Croatia.
Three major reasons are: i) low employment rate and short service period, ii) low effective
contribution rate, and iii) a significant segment of pension expenditures (special pensions,
transfers, social elements) financed from the general revenues. In 2022, Croatia had
employment rate of 65.9% for the age group 15-64 years, which was the fifth lowest in the
EU, with the EU average at 69.8%. As for the contribution rate, there is high variability across
the EU which includes various rates for different groups of contributors, and it appears that
Croatia is in the middle (EU 2024 Ageing Report, p. 154). Finally, the level of pension

48
   Asset transfers happened before (2014-2016) as well, when occupational groups with special pension benefits
were shifted from the second pillar back to the first pillar, but such asset transfers are disregarded in our analysis
as these were one-offs.
49
   According to the European System of National Accounts (ESA2010), European countries are obligated to
record a full set of position and flow data on pension obligations of all social insurance pension schemes.
Estimates of the pension entitlements are prepared with the methodology of accrued-to-date liabilities (ADL)
and reported by Eurostat and country’s statistical offices as Table 29.

                                                         68
expenditures financed from general revenues, described in the next section, exhausts almost
all of the pension deficit. Consequently, the low contribution-to-GDP ratio may indicate a
need for policies to broaden the contribution base further, but not to increase the overall
contribution rate.
Figure 5.2 Pension expenditure, contributions, and balance of the public scheme (% of GDP)

                20.0


                15.0


                10.0
     % of GDP




                 5.0


                 0.0


                -5.0
                       BG IT AT FR HR RO PL EL EE SK HU SE SI CZ DE IE ES CY NL LT LU FI LV MT PT

                             Contributions     Gross expenditure     Pension system balance

Source: EU 2024 Ageing Report (p. 35).



While the general pension contribution rate stands at 20% of gross wage, of which 5% is for
the second pillar, different types of income earners pay lower contributions relating to the
income earned either due to lower contribution rate or due to lower contribution base. The
contribution rate for other income earners (e.g., intellectual property contracts or service
contracts) stands at 10% of gross earnings in comparison to 20% for salaried workers. Given
the same pension formula, this leads to a situation where other income earners might have
the same pension benefits as the salaried workers who paid double the amount of
contribution for the same amount of income earned. That outcome depends on the amount
of income earned (in comparison to the average wage) as well as choice of the PAYG vs
combined pension.50 On the other hand, for several categories of income earners, mainly
related to independent activities such as craftsmen, self-employed in agriculture and forestry,
artists, sportspersons, different freelance activities, etc., the government each year
prescribes the contribution base calculated as a product of the amount of average salary and



50
   For example, in case of other income earned in the amount of average wage, the person is entitled to one
year of service and a value point of 1, the same as the salaried worker with average wage, despite paying only
half of the contributions. However, as the second pillar contribution for other income is 2.5% instead of 5%,
although income earner will receive the same basic pension as salaried worker, he/she will receive half as much
pension from the second pillar because the contributions are half as much as for the person who earns a salary.
In the case of 50% of average wage, salaried worker will end up with higher pension benefit as he/she will be
entitled to the minimum pension because of an average value point of 0.5, while a person who, for example,
earns only other income at the same level throughout his working life will not receive the minimum pension
because of an average value point equal to or greater than 1.

                                                          69
the coefficients for calculation which are prescribed by the Law on Contributions.51 Therefore,
many categories of workers are paying the prescribed minimum contributions irrespective of
the income they earn.52 For example, the minimum contribution base in 2024 (EUR 592.80) is
at 71% of the minimum gross wage (EUR 840), while the contribution base for craftsmen on
a flat rate regime (EUR 624) is at 74% of the gross minimum wage.53 No wonder some of these
activities gained popularity recently in the form of fake self-employment,54 as they pay lower
taxes and social security contributions in comparison to salaried workers for the same amount
of income earned. Moreover, a lower contribution base for different categories of income
earners often brings them into the category of minimum pension beneficiaries once they
retire. So, even if their contribution base was increased to, for example, the level of minimum
wage, they would still end up in the minimum pension category, leaving the adequacy of their
pension income unchanged.
Paying lower contributions for some type of income earners in comparison to salaried
workers not only distorts principles of fairness, reciprocity and solidarity of the pension
system, but it also leads to lower adequacy of their pension income at the old age. Although
engagement in new and flexible forms of employment, such as temporary jobs, professional,
and other freelance-type economic activities, usually leads to greater insecurity as well as
shorter and more fragmented careers, these persons should not be marginalized also in their
retirement by disregarding their social security contributions while working. Lower paid
contributions in relation to the earned income compared to salaried workers also raises
questions of fairness, reciprocity and solidarity of the system, as in this way salaried workers
pay much more in the proportion to their earned income that these other groups while in the
end they have similar pension rights. Of course, tax (contribution) deductions can and should
be applied for different self-employed categories, however, improving adequacy for some of
these categories requires changing the coefficients or introducing new ones, raising the
effective contribution base and, ideally, equalizing contribution burden for all types of
work/earnings - either by equalizing contribution rates or calculating pension points
according to the paid contribution instead of the gross income. In essence, income from all




51
   The average salary refers to the average monthly gross salary per employee in legal entities of the Republic of
Croatia for the period January-August of the previous year. These coefficients go from 0.38 (applied to, for
example, person insured based on agricultural activities), 0.65 (applied to, for example, craftsman who
determines income from self-employment), to 1.1 (applied to, for example, athlete who determines income
from self-employment). The highest contribution base is also defined in a similar manner – as the average salary
times the coefficient of 6.00.
52
   There are limitations as to the type of activities you can be registered for depending on your income (e.g., the
threshold for entering the VAT system currently stand at EUR 40,000).
53
   One category of income taxpayers that pays taxes according to the flat rate income – short-term renters in
tourism – do not pay contributions at all, just a per-bed or per-accommodation flat tax, depending on the local
government unit the accommodation is registered in.
54
   For example, IT engineers in some IT companies are being hired full time as self-employed in the form of crafts
for which tax is paid according to the flat rate income or alternative channels, instead of salaried employment.

                                                        70
work arrangements should be treated equally, as part of a shift from work-based to income-
based social security (Vukorepa et al., 2017). 55


5.2 PAYG deficit and transition cost of multipillar reform

Accounting treatment of different types of contribution revenues and calculation of PAYG
deficit is closely related to the widely debated issue of the transition cost of the multipillar
pension reform. Assessment of its impact on the PAYG deficit is subject to methodological
complexities which also involve the fact that pension expenditure reflects several
components of additional rights (e.g., minimum pension, survivor’s and disability pension,
higher pensions for war veterans and privileged groups of pensioners) which have stronger
impact on PAYG deficit compared to transition cost of the multipillar reform. Some of these
rights are social transfers which serve old-age poverty prevention, and some of them reflect
other collective preferences. They have in common the fact that these rights were not earned
– they are not proportional to individual contributions. Therefore, the large deficit of the
PAYG system does not automatically call for a higher rate of pension contribution if society is
willing to finance contribution-benefit gap for certain groups of pensioners from other
general fiscal revenues, on top of covering for the transition cost of the multipillar pension
reform. If the overall fiscal envelope allows it, transfers from the central government budget
instead of higher pension contributions may prevent additional intergenerational
redistribution and adverse effects on the competitiveness of labor of present and future
working generations.


      5.2.1 Pension Institute’s (HZMO) accounting approach to identification of transition
            cost


In HZMO’s accounts, the transition cost is based on the concept of residual fiscal transfer
from the central government budget. It is residual after accounting for other types of
identified fiscal transfers from the central government budget to HZMO. Transfers for
coverage of the PAYG deficit include three groups of transactions, of which the first two are
legally determined and itemized and the third is calculated as a residual: (i) transfers for
legislated pension supplements including the transfer according to the Law on increase of
pensions with a purpose of elimination of differences in pensions acquired in different
periods, (ii) transfer for higher pensions under special regulation and one-off assistances (e.g.,
transfers for Covid and energy assistance were recorded here), and (iii) the fiscal residual,
which includes the transition cost of the multipillar pension reform. While transfers (i) and (ii)
are calculated precisely based on excess pension rights over rights earned according to the
Law on Mandatory Pension Insurance, transition cost is a simple residual - a missing transfer

55
  For more detailed discussion on the relationship between earnings and pension rights of different categories
of self-employed workers in Croatia see Vukorepa et al. (2017) and Grgurev and Vukorepa (2018).


                                                     71
which is calculated as aggregate difference between total PAYG deficit and identified transfers
under (i) and (ii). This method of calculation of fiscal residual reflects cyclical component
because it is influenced by other cyclical factors besides transition cost itself. Therefore it is
not an appropriate measure of transition cost as transition cost is directly associated with a
contribution revenue forgone due to redirection of 5% of gross wages to the second pillar. An
economic approach to the analysis of the transition cost, which overcomes the limits of the
HZMO’s accounting approach, is required (and presented in Section 5.2.2.) for making sound
policy conclusions. Figure 5.3 shows that the transfer from the central budget to the pension
system stood between 4 and 5% of GDP in the past decade, with its share dipping below 4%
in 2022. The structure of the three components of the budget transfer has not changed
significantly over the years. Long-term average of the HZMO’s residual approximately equals
cash-flow based contribution revenue foregone; contribution payments to the second pillar
represented 1.6% of GDP in 2022 and 2023. Between 2.5% and 3.0% of GDP per annum are
other budgetary transfers related to the contribution-benefits gap. Such financing structure
reflects collective preferences and political decisions that point at no immediate necessity to
raise the first pillar rate of contribution.


Figure 5.3 Budget transfers for covering PAYG deficit in % of GDP 2012-2022

 6%

 5%
                1.6%
        1.4%                                                                          2.2%
 4%                         0.8%      1.5%
                                                1.9%
                                                            1.7%    1.5%    1.5%                 1.5%
                                                                                                         1.0%
 3%
        2.0%    2.1%        2.1%
                                      1.8%                                            1.6%
 2%                                             1.4%         1.4%    1.4%    1.4%                 1.5%    1.6%


 1%
        1.5%    1.5%        1.5%      1.4%      1.4%        1.3%    1.3%    1.3%      1.4%       1.3%    1.1%
 0%
       2012.   2013.       2014.     2015.     2016.        2017.   2018.   2019.    2020.       2021.   2022.

                       Residual including transition cost
                       Pensions under special regulation and one-off assistances
                       Pension supplements and elimination of differences in different periods

Source: HZMO




      5.2.2 Economic identification of the transition cost and fiscal effects of the reform


Economic model of the transition cost reflects the idea of the multipillar pension reform
generating fiscal savings over the long run, which would compensate for the Gross
Transition Cost (GTC), i.e., lost fiscal revenue associated with redirection of a part of pension
contributions from PAYG to the second pillar. However, the concept of fiscal savings in the
long run is not easily measured. Firstly, additional fiscal savings can arise through indirect
                                                              72
channels such as higher incentives for longer work and fiscal discipline, positive impact of
financial development on economic growth, etc. In short, there is a difference between gross
transition cost (contributions to PAYG foregone) and net transition cost (GTC minus indirect
benefits). Secondly, political decisions taken in the last 20 years in the PAYG system have led
to a reduced P2/P1 ratio compared to P2/P1 expected at the beginning of reform due to
higher P1. Because of it, new retirees who opt back to PAYG and set the limits on fiscal
savings56 do not violate the principle of expected fiscal savings established at the beginning
of the reform. However, the transition cost period is prolonged by political decisions on more
generous PAYG benefits taken after the reform's inception. Despite of it, significant fiscal
savings occur already due to payment of (lower) basic pension instead of full PAYG pension
for new retirees who choose default by remaining in the combined payout model P2 even
under new rules. Also, visibly positive fiscal effects arise related to transfer of the pension
savings to the state budget for members who opt-back. However, this is the current cash flow
perspective. The economic definition of fiscal savings in the case of opt-back requires that net
returns in the mandatory pension funds exceed cost of public debt accumulated due to
contribution revenues foregone.
Net transition cost (NTC) is smaller than gross transition cost (GTC) due to: (i) improved fiscal
compliance, (ii) improvements in functioning of financial markets and potential positive credit
ratings effects, and effects of financial development on economic growth (finance-growth
nexus), and (iii) stronger motivation for longer work due to tighter relationship between
contributions and total pension. Contribution to gross-net difference can also come from
optimization of size, structure, and quality of public expenditures. This is the point where the
proponents and critiques of the reform diverge most. Critiques claim that net socio-economic
effects of the multipillar pension reform may turn negative because present uses of additional
fiscal resources would be more efficient compared to its use for pension savings. Proponents
find it hard to fight this claim based on empirical arguments because it is hard to convincingly
isolate and measure adjacent positive effects. Although there is plenty of empirical evidence
showing inefficiencies in the use of public fiscal resources in Croatia,57 this debate is latent.
Inclusion of the cost of accumulated public debt (APD) in the calculation of gross transition
cost (GTC) is also an open issue. Using the concepts of contribution revenue foregone, labeled
CRF in the remainder of this section (5% redirected from PAYG before reform to the second
pillar), and cost of accumulated public debt, APD, as measured by long-term government
bond yield multiplying a part of gross public debt which was created by the accumulation of
CRF, we can define GTC as CRF + i x APD, where i stands for long-term government bond yield.
Net transition cost (NTC) is CRF + i x APD – X, where X denotes adjacent socio-economic
benefits described in the previous paragraph. It makes sense to include i x APD in the GTC
formula if measurement of fiscal savings related to the reform include the return on CRF
accumulated in the second pillar, which will be shown to be the case.
Accumulated public debt (APD) due to CRF is not perpetual as it declines in the long run:
visible fiscal savings begin to emerge when members of the second pillar begin to retire.

56
     Instead of basic component of combined multipillar pension only.
57
     See, for example, Sopek (2011), Bađun et al. (2014), Ahec-Šonje et al. (2018) or the World Bank (2020).

                                                         73
Visible long-term fiscal savings emerge because members of the second pillar consume fewer
fiscal resources after retirement. Their pension in a no-reform (or opt-back) scenario, P1, is
higher than the basic pension component of the multipillar pension. There is a static
difference (fiscal savings in the case of default choice of P2) between the two pensions in the
moment of retirement, and there is a dynamic difference which increases in time because of
indexation: applying the same rate of indexation on P1, which is higher than the basic pension,
leads to increase in the absolute amount of difference in time.
Both assets transferred to the pension insurance companies for annuity payments (ATP)
and assets transferred back to the government budget for members who activate opt-back
clause (OBC) include mandatory pension funds’ accumulated returns . So, conceptually, one
can think about multipillar pension reform net cost (MPRNC)58 in a very simplified way as
MPRNC = GTC – OBC – ATP – Y, where Y stands for (i) the dynamic difference related to
indexation described above, and (ii) eventual difference between the present value of true
future differentials between P1 and basic pension of second pillar members who remained in
the default payout mode, and ATP. If net transition cost could be measured, formula would
be MPRNC* = NTC – OBC – ATP – Y. MPRNC without Y component is used in the following
calculations, but reader should keep in mind that it overestimates the net cost of the reform
because it assumes GTC = NTC (X=0) and no additional fiscal savings related to indexation
effects (Y=0). Formula shows that MPRNC declines as members of the second pillar retire. The
question is: by how much?
Time perspective is critical for proper assessment and measurement of the multipillar
reform cost MPRNC due to time aspect of the definitions of transferred assets to the PAYG
(OBC) and PICs (ATP). Relation MPRNC = GTC – OBC - ATP shows that it is not possible to
assess long-term fiscal impact of the multipillar reform before the first generation that had to
enroll the second pillar in 2002 retires.59 Time aspect is also critical because OBC and ATP are
present value approximations of the future flows of pensions which are partly uncertain. Main
reason for uncertainty is related to unknown residual life expectancy of new retirees. Ideally,
in a fully developed micromodel and under assumption of perfect information, difference
between P1 and basic pension can be calculated for each person in the beginning of
retirement. Perfect information implies that residual life span for each person is known, so
the difference is calculated for each year into the future until person passes away. Here, a
simpler approach is used: ATP approximates the present value of future flow of the difference
P1-basic pension. ATP depends on annuity factors which contain protection of pension




58
  Distinguish MPRNC from the net transition cost NTC.
59
   Recall that population of new retirees who exited the accumulation stage in the second pillar so far is not
representative population of new retirees due to prevalence of early, survivors and disability pensions, and
disproportionally high shares of voluntary contributors who were at the age of 40-50y in 2002. In this population
there is probably a selection bias towards P1.

                                                       74
providers against longevity risk, which is probably high.60 Therefore, ATP should be a good
proxy for present value of future differences between P1 and basic pension.
There are two important aspects of the interpretation of the transition cost: (i) if assets
transferred back to the state budget (OBC) are interpreted as government revenue (including
principal and returns), then assets transferred to PICs (ATP), i.e., the present value of the
future second pillar pensions, which ideally equals aggregate fiscal savings, can be calculated
as a factor which immediately reduces transition cost; (ii) in the long run, if the mandatory
pension funds’ net returns are equal to interest on government bonds (CRF and principal
component of OBC offset), and if all members opt-back from the second pillar to PAYG in the
payout stage, all multipillar reform effects would cancel out. This is reform neutrality
condition. Effects would be neutral for society in the long run if the difference between gross
and net transition cost is disregarded. If the net cost is smaller, reform may have positive
effects even if long-term government bond yield is higher than the rate of net return of
mandatory pension funds.
In the economic model, cash flows should not be confused with economic effects that span
over long periods of time. When properly defined MPRNC begins to decline, it does not
necessarily decline on a cash-flow basis. Cash-flow effects will effectuate in the future
because of the present value logic. Method of measurement in Figure 5.4 presents MPRNC as
a measure of future dynamic (fiscal) path of the pension reform effects: rising MPRNC in time
implies lower probability of net fiscal effects turning positive (MPRNC turning negative) in the
foreseeable future, and vice versa. Before retirement of the first generation after
implementation of the reform, MPRNC will tend to grow by mere logic of demography as the
first generation still works.
Figure 5.4 Multipillar pension reform net fiscal cost (MPRNC) since inception of the reform in % of GDP,
with (MPRNC2) and without (MPRNC1) cost of accumulated portion of public debt




60
   Experiences in developed countries showed selection bias on the side of demand for pension annuities: people
with higher incomes are better educated and live longer on average. REGOS data on asset transfers point at high
likelihood that this is the case in Croatia as well: ratio of average assets transferred from mandatory pension
funds to PICs to average assets transferred back to the state budget ranges between 2.2 and 2.5. Therefore,
private pension providers most probably protect their equity with implementing buffers that protect them from
risk of longevity of their specific clients.


                                                      75
 2.5%



 2.0%



 1.5%



 1.0%



 0.5%



 0.0%
          2002

                 2003

                        2004

                               2005

                                      2006

                                             2007

                                                    2008

                                                           2009

                                                                   2010

                                                                          2011

                                                                                 2012

                                                                                        2013

                                                                                                2014

                                                                                                       2015

                                                                                                              2016

                                                                                                                     2017

                                                                                                                            2018

                                                                                                                                   2019

                                                                                                                                          2020

                                                                                                                                                 2021

                                                                                                                                                        2022
                                                                  MPRNC 1                      MPRNC 2

Note: CRF is true cash flow of gross contributions to the second pillar (including 0.5% entry fee). HANFA data are missing for
the first three years, so the Team made an approximation based on visual inspection of graphs having in mind the overall
envelope of cumulative data which is available (Tables 1.3. and 1.5. from HANFA Monthly Reports).
Source: HANFA for CRF, Croatian Statistical Office for GDP, REGOS for OBC and ATP (2020-2022 only) and Croatian National
Bank for average yield on long term (2y+) government bonds.

Maximum MPRNC without interest cost of APD included (MPRNC 1) was reached in 2019 at
around 1.5% of GDP (Figure 5.4). The decline began when the number of retirees from the
initial reform generation began to retire. This trend will continue by mere logic of
demography. Several years after the initial reform generation will become dominant among
new retirees (which will happen around 2028), MPRNC will show full dynamic impact. Impact
of interest component, which is reflected in MPRNC 2 in the Figure was at a maximum in 2012
(0.7% of GDP). It has been declining since then, largely reflecting volatility of government
bond yields which explains a spike in 2022. It is related to a general rise in interest rates from
previous historic lows, and it will stabilize in 2023-24. In general, dynamics of the higher line
will follow fundamental dynamics of the lower line in the long run.




                                                                                 76
6. Discussion on key policy issues

6.1 Multipillar pension system vs. PAYG-only: is there a need for systemic change
    again?

Croatian model of choosing the type of pension payout at retirement stands between the
two corner solutions: i) PAYG-only coupled with voluntary pension savings and ii)
mandatory payouts from the second pillar without the possibility to opt-back as was
envisaged at the inception of the multipillar pension reform. The first corner solution is
applied, for example, in Germany, Austria, Slovenia and Czech Republic, while the second one
is still implemented in Latvia, Romania, and several other non-EU European and Central Asia
countries. In Croatia, the concerns about pension adequacy prevailed over the concerns about
fiscal sustainability of the PAYG, which led to opt-back to PAYG if it means higher individual
pension adequacy.
 Over the last 15 years many of the CEE countries either completely dismantled their second
pillar or introduced the possibility for it to gradually reduce its role, considering exclusively
short-term effects. Hungary and Poland dismantled their second pillar, while Lithuania,
Estonia, Bulgaria, and Slovak Republic allowed the opt-out from second pillar in the
accumulation stage in a gradual policy scenario. Hungary dismantled the system without
gradualism in 2011. Common denominator of those changes is related to competition
between multipillar and PAYG-only pension system regarding expected pensions in the short-
run. PAYG is generally seen as a guarantor of pension adequacy and social protection. Table
6.1 summarizes policy options which emerged across the CEE in a wave of pension reform
reversals that started after inception of the Great Recession in 2009.
Table 6.1 Taxonomy of possible interactions between multipillar pension system and PAYG- in CEE EU
countries
 PAYG + voluntary pension        Multipillar with the            Multipillar with         Multipillar with
 savings                         second pillar open in the       second pillar open in    second pillar fully
                                 accumulation stage              the payout stage         separated from PAYG
 Czech Republic and Slovenia     Slovak Republic, Bulgaria,      Croatia,    due     to   Latvia, Romania.
 from the beginning (no          Lithuania, and Estonia          pension     adequacy
 multipillar reform based on     motivated by combination        concerns.
 three pillars), and Hungary     of a desire for control of
 and Poland after second         current     fiscal   deficit,
 pillar dissolutions motivated   dissatisfaction with funds’
 by fiscal problems and          returns and arguments in
 concerns about mandatory        favor of free individual
 pension funds’ net returns.     choice (less obligatory
                                 second pillar).


Reform reversal proposals have been based on a wide range of assumptions, starting from
a belief that the second pillar hurts credit ratings and fiscal space (due to transition cost –
loss of current revenue to the PAYG). High social discount factors (political myopia) led to
widespread beliefs that the PAYG system on its own can resolve pension adequacy – fiscal

                                                        77
sustainability trade-off despite the ageing of the population. Similar debates emerged in
Croatia, too. One of the main arguments in the camp of critiques of the second pillar is the
supposedly positive impact of dissolution on the reduction of current public debt and cost of
government finance.
Short-term fiscal argument against the second pillar is based on the idea that there is a
negative impact of the transition cost of the multipillar reform on the cost of government
financing and fiscal space, allegedly narrowing room for countercyclical fiscal policy .
Critiques of the second pillar emphasize a negative impact because higher fiscal deficit and
public debt to GDP ratio in the short run (partly due to pension reform transition cost) have
negative impact in terms of higher bond spreads.61 According to proponents of this view,
dissolution of the second pillar would immediately help government finances and economic
growth via larger fiscal space. The opposing view is that there is either no short-run
relationship between sovereign ratings and pension reform, or the relationship can be
positive if the reform produces positive policy signaling effects. Pro-pension reform view is
that accumulation of private savings for pensions may help fiscal sustainability of the PAYG
system and sovereign ratings in the long run when market participants will begin seeing the
multipillar pension reform as a signal of good policies and improved prospects for long-term
fiscal sustainability. However, long run effects cannot be persuasively proved or rejected by
empirical evidence because there are no long-term fiscal time series spanning over several
generations under different pension systems’ regimes.
The relationship between individual pension savings and sovereign credit rating is not
straightforward, as there are all sorts of factors that can affect sovereign credit ratings . In
general, countries with higher pension savings to GDP ratio have better sovereign credit
ratings on average.62 However, it does not mean that individual pension savings cause better
ratings. Many intervening variables, such as financial development and socio-economic
development in general, better policies, financial literacy, etc., can also explain positive
relationships (see Annex 1). In any case, multipillar pension reforms in the CEE countries
speeded up the increase of pension assets. Without government intervention in the form of
multipillar pension reform, several factors would slow down growth of private pension
savings and strengthen continued reliance on PAYG-only in an era of ageing population. The
fact is that in the CEE, including Croatia, high discount rates (shortsighted preference for
current individual and social spending), lack of developed financial markets, and relatively low
financial literacy back in the late 90s when the reforms have started, would have prevented
financial development and, possibly, slowed down overall economic development. Positive
impact was especially present in the government bond markets. The emergence of big
domestic institutional investors stabilized government bond yields. Home bias, which is


61
   From the outset of the pension reforms in CEE in the late 90s and early 00s, policy makers had the intention
to offset fiscal deficit and public debt impact of the transition cost (i.e., PAYG financing gap) by: (i) parametric
reforms in the PAYG (leading to lower replacement rates), and (ii) securing additional funding from fiscal reserve,
e.g., using part of privatization revenues for funding the PAYG financing gap. Neither proved to be the case in
the long run. So, it makes sense to assume that financing gaps had implications for fiscal deficit and public debt
(Egert, 2012).
62
   Relationship is nonlinear and R squared is relatively low (0.29), see Annex 1.

                                                        78
sometimes interpreted as a negative factor due to lack of international diversification of
portfolios, may have been a positive factor in this respect.63
Hungary is a case worth deeper consideration as it dismantled the second pillar in 2011 on
the grounds of short-term fiscal improvement; however, after more than a decade, the
effects do not seem so positive. Transfer of pension savings from the mandatory pension
funds to the PAYG (government budget) was recorded as government current revenue.
Hungarian fiscal deficit and public debt ratios indeed improved from 2011 to 2019. However,
Hungary did not perform well during pandemic and its after-shocks. Hungary’s sovereign
ratings remained on the edge of investment grade (BBB-), two notches below Croatian
sovereign ratings as of end 2023. Different fiscal and ratings trajectories in Hungary vs.
Croatia, Romania, and Bulgaria, may have emerged for different reasons. Nevertheless, it is
not possible to reject the hypothesis that dismantling of the second pillar in Hungary was not
helpful for subsequent fiscal path. It may have been detrimental. The following two
hypothesis are worth consideration in this respect: (i) dismantling of the second pillar in
advanced accumulation stage potentially creates fiscal complacency, softens market
discipline, and leaves the door open for increasing inefficient fiscal expenditure (this was the
essence of Egert's (2012) bleak prediction of the long-term impact of dismantling the second
pillar in Hungary);64 (ii) the existence of sizeable mandatory pension funds with mixed
portfolios of securities stabilizes overall demand for government bonds and securities in
general, and strengthens market discipline via punctuality of information revelation (hence
reduction of financial volatility with potential positive effect on growth).
In conclusion, dismantling the second pillar pension funds is always a policy option, as well
as the opposite – returning to mandatory payout from the second pillar for old-age
pensioners. The short-term political attractiveness of the first measure emerges from one-
time fiscal revenue, while the second measure has no political attractiveness. Therefore, the
first option will always gain higher weight in public political debates. However, the adverse
longer-term fiscal and social impact of such an option should not be overlooked. Hardships in
assessment of long-term impacts of radical policy scenarios and lack of background for
evidence-based policy making led us toward the conclusion that no radical policy scenario
should be implemented in Croatia. Rather, the existing system offers many opportunities for
balanced improvements of both pension adequacy and fiscal sustainability. These
opportunities are discussed in turn.


6.2 Indexation formula: effects of weight rotation and no-negative adjustment rules

There are three main elements of current adjustment rules that have almost equal
importance for the long-term impact of indexation formula on pension benefits: i)

63
   It can have a negative interpretation, too. If mandatory pension funds become leading buyers of domestic
government bonds in a concentrated government bond market, repeated interactions between the government
and fund managers, who do business in a highly regulated financial environment, may give birth to political
dependence and suspicions about proper execution of the fiduciary duty in the best interest of contributors.
64
   Egert, Balasz (2012).

                                                    79
indexation frequency; ii) rotation in weights, and iii) zero lower bound for adjustments. This
section aims to illustrate the importance of these rules on the relationship between regular
adjustment of pension benefits and wage growth. The results of implementation of the
current indexation rules are compared with the alternative rules implying different indexation
frequencies (quarterly and annual) and removing weight rotation and zero lower bound in
adjustments.
Indexation/valorization formula65 in the Croatian pension system is applied as adjustment
of the pension point value with the following rules:
    • Adjustment is semi-annual, on January 1 and July 1.
    • Adjustment is based on the gross wage growth and consumer price inflation, each
        calculated as a percentage change over average levels recorded in the previous half-
        year period.
    • Rotational 70%-30% adjustment: if wage growth is higher than inflation rate (for
        decision which one is higher, their absolute values are compared), then adjustment
        will take 70% of wage growth and 30% of inflation rate to get the rate of adjustment,
        but if wage growth is lower or equal to inflation rate, the opposite wage-inflation
        proportions are used.
    • Zero lower bound is applied meaning that there is no downward adjustment: if the
        abovementioned adjustment rules result in a negative percentage, there will be no
        change in point value.
    • No restrictions to upward adjustment: each time the rules point to a positive
        percentage of adjustment, it should be realized in full no matter if previous downward
        adjustments were skipped.
Testing a long-term impact of the current rules and their potential modifications is based
on the historical evidence on wage growth and inflation in Croatia between 1998 and 2023.
Although the current rules for valorization/indexation are in place from January 1, 2019, for
testing purposes we assumed that they were in place since January 1, 1999. The last 25 years
of data should give us sufficient real-life variability in wages and prices to better inform us on
the possible future developments after the application of alternative indexation formula.
Simulations show that higher frequency of adjustment, particularly a switch to quarterly
adjustment, can have a huge positive impact on the future evolution of the replacement
rate. Application of quarterly adjustment with unchanged other indexation rules would lead
to rates of pension adjustments that are higher than wage growth despite keeping 70:30
weights distribution. Based on historical data, the pension point value at the beginning of
2024 would be 245% higher compared to the value at the beginning of 1999 (Figure 6.1). For
comparison, the average gross wage in the fourth quarter of 2023 would be 196% higher than
in the fourth quarter of 1998. Two factors behind indexation at a higher rate than wage
growth had approximately the same importance: rotation in weights and zero lower bound
when negative rates of indexation occurred. Higher frequency of adjustment implies higher
variability in wage and inflation dynamics, hence: (i) more observations with very high both

65
 In Croatia, valorization (adjustment of the value of the first pension) and indexation (adjustment of the value
of pension in payments) are performed by using the same formula for adjustment of the pension point value.

                                                      80
positive and negative changes, (ii) more observations where inflation is higher than wage
growth, and (iii) more observations when negative change was abandoned due to zero-limit
rule.
Figure 6.1 Hypothetical evolution of the pension point value under current indexation rules applied to
quarterly, half-yearly and yearly adjustment frequency
390.0


340.0


290.0


240.0


190.0


140.0


 90.0
        4q1998


        1q2001




        2q2015
        3q1999
        2q2000

        4q2001
        3q2002
        2q2003
        1q2004
        4q2004
        3q2005
        2q2006
        1q2007
        4q2007
        3q2008
        2q2009
        1q2010
        4q2010
        3q2011
        2q2012
        1q2013
        4q2013
        3q2014

        1q2016
        4q2016
        3q2017
        2q2018
        1q2019
        4q2019
        3q2020
        2q2021
        1q2022
        4q2022
        3q2023
                          Wage level (4q1998=100)                             Point value (quarterly adj.)
                          Point value (yearly adj.)                           Point value (half-year adj.)

Notes: Current rules include 70%-30% rotational formula for indexation that begins in 1999. Actual gross wage growth and
CPI inflation in the period 1999-2023 is applied in calculations.



Combination of high frequency of adjustment based on consecutive growth rates
(quarterly), zero lower bound and rotation in weights should be avoided because it can lead
to an uncontrolled explosion of fiscal costs depending on unpredictable short-term
macroeconomic volatility. Quarterly adjustments with historical data show that rotation in
formula is activated 27 times out of 100 adjustment episodes of quarterly adjustments, while
downward adjustment is skipped 26 times. Based on historical data, it is estimated that
quarterly adjustments combined with other current rules will result in cumulative indexation
rate equal to cumulative wage growth if 57%:43% rotational formula was implemented.
Less frequent adjustments are likely to result in lower increases in the pension point value.
Simulation of annual adjustments with other rules in place showed an increase in the pension
point value of 155% after 25 years. That is less than wage growth in the same period.
However, due to rotation in weights, the increase is slightly higher than it would be with strict
adjustment of 70% with wage growth and 30% with inflation without rotation. Zero lower
bound was not activated in any yearly adjustment.
Semi-annual adjustment with current rules gives somewhat higher pension point value
increases than yearly adjustment. Semi-annual adjustment applied on the last 25 years of

                                                          81
data resulted in a 174% increase, which is somewhat lower than the wage growth of 196%
(Figure 6.1). Here, wage-inflation weight rotation in formula is applied in 16 out of 50 cases
of adjustment. On two occasions, a no-downward adjustment clause is activated, but the
negative adjustments would be rather small, even if effectuated.66 Rotation is the main cause
for point value adjustments being closer to wage growth than the 70%-30% adjustment rule
without weight rotation. Based on historical evidence, a full 100% wage adjustment without
any other rule is equivalent to semi-annual adjustments with 80%-20% rotational formula
under current rules. Indexation formula that applies a higher weight with wage growth may
result in pension point value adjustments above wage growth over certain periods if future
inflation and wage increases mimic their developments in the last 25 years.
Figure 6.2 Pension point indexation, CPI inflation and wage growth (January 2019 =100)


 145.0


 135.0


 125.0


 115.0


 105.0


     95.0
                              Jul-19




                                                                  Jul-20




                                                                                                       Jul-21




                                                                                                                                           Jul-22




                                                                                                                                                                                Jul-23
                                                         Apr-20
                     Apr-19




                                                                                             Apr-21




                                                                                                                                  Apr-22




                                                                                                                                                                       Apr-23
            Jan-19




                                                Jan-20




                                                                                    Jan-21




                                                                                                                         Jan-22




                                                                                                                                                              Jan-23




                                                                                                                                                                                                  Jan-24
                                       Oct-19




                                                                           Oct-20




                                                                                                                Oct-21




                                                                                                                                                     Oct-22




                                                                                                                                                                                         Oct-23


                                           Pension indexation                                         CPI inflation                                 Wage growth

Source: Own calculations based on HZMO and CBS data.



Current indexation rules provided a rather good protection of real and relative pension in
turbulent times 2020-2023 (Figure 6.2). Two features of indexation rules helped. First is
rotation in weights, second is half-year frequency. Cumulative increase in pension point value
(indexation) from the beginning of 2019 to the beginning of 2024 was 39%. At the same time,
the average gross wage increased 44% and consumer prices rose 26% (Figure 6.2). Although
pension indexation exhibits certain delay in upward adjustments compared to inflation and
wage growth, the adjustment of 8.42% in July 2023 and 4.19% in January 2024 resulted in
cumulative pension adjustment close to wage growth and significantly above inflation in the




66
  In the longer run, the abolishment of the no-downward adjustment clause gives the same results as leaving it
in place in combination with socially more acceptable rule of “compensating upward adjustment”. No-
downward adjustment rule can be left, but each upward adjustment should compensate for previous negative
adjustments that are skipped and only the remaining part of the increase is realized as an adjustment.

                                                                                                       82
first half of 2024, maintaining the pension benefit ratio.67 These findings can inform policy
discussion while proposals for indexation rule modification and simulations of its future
effects are presented in Section 7.


6.3 A proposal for pension-type-choice-neutral social transfers paid as pension
    benefits


Analysis presented so far has shown that pension benefits in the first pillar contain elements
that are not earnings-related, i.e., not proportional to contributions paid by the insured
person. They are based on social criteria or other merits instead of income level and
contributions paid from it. Such elements of the public pension system can be justified by the
solidarity principle. However, knowing that in recent years around 45% of pension
expenditures were financed by direct transfers from the state budget, question arises if these
benefit extensions should be treated as integral parts of the first pillar scheme or they should
be understood and regulated as elements of the social welfare system financed by the state
budget but administrated by the public pension scheme. If these benefits are treated as a part
of the social welfare system, then they should be neutral regarding the choice of P2 vs. P1.
The basic pension factor of 0.75 does not appropriately reflect the pension rights as many
of those benefits are financed by direct transfers from the state budget thus affecting the
choice between P1 and P2. The proportion of contributions paid to the first pillar of 0.75
would be a fair factor if and only if all pension expenditures in the first pillar are financed by
contributions. That is obviously not the case. Therefore, the basic pension factor of 0.75 does
not correctly reflect pension rights earned by payments to the first pillar: large portion of
PAYG benefits is financed by taxation, where tax obligations are equal for all, regardless of
pension type choice. If the basic pension factor is not fair in relative terms vis a vis full PAYG
pension, then the choice between P1 and P2 is not made on fair ground. It results in a bias
towards P1 pension as shown before.
The choice-neutral social transfer would eliminate the bias in the choice, increase pension
adequacy for those who choose P2 and probably increase share of new retirees who would
benefit from P2. One example can illustrate the issue. Mothers have the right to six months
of service to be added to the insurance period in PAYG. This measure, introduced in 2019, is
officially justified by lower income (contribution base) received by mothers in the period of
their parental leave. Also, it is said that this measure is an element of demographic
revitalization policy. In pension benefit calculation, mothers who take full PAYG benefit will
therefore get extra benefit from the first pillar in certain amount (e.g., EUR 10), while mothers
who take the combined pension will receive ¾ of that extra benefit (EUR 7,5). The


67
   The choice of the initial period for comparison may influence the results. We choose January 2019 because it
is the time when the current rules on pension indexation take effect. Selection of other basis for comparison
(average of the first semester of 2019, January 2021, or average of the first semester of 2021 - because surge in
inflation started in second semester of 2021) lead us to identical conclusions on relative movements between
inflation, wage growth and pension indexation.

                                                       83
demographic measures are financed by general taxation, and taxes paid are not dependent
on the pension insurance or the pension benefit type. Therefore, the benefit in terms of extra
pension should be available to all mothers under the same conditions irrespective of their
choice of P1 or P2. In our example, EUR 10 per child should be given to each mother. In that
way, extra basic pension would be treated as a social transfer, administrated by the first pillar
(HZMO), but actually paid from the government budget.
The same principle may apply to the 27% pension supplement and other social elements in
the pension system financed from the government budget. In this case, an equal absolute
amount of the pension supplement should be given to the beneficiary regardless of the
pension type chosen. For example, if PAYG pension according to the pension formula P1 is
EUR 500, and the pension supplement of 27% is EUR 135, then the total full PAYG pension
(P1) is EUR 635. If EUR 135 is treated as a social transfer, that amount should be added also
to the basic PAYG pension if the pensioner chooses the combined pension. The choice-neutral
social transfer approach can be extended to other social transfer elements in the pension
system, such as a portion of minimum pension which was not covered by payments of
contributions. As a result, it would increase the pension adequacy because it will increase the
basic PAYG portion of the pension for those who choose P2.
This measure should have been implemented in 2019 when the second pillar payout stage
was opened for opt-back, however, uncertain fiscal implications probably delayed its
consideration as choice-neutral social transfer approach seemingly goes with additional
fiscal cost in proportion to additional budgetary transfer for those who choose P2 . However,
there is an additional fiscal aspect that must be considered. With choice-neutral fiscal
transfers in P2, the number of new pensioners choosing P1 would be lower. Their transfers
would occur anyway, subject to opt-back under current rules. So, there is no additional fiscal
cost in proportion to number of members of the second pillar (i.e., in proportion to their
pension assets) who opt-back under current rules and who would accept default P2 under
choice-neutral transfer rules.
The proposal to treat social-based elements of pension benefit as choice-neutral lump-sum
transfers should be considered in two variants: a) broad coverage, and b) narrow coverage.
Broad coverage includes the minimum pension, the added insurance period for
mothers/parents and the 27% pension supplement. The narrow coverage is for the minimum
pension and the added insurance period.68 If applied, new treatment would increase pension
adequacy of the combined pensions as it would increase share of members of the second
pillar with P2>P1.69




68
   The same effect as alternative treatment of the added insurance period can be achieved by a small change in
the current legislation. It is enough to prescribe in Art. 32.a para. 2 of the Pension Insurance Act that insurance
period is added as it is earned before the introduction of the second pillar. Currently, the insurance period is
added to the longer of the pre-2002 or post-2002 insurance period. With such change, the result will be the
same as in our alternative approach.
69
   The possible effects are simulated, and the results presented in Section 7.

                                                        84
6.4 Second pillar pension payments and pension insurance companies


Many things can affect annuity factors in the long run, but as it is impossible to predict
which ones will prevail, a reasonable assumption is to leave annuity factors unchanged in
the long run simulations. Increase in annuity factors around 15% as of mid-2023 is expected
to last as it reflects: (i) improvement of inflation hedging techniques in pension insurance
companies (PICs) after Croatia's accession to the Euro Area in 2023, (ii) higher expected long-
term returns on fixed income instruments after normalization of ECB's interest rates, and (iii)
application of accounting standard IFRS 17 for calculation of PICs' technical reserves.70 Several
mutually conflicting factors will affect annuity factors in the long run: longer life expectancy
will have a negative impact, and growing business volumes and stronger competition in the
market for annuities, better asset management (PICs' net returns)71 and introduction of new
pension products will create positive impacts. It is impossible to predict which factors will
prevail in the long run. A reasonable assumption is to leave the annuity factors unchanged in
the long run simulations.
Limited business volumes in the annuities market are primarily related to a small share of
new pension beneficiaries who retire from the second pillar . Most new retirees so far have
been retiring from the first pillar because they were older than 40 in 2002. Out of those who
retire from the second pillar, accumulation period since 2002 was relatively short –
approximately half of working life. So, it is too early to assess adequacy of the second pillar
pension based on the experience of specific groups of retirees from the second pillar who
largely comprise early retirees (around 35%), disability and survivors’ pensions. As of end
2023, there were only 13,311 pension beneficiaries who received annuities from PICs. This is
1.1% of total pension beneficiaries in Croatia. PICs' total assets of EUR 384 million represented
around 0.5% of GDP. The small market also explains the small number of financial institutions
offering annuities. There are only two PICs in Croatia - one privately owned, and one
government owned. At the same time, mandatory pension funds' assets under management
in the accumulation phase represented around 27% of GDP. Therefore, PICs' assets under
management will record strong growth in the years to come. Growth of PIC's business
volumes will strengthen competition, however, contribution of stronger competition to
lowering of the administrative costs will be limited because regulatory limitations on fees are



70
   Before the implementation of IFRS 17, present value of PICs' future pension liabilities was not sensitive to
market parameters affecting valuation of assets such as interest rates. Asset/liability valuation mismatch created
substantial risk for PICs because they have to use own funds to cover the mismatch if value of eligible assets falls
below technical reserve. Both PICs had to employ own capital during bond market conundrum in 2022/2023. On
top of implementation of IFRS 17, recent regulatory changes (effective as of the 1st of January 2024) introduced
cyclically adjusted calculation of asset/liability mismatch. Payment from PICs' own funds can be delayed for 2
years which is a period long enough to smooth out transitory short-run extreme volatility effects. However, ban
on PICs' dividend payouts and obligation to set aside own funds for coverage of asset/liability mismatch in a
form of PICs' reserve will remain active during 2 years of delay.
71
   PICs' historical returns are around 1% per annum on average due to extremely high allocation to government
bonds. From end 2021 to end 2023 allocation to government bonds decreased from 84.2% to 71.5%, which still
leaves room for adding higher-yielding investments, which is further discussed below.

                                                        85
already active.72 Most positive long-term effects can be expected from liberalization of PICs'
investment.
Coupled with the introduction of new eligible asset classes, recent liberalization of PICs'
investment limits provided appropriate framework for improvement of pension portfolio
risk/return characteristics in the long run. Regulatory changes which are effective from
January 2024 lowered the floor for minimum allocation of PIC's eligible assets73 to public debt
instruments issued by OECD members and Croatia from 70% to 50%.74 At the same time, asset
allocation ceilings were increased for listed equities (from 20% to 35%), UCITS including ETFs
(from 20% to 35%), alternative and closed investment funds (from 10% to 15%), and for real
estate exposures (from 10% to 20%).
Assessment of the impact of recent legal changes should be carried out in a few years to
investigate further potential for liberalization of investment limits, while also evaluating
the probability of beneficiaries' participation in PIC’s profits. Beneficiaries’ profit
participation is optional when PIC's coverage ratio (ratio of eligible assets to technical reserve)
is within 110-115% interval. In this case, PICs can allocate excess return to increased second
pillar annuities, but don’t have to. When coverage ratio is higher than 115%, excess return
must be allocated to higher second pillar annuities.75 In both cases, ¾ of excess can be
allocated to pensions, and ¼ to the intervention reserve. Expected profit participation is
important, but very hard to measure due to risk and uncertainty which surround dynamic
implications of choice between P1 and P2. Also, there is no history (profit participation never
happened). Problem is even more complex when expected profit participation is weighed
against dynamic effects of residual life expectancy and indexation which works in favor of
expected P1 if expected real wage growth is positive.76 Future assessment of PICs’ risk/return
result should be carried out in conjunction with assessment of future differentiation of the
second pillar annuity products. The differentiation of the second pillar annuity products is
presently limited: PICs offer lifetime annuities only. Differentiation is related to the choice of
guaranteed vs. non-guaranteed annuities and joint vs. individual pension.77 Room for offering
variable lifetime annuities has not been used. Such a room may exist within existing regulatory
solutions according to industry experts, but this room was never tested. It is reasonable to




72
   PICs charge one-off entry fee up to 1.5% (for asset transfer from obligatory pension funds), and up to 0.17%
annual payout fee. Asset management fee is limited at 0.6% and declining by 10% per annum to reach 0.3% by
2028.
73
   Up to 7% can be invested in covered bonds issued by credit institutions and 3% can be invested in unlisted
securities.
74
   Article 93 of the Law on Pension Insurance Companies. This section is not intended to provide comprehensive
review, as it presents only the most important changes which may materially affect PIC's asset management
performance in the long run.
75
   No profit distribution happened so far.
76
   Under assumption of 1.5% annualized real wage growth in the long run and 15% of combined pension coming
from the second pillar annuity, dynamic indexation effect in favor of P1 can be eliminated by long-term profit
participation equal to annualized long-term rate of 0.16%.
77
   Guaranteed products are offered with lower annuity factors due to different account for mortality risk because
remaining pension assets are paid to inheritors when beneficiary passes away.

                                                       86
expect that the market will test it in the coming years when market size increases, especially
after recent regulatory changes allowed the offering of non-indexed second pillar annuities.78
There is no need to implement additional regulatory changes in the payout stage for the
next several years before the recent changes show the results. Recent regulatory discussions
involved the debate about mandatory institutional integration of PICs and mandatory pension
fund management companies. This change was not implemented. There is no need for radical
institutional reforms in the payout stage before in-depth assessment of effects of the 2024
regulatory changes after a couple of years. Frequent and deep regulatory changes without
evidence-based backup may increase regulatory cost and risk. At present, it is not clear how
institutional integration of accumulation and payout stage of the second pillar would affect
long-term pension adequacy and fiscal sustainability. The same holds for radical liberalization
of pension annuity products. Instead, the policy focus should shift to the accumulation stage
because contribution of the second pillar annuity to a combined pension primarily depends
on the value of pension savings in the mandatory pension fund by the end of an individual
career.


6.5 Investment policy in mandatory pension funds


There are five factors affecting the investment returns in the second pillar: (i) long-run
performance of financial markets, (ii) administrative cost of the second pillar, (iii) fund
managers' success within market and regulatory constraints, (iv) regulatory investment
limits, and (v) lifecycle portfolio allocation. The fifth factor was introduced in August 2014
by splitting previous single-portfolio funds into three separate risk categories - A (high risk), B
(balanced) and C (low risk – fixed income).79 This section is focused on factors (iv) and (v) after
the financial market turmoil in 2022 motivated regulatory changes implemented in 2024.
The most important regulatory change launched in 2024 is related to regulatory rules for
default lifecycle allocation which are applied to members of the second pillar who do not
make active decisions about the lifecycle portfolio. More than 95% of members in the second
pillar are passive members who follow default allocation set by regulation. Before the recent
changes, the following rules were in place: (i) automatic enrolment in A, (ii) default shift from
A to B after 10 years (active members could stay in A until 10 years before statutory
retirement age), (iii) default obligatory shift from B to C 5 years before statutory retirement
age. The third rule was based on the theoretical expectation of low volatility in C in the short
period before retirement. The rule disregarded the long period of historic lows of interest
rates which suddenly reversed due to outburst of high inflation in 2022 and 2023.
Consequentially, the pension accounts of members 5 years before retirement shifted to
portfolio C before a strong rise in interest rates in 2022 coincided with a rapid fall in bond


78
   Provisions contain appropriate obligatory information for beneficiaries with respect to understanding of
inflation risk.
79
   Before August 2014 second pillar assets were invested in conservative balanced portfolio which is currently
labeled as B.

                                                     87
prices. The values plunged for a relatively short period until retirement, leaving no time for
recovery of C portfolios. Inflation added to the fuel to the fire, which is reflected in an
unprecedented drop in real net returns. Portfolio B’s, and especially A’s returns remained
higher and more resilient. Table 6.2 distinguishes between previous normal inflation period
until end 2021 from high inflation period in 2022 and 2023 showing outcomes which
prompted three regulatory changes in 2024: (i) default reallocation from A to B is triggered
after 15 years instead of 10 years as before; (ii) active members can cancel automatic
reallocation from A to B and remain in A until 10 years before statutory retirement age, while,
on top of it, membership in A funds can last until 5 years before statutory retirement age
subject to additional cancelation of reallocation 10 years before retirement; (iii) previous
default reallocation from B to C 5 years prior to statutory retirement age is annulled unless
member requests reallocation of pension account to C.
Table 6.2 MIREX* based net annualized returns for A, B and C portfolios
                                            Annualized nominal net return               Annualized real* net return
                                              Until end      Until end                   Until end       Until end
                                               2021             2023                      2021             2023
        A from August 2014                     7.78%           7.20%                      6.72%           4.48%
    B from inception of reform                 5.51%           5.21%                      3.28%           2.64%
        B from August 2014                     5.75%           4.51%                      4.71%           1.85%
        C from August 2014                     4.48%           3.36%                      3.45%           0.73%
*MIREX is asset-weighted index calculated by regulator HANFA per each risk category. Real net return is calculated by
application of CPI. August 2014 is the 21st of August which was the first day of operation of lifecycle mandatory pension funds
when unit values were set equal to 100. The last 10 days of August 2014 were ignored in inflation adjusted calculation.
Source: HANFA for MIREX, Croatian Bureau of Statistics for CPI, own calculation.

The latest regulatory changes represent an improvement, but they do not eliminate
fundamental problems related to cascading and radical portfolio shifts which happen in the
work life cycle of passive members. If radical cascading shifts between portfolios happen
when markets are in extreme conditions, effects of such shifts may be hard to reverse. It can
create lucky and unlucky generations. As such shifts happen due to regulatory rules, people
hold government responsible for its effects, especially when such effects are negative. This
puts the reputation and credibility of the multipillar pension system in jeopardy. Finally, as
portfolio composition switches are rare, and they do not involve members’ participation in
the decision making, people are poorly informed about it. It does not create incentives for
more active participation. Such flaws of the regulatory cascading system can be overcome by
a more refined lifecycle allocation rules which will nurture incentives to gather and process
information about the second pillar, increase financial literacy and invite informed members
to take active participation in the decision making.
Authorities may consider elimination of regulatory cascades by application of smoothing
lifecycle rule labeled as Incremental Allocation System (IAS). In IAS example presented in
Table 6.3, every three years starting from the 30 years before statutory retirement age,
exposure to high-risk A portfolio is reduced by an increment (8-10pp) which leads to gradual
reduction of exposure to high risk with time. After advancement of the automatic reallocation
period, each member of the second pillar will have pension accounts opened with 2 or 3 fund



                                                             88
categories, which is technically feasible.80 Also, with time, effective high-risk exposure for
passive members shall remain relatively high because portfolio B also incorporates part of
high-risk investments. Therefore, final design of key IAS parameters, such as (i) starting date
for a first incremental shift, (ii) desired allocation at the endpoint, (iii) duration of time-
increments (3 years is arbitrary choice), and (iv) percent change-increments, should take into
consideration structure of portfolios and management practice in each individual category of
mandatory pension funds. Final design should also involve in-depth consultations with
stakeholders including industry and experts in behavioral finance.
Table 6.3 An Example of IAS
                                                                                             YEARS BEFORE STATUTORY RETIREMENT AGE
                                                                   Up to 30th    30    27      24       21        18       15      12     9     6     3
Maximum exposure to A                                                   100%    90%   80%     70%      60%      50%      40%      30%   20%   10%    0%
Minimum exposure to C                                                                                            8%      16%      24%   32%   40%   48%
Residual exposure to B for passive members                                0%    10%   20%     30%      40%      42%      44%      46%   48%   50%   52%
Effective allocation to high-risk for passive members                    85%    82%   78%     75%      71%      64%      56%      49%   41%   34%   26%
Effective allocation to high-risk for pesimists                           0%     0%    0%      0%       0%       0%       0%       0%    0%    0%    0%
Effective allocation to high-risk for optimists (opt-out to B at 12)     85%    82%   78%     75%      71%      64%      56%      50%   50%   50%   50%
Average allocation to high-risk assets in A                              85%    85%   85%     85%      85%      85%      85%      85%   85%   85%   85%
Average allocation to high-risk assets in B                              50%    50%   50%     50%      50%      50%      50%      50%   50%   50%   50%
Average allocation to high-risk assets in C                               0%     0%    0%      0%       0%       0%       0%       0%    0%    0%    0%

Source: WB team.



If properly designed, IAS may have profound implications for financial literacy and
members’ financial behavior because each incremental change in allocation requires
delivery of official information. If properly designed (e.g., framed with a modern software
application potentially involving AI assistance), IAS may produce side benefits in terms of a
growing share of active participants and increased market discipline.81 For example, in a short
time window around the day of default incremental change, contributor to the second pillar
may be invited to change default allocation by choosing either higher or lower risk exposure
compared to default allocation within wider regulatory limits. There is no need to set
regulatory limits for more conservative desired allocations, but authorities may allow one or
several incremental steps back (towards higher risk exposure) compared to default. Taking
steps higher on the risk stairs may create discomfort among regulators, especially if the
remaining period before individual’s retirement is relatively short. However, the recent
elimination of cascading shift to C 5 years before retirement already incorporates this kind of
risk.
Adopting IAS makes sense, given that the pension system has already implemented a social
buffer in terms of opting-back to the full PAYG pension. With lower limits on pensions in
place (opting-back to the full PAYG pension), riskier allocation of assets in the second pillar

80
   Note that, in the long run, legacy of B portfolios currently representing 88.7% of net asset value in the second
pillar can be overcome as any allocation can be represented as linear combination of A and C.
81
   Authorities are aware of the need to increase market discipline. Recent legislative changes introduced the
option to change the risk category and/or fund manager once a year at any time, whereas previously such
changes were allowed in the month of the birthday date only.

                                                                                        89
transforms the system into an option-like financial instrument which can deliver higher
returns in the long run.
Implementation of IAS may lead to specialization of A and C funds in terms of (potentially)
100% allocation to respective risk categories, which can improve quality of pension asset
management. In theory, any desired allocation can be achieved as linear combination of A
and C, where both lifecycle portfolios are 100% invested in respective asset classes. Such
specialization can increase quality of asset management. However, implementation of IAS will
require a long period of transition because of the system’s legacy: in almost 10 years since
the split of 100% B portfolio in August 2014, share of A portfolio in terms of assets increased
to only 2.1%, and share of C to 9.2% by the end of 2023. B-class funds are dominant and will
remain dominant for many years to come. Portfolio structures inside each risk category also
change slowly: share of government bonds in B decreased from 72.0% at end- -2014 to 60.2%
at end-2023. The share of government bonds in A is still relatively high at 36.2% as of end-
2023. This concentration in government bonds (largely domestic) is a legacy of the past as the
pension reform started with government bond asset allocation floor set at 70%. Therefore,
investment rules and limits in portfolio asset classes will remain important for a long time
even if IAS is implemented. However, implementation of IAS should lead to further
liberalization of investment limits to bring A-category closer to its natural high-risk exposure.
Several subsequent cycles of liberalization were primarily prompted by the EU accession .
Each liberalization step represented an attempt to balance freedom of managers’ choices and
stability in each lifecycle category separately, without taking account of interactions and
transitions during individual lifecycle. This created a dense web of complex investment limits
for individual risk categories which reflect regulatory fear from risk. It also created inertia in
pension funds’ portfolios A and B. This is not optimal for portfolios with long investment
horizons: existing limits would be considered strict even if applied on open-end UCITS with
volatile liabilities. In this sense, IAS can provide an umbrella for long-term optimization of the
risk exposure in the second pillar - a unifying framework for gradual consideration of final
steps in the long process of liberalization of investment limits.
Recent changes in investment limits, effective from 2024, represent additional, although a
modest step in the right direction. The upper limit for direct investment in listed equity
securities in A was raised from 65% to 70%. Setting the upper limit for higher risk exposures
in the riskiest category reflects continued regulatory preference to keep A’s portfolio
structure closer to balanced portfolios. Preference for conservative allocations is also
reflected in continuation of 40% upper allocation limit on investment in listed equity
securities in B.82 Nevertheless, technical improvements in asset management may come from
increased ceilings for investment in eligible UCITS, including ETFs, from 30% in A and B to 45%
in A, 35% in B and 10% in C. However, it is not clear why such aggregate limits for UCITS/ETFs




82
  Within total listed equity limit, legislative changes introduced a smaller sub-limit of 10% for investment in
equity securities listed on multilateral trading platforms and 0.2% for investment in securities listed on
multilateral crowdfunding platforms.

                                                      90
exist, and why are they different for different categories of pension funds.83 In addition, floors
for government bonds were seemingly lowered by 5 percentage points from previous 30% in
A and 50% in B. However, 5% is intended for alternative investment funds with principal
guarantee issued by the OECD member state or multilateral institution. If no such AIFs exist,
previous limits of 30% and 50% must still be met by purchases of government bonds. As there
is no experience with AIFs with government backup, there will be no effective impact on the
side of government bond allocation for some time. Eventual small long-term impact in
proportion to 5% minimum allocation may arise after many years if such AIFs will be offered
in the market. Even in this case it is not clear why 5% allocation is set as a floor. It can create
opportunity costs if early birds AIFs with government guarantee will deliver principle
guarantee only.
Some improvements of risk/return profiles can be expected from direct investment in real
estate which were not allowed before 2024. Currently, temporary upper limits of 5% in A
and 4% in B were introduced with upside to ceilings of 10% and 7%, respectively, when proven
management capacity conditions for managing real estate portfolio will be met.84 In addition,
the aggregate limit for investment in AIFs was raised (equalized with limit for listed equities)
and a new limit for unlisted covered bonds of 10% in A, 7% in B and 5% in C was introduced.85
Upper limits for securitized exposures related to government infrastructure investments of
55% in A, 35% in B and 10% in C were left unchanged. Finally, category C is now opened for
riskier investment, as a ceiling of 10% for listed equities was introduced.
Aggregate investment limits should be simplified, and IAS can provide a unifying framework
for capital markets development and nurturing positive loops among fund managers’
competition, investment returns, and members’ active investment behavior. Simplification
of investment rules may include: (i) elimination of upper limits for UCITS including ETFs, (ii)
elimination of upper limit for equity exposures and floor for government bonds in A (with
potential introduction of government bond ceiling to preserve high risk nature of A), (iii)
change and reversal of equity and government bond limits in B, so that a floor is set for
equities at 25% or 30% and ceiling for government bonds at 70% or 75%, (iv) elimination of
obligatory floor of 5% for AIFs with government guarantee, and (v) a very limited window for
investment in low credit ratings instruments with high expected return. Demand from
mandatory pension funds may play a critical role in providing financing for government
investment projects, but it is important to eliminate any regulatory obligation from
investment rules. Strengthening managers’ independence and professionalism to perform
their fiduciary duty freely is very important for overall credibility of the multipillar pension
system.



83
   The Government's analysis of 17 comparable regulatory frameworks showed that there is no such aggregate
limit in majority of funded pension systems (cro. Prijedlog Zakona o izmjenama i dopunama Zakona o obveznim
mirovinskim fondovima, p. 9, 2023).
84
   Aggregate real estate exposure limit including indirect exposures via real estate AIFs may increase up to 20%
in A and 10% in B.
85
   Additional aggregate limit for low liquidity alternative asset classes of 20% in A and 15% in B was also
introduced.

                                                      91
It is equally important for authorities to engage in direct policies for capital markets
development along the lines recommended by the OECD, especially after the availability of
EU funds narrows. The main OECD recommendations related to capital markets are: (i)
listings (IPOs) of large state-owned companies and (ii) development of capacities for issuing
securitized financial instruments for financing public investment based on predictable cash
flows. In the last few years, incentives for such debt issues were low as the supply of EU funds
was greater than demand for financing of government investment projects. However,
availability of EU funds will diminish in the not-so-distant future. The first reduction of inflows
of EU funds will happen in 2027 after the closing of Recovery and Resilience Facility at the end
of 2026. Second cascade will come in 2030 when Croatia’s large allocation in Multiannual
Financial Framework 2021-2027 will be utilized. This leaves a window of a few years for the
government to develop know-how and capacities for issuing complex securitized instruments
for financing public investment. It will be critical for maintaining continuity of public
investment in the medium and long run when the importance of EU funds will decline. It is
recommended that administration capacity building and pilot projects start as soon as
possible to overcome potential cliff effects.
Room for improvement of pension adequacy through lower administrative costs in the
second pillar is narrow, as the combined effect of cancelation of entry fee and reduction of
management fee will lead to an increase in pension savings after 40 years by an estimated
maximum of 1.6%.86 However, permanent cancelation of entry fee (previously 0.5%) in 2024
will have a limited impact on pension savings because this effect is partly offset by
enhancement of eligible administrative costs that can be charged against assets under
management.87 Mandatory pension funds’ management fee was set by the Law at 0.27% for
2023.88 The effective management fee is lower as there is a limit for application of this rate in
large mandatory pension funds. Limit is set as a maximum value of assets under management
in proportion to market share of 20%. Any excess amount is paid back proportionally to the
pension accounts of all members of the second pillar. Two larger fund managers out of four
have market shares above 20% each. Their joint excess market shares above 40% is around
26 percentage points. Therefore, effective management fee is (100%-26%) times 0.27%, i.e.,
around 0.20%. This is around 4% of the historical net nominal return. It implies a small impact
of management fee on cumulative return and choice of payout. Recent legislative changes
introduced further gradual downward adjustment of the management fee to 0.20% until 2029
(around 0.15% effectively after taking market shares into account). With close to 27% of GDP
of assets under management in the second pillar, even a small fee is enough to make the fund
management a profitable business with 15-30% return on fund managers’ equity (ROE), but


86
   Maximum effect is calculated under assumption of 2.2% real net return (constant, annualized) for equal annual
contribution payment for 40 years. 0.5% is proportional to a maximum effect of full abandoning of entry fee.
The rest of 1.1 percent reflects cumulative effect of the long-term reduction of management fee by 5 basis points
below 0.20% envisaged in current regulatory rules.
87
   According to the largest pension asset manager annual report for 2022, costs of REGOS and HANFA taken
together represent 300% of revenue from the second pillar entry fee. Therefore, regulatory costs are mostly
covered from management fee.
88
   There was a declining schedule in annual steps defined in proportion of expected evolution of assets under
management, which started from 0.338% and ended with 0.27% in 2022 and 2023.

                                                       92
on the average, 60-70% of managers’ profits come from management of the second pillar
pension funds which is a lower proportion compared to the proportion of assets under
management (94%). The reason behind it is related to the fact that the same management
companies manage voluntary pension funds: fees from the third pillar make disproportional
contribution to fund managers’ profits, estimated at 30-40%. There are two explanations for
this disproportion: (i) the second pillar line of business consumes a higher percentage of
managers’ equity89 and (ii) voluntary pension funds have significantly higher management
fees, which is discussed in the next section. Given direct regulation of fees, setting the
management fee rates too low in the second pillar may drive smaller managers out of
business and reduce competition. On the other hand, elimination of limited return-risk
guarantee90 may provide some room for further reduction of administrative fees. Generally,
there is no substantial contribution to be expected from lower fees in the second pillar. More
important positive long-term effects may emanate from optimization of lifecycle portfolio
allocation (IAS) and liberalization of investment rules.


6.6 Contribution of voluntary pensions to pension adequacy


Voluntary third pillar pension funds were introduced in parallel with second pillar
mandatory pension funds in Croatia in 2002; however, contribution of voluntary savings to
the pension adequacy is still very small on average. As of end 2023, there were 443 thousand
members with accounts open in the third pillar. This represents around 28% of total
employment and around 14% of the population 20y+. Total net assets under management in
8 open voluntary pension funds and 21 closed (sponsored) were EUR 1.26 billion i.e., around
1.6% of GDP.91 Average voluntary pension saving per member was around EUR 2840 vs. EUR
9000 in the second pillar. Fixed annuity for a 20-year limited payout period based on transfer
of EUR 2840 to the Pension Insurance Company92 is EUR 14.10 per month. Fixed monthly
annuity for life for a 65-year-old is EUR 12.79. Annuities are not inflation-linked. Despite more
than two decades of stable growth of membership and assets under management, the
contribution of voluntary savings to the pension adequacy is still very small on average.
Voluntary scheme is based on the web of six incentives: (i) government subsidy up to around
EUR 100 for individual savings up to EUR 664 per year with effective subsidy rate of 15% up
to the limit, (ii) tax deduction from corporate income tax base up to EUR 796 per year per

89
    Assessment of fund managers’ ROE should consider the risk of guaranteed return which is paid by fund
management company if net returns of fund(s) under management deviate from the average (represented by
index MIREX). The following rules apply: gap versus average return is paid from managers’ own funds to
individual pension accounts if net return is (i) more than 12pp lower that the average in A, (ii) more than 6pp
lower than the average in B, and (iii) more than 3pp lower than the average in C, calculated at the basis of 3-
year annualized net return. Minimum guaranteed return rule was nev er activated because funds’ portfolio
structures are similar, and it is unlikely that it will be activated in the future.
90
   Ibid.
91
    Based on the sum of assets in open-ended voluntary pension funds that manage 83% of assets under
management in the third pillar and closed-end voluntary pension funds, which have corporate and individual
sponsors for own employees, that manage 17% of assets under management in the third pillar.
92
   Based on HRMOD annuity calculator on its web page as of April 20th 2024.

                                                      93
employee if employer pays contribution to workers’ pension account in the third pillar,93 (iii)
returns on voluntary pension savings are tax-free (no income taxation in the payout stage),
(iv) commencement of payout is independent from PAYG retirement: individuals with 55
years of age or more94 can initiate payout with a right to withdraw up to 30% of voluntary
pension savings in cash, (v) members can continue to save and receive government subsidies
irrespective of the statutory retirement age and/or formal retirement status, and (vi)
members who decide to retire with less than EUR 13,272 of voluntary pension savings after
one-off withdrawal of 30% of savings in cash can agree payments with pension fund
management company in regular monthly tranches in terms of number of units with variable
value (mimicking variable annuities). There is a higher threshold of EUR 17,254 which applies
for limited-period payouts. For voluntary pension savings above the threshold, amounts must
be transferred to the pension insurance company for payout. Below this threshold, a member
can also agree fixed annuity payment with PIC, but transferred amount of savings can be no
less than EUR 3,982. At the end of 2023, there were 4,539 payout contracts with PICs in the
third pillar. This number is declining (there were 7,908 such contracts in 2019). There is a
strong preference for payout via fund management companies due to simplicity and
acceptance of risk related to variation of value remaining under management until payment.
This also indirectly shows that the complex web of six incentives works in favor of individuals
with higher incomes who have the capacity to save, who will enjoy decent pensions anyway,
and who have some risk preference.95
Better targeting of incentives may attract individuals with below average incomes to the
third pillar and contribute to adequacy of their pensions . The fiscal cost of the existing
voluntary pension incentives is low. Croatia has fiscal capacity to increase the total amount of
incentives under condition of better targeting to improve pension adequacy for lower income
earners.96 Non-linear individual incentive scheme can improve rewards for individuals with
limited capacity to save (e.g., 30% subsidy up to savings of EUR 150 p.a., 15% on EUR 150-300
p.a., and 7.5% for EUR 300-600 p.a.). Such a scheme can be made even more degressive to
better target low-income earners. The final shape of the cascade should be determined after
closer examination of the existing distribution of contribution payments.
Employers’ incentive scheme also benefits high-income workers as voluntary pension
payments by employers serve for retention/attraction of workers in a tight labor market,
which is afforded by first-tier employers who most likely also pay higher wages. There is
incentives-related selection bias as high-quality employers, which offer superior
combinations of wages, promotion potentials, workers’ education and safety of work
contracts would find a way to reward quality workers anyway.97 To better target employers’

93
   Individual subsidy and corporate income tax are cumulative i.e., not mutually exclusive.
94
   Age limit was lifted from 50 to 55y in 2019.
95
   Relatively lax conditions for exit from the 3rd pillar raises suspicion that voluntary pension scheme cannibalizes
UCITS industry. There are more than 100 UCITS in Croatia with all sorts of investment policies, but their assets
under management amount to 2.9% of GDP only, which is barely 70% more compared to assets accumulated in
the VPFs. This hypothetical substitution effect also points to the possibility that the third pillar incentive scheme
incentivizes people who would save long-term anyway.
96
   In 2021, fiscal cost of voluntary pension scheme individual incentives amounted to 0.02% of GDP.
97
   And quality workers in recessions have higher probability to keep their jobs.

                                                         94
incentives, deduction from corporate income tax base can be enhanced depending on the
wage bracket of respective employee.
There are several additional regulatory adjustments that may improve policy targeting in
the third pillar. First, incentive thresholds (amounts) should be linked to inflation to keep the
real incentives unchanged. Second, the age limit for the exit, which is currently set at
minimum age of 55y, should be lifted closer to statutory retirement age (at least 60y,
potentially 65y). Third, government incentives for voluntary pension savings after statutory
retirement age can be abolished because justification for additional fiscal cost for motivating
additional individual savings in retirement is not clear. Probably, only a narrow group of
relatively wealthy beneficiaries can afford to save during retirement. They would most likely
continue to save without government subsidy. Fourth, the payout stage can be liberalized and
simplified. It is hard to justify regulatory limitations on the payout method, i.e., transfer of
assets to PICs versus payments of units with variable values from voluntary pension funds. At
the beginning of the multipillar pension reform there was a hope that the third pillar payouts
would boost PICs’ business volume before most new retirees begin exiting the second pillar.
This did not happen. Relative importance of the third pillar for PICs’ payout volume is
declining. Out of total payouts from the third pillar pension funds for various reasons in March
2024, only 16.7% was transferred to PICs. The volume of PICs’ business related to exits from
the second pillar accumulation stage will increase substantially in the coming years, so there
is no reason to limit payouts of voluntary pension savings via fund management companies.
It is enough to implement regulation on adequate information for beneficiaries who must be
aware that payout via PICs implies lower risk.
The cash withdrawal option at the moment of beginning of payout should also be
reconsidered, subject to an increase of the age limit for the beginning of payout closer to
statutory retirement age, making it more related to the retirement age. Earlier withdrawals
of cash could be allowed, but after payback of government subsidies with return in case of
exit from voluntary savings before retirement (except in the case of permanent work disability
and similar conditionalities).98 To allow some flexibility, gradual payback of government
subsidies plus return may apply to accommodate individual preferences and unfortunate life
events. If there was no unfortunate life event, cash payout before statutory retirement age
can start with 100% subsidy plus return payback at the age of 55, reaching 50% at the age of
60, and 0% at retirement age of 65 (with gradual adjustment proportional to the adjustment
of statutory retirement age for women). This would bring the third pillar in line with other
measures stimulating longer work. With this package in place, effectiveness of voluntary pillar
will not be assessed in terms of mere number of members in the third pillar pension funds
and value of total assets under management, but rather in terms of better targeting and
attraction of savings by people with lower incomes who would otherwise be more exposed
to the pension adequacy problem.
The administration cost is relatively high in the third pillar, whereas further lowering of fees
will depend on market size, i.e., assets under management. The regulatory cap on the
management fee is currently set at 3%. Competition pressures drove the fees down: asset-

98
     See further elaboration of withdrawal conditions in the section on autoenrollment.

                                                        95
based weighted average management fee for eight open-end voluntary pension funds
currently stands at 1.78% for largely balanced portfolios.99 However, growth of total assets
under management is modest, which does not promise stronger competition and lower fees.
A strong push for voluntary pension savings may be created by auto-enrollment program that
may involve incentives for staying with a default proposal of paying additional contributions.


6.7 Autoenrollment

Auto-enrollment comprises measures aiming at boosting (mainly) voluntary pension
savings based on the concept of nudges borrowed from behavioral economics. Individual is
informed about the default choice made by the government, and/or employer as an agent:
net personal income is reduced by an amount or rate X monthly, and X is paid to an individual
pension account, usually with an added contribution by an employer and/or the government.
A worker can change X or opt out and cancel the choice during a limited time window after
automatic enrollment, usually for about 30 days. After opting out, previous conditions remain
unchanged as if nothing happened. If default choice is accepted, it is locked-in, and restrictive
criteria for early cash withdrawals are applied until retirement.
Auto-enrollment is based on the experience which shows that a lower share of people opts
out from a default choice than opts in based on own active decision, and the difference is
particularly present among low-income earners.100 This is explained by low propensity to
save. However, low savings are not always driven by low incomes. It also involves high
subjective discount rates associated with financial illiteracy and psychological biases.
Therefore, default choice is framed in a positive information campaign. Campaigns also
include information about opting out which should be accessible - cheap, fast, and
anonymous - to avoid any potential complaints about information manipulation and lack of
freedom of choice.
Auto-enrollment programs gained popularity in the past decade after the successful launch
of the NEST program in the UK, but program designs, goals and outcomes are very diverse,
so it is not possible to easily identify the best practice. Still, there are several common
denominators: (i) Employers play important roles as distributors of information to jobholders,
and/or agents for contribution payments101 and/or, like in the UK, they select occupational
pension scheme in the market on behalf of employees. In the UK, scheme selection included
a government pension savings provider NEST which became the largest occupational pension
provider in direct competition with private pension savings schemes (Box 6.1). (ii) Program
usually starts with a pilot wave, which includes employers who participate voluntarily. The
purpose is to learn and adjust the next compulsory waves. (iii) Auto-enrollment is usually
applied in several successive compulsory waves depending on the company size: large



99
   Two funds with conservative investment policies have lower management fees of 0.75% and 1.15%.
100
    Rudolph, H. P. (2016).
101
    A specialized institution can act as a contribution paying agent on behalf of employer.

                                                   96
companies opting in early, and small companies opting in at later stages.102 (iv) Program
targets workers not covered, or not adequately covered by existing pension savings schemes,
and (v) existing pension institutions and expertise are used to avoid duplication and provide
uniform information to members of the pension scheme(s).


  Box 6.1 UK's auto-enrollment scheme

  UK scheme is considered to be a big success as 94% of automatically enrolled members accepted the
  default choice. Employers were obliged to join the program if they employed jobholders older than 22y,
  and younger than statutory pension age, who were not covered by any prior qualified occupational
  pension scheme (eligible jobholders). Non-eligible jobholders could opt in. Employers which previously
  employed qualified occupational pension scheme could stay with an existing scheme or opt in the program
  by replacing the old scheme with a new one. Eligible pension schemes are schemes administered in the
  European Economic Area (EEA) subject to qualifying conditions such as, for example, no exit before the
  age of 55 in case of DC schemes. Both DB and DC schemes were eligible, but eligibility conditions differ
  (e.g., exit age from DB schemes has to be in line with statutory retirement age and adjusted accordingly
  upwards if legal provisions for PAYG pension change). Employers are responsible for checking continued
  qualification criteria which have to be met by a selected pension scheme. Regulators expected that many
  employers would face difficulties in meeting and tracking scheme selection criteria, and thus it led to
  launching of NEST – National Employment Saving Trust, which is eligible by default. NEST is a government
  sponsored scheme which offers pension asset management services to employees via their employers,
  including target date funds. In a relatively short period, NEST became the largest occupational pension
  fund in the UK. In other similar auto-enrollment schemes (New Zealand, Canada/Quebec), there was no
  government sponsored centralized management entity. Employers were selecting a pension scheme
  among providers in a free market subject to prior licensing by the regulator (which is renewed every 7
  years in New Zealand). NEST started after four years of research and preparations (2008-2012) based on
  70 surveys and 6,000 interviews with stakeholders.

  In the UK, minimum total contribution rate was set at 8% of eligible earnings with employer's minimum
  contribution at 3% (contained in 8% of total). Target contributions were reached after gradual phasing-in
  since NEST's inception in 2012 until 2018. Initial contributions were 1%+1% (or higher, at will) for five years,
  until all employers were covered by the scheme in four waves until 2018. In 2017/2018 minimum
  contributions were set at 2% employer's + 3% member's contribution. Since October 2018, 8% minimum
  applies, hence 5% employee's contribution is a maximum. Base for contribution calculation includes all
  (eligible) earnings, not only salaries and wages (e.g., bonuses, sick pays, maternity leave compensations,
  etc. – employer makes assessment of qualifying earnings). Auto-enrollment is based on the contract
  between employer and provider of the qualified pension scheme. If employer provides full 8%
  contribution, no agreement with jobholder is required; if jobholder participates in 8% with individual
  contribution, there has to be a separate agreement between employer and a jobholder. A jobholder could
  opt out within 30 days after receiving a notice that the employer joined the scheme. Employers have to
  offer once a year a short time window when jobholders who opted out can rejoin. Those who opted out
  are automatically enrolled back every three years, but they can opt out again.




102
   NEST started with employers employing 250 or more workers in 2012-2014. Enrollment stage ended four
years later with employers employing less than 30 employees. Thresholds (4-6 thresholds in obligatory waves
depending on company size) generally depend on the size of target population. E.g., OregonSaves plan in the
State of Oregon was effectuated in six obligatory waves from 2017 to 2020 starting with employers employing
more than 100 employees ending with employers employing less than 4.

                                                         97
Poland is a very interesting case of a less successful auto-enrollment scheme. Polish pension
reform was similar to Croatia’s until 2013 when a gradual dissolution of the multipillar reform ,
which took six years to complete, embarked policy thinking about voluntary occupational
scheme. Auto-enrollment program started by creating voluntary occupational pension funds
(PPK – Employee Capital Plans) in 2019. PPKs manage contributors' portfolios on a
generational basis following the life-cycle logic. There is a regulatory cap on the management
fee: 0.5% per annum is a maximum for a flat part, and 0.1% is a maximum for a success fee.
Employer selects PPK on behalf of employees. In this sense, the program is obligatory for
employers and voluntary for employees who can opt out. PPKs are funded by simultaneous
employer-employee contributions. Employee contribution rate is 2% of gross wage (net wage
is reduced proportionally), but employees with incomes below 120% of minimum wage can
contribute at a lower rate of 0.5%. The minimum employer's added contribution is 1.5% for
all. Both contributions can be increased up to 4% each. Companies with more than 250
employees started auto-enrollment in 2019, followed by mid-sized companies in 2020. Small
companies and public entities followed in 2021. Employees aged 18-60 in legal persons were
included in the scheme, estimated at 11 million out of 18 million total of employed persons
in Poland as of 2019. Money can be withdrawn in full amount at the age of 60, or at any time
before the 60th birthday. This contrasts with UK's autoenrollment scheme which is not open
for early exit and the default choice is reinforced every three years for members who opt
out.103 However, approximately 60-70% of automatically enrolled employees decided to opt
out in Poland.
Extended opt out provision on top of the lack of employers’ interest transformed the Polish
PPK scheme into an alternative savings product, rather than a pension scheme. Together
with an earlier dismantling of the multipillar reform, extended opt out provision is probably
the main reason for a high opt out rate in Poland. Lack of interest among employers is
associated with the fact that a worker who accepts the default choice poses a higher cost to
employer because subsidies/incentives do not compensate for the present value of
employer's minimum contribution (1.5%). The government participated in the scheme by
adding a welcome transfer of EUR 54 to individual pension account, and there were no strong
enough financial incentives for employers. So, in weighing potentially positive contribution to
human resource management and a negative contribution in terms of higher labor cost,
employers felt the latter was stronger.104 Lessons learnt from Poland is that radical changes in
the pension system diminish pension reform credibility: new policy solutions should be built
on the shoulders of the existing ones after careful preparations and analysis including
assessment of the balanced incentives for stakeholders involved. The absence of balanced
incentives is an obstacle for successful reform.
Additional explanations for a high opt out rate in Poland may be related to lower incomes,
lack of financial literacy and limited experiences with capital markets instruments
compared to countries where auto-enrollment was successful (UK, New Zealand, US, and


103
    Early payouts from voluntary pension schemes are usually limited to special circumstances such as serious
health issues, permanent disability for work, and serious financial difficulties.
104
    Sczepanski and Kolodzieyczik (2021).

                                                     98
Canada). In a country like Poland, as well as in Croatia - where limited scope of demand for
voluntary pension savings is well documented - inability to save due to hardships in meeting
basic financial needs in conjunction with weak financial literacy and high subjective discount
rates may prompt more frequent opting out, especially among lower-income earners. It is less
likely that other savings instruments, such as life insurance products, mutual funds and direct
investment in securities serve as substitutes for savings for retirement income; they do not
represent significant portion of households' portfolio of assets.105 However, strong
preference for real estate ownership probably crowds out other forms of long-term savings,
so it may contribute to opting out.106


Real estate preference among Croatian population must be considered in setting realistic
goals and proper design of an auto-enrollment scheme. Widespread lifetime real estate
ownership is both a blessing and a curse: absence of pecuniary rent payments is helping the
living standard in the old-age,107 but costs of capital and maintenance/depreciation during
service years may push the opportunity cost of life-long real estate investment high. Due to
lack of ultra-long time series, it is not possible to provide an expert-based opinion on the long-
term superiority of any asset class in Croatia beyond observation that concentration of assets
in one asset class such as real estate increases the risk due to lack of portfolio diversification.
A possible solution is to open voluntary pension savings for a limited one-time withdrawal
(50-90%) in case of purchase of the first residential real estate before reaching a certain age.
The age of 42 may be considered young enough to leave the rest of career long enough to
accumulate significant voluntary pension savings. Age-limited-big-percent (100% is also a
possibility) withdrawal option would nudge younger people towards accepting the default
choice. Accepting default at a young age is beneficial per se, as young workers are getting
acquainted with long term savings for pension.108
Combination of relatively high income, real estate ownership, and higher adequacy of
combined pension P2 may incentivize older cohorts in higher-income groups to opt out from
the auto-enrollment scheme more frequently. Widespread home ownership in conjunction
with existence of the second pillar has two additional implications for the design of an
effective auto-enrollment voluntary pension scheme in Croatia. Firstly, as existing evidence
suggests, and additional research should confirm, a large majority of existing members of the

105
     According to results of the European Central Bank's Households Financial and Consumption Survey 2021,
lowest share of households' wealth in financial instruments was in Croatia (Table C.1, Statistical Appendix, p.
15). However, with 80% of surveyed households being real estate owners (official statistics brings this
percentage closer to 90%, the second highest in the EU after Romania), Croatia is among the countries with
highest real estate ownership. Moreover, 86% of residence owners in the bottom pentile of the income
distribution, which largely consists of elderly population, is by far the highest share among surveyed EU member
states (Table B.3).
106
    Ibid.
107
    It also leads to depreciation of the housing stock due to minimum maintenance, but it is reasonable to assume
that the cost of depreciation is substantially smaller than out of pocket expense for rent and cost of risk
associated with it in the old-age.
108
    Solution with upper age limit for one-time withdrawal related to purchase of the first residential real estate
was applied in New Zealand.

                                                       99
voluntary third pillar are on average higher-income earners who are also relatively real estate
rich. This was the main argument for a proposal to better target the third pillar incentives
towards lower-income earners. The third pillar reform described in Section 6.6 should be
designed with having in mind a potential auto-enrollment scheme oriented towards lower-
income earners. Secondly, higher-income earners are more likely to reap the benefits in terms
of combined P2 in the long run (P2 > P1). If these two assumptions are true, combination of
relatively high income, real estate ownership, and higher adequacy of combined pension P2
may incentivize older people in higher-income groups to opt out from the auto-enrollment
scheme.109 However, if it happens at all, it should not be perceived as a public policy problem.
Incentives should not be designed to achieve an overall low opt out rate. Higher share of opt
outs in higher-income groups can be sacrificed if scheme's design will be tilted towards
accepting the default choice among lower-income earners who are more likely to opt back to
P1, who are probably rare among the third pillar participants right now, and who will be more
exposed to risk of having no adequate pension.110
Auto-enrollment scheme in Croatia should aim to attract lower-income earners who are
more likely to face pension adequacy problem, and criteria for withdrawals should be set
accordingly. Limited withdrawal options should be designed with this target group in mind.
Besides early withdrawals in case of serious illness, permanent work disability and purchase
of the first residential real estate, scheme should also consider criteria for early withdrawals
in case of serious financial difficulties which are more frequent among target population.
Without this option, auto-enrollment might miss the target group.
With the second and the third pillar in place for longer than two decades, Croatia may
introduce auto-enrollment relatively easy without major prior regulatory reform and
without a need for implementation in successive waves from big to micro enterprises.
REGOS can easily administer the auto-enrollment program which can target all employees up
to the age of 60 at once. Still, the key institutional question about relations between the
second and the third pillar in implementation of the auto-enrollment scheme in Croatia
remains. Involving the second pillar pension funds in the auto-enrollment scheme could
create confusion about what is obligatory, and what is voluntary. Opting out from the second
pillar auto-enrollment program would increase demands for opting out from existing
obligatory 5%, and potentially lead to dilution of the second pillar. Also, mandatory pension
funds face more restrictive investment limits and involve no financial incentives.
Simplicity, broad understanding, building upon the structure of the existing system, and
ease of implementation suggest that auto-enrollment should involve voluntary pension
funds only. Auto-enrollment should come after the reform of incentives as proposed in

109
    Subject to ceteris-paribus clause. Other factors such as excess savings among very wealthy may lower
incentives for opt out. On the other hand, higher financial literacy among higher-income earners may increase
opt out due to wish of confident people to make free individual choice.
110
    Auto-enrollment program is also the solution to the so-called “sales force problem” (Rudolph, 2016). Sales
force is critical for informing and incentivizing potential participants in the voluntary third pillar. However, fixed
cost of sales for private pension savings providers acts as a selection mechanism which is tilted towards higher-
income individuals as sales targeting strategies tend to maximize ratio of expected assets under management
over fixed cost of sales.

                                                        100
Section 6.6. It would also pave the way for a fundamental third pillar reform aiming at
increasing competition and pension products diversification, lowering administration costs in
the third pillar (management fees), and increasing individuals' early interest in savings for
pension (active choice) which we discuss in turn.
With an additional 300-700 thousand individual voluntary pension accounts for low wage
earners, additional annual inflows of voluntary pension savings may increase between EUR
84 and 197 million, i.e., between 0.11% and 0.26% of 2023 GDP. Under the assumption of
contribution rate of 2% applied on average gross income below median wage (fourth
decile),111 additional 300-700 thousand individual voluntary pension accounts would produce
an increase of voluntary pension savings inflow in the range of EUR 84 and 197 million per
year. This is equal to: (i) between 0.11% and 0.26% of 2023 GDP, (ii) between 69% and 160%
of annual contribution payments to the third pillar in 2023, and (iii) between 7% and 16% of
annual contribution payments to the second pillar in 2023. Under the assumption of 2.5%
long-term annualized constant real return and no payouts, such inflow would create
additional assets under management in the third pillar between 1.3% and 2.9% of GDP after
10 years, and between 7.5% and 17.6% of GDP after 40 years. The expected increase of the
total real net benefit ratio112 would be significant: 1.6% after 10 years, and 9.5% after 40
years, coming solely from additional contribution of 2% under auto-enrollment program.
Figures serve for illustrative purposes only, as realistic expectations about potential effects of
auto-enrollment should be built on evidence, surveys, research, and modelling including
payouts.
The most important takeaways from the initial consideration of a voluntary auto-
enrollment program in the case of Croatia are: (i) probable increase of pension adequacy
for low-income earners, (ii) increased competition, portfolio diversification and quality of
pension asset management, and (iii) the decrease of administrative costs in the third pillar
associated with higher business volumes. Firstly, probable increase of pension adequacy is
the reason to include broader spectrum of eligible incomes and forms of employment in
autoenrollment, not only wages and salaries; such as self-employed, craftsmen, workers in
agriculture and lump-sum income taxpayers who earn from renting apartments. The final
scope of types of income would depend on the Tax Administration’s capacity to administer
synthetic income data and transfer it to the scheme’s public administrator (probably REGOS).
Secondly, the expected additional growth of assets under management opens room for
increasing competition by broadening the spectrum of pension savings products. Existing
providers of voluntary pension savings products may be incentivized or obliged to create life-
cycle portfolios and/or target date funds. Market access can be opened for pan-European
pension products and qualified EU-licensed providers of pension savings services who will
strengthen competition and contribute to international portfolio diversification and quality



111
    Fourth decile gross wage for December 2023 was used in the calculation (EUR 1,172). Below median wage is
used due to targeting of low-income earners.
112
    Individual lifetime annuity without withdrawals based on HRMOD calculator divided by 4th decile net wage
in end 2023.

                                                    101
of pension asset management.113 Thirdly, predictable growth of business volumes will open
room for cascading decrease of maximum rates of management fees in the third pillar. It
would contribute to pension adequacy per se, and indirectly benefit existing members of the
voluntary pension scheme.
Financial incentives and choice of default contribution rate are key elements of the auto-
enrollment program. A higher default contribution rate can magnify the expected
improvement of total multipillar replacement rate, as illustrated above. However, higher
default is a double-edged sword. Overly high rate of jobholders' default contribution may
increase opting out as some people will employ binary choice (accept default or opt out). This
may happen even if there is no lower limit on individual choice of the contribution, so that
auto-enrolled individual can reduce the default contribution rate. Simplicity suggests framing
of choice within a range of 1pp increments (e.g., 1%, 2%, 3% or 4%) with 2% as a reasonably
balanced starting point (default).114 On top of it, default contribution rate should be
considered in conjunction with another behavioral nudge: delay of pain associated with
today's choice. E.g., a 0% contribution can be offered as (i) opting out, or (ii) starting point
which leads to automatic activation of 2% contribution after a year. It is easy to imagine an
app where incremental range between 0% and 4% is offered, and after a click on 0%, a new
screen pops up offering alternatives: opt out (final cancelation) or delay of pain (0% for a year,
and choice of 1%-4% thereafter). Final design should be dependent on the results of ex ante
research.
Incentives are associated with contributions, and incentives imply fiscal cost; however,
fiscal cost can be controlled by shaping incentives in favor of low-income earners who
represent target population in the program. There is a way to achieve a high participation
rate among target population and a low fiscal cost by applying nonlinear individual
subsidization scheme of the kind proposed in Section 6.6 before implementation of auto-
enrollment. Following our illustrative calculation based on 300-700 thousand new accounts
opened, existing individual incentive scheme (involving 15% government subsidy up to a limit
close to EUR 700 per year) with 2% average contribution applied on the fourth percentile
average gross wage implies a very small additional fiscal cost between 0.017% and 0.039% of
GDP per year. Implementation of the nonlinear scheme which is tilted towards low-income
earners (30% subsidy up to annual contribution payment of EUR 150, 15% between EUR 150
and 300, and 7.5% for EUR 300-600) does not change the conclusion as the range is between
0.025% and 0.059% of GDP. Nonlinear cascading scheme increases fiscal cost in proportion to
participation of low-income earners who pay smaller amounts, but fiscal compensation may
come from smaller expenditure for subsidies on voluntary pension savings for higher-income
earners.115 Following this logic, auto-enrollment may lead to consideration of an even more

113
    EU passport only is not enough for automatic eligibility under auto-enrollment program. Additional eligibility
requirements should apply to meet the reform objectives (e.g., restrictive withdrawal criteria and exit
conditions).
114
    There is no need to have different defaults depending on the wage level like in Poland. It is enough to inform
workers that they can stay in with a lower rate. Increments in decimal points (0.5pp or even 0.1pp) can be
considered as well subject to opinions of behavioral finance specialists.
115
    Exact calculation requires further in-depth research of Ministry of Finance and VPFs' subsidy and contribution
data.

                                                      102
degressive individual incentive scheme compared to one proposed in Section 6.6: for
example, omitting the third bracket would imply fiscal savings, orient the third pillar towards
low-income earners, and still leave a decent incentive for savers who can afford to contribute
more.116 Policy makers may also consider extreme solutions such as subsidizing strongly only
up to EUR 150 contribution payments per year.117 In general, minor fiscal cost is a strong
argument in favor of no differentiation between the reformed regular fiscal incentives for
individuals in the third pillar and the incentives applied in the auto-enrollment program. This
condition is primarily based on arguments of simplicity, transparency, low administration
cost, and ease of understanding. The final choice of the individual incentives cascade should
depend on the results of further research based on surveys and interviews.
Table 6.4 Illustrative calculation of auto-enrollment effects
 Gross wage, 4th decile                 1172
 Net wage, 4th decile                    903
 Default contribution rate             2.00%
 Real rate of return, net              2.50%
 Annual contribution, EUR             281.28
 Average assets after 10y           3,151.29

 Average assets after 40y          18,958.99
 Lifetime mo. annuity, 40y             85.38
 Number of new pension
 accounts opened                     300,000      400,000       500,000       600,000       700,000
 Existing fiscal cost (15%
 individual subsidy) in % of
 GDP 2023                            0.017%        0.022%       0.028%        0.033%        0.039%
 Nonlinear incentive (30% up
 to EUR 150, 15% above) in %
 of GDP                              0.025%        0.034%       0.042%        0.051%        0.059%
 Maximum        fiscal     cost
 associated with employers'
 payments in % of GDP                0.020%        0.026%       0.033%        0.040%        0.046%
 AUM in % of GDP, 10y*                  1.3%         1.7%          2.1%         2.5%           2.9%
 AUM in % of GDP, 40y*                  7.5%        10.0%         12.6%        15.1%          17.6%
*No payouts



Employers’ incentives are equally important, and they should also favor low -income
earners and balance this goal with fiscal prudence. Table 6.4 shows maximum fiscal cost
associated with employers' incentives under assumption that all contributions will be paid by
employers under present employers’ incentive scheme. It is not realistic, but it is illustrative
as it shows an extreme outcome. At present, all voluntary contributions paid by employers
are tax free and deductible from corporate income tax base. The incentive (fiscal cost) is equal

116
    Individual who saves EUR 1,000 per year would receive effective weighted rate of subsidy of 6.75%, which is
still high. Additional incentive may still come from employers' payments of contributions.
117
    Effective subsidy return rate with 30% up to EUR 150 would be 4.5% per EUR 1,000 annual contribution.

                                                     103
to 18% of the contribution cost for big employers who pay 18% corporate income tax rate,
and 10% of the contribution cost for small employers who pay corporate income tax at 10%.118
This is corporate income tax foregone. When figures from our illustrative calculation above
are applied under extreme assumption that all contributions will be borne by big employers
who pay the tax rate of 18%, maximum additional fiscal cost is between 0.020% and 0.046%
of GDP. Note that fiscal cost is cumulative as employers' benefits are applied on top of
individuals' benefits, but this sum is still small: around 0.1% of GDP per year in the case of 700
thousand new contributors. It does not follow that nothing should change with respect to
employers’ incentives in auto-enrollment compared to the existing system.
Also, it does not follow that employers’ contributions should be strictly obligatory for
workers who accept default. As witnessed in the case of Poland, unnecessary imposition of
additional cost of labor may deter cost competitiveness of enterprises and create perverse
incentives for employers who will act passively or even discourage workers' acceptance of
default choice. Active employers armed with a positive attitude and motivation may be critical
for success as they act as information brokers for their employees, and their additional
contribution is an added motive for accepting the default. As a matter of principle, employers’
incentive should also follow non-linear cascade which is tilted in favor of low-income earners.
For example, employers’ additional contribution can be obligatory only below certain wage
threshold, and/or tax shield can be enhanced for contribution payments for workers with
lower wages hence eliminating negative effect of obligatory adding of contribution. The final
design of employers' incentive should be based on in-depth consultations, research, and
modelling, having in mind a need to employ occasional repetition of auto-enrollment (e.g.,
every 3 years like in the UK) for workers who decide to opt out. Also, the concept of below
certain wage threshold can be implemented with having in mind the interactions with the
second pillar as our analysis showed a small likelihood that persons with salaries below 81%
of average (minimum pension beneficiaries = 1.03/1.27) will choose P2.
Two potential pitfalls, namely, crowding out of other investment products and no positive
feedback loop on fiscal sustainability of PAYG, should be considered and minimized. Firstly,
success of the voluntary auto-enrollment may cannibalize other investment products offered
in the free market. This is controlled by setting incentives to target low-income workers who
largely do not purchase other investment products. Additional mechanism of control is based
on imposing restrictive conditions for early cash withdrawals after setting the exit age from
the voluntary accumulation stage closer to or at statutory retirement age. This restriction will
effectively establish a firewall between pension savings products and other products aiming
at shorter time horizons. Secondly, we showed that auto-enrollment will imply (small)
increase in the fiscal cost of the third pillar, but there is no positive fiscal feedback loop in
terms of impact on PAYG fiscal sustainability. Key question is: can positive fiscal loop be
established with respect to the second pillar? A radical solution would be to give back
individual subsidies with added returns from the third pillar to the government in case of
choice of P1. This solution would complicate information required for making an informed
choice and disincentivize people with low probability of P2>P1 who belong to the target

118
      Threshold is at EUR 1 million of annual revenue.

                                                         104
group. This is not feasible. However, if program will follow prior implementation of our
benefit-as-lump-sum-transfers proposals (choice-neutral social transfers elaborated in
Section 6.3) which positively affect the frequency of P2>P1, a proposed radical solution may
be added to the list of policy options to be considered. Alternatively, obligation to return
subsidies in case of P1 selection can be applied only to higher-income earners who opted in.
Finally, the launch of the auto-enrollment program is an opportunity for a leap forward in
terms of promoting active individual choice of pension savings. The promotion can be done
via information campaigns, distribution of leaflets, and design of web services and
applications with new features such as bots and AI. It would particularly serve the younger
generation, helping the development of their financial literacy with long-term financial
benefits in mind. In combination with the second pillar reform (e.g., implementation of the
Incremental Allocation System, IAS, see Section 6.5) it can boost senses of freedom to choose
and of individual responsibility for one's own financial future. At a minimum, new entrants to
the pension system would have to simultaneously consider opting out from the voluntary
system and a choice of the mandatory second pillar fund: it may incentivize more new
entrants to make an active choice in the second pillar and learn more about the pension
system at the beginning of their career. At a maximum, if auto-enrollment were implemented
in parallel with IAS, jobholders would face a complex set of pension issues at the same time.
It can speed up the learning curve. On the other hand, too much information and decision
requirements may crowd out relevant information and cement passivity (freezing effect in
front of the shelf with too many differentiated but similar products). Final saying on this
matter should be left to psychologists, behavioral economists, financial literacy, and
communication experts based on the results of thoroughly designed research. The
combination of good communication, optimum timing and sequencing of carefully prepared
reforms will be critical for success.119




119
   It also requires good communication preparation in terms of addressing critiques of the existing system,
which mostly aim at allegedly low net returns and factually low share of the second pillar contributors who
choose P2 over P1. These topics should be addressed before launch of the auto-enrollment program.

                                                   105
7. Policy options to improve pension adequacy and fiscal
   sustainability

7.1 The proposed policy measures – initial list

In this section we summarize analyses and policy considerations discussed so far and
present our initial list of policy measures for improvement of pension adequacy and/or
fiscal sustainability, some of which will be analyzed by microsimulation Croatia PROMIS
model. First, we provide the list of potential measures with short explanations where
needed.120 Second, we estimate by the individual TRR model the impact of selected policy
measures on pension adequacy for hypothetical individuals. Third, we present the impact of
the selected policy measures on the average pension adequacy and fiscal sustainability
estimated by Croatia PROMIS model. Fourth, policy proposals are finally bundled in policy
packages which are applied in both TRR and Croatia PROMIS models where appropriate. It
should be emphasized that policy proposals that could have not been simulated by the
models, e.g., third pillar policy measures, are presented in the policy lists below for
consideration by policy makers based on the analysis presented in the text.
The main adjustment paths of the proposed measures include:
(i)      Slowing down the inflow of new pensioners by adjusting the pension eligibility
         conditions, mandating and/or encouraging later retirement, and by narrowing
         pathways to early retirement. These measures will automatically translate to higher
         pensions and open fiscal space for targeted interventions.

(ii)     Limiting the announced general increase of pension expenditures by keeping below-
         the-wage valorization and indexation of pensions.

(iii)    Strengthening of the funded pension schemes to better complement the public PAYG
         scheme in reaching a higher pension adequacy.


        7.1.1 Measures for improvement of pension adequacy and/or fiscal sustainability in
              the PAYG


1. Gradual increase of early retirement age for men from 60 to 62 years in the period 2025-
   2032 (or 2026-2033), by 3 months per year. If standard retirement age for men is
   increased in the short run, raise the early retirement age by 3 months in addition to
   standard retirement age increase.


120
   The list of measures discussed and analyzed combines the reform plans of the client and the team’s
diagnostics presented in the report. The Terms of Reference provided guidance regarding policies stated in the
NRRP and those likely to be considered by the client. The client’s intended policy concepts implied i) no reversals
of previously introduced benefits, ii) higher valorization and indexation rate, iii) no increase in overall
contribution rate, and iv) no “radical” overhauls of the pension system.

                                                       106
2. Increase of age for old-age pension for long-term insured from 60 to 62 years for men and
   women in the period 2025-2032 (or 2026-2033), by 3 months per year.

3. Increase of insurance period for old-age pension for long-term insured from 41 to 42 years,
   from January 2025 (or January2026).

4. Increase of penalty for early retirement pension from 0.2% to 0.3% or 0.4% per month of
   early retirement, from January 2025 (or January 2026).

5. After expiry of the current transitional period for old-age and early retirement pensions
   for women in 2030, increase of early retirement age for women from 60 to 62 years in the
   period 2031-2038, by 3 months per year. If the projections indicate more fiscal space is
   needed in the short run, raise early retirement age for women earlier in the same
   increment as standard retirement age.

6. Increase of statutory retirement ages for old age pension, early retirement pension and
   old age pension for long-term insured by at least 1 month per year, from January 2039 or
   earlier if the standard or early retirement age adjustment period accelerates.121


       7.1.2 Measures for improvement of pension adequacy in PAYG only

1. Valorization/Indexation based on 85%:15% wage-inflation rotational formula without
   negative adjustment (current adjustment rules with new weights).
      The change in weights was the subject of public policy discussions for some time, whereas
      measures such as separation of the valorization formula from the indexation formula
      were treated as being out of reality in the current circumstances. The proposed weights
      provide a good balance between adequacy and sustainability requirements.
2. Increase of disability and survivor pensions by extending the added period.

      Increase of disability and survivor’s pensions by extension of period for calculation of
      added period by 5 years, to 65 years of age.
      According to the current legislation, the added period is calculated as 2/3 of the period
      between the age at the occurrence of disability/death and 55 years, plus ½ of the period
      between 55 and 60 years of age. Despite the added period, the average disability pension
      lags the average old age pension and average survivor’s pension (see section 3.3.5)
      By prolonging the period for calculation of added period, around 100 thousand current
      disability pensions and 20 thousand current survivor’s pensions would be recalculated due
      to added period prolonged by average 2.3 years, resulting in average increase of these
      pensions of around 7%. New disability and survivor’s pensions would be increased


121
   The increase of one month annually is close to 2/3 increase of life expectancy (LE) at the age 65 for men and
women. Predetermined increase of retirement ages is more suitable than the automated adjustments to
changes in LE, as it is more transparent and known in advance, giving the time to insured persons to adapt and
plan their retirement.

                                                     107
      accordingly. Consequently, the risk of poverty for these beneficiaries would decrease and
      the levels of different categories of pensions would be more aligned.122



3. Abandon the requirement of 35 years of insurance period for increase of initial factor for
   old-age pension.

      Currently, for persons that delay retirement after the statutory retirement age, initial
      factor increases by 0.45% per month, but only in case the insurance period of at least 35
      years, which negatively affects old-age pensioners that have less than 35 years of
      insurance.

4. Increase of pension supplement from the current 4%-27% to 7%-30%

      The increase of pension supplement to 7%-30% would establish a more appropriate
      relation between pre-reform pensions realized until 1998 and post-reform pensions,
      which lag the pre-reform pensions. An increase of pension supplement would also
      decrease the gap between pensions realized according to general legislation and pensions
      realized under more favorable special schemes regimes.

5. The increase of minimum pension from 103% to 106% of APV per year of insurance from
   January 2026 is in line with the three percentage points increase of the pension
   supplement mentioned above.

      As the relative increase of minimum pensions would be the same as the increase of
      pension supplement, the current and prospective number of minimum pension
      beneficiaries should not be affected.


       7.1.3 Measures for improvement of the contribution to benefit ratio


1. Apply the increase of the initial factor for long-term insured for maximum of 60 months.

2. Initial factor would increase only in the case of retirement after the statutory retirement
   age for persons that completed insurance period in increased duration (beneficirani staž).


3. Equalization of contribution rate for other income with the general contribution rate.
      By increasing the pension insurance contribution rate for other income from 10% to 20%,
      tax burden on all types of personal income would be similar, contribution revenues would
      increase, and distortions in the labor market due to lower contributions for some types of
      personal income would decrease.


122
   This proposed measure requires administrative effort to recalculate and send individual notices or legal
decisions to the affected 120 thousand beneficiaries.

                                                   108
 7.1.4 Measures for improvement of second pillar and/or combined pensions


1. Increase the 2nd pillar contribution to 7% when the social and fiscal conditions are
   appropriate and in balance.
   Analysis of the options for further substitution of the pension contribution rates
   (raising the rate for second and reducing for PAYG) should precede such a policy
   decision. The timing of its potential implementation should be carefully tailored. The
   analysis should also provide policy guidance for setting the basic pension factor –
   proportionally reducing it or leaving it unchanged at 0.75.
2. Apply pension-type-choice-neutral transfer scheme: the same choice-neutral absolute
   amount of social benefit/transfer from the first pillar received irrespective of choice
   of the pension payout model.
3. Promotion of active choice of investment allocation and implementation of
   Incremental Allocation System (IAS) for smooth transition between different risk-
   portfolios during lifecycle.
4. Elimination or further relaxation of cash withdrawal condition (basic pension to
   minimum pension) after implementation of research to test the hypothesis that risk
   of choosing P2 by beneficiaries with P1>P2 is small.
5. Enrich the information sheet for new retirees by showing not only the first P1 and P2
   pensions, but also future expected pensions to reflect effects of different indexation
   and profit participation in P2.
6. Further liberalization (simplification) of investment limits.
7. Implementing capital markets development program including listings of SOEs.
8. Issuing securitized financial instruments for financing public investment.


 7.1.5 Measures for improvement of voluntary pension savings


1. Apply non-linear cascading incentive scheme for better targeting low-income workers.
2. Inflation indexation of thresholds for government incentives.
3. Increase exit age from 55 closer to statutory retirement age and consider early
   withdrawal option activated at 55, subject to return of government incentives with
   returns, whereas share of returned incentives decreases with approaching statutory
   retirement age.
4. Increase existing cash withdrawal option from 30% to 50%, but not earlier than
   statutory retirement age.
5. Free choice of payout method irrespective of amount of pension savings (payout via
   management company vs. payout via PIC: allow variable annuity).
6. Further liberalization of investment limits.
7. Abolish fiscal incentives for people who pay contributions to the third pillar after
   statutory retirement age.

                                         109
      8. Auto-enrollment after implementation of measure (1) for targeting lower-income
         workers:
            a. Include all workers under the age of 60 (older can opt in) in one wave and
               target lower income workers.
            b. Set default contribution rate around 2%, to be paid from gross wage (net wage
               is diminished accordingly) after thorough research based on surveys and
               interviews and set upper contribution limit at 4%.
            c. Implement delay of pain: option to set contribution at 0% initially (instead of
               opting out) and increase later.
            d. Re-enroll workers who opted out every 3 years.
            e. Set early withdrawal conditions to better target lower-income and younger
               workers by, e.g., including withdrawal options in case of severe financial
               difficulties and in case of purchase of the first residential real estate under the
               age of 42.
            f. Design employers’ contribution scheme to maximize their participation and
               minimize increase in labor cost.123
            g. Strengthen competition, broaden the spectrum of voluntary pension savings
               products including life cycle and target date funds, and pan-European pension
               products which meet other auto-enrollment conditions (introduce concept of
               eligible scheme).
            h. Set downward cascade of maximum management fees depending on assets
               under management like in the second pillar (but with higher rates due to lower
               expected AUM than in the second pillar).
            i. Design and implement information-education campaign based on web
               services and apps involving AI, linking it with a sister-campaign for promotion
               of IAS; coordinate two projects closely.


7.2 The impact of selected policy measures estimated by the TRR model

       7.2.1 Change in indexation/valorization formula


Valorization/indexation formula is one of the most crucial elements of the evolution of the
replacement rates in the long run. Figure 7.1 a illustrates the outcome of two extreme
indexation options, 100% with inflation and 100% with wage growth shown. With a 100%-
inflation indexation in the case of M65/40, the gross replacement rate declines from 38% to
21% in the case of full PAYG pension (P1). In the case of 100%-wage indexation the
replacement rates for full PAYG pension remain unchanged over the entire projection
horizon.



123
  The policies could help reach high coverage through diversifying the employers’ responsibilities across
employees’ income levels and tax treatment of default contributions.

                                                  110
Not only is the combined pension (P2) projected to rise at somewhat higher rates than the
full PAYG pension, but it is also expected it will become higher than the full-PAYG pension
by 2030 even in the case of 100%-wage indexation. With such indexation, the future
replacement rate of P2 will be higher than today. In our illustrative case of M65/40, the gross
replacement rate is expected to increase from 38% to 40% in the next 20 years and further to
40.5% by 2064 (Figure 7.1 a). In the case of 100%-inflation indexation, P2 will reach P1 faster
than in case of full wage indexation and stay well above P1 in the future, although it will face
declining replacement rates.
Our policy proposal of indexation 85%:15% by wage growth and inflation should stabilize
future replacement rates for combined pensions and slow down the reduction in
replacement rates for full PAYG pension (Figure 7.1 b). Improvement in replacement rates
over those with the 70%:30% wage-inflation rule is notable.
The effects of changes in indexation formula to 85%:15% rule require careful estimates of
fiscal costs and long-run fiscal sustainability (see Section 9.1).


Figure 7.1 The impact of indexation rules on the gross TRR
                  a) 100% wage indexation and 100% inflation valorization/indexation
42.0%


37.0%


32.0%


27.0%


22.0%


17.0%
        202420262028203020322034203620382040204220442046204820502052205420562058206020622064
                                 Full PAYG pension - 100% inflation indexation
                                 Combined pension - 100% inflation indexation
                                 Full PAYG pension -100% wage indexation
                                 Combined pension - 100% wage indexation


                    b) 70%:30% and 85%:15% wage/inflation valorization/indexation




                                                   111
40.0%

39.0%

38.0%

37.0%

36.0%

35.0%

34.0%

33.0%

32.0%
        202420262028203020322034203620382040204220442046204820502052205420562058206020622064
                               Full PAYG pension - 70% wage 30% inflation indexation
                               Combined pension - 70% wage 30% inflation indexation
                               Full PAYG pension - 85% wage 15% inflation indexation
                               Combined pension - 85% wage 15% inflation indexation
Note: The results refer to the average wage earner, M65/40, passive member with flat wage profile.
Source: WB Team




                                                          112
       7.2.2 Increase in minimum age for early retirement


The proposed increase in minimum age for early retirement from 60 to 62 in 2031-2038
period without changing minimum requirement of service period is likely to lead to
significant increases in replacement rates in the transition period. Figure 7.2 shows the
impact of rising minimum age for early retirement in cases of women retired immediately
after fulfilling minimum conditions for early retirement. In 2024 the minimum age is 58 years
and 6 months, while minimum length of service is 33 years and 6 months. If these “first early
retirement conditions” are raised in line with the current legislation up to 2030, and then up
to 2038 in line with our policy proposal, the gross replacement rate for the average wage
earner, women, increases by 1 percentage point, from 28.3% to 29.3% between 2024 and
2038 for full PAYG pensions. The increase in the case of the combined pension is even larger,
from 28.3% to 30.5%. If there will not be further adjustment in minimum age for early
retirement, then after 2038 replacement rates for both full PAYG and combined pension will
be on a continuous decline.
Figure 7.2 Gross TRR with higher minimum age for early retirement (women, retirement with the first
conditions for early retirement)
  31.0%

  30.0%

  29.0%

  28.0%

  27.0%

  26.0%

  25.0%

  24.0%



           Full PAYG pension - pre-reform                           Full PAYG pension - increase in minimum age
           Combined pension - pre-reform                            Combined pension - increase in minimum age
Note: The result refers to women, average wage earner, flat wage profile, passive member, up for retirement as soon as first
conditions for early retirement are met. Up to 2030 currently enacted increase in age and service period for women is taken.
In 2031-2038 it is assumed that minimum early retirement age increases from 60 to 62, by 3 months each year, while
minimum service period remains at 35 years.
Source: WB Team.


       7.2.3 Increase in statutory retirement age

Our reform proposal to increase statutory retirement age by 1 month each year is a
powerful tool for stabilization of the replacement rates. If the increase of statutory
retirement age is implemented as early as possible it can lead to an increase in replacement

                                                           113
rates for combined pensions (Figure 7.3 a). The increase is possible if annuity factors remain
the same as they are today. In our policy discussion earlier in the text we presented arguments
about why these factors can stay stable over the medium-term despite rising life expectancy.
But, even if annuity factors are adjusted in line with the increases in the statutory retirement
age, the replacement rates for combined pensions will not decline, they are more likely to
hover around the current values.

If the process of linking the statutory retirement age with life expectancy starts in 2039 as
proposed, the stabilization of replacement rates for combined pension will happen on a
much lower level (Figure 7.3 b). The postposed reform is proposed to first finalize the
equalization of retirement conditions for men and women and then to complete the process
of rising minimum condition for early retirement for two years.

Figure 7.3 Gross TRR assuming rising statutory retirement age by one month each year
                                        a) Reform starts in 2025
41.0%

40.0%

39.0%

38.0%

37.0%

36.0%

35.0%

34.0%

33.0%

32.0%
        202420262028203020322034203620382040204220442046204820502052205420562058206020622064
                         Full PAYG pension - pre-reform
                         Full PAYG pension - increase in statutory retirement age
                         Combined pension - pre-reform
                         Combined pension - increase in statutory retirement age
                         Combined pension - increase in retirement age & revised annuity rates




                                                114
                                                   b) Reform starts in 2039
 40.0%

 39.0%

 38.0%

 37.0%

 36.0%

 35.0%

 34.0%

 33.0%

 32.0%


                                Full PAYG pension - pre-reform
                                Full PAYG pension - increase in statutory retirement age
                                Combined pension - pre-reform
                                Combined pension - increase in statutory retirement age
                                Combined pension - increase in retirement age & revised annuity rates
Note: The result refers to men, average wage earner, flat wage profile, passive member, retirement at statutory age, service
period increases in line with the increase in statutory retirement age, starting from 40 years. Baseline assumption (pre- and
post-reform) is that annuity factors applied by MODs are not changing. In scenario with revised annuity rates, we assume
that annuities are adjusted continuously starting from the first year of the reform and in line with the increase in the statutory
retirement age, meaning that, other things being equal, retirement at statutory age will face the same annuity factors over
the entire projection horizon.
Source: WB Team.


       7.2.4 Increase in pension supplement and minimum pension


Proposed increase in the pension supplement from 27% to 30% and in minimum pension
point value from 103% to 106% of the general pension point value will result in one-off
increase in old-age pension benefits and replacement rates but will not influence longer
term trend. The increase in pension benefits will be similar for minimum pension beneficiaries
and all others whose pensions will increase due to pension supplement. Figures 7.4 a and
Figure 7.4 b show that on examples of two M65/40 careers, one with individual-to-average
wage ratio of 50% and the other with the ratio of 100%.
Increase in the pension supplement and minimum pension will also have an impact on the
P2/P1 ratio because P1 pensions will increase by more than P2 pensions in the initial period.
As trends are not changed by the proposed measure, P2 will only need more time to reach
P1. In 2024 P2 is below P1 for M65/40 (average wage earner) career, and it will take 3 years
to reach P1 without the reform and 5 years with reform (Figure 7.4 b and Figure 7.5).




                                                              115
Figure 7.4 Gross TRR after increase of the pension supplement to 30% and the minimum pension by 3%
(M65/40, flat wage profile)
                                    a) Individual-to-average wage ratio of 50%

66.0%

64.0%

62.0%

60.0%

58.0%

56.0%

54.0%

52.0%

50.0%
        202420262028203020322034203620382040204220442046204820502052205420562058206020622064

                   Full PAYG pension - pre-reform                 Full PAYG pension - post-reform
                   Combined pension - pre-reform                  Combined pension -post-reform


                                   b) Individual-to-average wage ratio of 100%

40.0%

39.0%

38.0%

37.0%

36.0%

35.0%

34.0%

33.0%

32.0%
        202420262028203020322034203620382040204220442046204820502052205420562058206020622064
                   Full PAYG pension - pre-reform                 Full PAYG pension - post-reform
                   Combined pension - pre-reform                  Combined pension -post-reform
Source: WB Team.



        7.2.5 Pension-type-choice-neutral lump-sum social transfers paid as pension benefits

The proposed alternative calculation of social transfers aims to enable policy changes in
social elements of first pillar pensions benefits without having the impact on the choice

                                                     116
between P1 and P2. 124 For example, if there is a decision to increase the amount of minimum
pension, then the result should not have an impact on the P1-P2 choice.

The proposed alternative calculation will result in higher basic PAYG pensions, hence also
higher combined pension. The impact depends on the coverage of social transfers that are
included in the measure. In the narrow coverage, including the added service period for
children and minimum pension, the increase in P2 is limited. Figure 7.6 shows the resulting
replacement rates before and after the reform for women with two children, retired at the
age of 60 with 35 years of service. The additional pension benefit from the first pillar for
children is given in the same amount for the full PAYG pension and basic PAYG pension, which
resulted in a higher PAYG pension and therefore higher combined pension. The impact is
relatively small. However, any possible future policy change in added service period will result
in equal change in P1 and P2.

The broad coverage of social transfers selected for alternative calculation of basic PAYG
pension will result in significant increase of the combined pension (Figure 7.5). The 27%
pension supplement is covered by alternative calculation. The advantage of the proposal is
that any future change in the amount or percentage of the supplement (for example increase
from 27% to 30%) will result in equal increase in P1 and P2 and therefore will not influence
the choice of pension type.

Figure 7.5 Gross TRR for alternative calculation of social transfers (W 60/35, two children)
  33.0%


  31.0%


  29.0%


  27.0%


  25.0%


           Full PAYG pension - pre-reform
           Combined pension (basic PAYG + 2nd pillar annuity) - pre-reform
           Combined pension (NARROW alternative model for minimum pension and added service for children)
           Combined pension (BROAD alternative model for pension supplement, minimum pension and added service for children)

Note: The results refer to the average wage earner, flat wage profile in the career, and passive pension fund member.
Source: WB Team.


The proposal of choice-neutral lump-sum social transfers paid as pension benefits will lead
to significantly higher combined pension for low-wage earners, however, the proposed
measure requires fine tuning with respect to coverage in order to achieve desirable


124
      See more on the proposal in Section 6.3 and Section 7.2.4

                                                                 117
outcomes. In 2024 the narrow coverage of the measure is not likely to change the choice
between P1 and P2 a lot (Figure 7.6). In our case of a woman 60/35, with two children, the
average wage earner, flat wage profile in the career, and passive pension fund member, P1
will be higher with and without the reform. Broad coverage of the reform measure that
includes the pension supplement is likely to increase the combined pensions above full PAYG
pension for all but a group of pensioners with high wage history. By 2034, the proposed
measure in both of its forms (narrow and broad) will result in P2 higher than P1 for all. In that
case, minimum pension beneficiaries will also be more likely to choose P2.

Figure 7.6 Combined-to-full PAYG pension ratio (P2/P1) for alternative calculation of social transfers,
2024 and 2034

 1.080

 1.060

 1.040

 1.020

 1.000

 0.980

 0.960

 0.940

 0.920
           80%




          210%
           50%
           60%
           70%

           90%
          100%
          110%
          120%
          130%
          140%
          150%
          160%
          170%
          180%
          190%
          200%

          220%
          230%
          240%
          250%
          260%
          270%
          280%
          290%
          300%
          310%
          320%
          330%
                                   Career-wide individual-to-average wage ratio
                                      Baseline 2024
                                      Baseline 2034
                                      NARROW alternative model for social transfers -2024
                                      NARROW alternative model for social transfers -2034
                                      BROAD alternative model for social transfers -2024
                                      BROAD alternative model for social transfers -2034

Note: The results refer to women 60/35, two children, the average wage earner, flat wage profile in the career, and passive
pension fund member.
Source: WB Team.



       7.2.6 Increase in the second pillar contribution rate


Pension adequacy of P2 can increase by application of higher second pillar contribution
rate. There are several ways to implement a higher contribution rate. First, continuing with
gradual substitution of first with second pillar rate, e.g., 13%+7% and decreasing the basic
pension factor in the calculation of the basic PAYG pension proportionally. Second, substitute
the contribution rates in the same manner without changing the multipillar formula and
keeping the basic pension factor (at 3/4) unchanged. Third, implementing the increase of the
second pillar contribution rate without changing contributions in the first pillar and without
changing basic pension factor. Fourth, introducing a voluntary autoenrollment model that

                                                           118
may or may not be connected to the second pillar. We have tested by TRR calculations the
impact of the first and the second option for the distribution of the contribution rates as
13%+7%.
Simulations for old age pension with 40 years of service show a mild long-term impact of
additional 2pp of contribution for the second pillar on the multipillar pension if basic
pension factor is corrected to reflect new ratio of contributions. We have simulated the
impact on replacement rate on full PAYG and combined pension in the case of switching to
13%-7% contribution rates for the first and the second pillar with two options considered,
with and without changes in the basic pension factor. From pension adequacy perspective,
what one gets in the form of a higher second pillar pension is partly lost due to lower basic
PAYG pension: after 40 years of implementation (2064) and correction of the basic pension
factor to 13/20, gross replacement rate for average wage earner is expected to reach 38.2%
compared to 36.3% in a no-additional contribution scenario (Table 7.1).125
Alternative solution without correction of contribution ratio (remaining at 15/20) would
lead to significantly higher pension adequacy. At the same time, it would worsen fiscal
sustainability. Given that the value of the basic pension factor is a political decision, an option
with unchanged basic pension factor seems more feasible than the one that brought higher
pension adequacy.126
Table 7.1 Gross replacement rates in case of increasing contribution rate for the second pillar to 7% and
decreasing contribution rate for the first pillar to 13% since January 1, 2025 (old-age pension with 40
years of service for average wage earner, men)

                                                2024              2026             2044              2064
 Full PAYG pension                               38.4%             38.1%            35.2%             33.2%
 Combined pension – baseline (no-
 reform)                                         38.1%             37.9%            37.4%             36.3%

 Combined pensions – reform with
 corrected basic pension factor                  38.1%             37.9%            37.7%             37.1%

      Corrected basic pension factor            0.7500            0.7457            0.7025           0.6525
 Combined pensions – reform with
 unchanged basic pension factor                  38.1%             38.0%            39.4%             40.2%

Source: WB Team.




125
    It is assumed that contribution rate for the second pillar increases on January 1, 2025, while other pension
system parameters remain as on January 1, 2024.
126
    The current basic pension factor of 0.75 does not reflect well the structure of financing the PAYG scheme. A
deeper insight into revenues and expenditures of the scheme is given in Annex 2 with the conclusion that a fair
basic pension factor is more likely to be around 0.80.

                                                       119
8. Baseline (Status Quo) Projection of Croatian Pension System Using
   PROMIS MODEL

A status-quo (baseline) simulation, assuming no changes in current parameters and
legislation, was prepared with World Bank’s Pension Reform Options Microsimulation
Toolkit (PROMIS) model adjusted to fit Croatia’s pension system parameters.127 The PROMIS
model incorporates Croatia’s actual data on contributors and beneficiaries in 2020, most
recent demographic data (EUROSTAT population projection based on 2021 Census), and
macroeconomic assumptions for the period until 2065, which is the simulation horizon for
the baseline and all simulations.
Croatian population, already among the oldest in South and Eastern Europe, is projected to
continue aging rapidly. Croatian population is expected to decline by 18% by 2060s with the
share of men converging to the share of women by late 2040s and slightly overrunning it by
the end of projection period (Figure 8.1). The population projection in the PROMIS model
assumes no change in net migration. Baseline projection suggests that a consequence of these
trends would be worsening of the old-age dependency ratio (ratio of population 65+ to
working age population) from 23% in 2023 to 32.7% in 2065 (Figure 8.2).


Figure 8.1 Croatia – Population Projection (2023-2065)




Sources: EU Population Projections; Croatia Population Census 2021; World Bank PROMIS model for Croatia



127
   PROMIS is a World Bank’s simulation tool used to analyze the pension syste m features and outlooks based
on actual individual data in the analyzed country. The microsimulation model enables the testing of effects of
policy scenarios on different groups and segments of contributors and beneficiaries instead of on average ones,
commonly addressed by cohort level models. The description and features of Croatia PROMIS model are
presented in Annex 5.

                                                        120
Figure 8.2 Croatia – Demographic Structure Projection (2023-2065)




Source: World Bank PROMIS model for Croatia


The macroeconomic assumptions needed for projections with the model include real wage,
real GDP growth, consumer price index (CPI), and real rates of return on assets accumulated
in three lifecycle portfolios in the second pillar (A, B and C). The assumptions include the
actual data in 2021 and 2022 and the estimated levels for 2023. Real GDP, after the post-
pandemic recovery hike in 2021 and 2022 and the moderate slowdown in 2023 and 2024, is
expected to grow by 2.5% in 2025 and gradually decline to its long-run real growth rate of 1%.
After the slump in 2022, real wage growth is expected to gradually rise from the current 1.5%
to 2.1% per year in 2027, slightly above the productivity growth in that period, and then
steadily fall to 1.5% per year by the end of the projection period. Similarly, the inflation rate,
after the 2022-23 high inflation episode, is assumed to converge towards its long-term level
of 2% (Figure 8.3). In combination with the assumptions for the rates of return on second
pillar portfolios, presented in Section 3.6, these assumptions may be characterized as
moderate.

The demand for labor is expected to remain strong in the short run and continue to boost
employment and consequently the number of contributors. With the assumption of a
modest rise in productivity, the economic growth would be generated by a fast growth in
labor participation from 63% to 68% by 2025, which would then continue to rise more
gradually towards 75% until the end of the projection period (Figure 8.4), compensating for
the negative effect of declining working age population on the pension system, as elaborated
in section 3.4.4.




                                               121
Figure 8.3 Macroeconomic assumptions in Croatia PROMIS model 2023-2065




Source: World Bank PROMIS model for Croatia


Figure 8.4 Share of Contributors in Working Age Population - PROMIS model 2023-2065




Source: World Bank PROMIS model for Croatia



Croatia PROMIS model incorporates the same long run macroeconomic assumptions as the
Croatian model in the Aging Report 2024. Comparison of the two models is summarized in
Table 8.1. The only notable difference is the higher GDP growth rate and negative real wage
growth in 2022 given that PROMIS model incorporates actual macroeconomic numbers in

                                              122
2021-2022 period. The difference would result in different initial year indicator values (such
as pension expenditure to GDP ratio) but would not significantly affect the model results and
policy conclusions.
Table 8.1 Comparison of macroeconomic assumptions PROMIS- AR24
                  2022               2030               2040             2050     2060
                                      Real GDP, rate of change (%)
 PROMIS               6.2               1.5                 1.7             1.4     1.0
 AR24                 3.2               1.5                 1.7             1.4     1.1
                              Real average wage growth, rate of change (%)
 PROMIS               -2.2              2.1                 2.2             2.1     1.7
 AR24                 1.8               2.1                 2.2             2.1     1.7
                                          Inflation rate, % yoy
 PROMIS               10.8              2.0                 2.0             2.0     2.0
 AR24                 10.7              2.0                 2.0             2.0     2.0
Source: World Bank PROMIS model for Croatia and AR24


In the absence of further reforms, Croatian pension system will face a stubborn long-term
worsening of its support ratio (ratio of contributors to beneficiaries, SR). As the result of
demographic and labor market assumptions, after the current vibrant employment period
ends in a few years (described in section 3.4.4), the SR is expected to start contracting in
parallel with the population decline. On the other hand, the number of beneficiaries would
remain stable over the next decade and start rising in 2040s as a consequence of unchanged
retirement age and slightly larger contributors’ cohorts retiring. The resulting SR deteriorates
until mid-2050s (Figure 8.5). Erosion of SR, although not as severe as in some other countries,
exerts pressure on pension system’s capacity to maintain the adequacy level with current
financing patterns and adds to the required strength of the needed compensatory policy
measures.
Figure 8.5 Croatia – Pension Support Ratio, Baseline simulation (2023-2065)




Source: World Bank PROMIS model for Croatia

                                                       123
The share of new pensioners with second pillar account and the option to choose the type
of pension at retirement will rapidly increase in the medium run. As mentioned earlier, a
specific feature of the Croatian pension system is the possibility of individual choice between
two pensions at retirement – 100% PAYG pension (P1) with the handover of the second pillar
savings to the state and a combined pension (P2) consisting of basic PAYG pension plus second
pillar lifetime annuity with an option to withdraw a lump sum up to 20%.128 At retirement the
individuals make a choice based on the assessed first nominal pensions for both options, with
and without the withdrawal of the 20% lumpsum.129 In the baseline simulation the choice is
modelled on the basis of the higher of the two initial nominal pensions without the 20% lump
sum withdrawal.130 Only the contributors with second pillar account have the choice between
two pensions at retirement. Generations currently retiring are still largely those without the
second pillar account. Their share in the initial simulation years is higher than the share of
individuals with second pillar (Figure 8.6). In 2023 it is estimated to stand at 50.5% for retirees
from active and 80% for retirees from dormant status.131 Currently retiring cohorts with
second pillar savings comprise the voluntary second pillar participants (those that were older
than 40 in 2002) and early retirees who were in the second pillar mandatorily since 2002. In
2027, when the first mandatory second pillar cohort (age 40 in 2002) retires, the share of new
old age pensioners without second pillar will plummet. For disability pensioners, the share of
new beneficiaries with second pillar is higher since they emerge from all cohorts with average
age lower than the old age pensioners.
Figure 8.6 Share of new old-age and disability pensioners without second pillar in total
                  Actives                                              Dormants




Source: World Bank PROMIS model for Croatia




128
    The lump withdrawal of 20% of assets accumulated at the individual account has been effective since January
2024. Until January 2024 the lump sum withdrawal was capped at 15%.
129
    The initial pension assessment is prepared by REGOS and HZMO and sent to future beneficiaries in the process
of pension determination.
130
    As elaborated earlier in the report, the option of a 20% lump sum withdrawal does not appear to be a
significant push for the decision to take the combined pension if initially lower than PAYG only.
131
    The dormant status (individuals without active contribution in the last 12 months) is explained in Annex 4.

                                                     124
Replacement rates will improve in the short run, stabilize in the medium run, but erode in
the long run with replacement rates for combined pensions increasingly exceeding PAYG-
only. Figure 8.7 presents baseline gross average replacement rates for three groups of old
age pensioners: i) those who chose combined pension P2 based on the first pension amount,
ii) those who chose PAYG pension P1 and iii) new beneficiaries without second pillar savings
at retirement.132 The projected replacement rates are initially dominated by strongly growing
length of service at retirement (Figure 8.8) resulting from demographic and labor market
assumptions. The growth of average replacement rate for pensioners without second pillar
until 2030 picks up the higher length of service and higher average point value features of
these retiring cohorts. Afterwards, their replacement rate gradually declines for the shrinking
number of PAYG-only individuals.133 Replacement rates for retirees who choose between
PAYG-only and combined pension show stable trends in the 2025-2035 period after the initial
hike. The PAYG-only replacement rate is outgrown by the combined pension replacement rate
throughout the projection period. The choice of a combined pension has been consumed by
the higher-income earners, the PAYG-only largely by lower-income earners.134 However,
after 2035 the old-age average pension replacement rate gradually declines for both groups
in the longer run (Figure 8.7). Female replacement rates converge to men in the longer run
due to female average length of service exceeding that of men by around one year (Figure
8.9). Longer female length of service adds more points and contributes to closing the impact
of gender wage gap on pensions.


Figure 8.7 Baseline replacement rates for new old age pensioners with and without benefit choice




Source: World Bank PROMIS model for Croatia



132
    Replacement rate is presented in terms of average gross national wage here, not in terms of individual gross
wage. The pensions represented in the average gross replacement rate incorporate all pensions after the
individual choice at retirement. The average replacement rate is calculated by dividing their average pension
resulting from the simulation by national gross average wage.
133
    Such individuals would, for example, be the 40+ year olds that started with first employment after 2002.
134
    In both cases, however, the share of minimum pensioners in total remains relatively stable around 50%.

                                                     125
Figure 8.8 Length of service of new old age pensioners by gender




Source: World Bank PROMIS model for Croatia


Figure 8.9 Baseline replacement rates for new old age pensioners who chose PAYG-only pensions and
combined pensions by gender




Source: World Bank PROMIS model for Croatia


The main factor affecting the steady decline of new replacement rates is the valorization
formula applying 70% of wage growth. If real wages grow, the real pension grows (30%)
slower, and the replacement rate deteriorates with other parameters equal. However, in case
of real wage fall, the formula would include only 30% of wage fall and 70% of inflation rate,
benefiting the new pensioners through higher replacement rates, as it happened in 2022 and
2023. The decline in replacement rate in our model is probably overrated due to the

                                                 126
assumption of continuous real wage growth. In actual life situations where episodes of real
wage growth are followed by episodes of real wage decline, the replacement rates will note
decline as much as projected. With “rotational” valorization and indexation formula, the
replacement rates erode slower than without rotation, and slower than with valorization
formulas used in the past, including the “Swiss formula” (average of wage and price
inflations).

Projected replacement rates for disability and survivors’ pensioners in the baseline remain
low and decline at a similar pace as other replacement rates. Survivors’ and disability
pensions are modelled as PAYG-only, based on the negligible share (less than 0.5 percent) of
basic pensions among new retirees in these two retirement types in the base year 2020. The
model assigns a disability status (occurrence of disability) to new pensioners based on
probabilities established in the base year, while the number of survivors is determined from
the number of deceased members (including contributors, old age pensioners, and disabled)
and using base year data on the number of dependents per deceased. It then calculates the
disability pension according to individual wage history, added service period and other
parameters of a contributor in line with the disability pension formula (Section 2). For
survivors, the model calculates the pension as 70% of the average new old age pension benefit
(based on summary statistics derived from the individual records of new pensioners in
2020).135 The model does not incorporate the survivor’s option to receive 27% of the
deceased spouse’s or partner’s pension in addition to survivor’s own pension if the eligibility
criteria are met.136 The gross disability and survivors’ replacement rates decline from the level
of around 25% by 5 percentage points in the long run and remain low mainly due to the
spectrum of disability pensions and the pension formula which imputes the pension service
generating relatively low assumed service period (Figure 8.10).137 On the other hand, the age
criteria for survivors’ pension are more generous than in other EU countries as well as the
mentioned “survivor’s pension inheritance”, generating higher PAYG costs, poorer incentives
for activating younger spouses, and higher poverty risk for the disabled.




135
    At the individual level, the pension factor that eq uals to 77% of the deceased member’s pension for a single
survivor has been in force since January 2023. In the case of multiple survivors, the survivor benefit formula
factors in higher pension factors, which are accounted for by using the average new survivor replacement rate
equal to 70% of the average old age pension.
136
    The model, however, incorporates the aggregate fiscal effect of this option. Conditions for eligibility to 27%
of spouse’s pension (colloquially called the “survivor’s pension inheritance”) depend on the couple’s combined
income and are described in Section 2. Upon establishing the eligibility, the survivor’s pension is not reduced
which results in survivors with spouses qualifying for higher pensions than the single pensioners. Apart from
differentiating between individuals based on the marital status, non-reduction of the own pension favors the
PAYG-only pension. Namely, in the second pillar, if spouses wish to insure their pension for case of spouses'
death, they must do so through the joint-and-survivor pension which reduces initial own pension by an
actuarially fair proportion (currently it is reduced by about 13% in case of selected survivor's replacement of
60% of the deceased's pension).
137
    The low assumed service period holds for full work incapacity due to general health conditions. On the
contrary, for disability caused by work injury of professional disease it assumes a service period of at least 40
years.

                                                      127
Figure 8.10 Average baseline replacement rate for old-age, disability and survivor pension




Source: World Bank PROMIS model for Croatia


The choice of the combined pension over the PAYG-only in the baseline rises over time. The
share of old age retirees with the option of choosing the combined pension declined from
29% in 2021 to 15% in 2023 as elaborated in the first part of the Report. In Q1 2024 it reached
21%. The share of old-age pensioners choosing the combined pension in 2024-2033 period is
targeted to gradually rise from the Q1 2024 level to 30%.138 In the remaining period, the
model projects that share of those choosing the combined pension would steadily increase
to 70% in line with findings based on individual model projections (Figure 8.11).139 New
pensioners currently choosing the combined pension own an over-proportional share of
second pillar assets as presented previously in the Report. However, the baseline simulation
shows that the choice converges closer to average as the system matures. In time, the
Croatian second pillar would be chosen by a wider range of income earners, not only the
wealthiest ones.




138
    For the 2024-2033 period, the PROMIS model initially generated the shares of old-age pensioners choosing
combined pension that were significantly different from the actuals realized in 2020-2024 (2024 data are
preliminary). The main reason for these differences was the lack of reliable information on wage profiles of
retirees with second pillar accounts between 2002 and the year of retirement. Also, pensioners often chose
between the combined and PAYG-only pensions based not on calculated pension amounts but because they
preferred the option with a lump sum payout. Thus, for the mentioned period, we imputed the second pillar
balances as of initial year of the projection (i.e. 2020) using load factors that imply the wage profile which is
rising faster than nominal wage growth in the economy, and relatively higher earnings of those who chose
combined pensions correspondingly.
139
    The share comprises old-age pensioners retiring at standard retirement age, early old-age pensioners and
long-service old-age pensioners choosing the combined pension.

                                                      128
Figure 8.11 Share of new old-age pensioners choosing combined pension




Source: World Bank PROMIS model for Croatia


Although the combined pension in 2023-2033 period is programmed to be chosen more
frequently by standard old age pensioners, in the long run the choice for other categories
converges to standard. The share of choosing the combined pension differs across the old
age pensioner groups – standard age, early and long-service retirees. Standard active old age
pensioners, retiring at statutory retirement age are initially likely to choose the combined
pension (Figure 8.12).140 For early retirees from actives the combined choice share explodes
in 2032 and 2033 and reaches the one for regular pensioners. According to the model, it
results due to expanding combined P2 pension stemming from longer accumulation periods
superseding the currently favorable PAYG pension decrement of 0.2% per month of early
retirement.141 For active long service pensioners, the model predicts fewer combined
pension choices until 2035. For this category, the PAYG benefit is determined without
decrement regardless of the age at retirement. On the other side, that benefit cannot be
consumed for free in the second pillar. The second pillar accumulation is spread over a longer
retirement period, yielding lower annuities and making the combined pension less attractive.
For all categories the choice of combined pensions strengthens over time to 60-70% levels in
the long run.




140
    The shares of those choosing combined pension is presented for active regular, long-service and early old age
retirees. The share for all includes the dormants who choose the PAYG-only at a larger scale.
141
    Current PAYG early retirement decrement is administratively set below its actuarially neutral level. On the
other hand, second pillar annuity factors depend on demographic factors, assumed rates of return in the payout
stage and the administrative fees. Current early retirement decrement in Croatian annuity factors (HRMOD)
stands around 3.5% per year of early retirement.

                                                      129
Figure 8.12 Share of old age pension categories retiring at statutory retirement age choosing combined
pension




Source: World Bank PROMIS model for Croatia


Both the stock of second pillar members and their share in all pensioners increases over
time. As a result of growing share of new retirees choosing the combined pension, the
baseline stock of combined pensioners in total number of pension recipients (old age,
disability, and survivors) gradually grows from 1% in 2023 to 38% by the end of projection
period (Figure 8.13). In the initial pension reform design from 1998, the participation in the
(mandatory) second pillar by 2065 was programmed to reach around 80-90% of all pensioners
(100% of eligible old age pensioners and a share of disability and survivors’ pensioners).
According to the baseline results, the choice between two pensions introduced in 2019 in the
long run results instead in approximately the equal split - 50% of combined old age pensioners
and 50% of those old age pensioners that decide to take full PAYG pension. While the
opportunity to choose the higher pension undoubtedly raises pension adequacy of those who
switch back to PAYG-only, it comes with additional fiscal cost for those that have to finance it
in the future, i.e., current and future generations of contributors and taxpayers as shown in
the next sections.




                                                 130
Figure 8.13 Combined pension beneficiaries – share in total pensioners 2023- 2065




Source: World Bank PROMIS model for Croatia


According to baseline projection, the adequacy of all pensions in payment would worsen
over time. Declining replacement rates for new pensioners shown above and rotational
indexation pattern in the real wage steady-state drive the pension benefit ratios142 down.
Baseline simulation shows that the benefit ratio of the PAYG-only-choosing pensioners
declines at similar pace as the benefit ratio of the combined pensioners (Figure 8.14). Due to
identical valorization and indexation patterns, the resulting adequacy of PAYG-only pensions
in payment, measured by benefit ratio, declines at the same pace as the adequacy of new
PAYG-only pensions, measured by the replacement rate (Figures 8.7 and 8.9 above). For
combined pensions, the basic pension segment of the combined pension follows the same
trend as the PAYG-only pension. Although the second pillar pension of current combined
pensioners erodes faster than the PAYG-only due its 100% price indexation, average second
pillar annuity benefit ratio gradually increases with the inflow of new and higher combined
pensions. Overall, pension adequacy of both PAYG-only and combined pensions in payment
erodes in the medium and long run and calls for immediate policy measures to gradually
improve the outcome over time.




142
   Pension benefit ratio is defined as a share of average pension of all pensioners in average wage in the
economy. In this report we use the gross benefit ratio.

                                                   131
Figure 8.14 Baseline benefit ratio of combined pensioners vs. PAYG-only pensioners




Source: World Bank PROMIS model for Croatia



In the absence of further reforms, Croatian multipillar pension system would continue to
require significant public pension financing. Despite low adequacy and eroding replacement
rates, the projected funds required to finance public pension expenditures do not decline.
The PAYG pension expenditures, consisting of full PAYG pension of those without second
pillar, those that chose to return to the PAYG pillar, and the basic pension for the combined
pensioners, absorb between 9% and 9.5% of GDP throughout the projection period (Figure
8.15 a).
The accompanying PAYG current deficit143 is expected to remain stubbornly stable between
4% and 5% of GDP until narrowing to 3% of GDP only at the end of the projection period . If
the extraordinary revenue from asset transfer from second to PAYG pillar for pensioners who
chose the PAYG pension is incorporated in revenues, the deficit is lower by around 1
percentage point of GDP, however, still at fiscally concerning levels (Figure 8.15 b).144




143
    Current deficit is calculated as a difference between contribution revenues and PAYG expenditures. It
excludes the mandatory transfers from the central budget (legislated supplements, benefits for veterans and
other special categories, reimbursement for contributions diverted to the second pillar) and residual deficit
coverage.
144
    Individual choice of PAYG-only pension provides the government with short-term and one-off revenue but
generates additional long-term PAYG costs. It is controversial to count the asset transfer for those that switch
back to PAYG as current revenue. The transfer is effectively the refund for a lifetime full PAYG pension above
the basic pension that should have been financed by a full 20% contribution rate. The difference has, however,
already been compensated by the central budget, so the transfer should be accounted as extraordinary revenue
of the central budget and not the pension system.

                                                     132
Figure 8.15 Baseline PAYG expenditures and pension system deficit
                           (a)                                   (b)




Source: World Bank PROMIS model for Croatia



Figure 8.16 Second pillar payouts will remain a small share of future pension expenditures




Source: World Bank PROMIS model for Croatia


In the baseline - status quo - scenario the second pillar develops slowly, both in terms of
participation at retirement as well as the share of funded pensions in total pension benefit.
The introduction of the choice between two pensions slowed down the pace of second pillar
growth and reverted a large segment of pension liabilities to PAYG. Figure 8.16 gives a preview
comparison between baseline pension expenditure path and the border options of “all
receiving PAYG-only pension” and initially (in 2002) legislated “all receiving a combined
pension”.145 Instead of PAYG expenditures sliding down the gray-dotted line with a significant
share of pension covered by own savings from second pillar, Croatian pension system has
been standing closer to the other extreme of all in PAYG-only (blue-dotted) line, with a much

145
      These simulations are presented in more details later.

                                                        133
larger share paid for by the public PAYG system (redline) and a small part by the second pillar
(difference between red and yellow lines). Pension policies implemented in the previous
decade result in smaller and slower diversification of retirement incomes in Croatia than
initially planned by the pension reform of 1998.




  Without new policy measures the already low pension adequacy in Croatia will gradually
  erode despite expectedly rising length of service until 2031-32 and favorable second pillar
  returns. Public pension system will continue to absorb high proportion of GDP devoted to
  pension expenditures and generate stubbornly high pension deficit. Improvement of
  pension adequacy in the fiscally sustainable manner will require implementing additional
  and courageous set of policy measures.

In the following sections we consider and simulate policies aiming at improving pension
adequacy and strengthening Croatian pension system proposed in Section 7 and estimate
their short and long-run fiscal sustainability.




                                             134
9. Simulation of policy measures

The objective of simulating pension policies identified and recommended using individual
pension model is to assess their potential impact on the overall system pension adequacy,
aggregate fiscal sustainability and other relevant aspects of the pension system. Policy
measures identified in previous sections and recommended in Section 7, selected to be tested
by the Croatia PROMIS are listed in Table 9.1. It starts with individual policy simulations which
assess the strength and impact of a specific individual policy only. Integral policy packages
combine individual policies and explore their synergies in addition to individual effects.
Selection of individual policy simulations performed with Croatia PROMIS model and
elaborated in this Report has been limited to key pension policies.146 The analysis of policy
measures not simulated with PROMIS model rests on individual pension model with
recommendations described in Section 7.
Table 9.1 Individual Pension Policies and Policy Packages Considered and/or Simulated with Croatia
PROMIS Model
 Valorization and indexation patterns
     1. 85%W+15%P         New pension valorization and current pension indexation with 85%
                          wage growth and 15% consumer price index from 2025
 First pillar framework
     2. SUPPL30%+         PAYG pension supplement raised from 4%-27% to 7%-30% and
           MIN106%        minimum pension per year of service raised from 103% to 106% of
                          actual point value.

     3. EQCHOICE          PAYG supplement introduced to assure equality of pension choice
 Second pillar framework
     4. 2PILL7%           Contribution rate for the second pillar 7%, for PAYG 13%. Basic pension
                          factor remains at 0.75 in lieu of equal choice
 Later retirement and narrower early retirement window
     5. RETAGE SLOW Statutory retirement increases from 2039; early retirement in 2025;
                          long service age increase in 2025
     6. RETAGE FAST       Statutory retirement increases from 2025, early retirement faster from
                          2025, long service age increases faster from 2025
 Integral policy packages
     7. PACK SLOW                   1+2+3+5
     8. PACK FAST                   1+2+3+6
     9. PACK                        1+2+5
         SLOWNOEQ
     10. PACK                       1+2+6
         FASTNOEQ




146
  Selection was limited due to short Report delivery time and longer-than-expected process of Croatia PROMIS
model development.

                                                    135
9.1 Valorization and indexation policy

Croatian pension system features identical formulas for valorization and indexation of
pensions. Both are regulated in the Pension Insurance Law via adjustment of the Actual Point
Value (APV) through the rotational pattern. Such a pattern equalizes replacement rates with
benefit ratios for the same retirement conditions and assures same nominal pensions of new
and current retirees for the same service and average earnings. Public and political support
to maintain this nominal equality feature of the pension system has remained strong and
almost carved-in stone in the pension system parameters in Croatia. The argument that
delinking the valorization and indexation parts of the formula, with initial pensions valorized
with wages and pensions with CPI or below wages, improves the intergenerational equity has
been perpetually rejected in the last two decades by both the left and right political options
in the office. Based on the Client’s statement that such policy measure would not be
considered, it was agreed with the Ministry not to consider it in the simulations.
There have been strong political pressures to expand the rotational indexation pattern to
the higher of 85% wage or price rate of change plus 15% of the lower of the two, or even to
maximum - higher of 100% wage or price rate of change valorization and indexation.147
Accordingly, the Ministry has requested from the team to simulate and analyze the impact of
these policy measures on adequacy, fiscal sustainability, and pension pillar structure. In this
Report we simulate the 85%W+15%P indexation and valorization pattern and compare it with
the baseline. In this simulation the Actual Point Value is adjusted twice a year, starting from
2025, with the rate of change of 85% of nominal wages and 15% of the CPI. Valorization and
indexation are effectively channeling 85% of real wages into real pensions.
In all the simulations real wage is assumed to grow so the 85%W+15%P scenario effectively
shows only the real wage growth aspect of the impact of the rotational indexation formula.
In case of real wage fall (or a slower growth than of prices), the rotational formula switches
to 85%P+15%W and transforms only 15% of real wage fall into real pensions. Due to the
absence of that effect in the simulations and its potential positive impact on pension
adequacy but also higher fiscal cost, the Croatian model in the Ageing Report emulates it
through setting the current valorization and indexation pattern at 80%W+20%P (instead of
70%W:30%P). In our simulations we will not tweak the baseline or the 85%W+15%P to
capture the impact of the rotational indexation effect because tweaking both would not affect
their relative ratio and policy conclusions.
Higher wage weight in the valorization formula results in higher pension adequacy
measured by the replacement rate of both PAYG-only and combined pensioners than in the

147
   The higher of 100% nominal wage or price valorization and indexation pattern results in constant replacement
rate if the real wages grow (for the same individual characteristics) and growing replacement rates if the real
wages fall. Pensioners are always better off with such a “double lock” as commonly addressed. Expectedly, such
a pattern is fiscally challenging and requires support of other measures to remain fiscally sustainable. In the UK
the “triple lock” indexation adds a floor of minimum 2.5% indexation to the double-lock conditions. In case when
both nominal wages and inflation fall or rise between -2.5% and +2.5%, the indexation floor sets the indexation
rate at 2.5%. Tripple pension lock adds a strong dose of fiscal certainty and implicit fiscal liability to the national
finances.

                                                         136
baseline (Figures 9.1 a and b). With the 85%W+15%P valorization pattern, the replacement
rates erode slower than in the baseline. For the PAYG-only old age pensioners the
improvement in adequacy takes off in 2028 and gradually grows almost 5 pps above the
baseline. The combined replacement rate in 85%W+15%P starts raising above the baseline in
2041 and grows by almost 5 p.p. by 2065, maintaining it at a very stable level. PAYG-only old
age pension grows slightly faster and slightly stronger than the combined as the change in
valorization pattern would affect only the PAYG segment of pensions, which is larger in PAYG
only system. The average replacement rate (Figure 9.1 c) gives a full picture of the effect of
85%W+15%P on average replacement rates of all old age pensioners. The indexation measure
raises the replacement rates until 2035 with the help of strongly increasing service periods
(Figure 8.8) and stabilizes it afterwards.148
Figure 9.1 Simulation of 85%W+15%P valorization and indexation patterns – replacement rates
                             a                                                      b




                                                 c




                         Source: World Bank PROMIS model for Croatia



Benefit ratios for pensions in payment in 85%W+15%P simulation improve substantially.
Due to indexation pattern identical to valorization, the average old-age benefit ratios (share
of average pension in average wage) show a pattern similar to replacement rates (Figure 9.2).
Pension indexation affects only the PAYG segment of the benefit. A higher wage segment in
the indexation formula in 85%W+15%P pattern generates higher pensions in payment, thus


148
  Comparison of replacement rates by groups of P1 and P2 choosers is complicated and biased by the fact that
the samples compared are not the same. In comparison of other simulations, we will thus for presentational
purposes be using only the average replacement rate.

                                                     137
higher adequacy. The effect is substantially stronger with PAYG-only pension because it
affects full pension, not just its basic part as in the combined pension. The adequacy boost
starts early with PAYG-only pension, in late 2020s, a decade later with combined pensions. In
the long run, the adequacy of both PAYG-only and combined pensions improves by very high
5 percentage points. Same as with replacement rate, the full effect is seen in average benefit
ratios in this simulation showing a twice slower decline in benefit ratio than the baseline
(Figure 9.2 c).149
Figure 9.2 Simulation of 85%W+15%P valorization and indexation patterns – benefit ratios
                             a                                                      b




                                                     c




                   Source: World Bank PROMIS model for Croatia



Pension choice strongly varies with the impact of the indexation pattern on adequacy of P1
and P2. Alternative valorization and indexation patterns significantly affect the choice
between two pensions P1 and P2 at retirement. With higher wage weight in the valorization
and indexation formula, fewer new retirees choose the combined pension P2. In 85%W+15%P
PAYG-only pays more, adjusts faster than the combined, and motivates more individuals to
select PAYG-only. After 2030 the share of combined P2 choosers drops almost by one half
compared to baseline (Figure 9.3 a). In the following years the choice of P2 recovers, however,
without reaching the baseline levels. Consequently, with 85% wage weight instead of 70% the

149
   Same as with replacement rates, the comparison of benefit ratios for P1 and P2 groups selecting PAYG-only
or a combined pension is biased by the fact that the samples compared are not the same. In the remaining
simulations we will use only the average benefit ratios for comparison.

                                                     138
stock of combined old-age pensioners in the long run drops to 36% of all old-age pensioners,
substantially below the 50% reached in the baseline (Figure 9.3 b).


Figure 9.3 Simulation of 85%W+15%P valorization and indexation patterns – choice of P2, flow and stock
                              a                                           b




Source: World Bank PROMIS model for Croatia



Stronger valorization and indexation pattern comes at a high fiscal cost. More generous
wage indexation gradually generates higher expenditures and widens the pension deficit.
Initially, the additional fiscal cost is modest. 85%W+15%P indexation would require
additionally only 1% of GDP in total by 2032. In the rest of 2030s and in 2040s the difference
from the baseline accelerates. In early 2040s the 85%W+15%P indexation would exhaust
additional 0.5 pp of GDP per year in pension expenditures, by 2065 it raises to 1.5 pp of GDP
annually (Figure 9.4). Total additional fiscal requirement for this policy measure in the next
40 years of the projection period stands at 33.8% of GDP (see Section 9.5 with policy packages
and summary simulation comparison). While the pension adequacy improves, the PAYG
pension expenditures in this simulation reaches 11% of GDP by 2050s. The gross PAYG deficit
it generates expands above the baseline by the same amount (Figure 9.5 a) and rises above
the 5% mark in 2040. Due to more individuals choosing PAYG pension P1, the extraordinary
choice revenue from transferring their second pillar accumulations to PAYG covers the
additional expenditures by late 2030s, keeping the net deficit at the same level as the baseline
(Figure 9.5 b). However, from 2040s the additional current PAYG expenditures overrun, and
additional net deficit reaches 1.3 pp of GDP by 2065, leaving the choice revenue financing
only 0.2% of GDP of this additional expenditure.




                                                 139
Figure 9.4 Simulation of 85%W+15%P valorization and indexation patterns – PAYG expenditures




Source: World Bank PROMIS model for Croatia



Figure 9.5 Simulation of 85%W+15%P valorization and indexation patterns – PAYG gross and net deficit
                            a                                          b




Source: World Bank PROMIS model for Croatia




 Stronger pension valorization and indexation raises pension adequacy of new PAYG
 pensions as well as those in payment, but it also significantly boosts PAYG expenditures.
 Unless there are countermeasures to reduce the costs elsewhere or collect additional
 pension revenues, it fuels into the pension deficit strongly and stubbornly at the cost of
 future generations. Appropriate valorization and indexation pattern needs to be
 determined in sync with other policy measures and taking into account its short, medium
 and long-term impacts.




                                                140
9.2 First Pillar reform framework

      a. Three PAYG-oriented policies recommended in Section 7 and elaborated in the
         report are tested in two policy simulations using the PROMIS model: Raising the
         pension supplement 4%-27% to 7%-30% and increasing the minimum pension per one
         year of service from 103% to 106% of actual point value, (SUPPL30%+MIN106%). This
         simulation incorporates two complementary policies aiming to raise the adequacy of
         all current and new minimum pensions, PAYG pensions and basic pensions in
         proportion to pension level.
      b. Extracting and setting the PAYG pension social protection supplement, thus
         strengthening the equality of choice between PAYG-only and combined pension
         (EQCHOICE).
Both simulated policies policy scenarios tackle a certain group of PAYG pension recipients
raising their benefit level or expanding their eligibility criteria. SUPPL30%+MIN106% raise is
applied to all current and new beneficiaries that are eligible for the pension supplement and
the minimum pension. EQCHOICE introduces the pension choice-neutral supplement
determined in the same amount per beneficiary regardless of the selection of the PAYG or a
combined pension (Section 7).150 We use the PROMIS microsimulation model to estimate the
potential magnitude of these policies’ impact on key system parameters, primarily the fiscal
costs and deficit. We will compare and present their results simultaneously.
Both policies improve pension adequacy of new and current pensioners. Figure 9.6
summarizes the individual impact of these two pension policies on average replacement rates
and average benefit ratios. Both simulations show the policies resulting in higher adequacy of
new pensions (replacement rates) and all current pensions (benefit ratios). For new pensions
the EQCHOICE has a stronger impact. In the first decade, the adequacy of new pensions is 1
percentage point higher while with SUPPL30%+MIN106% it comes up by 0.4 pp per year (see
Section 9.5 comparing all policy simulations). The impact of EQCHOICE grows over time with
more individuals opting for P2. The impact of higher supplement and minimum pension
remains relatively constant over time, steadily improving the adequacy of every new
generation’s initial pensions. On the other hand, SUPPL30%+MIN106% has a tremendous
impact on adequacy of all current pensions in the first years of the simulation period. In 2025
alone the SUPPL30%+MIN106% raises the benefit ratio by 0.66 pp and then gradually
declines, handing over the longer run effects to the EQCHOICE policy.
Figure 9.6 Simulation of higher supplement, minimum pension and choice-neutral transfer – average
replacement rates and benefit ratios



150
    The formulation of the EQCHOICE policy in PROMIS model is made through a following procedure of
expanding the basic pension or minimum basic pension aiming to neutralize the bias in the pension choice:
  - If Pension Benefit (PB) by formula exceeds the Minimum Pension (MP) (i.e., if PB>=MP), then PB_Basic_New
  = PB /1.27 * 1.02 = PB * 0.80315 (instead of PB * 0.75) = PB_Basic * 1.070866. If Minimum exceeds Formula
  (i.e., if MP>PB), then PB_Basic_New = PB_Basic * 1.070866 + (MP-PB).


                                                   141
                              a                                              b




Source: World Bank PROMIS model for Croatia




EQCHOICE and SUPPL30%+MIN106% have diametrically different impact of the choice of
combined pension. Expectedly, the introduction of the choice-neutral transfer eliminates the
bias in the choice and makes the combined pension attractive to the vast majority of new
beneficiaries (Figure 9.7 a). Instead of creating the gap between two pensions at retirement,
the transfer allows all beneficiaries to reap the benefits of the combined pension, including
the minimum pension beneficiaries and women. SUPPL30%+MIN106%, on the other hand,
raises the PAYG segment of the pension and minimum pension and makes the PAYG only
pension more attractive than the combined. Figure 9.7 a shows that fewer new beneficiaries
choose combined pension in the first 15 years. Afterwards, the choice rate converges to the
baseline. The stock of combined pensioners develops in sync with the new pensioners’ choice.
With EQCHOICE the share of combined pensioners converges to 100%, with
SUPPL30%+MIN106% to the baseline share of around 50% of beneficiaries (Figure 9.7b).



Figure 9.7 Simulation of higher supplement, minimum pension and choice-neutral transfer – choice of
combined pension of new pensioners and all pensioners
                              a                                              b




Source: World Bank PROMIS model for Croatia

SUPPL30%+MIN106% generates higher and EQCHOICE lower expenditures than the
baseline. Fiscal outturns of these two policy simulations are also different.
SUPPL30%+MIN106% generates additional costs of 0.22% of GDP in 2025 and a cumulative

                                               142
additional expenditure of 2.23% of GDP by 2034 (Figure 9.8 a and Section 9.5). Accordingly, it
generates an additional deficit, a larger in gross and slightly lower in net terms due to more
individuals choosing PAYG pension and larger transfers of second pillar accumulations (Figure
9.8 b and 9.8 c). EQCHOICE yields fiscal savings, a cumulative of 0.84% of GDP by 2034. The
savings are more modest because of additional fiscal costs for the equal choice transfer
included in the basic pension. Nevertheless, the residual pension financed by second pillar
annuity rather than PAYG is still generating fiscal savings. However, the gross deficit declines
in comparison with the baseline, but the net deficit increases due to less inflow from
individual accounts who all end annuitized. This difference accounts for 0.3% of GDP in 2025
and reaches a level of 0.5% per year in the next two decades. (Figure 9.8 c). It should be
emphasized that the larger net deficit should not be observed as fiscal shortage of the first
pillar due to fewer opt-outs to PAYG pillar. Instead, it stops the practice of generating
additional revenue for the PAYG pillar by unfair pension choice conditions. The EQCHOICE
brings back order and balance among the pension pillars.
Figure 9.8 Simulation of higher supplement, minimum pension and choice-neutral transfer – PAYG
expenditures, gross and net deficit
                                            a




Source: World Bank PROMIS model for Croatia

                           b                                        c




Source: World Bank PROMIS model for Croatia



                                              143
  PAYG measures tackling pension adequacy immediately, in the form of supplement of
  30%, minimum pension of 106% of APV per year and neutral-choice supplement, improve
  pension adequacy with the opposite fiscal outcome. Higher supplement and minimum
  pension raise the expenditures and deficit, while the neutral-choice supplement curbs the
  gross deficit but expands the net one due to insured individuals’ choosing the combined
  pension instead of PAYG. These measures have strong effect on pension adequacy and
  should be included in the policy package along with balancing measures bringing the
  system to a steady state.



9.3 Second pillar reform framework




Simulation 2PILL7% illustrates the potential impact of increasing the contribution rate for
the second pillar from 5% to 7% with simultaneously decreasing PAYG rate from 15% to
13%. At the same time the basic pension is projected to stay at 75% of PAYG pension instead
of being reduced to 13/20 in proportion to rate reduction. The reason for departing from the
proportionality of basic pension to PAYG contribution rate in this individual policy scenario is
to compensate for the non-neutrality of the choice caused by the financing the PAYG pensions
from the budget in excess of pension contributions. Although such choice is likely to over-
compensate for the non-neutrality, as it is assessed to be around 7% (see Annex 2, where it is
estimated that the current basic pension factor should be 0.80 instead of 0.75), we wanted
to present as simple as possible policy proposals that show a possible impact and did not
precisely assess future non-neutrality that should be compensated for. Furthermore, based
on recent pension policy plans presented in the Program of the Government of the Republic
of Croatia 2024-2028,151 some of the measures likely to be implemented would incur further
disturbance of the choice neutrality principle (introduction of the annual supplement if
determined by applying basic pension factor, a longer insurance period added to mothers,
and a higher bonus for late retirement if set above actuarial neutrality). It seems reasonable
to have a policy proposal that anticipates and takes into account such possible developments.


151
   Among other things, the Program states that (i) by the end of the mandate, pensions will be increased by at
least 30%, with an average overall pension of at least EUR 750 at the end of the term; (ii) the model of adjusting
the level of pensions relative to the increase in wages and consumer prices will be modified by changing the
formula for adjusting pensions in the ratio of 85:15 to a more favorable rate, in order to ensure a higher ratio of
the average overall pension over the average wage; (iii) an annual supplement for pensioners will be introduced
by taking into account the balance between solidarity-based reallocation of funds and previous payments to the
pension system; (iv) voluntary longer stay in the world of work after the age of 65 will continue to be encouraged
through a redefined pension increase model for each month of later retirement (bonification); (v) students
working on students’ contracts will be able to recognize their work in pension insurance ; (vi) for each child born,
the mother will receive one year of pensionable service: both for lowering the age limit and for calculating the
pension benefit; (vii) the amendments to the Law will allow part-time work for retired craftsmen, while other
independent taxpayers will be able to pay contributions simultaneously while receiving their pension benefits in
order to enable as many pensioners as possible to earn additional income.

                                                       144
A higher second pillar contribution rate and basic pension determined at 75% of full PAYG
make the combined pension more attractive. The choice of the combined pension increases
quickly in comparison to baseline. The share of new pensioners choosing combined pension
reaches 50% in 2035, and in the long run it grows to 90% (Figure 9.9 a). As a consequence,
the share of combined pensioners in total grows faster and reaches 73% by the end of
projection period (Figure 9.9 b).


Figure 9.9 Simulation of higher second pillar contribution rate (7%) – choice of combined pension for
new pensioners and all pensioners
                              a                                               b




Source: World Bank PROMIS model for Croatia

2PILL7% improves adequacy in the longer run with lower share of pension expenditures in
GDP. The higher second pillar contribution rate with the current basic pension formula
increases replacement rates and benefit ratios in the medium and long run. The improvement
builds up gradually in parallel with second pillar accumulations (Figures 9.10 a and b). Higher
adequacy of new and current pensions emerges from more persons participating in the
second pillar at higher contribution rates and drawing a larger share of their pensions from
their second pillar savings rather than PAYG. As a consequence, the share of GDP required to
finance the PAYG pensions declines (Figure 9.11 a). The cumulative reduction in PAYG
expenditures in this simulation stands at 10% of GDP in the whole projection period (Table
9.3 in Section 9.5).
However, in the short and medium run larger second pillar contribution rate generates high
fiscal deficit. More contributions flowing into the second pillar brings less contributions into
the PAYG. Gross deficit with 2PILL7% grows to 5% of GDP in 2025 immediately with net deficit
expanding even more due to fewer beneficiaries choosing PAYG pension and returning their
savings to the state budget (Figure 9.11 b). In the longer run the transition cost fuels into
higher adequacy and phases out, which illustrates the expected outcome of the transition
cost in the substitutive second funded pillar systems. There are long-term gains in higher
adequacy and lower PAYG costs for future generations, but the current generation needs to
make a decision and bear the transition cost.




                                                145
Figure 9.10 Simulation of higher second pillar contribution rate (7%) – average replacement rates and
benefit ratios
                              a                                                  b




Source: World Bank PROMIS model for Croatia




Figure 9.11 Simulation of higher second pillar contribution rate (7%) – total PAYG expenditures and gross
deficit
                  a                                               b




Source: World Bank PROMIS model for Croatia




                                                  146
 Higher second pillar contribution rate, such as simulated 7% substituting the two
 percentage points of the PAYG rate, has long run gains but short run (transition) costs.
 From the aspect of fulfilling the social, fiscal and intergenerational objectives of the
 pension system, that cost is worth taking but it should be done when the social and fiscal
 conditions are fulfilled and in balance. Large PAYG deficit and most new pensioners
 choosing PAYG pension currently form the unfavorable setting for further second pillar
 rate substitution. Besides the fiscal and social sustainability arguments, the policymakers
 need to factor in political economy elements and assess the national priorities before
 deciding to raise the contribution rate for the second pillar.



9.4 Later retirement and shorter early retirement window

Pensionable service periods in Croatia are shorter than it most of EU countries. As
elaborated in previous sections, the effective service period at retirement in Croatia is among
the lowest in the EU. While the mandatory retirement age, 65 for both men and women (age
for women transitioning towards 65), is comparable to countries with similar demographics,
the early retirement opportunities are more abundant than elsewhere. As indicated in Section
3, the average exit age from labor market is below the EU average (Figure 3.18), and the
average service period at exit is among the lowest in the EU (Figure 3.20). Regular early
retirement 5 years prior to statutory, long service retirement, and extended service period
(“beneficirani staž”) effectively reducing retirement age and adding service for hazardous
occupations all contribute to shorter service periods and below-potential pension levels.
The policy of extension of retirement age and narrowing early retirement options is tested
through two simulations:

   •   RETAGE SLOW: setting the standard, early and long service parameters as
       recommended in Section 7:
          o Early retirement for men increases by 3 month per year from 2025-2032 to
             reach 62 years, narrowing the early retirement window to 3 years.
          o Early retirement for women continues raising by 3 months per year 2031-2038
             reaching 62 years.
          o Long service retirement age increases from 2025-2032 by 3 months per year
             from 60 to 62.
          o Service requirement for long service increases from 41 to 42 years in 2025.




                                             147
             o From 2039 standard retirement age, early retirement age and long service
               retirement age increase by 1 month per year (2/3 of life expectancy).152

      •   RETAGE FAST: the FAST retirement age policy assumes much faster retirement age
          increases:
              o Early retirement for men increases by 4 month per year in the period 2025-
                 2039 to reach 65 years. After 2039 the early retirement age for men raises by
                 2 months per year.
              o Early retirement for women raises by 7 months per year until 2030 to equalize
                 with early retirement for men at 62. Afterwards, it raises by 4 month per year,
                 same as for men.
              o Long service retirement age increases in the period 2025-2030 by 3 months
                 per year and continues to increase by 2 months afterwards.
              o Service requirement for long service increases from 41 to 42 years in 2025.
              o Standard retirement age for men increases by 3 months per year, by 6 months
                 per year for women in the period 2025-30 to reach 66,5 years for both. From
                 2031 the retirement age will be raised by 2 months per year.153 Retirement age
                 increases stop at the age of 72 in 2063.


The objective of this simulation is to identify potential implication on adequacy and fiscal
sustainability and its capacity to offset the fiscal costs of more expensive measures such as
stronger indexation of pensions. Figure 9.12 illustrates the tempo of retirement age increase
in SLOW policy scenario. The scenario adjusts the retirement ages slowly, by 2/3 of life
expectancy, starting from 2025, ignoring the previous period of retirement ages trailing
behind the life expectancy. In RETAGE SLOW the retirement age for men and women by 2065
reaches 67 years and 3 months, while the early retirement age gets to 64 years and 3 months.
Figure 9.13 shows the intensive pace of retirement age increase in RETAGE FAST policy
scenario which, in addition to retirement age increase in SLOW, gradually compensates for
retirement age not being raised in the period 2011-2022 when life expectancy increased by
two calendar years. In RETAGE FAST the retirement age for men and women by 2065 reaches
72 and the early retirement age hits 69 years. The estimated healthy life expectancy at 65 in
Croatia stood at 5.2 years in 2021, while in EU it was 9.9 years.154 This implies that an average
pensioner in Croatia spends only one third of life in retirement in good health. There are no

152
    Increase in retirement age by 1 month per year approximately stands for 2/3 of the increase in life expectancy.
According to EUROSTAT’s life expectancy projections for Croatia applied in the 2024 Ageing Report (and used
in PROMIS), between 2024 and 2065 the life expectancy at the age of 65 will increase by 60 months on average
for both genders. Over the forty-year period, 2025-2065, the increase is 60/40 = 1.5 months. The average life
expectancy at 65 is presented as (65 + average years of life expectancy at 65)
153
     This assumed increase in life expectancy in RETAGE FAST goes faster than life expectancy in 2025
compensating for the retirement age being behind life expectancy in the last 15 years. Compared to life
expectancy, the retirement age should have been raised by 12 months in the 2019-2024 period and another 12
months in 2010s. The stronger part of compensation takes place until 2030, and more modest afterwards. The
increase in retirement age of two months per year would exhaust entire life expectancy increase and
compensate for remaining retirement age non-adjustment.
154
    https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Healthy_life_years_statistics

                                                       148
projections of healthy life expectancy for the next 40 years, but it can be assumed that the
share of healthy life expectancy would not decline below the current level.155


Figure 9.12 Assumed increases in retirement age and early retirement age 2025-65 and increase in life
expectancy according to EU projections, RETAGE SLOW




Source: World Bank PROMIS model for Croatia.Note: average life expectancy at 65 is presented as (65 + average years of
life expectancy at 65)




Figure 9.13 Assumed increases in retirement age and early retirement age 2025-65, and increase in life
expectancy according to EU projections, RETAGE FAST




155
   There is high subjectivity accompanying the compilation of this indicator resting on a survey with self-
assessment of health condition. It is not recommended to use this indicator for determination of policy measures
without supplementary analysis of health status of population over 65.

                                                          149
Source: World Bank PROMIS model for Croatia.
Note: average life expectancy at 65 is presented as (65 + average years of life expectancy at 65)




Extending the mandatory retirement ages improves the support ratio. The support ratio
improves in both SLOW and FAST RETAGE simulations (Figure 9.14). Later retirement leads to
slower inflow of new pensioners and longer stay in the labor force. With RETAGE SLOW the
support ratio improves by about 10% in comparison to baseline, while RETAGE FAST almost
stabilizes the support ratio at current level of 1.3. No other pension policy measure has the
power to achieve such an improvement in the support ratio.


Figure 9.14 Retirement age increase, RETAGE SLOW and RETAGE FAST, support ratio




Source: World Bank PROMIS model for Croatia




                                                            150
Retirement age pushes up the effective service periods at retirement. The transfer rate from
retirement age increase into the service period is assumed to stand at 70%, following the rate
identified for women during the 2017-2022 period of retirement age increase (Section 4).
Figure 9.15 shows that the retirement age increase in RETAGE SLOW would result in about 2
more years of service, while the RETAGE FAST would raise by 6 years (up to 42 years) until
2065. Although that may look overly optimistic from the current perspective, the average
service years of 40 are already registered in some of the EU countries, as shown in Section 3.


Figure 9.15 Retirement age increase, RETAGE SLOW and RETAGE FAST, length of service at retirement




Source: World Bank PROMIS model for Croatia




Retirement age increases significantly improve the pension adequacy. Given the impact on
effective service periods at retirement, it is not surprising that higher retirement ages improve
the adequacy of new pensions via higher replacement rates and then gradually the benefit
ratios. While RETAGE SLOW increases the replacement rate after 2039 between 1 and 2
percentage points, the RETAGE FAST result is quite impressive (Figure 9.16 a). Not only the
fast retirement age increase boosts the replacement rates by 4 percentage points, but it also
keeps them stable regardless of below-wage indexation. Positive effect on replacement rate
gradually flows into the benefit ratios, almost annulling its deterioration due to below-wage
indexation (Figure 9.16 b).




Figure 9.16 Retirement age increase, RETAGE SLOW and RETAGE FAST, pension adequacy

                                               151
                           a                                                b




Source: World Bank PROMIS model for Croatia




Retirement age increase substantially improves the fiscal outlook of the PAYG pension
system. Besides fewer new pensioners in the retiring cohort, longer stay at work also brings
longer contribution periods and more pension revenues. The potential magnitude of our two
simulations is seen in Figures 9.17. a-c. Expenditures under RETAGE SLOW are cumulatively
lower by 1.4% of GDP by 2035, and as much as 3.9% with RETAGE FAST. For the entire
projection period the savings are 16.8% and 41.4% of GDP, respectively. The resulting
reduction in gross deficit is impressive. By 2034 RETAGE SLOW reduces the cumulative deficit
by about 1% of GDP, while the reduction with RETAGE FAST stands at high 3% of GDP. Over
the entire projection period, the total cumulative savings reaches 24% of GDP for RETAGE
SLOW and 60% of GDP for RETAGE FAST, respectively (see Section 9.5).156 The pattern for net
deficit is similar because the RETAGE simulations do not significantly alter the choice of
combined vs. PAYG only pension (Figure 9.18).157


Figure 9.17 Retirement age increase, RETAGE SLOW and RETAGE FAST, expenditures and deficit
                           a                                                b




156
    Cumulative savings here result from lower pension expenditures and higher pension contributions. For
RETAGE FAST the lower expenditures share is the mentioned 41.2% of GDP, while the higher contribution 18.8%
of GDP.
157
    The spikes in some results of RETAGE simulations are due to PROMIS' raising the retirement age
discontinuously.

                                                   152
                                                      c




                           Source: World Bank PROMIS model for Croatia



Figure 9.18 Retirement age increase, RETAGE SLOW and RETAGE FAST, choice of P2




Source: World Bank PROMIS model for Croatia




 Raising retirement ages and narrowing early retirement opportunities are the most
 powerful policy measures to improve pension adequacy and fiscal sustainability
 simultaneously. There is no other pension policy measure with the same outcome.
 Furthermore, it strengthens intergenerational sustainability and equity by preserving the
 ratio of work and retirement periods across generations. If you live longer, you should
 work longer to earn the same relative pension. Retirement age policy must be included
 in every serious pension policy reform.



9.5 Policy packages and its adequacy and fiscal outcomes



                                                      153
Four policy packages are simulated combining the selected individual policies:

   •   PACK SLOW: combines the policies 85%W+15%P; SUPPL30%+MIN106%, EQCHOICE,
       and RETAGE SLOW. Its main objective is to test to what extent the slow retirement
       age increase adds to pension adequacy and offsets the fiscal costs generated through
       PAYG-expanding policies.
   •   PACK FAST: combines the same policies (85%W+15%P; SUPPL30%+MIN106%,
       EQCHOICE), but with RETAGE FAST with the same objective as previous.
   •   PACK SLOW NOEQ: tests the impacts of PACK SLOW without EQCHOICE. The main
       objective is to investigate if closing the wide net deficit generated by equal choice
       transfer would affect pension adequacy.
   •   PACK FAST NOEQ: extracts EQCHOICE from PACK FAST with the same objective as
       previous scenario.


All simulated policy packages significantly improve pension adequacy in Croatia. The
combination of individual policies in all four policy packages stops the erosion of pension
adequacy for new pensioners (replacement rates on Figure 9.19 a). Improvements with both
PACK SLOW and PACK FAST are tremendous. The average replacement rate, in the first
decade higher by 2.2 pp with SLOW and by 3.6 p.p. with FAST package (Table 9.2), would by
the end of the projection period be higher by 7.3 and 11.2 percentage points on average,
respectively. If the equal choice transfer is excluded from the package, the results are more
modest than with the transfer. PACK SLOW NOEQ rises the replacement rates in the first
decade by 1.1 pp and at the end of the period by 5.3 pp. With PACK FAST NOEQ the rates
stand higher, at 2.5 pp and 8.8, respectively. A similar conclusion is drawn with benefit ratios
(Figure 9.19 b and Table 9.2). While the baseline benefit ratio continues to erode, policy
packages either stabilize the benefit ratio at around 35% gross (PACK SLOW and PACK SLOW
NOEQ) or make it rise above that level (PACK FAST and PACKFAST NOEQ), which is the
ultimate goal of the pension reform. PACK FAST is the most effective in both replacement rate
and benefit ratio increases due to its tighter effect on new pensioners and their replacement
rates. Excluding the equal choice reduces the positive impact of both FAST and SLOW
retirement age policies on pension adequacy.


Figure 9.19 Policy packages SLOW, FAST – replacement rates and benefit ratios
                       a                                               b




                                                154
Source: World Bank PROMIS model for Croatia



Table 9.2 Average annual differences to baseline by 5-year periods and decades in pension adequacy –
policy packages
                         Change in replacement rates (percentage points)
                           2025-29        2030-34    2035-39   2040-44     2045-54     2055-65
    85%W+15%P                0.3              0.8      1.2       1.7         2.3         2.9
       2PILL7%               0.1              0.4      0.7       1.0         1.6         2.7
 SUPPL30%+MIN106%            0.4              0.4      0.4       0.3         0.3         0.3
      EQCHOICE               0.9              1.1      1.3       1.5         1.6         1.7
   RETAGE SLOW               0.3              0.0      0.3       1.3         1.4         2.0
    RETAGE FAST              1.3              1.8      2.8       3.4         4.3         5.4
     PACK SLOW               1.9              2.4      3.3       4.9         5.7         7.3
     PACK FAST               2.9              4.3      6.1       7.3         9.1         11.2
 PACK SLOW NOEQ              1.0              1.2      2.0       3.4         4.0         5.3
  PACK FAST NOEQ             2.0              3.0      4.5       5.5         7.0         8.8


                            Change in benefit ratios (percentage points)
                           2025-29        2030-34    2035-39   2040-44     2045-54     2055-65
    85%W+15%P                0.3              0.7      1.2       1.7         2.3         3.0
       2PILL7%               0.0              0.1      0.2       0.4         0.6         1.1
 SUPPL30%+MIN106%            0.6              0.5      0.4       0.4         0.4         0.3
      EQCHOICE               0.4              0.6      0.7       0.8         0.9         1.0
   RETAGE SLOW               0.3              0.4      0.4       0.7         1.1         1.7
    RETAGE FAST              0.7              1.2      1.6       2.1         2.9         4.0
     PACK SLOW               1.6              2.2      2.7       3.5         4.7         6.2


                                                    155
     PACK FAST               2.0              3.1     4.0      5.0       6.7         8.9
 PACK SLOW NOEQ              1.2              1.6     2.1      2.9       4.0         5.3
  PACK FAST NOEQ             1.6              2.5     3.2      4.1       5.6         7.4
Source: World Bank PROMIS model for Croatia




Simulated policy packages raise expenditures and deficits in the short run compared to
baseline. Raising adequacy via upgrading the pension supplement to 30%, minimum pension
per year to 106% of APV, equal-choice transfer, and introducing pension valorization and
indexation with 85% wages and 15% prices (if real wage grows constantly) brings immediate
increase in expenditures (Figure 9.20). Table 9.3, detailing the cumulative additional costs
generated by different measures, reveals that raising the supplement and minimum pension
(SUPPL30%+MIN106%) has the strongest effect in the short run, generating cumulative
expenditures of 1.3% of GDP by 2030. In the same period the additional costs of the stronger
indexation (85%W+15%P) stand at 0.6% of GDP. Retirement age increase cannot
counterbalance in the short run, so the net additional expenditures resulting from PACK SLOW
by 2030 is 1.1% of GDP, and from PACK FAST 0.4%. Without the equal choice transfer, the
resulting growth in expenditures is even higher, 1.4% for PACK SLOW NOEQ and 0.6% for
PACK FAST NOEQ.


In the longer run the expenditures path depends on the strength of retirement age increase.
In the remaining projection period PACK SLOW and especially PACK SLOWNOEQ fail to bring
the PAYG expenditures below the baseline. On the contrary, PACK FAST and PACK FAST
NOEQ’s savings grow stronger. By 2040 under PACK FAST, 2.7% of GDP is less spent
cumulatively, while for the entire projection period it reaches 15.1% of GDP. The cumulative
savings with PACK FAST NOEQ is slightly smaller, 10.7%. Table 9.3 reveals the additional fiscal
requirements of the indexation measure and the fiscal savings of the retirement age measure.
In the projection period, the simulation 85%W+15%P generates additional expenditures of
33.8% of GDP cumulatively, while the RETAGE FAST saves 41.4% of GDP. Savings with RETAGE
SLOW is 16.8%, insufficient to offset the PAYG adequacy boosting measures.




Figure 9.20 Policy packages SLOW, FAST – Total PAYG expenditures




                                                    156
Source: World Bank PROMIS model for Croatia



Table 9.3 Cumulative differences to baseline TOTAL PAYG EXPENDITURES – individual policies and policy
packages, % of GDP
                         2025      2026       2027    2028    2029    2030    2040    2050     2060     2065

         85%W+15%P        0.0%     0.1%       0.2%    0.3%    0.4%    0.6%    4.6%    13.4%    26.5%    33.8%

             2PILL7%      0.0%     0.0%       0.0%    0.0%    -0.1%   -0.1%   -1.1%    -3.7%    -7.8%   -10.0%

 SUPPL30%+MIN106%         0.2%     0.4%       0.7%    0.9%    1.1%    1.3%    3.7%     5.8%     7.6%     8.3%

           EQCHOICE       0.0%     0.0%       -0.1%   -0.1%   -0.2%   -0.3%   -2.1%    -5.2%    -8.3%    -9.4%

        RETAGE SLOW      -0.1%     -0.1%      -0.2%   -0.3%   -0.4%   -0.5%   -2.9%    -8.0%   -14.2%   -16.8%

         RETAGE FAST     -0.1%     -0.2%      -0.4%   -0.6%   -0.9%   -1.2%   -8.0%   -20.0%   -34.5%   -41.4%

          PACK SLOW       0.2%     0.4%       0.6%    0.7%    0.9%    1.1%    1.9%     2.2%     4.7%     7.8%

           PACK FAST      0.1%     0.3%       0.4%    0.4%    0.4%    0.4%    -2.7%    -8.7%   -14.0%   -15.1%

    PACK SLOW NOEQ        0.2%     0.4%       0.7%    0.9%    1.2%    1.4%    5.7%    12.5%    20.6%    25.4%

     PACK FAST NOEQ       0.2%     0.3%       0.4%    0.6%    0.6%    0.6%    -1.2%    -5.6%    -9.7%   -10.7%

Source: World Bank PROMIS model for Croatia




Similarly, gross PAYG deficit rises in the short run, but curbs in the medium and longer run.
The adequacy-enhancing pension measures, primarily the pension supplement and minimum

                                                      157
pension, and then the more abundant indexation, boost the expenditures in the short run.
Without raising the contribution rate, the only fiscal-balancing response comes through
higher retirement ages, which need time to effectuate. The gross deficit with the slow
retirement age policy hovers around the baseline but significant savings take off only with the
fast package (Figure 9.21). The gross deficit in PACK SLOW grows by about 0.2% of GDP in
each year until 2030 and requires supplementary financing in that amount. With PACK FAST,
the additional financial requirements by 2027 stand at 0.1% of GDP annually, with savings
generated afterwards (Table 9.4). Packages without equal choice transfer (SLOW PACK NOEQ
and FAST PACK NOEQ) generate higher gross deficits than their peer simulations for the
amount of savings that EQCHOICE generates throughout the projection period (Table 9.4).
The calculations do not take into account potential revenues from rising the contribution rate
for other income.158 Unless the additional temporary deficit is financed by supplementary
transfer or contribution revenue, the adequacy boosting policy would have to be postponed.


Figure 9.21 Policy packages SLOW, FAST – GROSS PAYG deficit




Source: World Bank PROMIS model for Croatia



Table 9.4 Annual differences to baseline GROSS PAYG DEFICIT – individual policies and policy packages,
% of GDP
                          2025     2026       2027   2028   2029     2030     2040    2050     2060     2065




158
  Crude calculations indicate that additional contribution revenues would have reached 0.2% of GDP in 2023 if
implemented decisively and without the induced crowding-out effect.

                                                     158
         85%W+15%P         0.0%    -0.1%      -0.1%   -0.1%   -0.2%   -0.2%   -0.6%   -1.1%   -1.4%   -1.5%

              2PILL7%     -0.6%    -0.6%      -0.6%   -0.6%   -0.6%   -0.6%   -0.4%   -0.3%   -0.2%   -0.2%

 SUPPL30%+MIN106%         -0.2%    -0.2%      -0.2%   -0.2%   -0.2%   -0.2%   -0.2%   -0.2%   -0.2%   -0.1%

           EQCHOICE        0.0%     0.0%      0.0%    0.1%    0.1%    0.1%    0.3%    0.3%    0.3%    0.2%

       RETAGE SLOW         0.0%     0.0%      0.1%    0.1%    0.1%    0.2%    0.6%    0.7%    0.8%    0.8%

        RETAGE FAST        0.1%     0.2%      0.2%    0.3%    0.4%    0.5%    1.3%    1.9%    2.3%    2.0%

          PACK SLOW       -0.2%    -0.2%      -0.1%   -0.2%   -0.1%   -0.1%   0.2%    0.0%    -0.2%   -0.4%

           PACK FAST      -0.1%    -0.1%      -0.1%   0.0%    0.1%    0.2%    0.9%    1.2%    1.2%    0.8%

   PACK SLOW NOEQ         -0.2%    -0.2%      -0.2%   -0.2%   -0.2%   -0.3%   -0.3%   -0.7%   -0.7%   -0.7%

    PACK FAST NOEQ        -0.1%    -0.1%      -0.1%   0.0%    0.0%    0.1%    0.7%    1.0%    1.2%    0.8%

Source: World Bank PROMIS model for Croatia




Due to lower revenue from fewer P1 choosers, the financing requirements stand higher,
but stronger retirement age increase still manages to compensate for the incurred cost. The
net deficit (including the extraordinary revenues due to transfer of accumulation from the
second pillar to PAYG) is substantially higher than gross deficit (Figure 9.22) due to
introduction of the equal-choice transfer (EQCHOICE). EQCHOICE reverses the balance in the
choice between two pensions and stimulates most individuals to choose the combined
pension (Figure 9.23). According to the law, the lack of “returned” funds needs to be
replenished by the budget as a part of the transition cost. The net cost of equalization transfer
(loss of returned accumulation) as individual measure stands around 0.4% of GDP annually
(Table 9.5). Slow retirement age package does not manage to compensate in the whole
projection period, but the PACK FAST does. Elimination of the equal transfer from the policy
package brings the net deficit closer to the baseline because some choose PAYG only pension
and hand over their second pillar savings to the state. Interestingly, the simulation PACK
SLOW NOEQ results in diametrically opposite situation with almost everyone choosing PAYG
only pension (Figure 9.23) in the period 2028-2041. This is the model result of the
combination of PAYG boosting measures (indexation 85%-15%; supplement 30%; and
minimum pension increase) which, combined with slow retirement age increase in the
PROMIS model made the PAYG only pension more attractive. As a result, the net deficit does
not worsen in comparison to baseline, but the gross deficit does, reflecting the outcome of
more extraordinary revenues with the former instead of more transfers from the budget in
the latter case. With PACK FAST NOEQ the net deficit starts lower but after equating with the
baseline in 2033 reduces quicky because of significant number of (declining) new old age
beneficiaries still choose the combined pension.
Figure 9.22 Policy packages SLOW, FAST – NET PAYG deficit



                                                      159
Source: World Bank PROMIS model for Croatia

Table 9.5 Annual differences to baseline NET PAYG DEFICIT – individual policies and policy packages, %
of GDP
                          2025      2026      2027    2028    2029    2030    2041    2050    2060    2065

     85%W+15%P            0.0%      0.0%      -0.1%   0.0%    0.0%    -0.1%   -0.2%   -0.7%   -1.2%   -1.3%

       2PILL7%            -0.7%    -0.7%      -0.7%   -0.7%   -0.7%   -0.7%   -0.8%   -0.7%   -0.5%   -0.4%

 SUPPL30%+MIN106%         -0.1%    -0.1%      -0.1%   -0.1%   -0.1%   -0.2%   -0.2%   -0.2%   -0.2%   -0.1%

      EQCHOICE            -0.3%    -0.4%      -0.4%   -0.5%   -0.5%   -0.5%   -0.5%   -0.4%   -0.2%   -0.1%

    RETAGE SLOW           0.0%      0.0%      0.0%    0.1%    0.1%    0.1%    0.4%    0.5%    0.8%    0.8%

     RETAGE FAST          0.0%      0.0%      0.0%    0.1%    0.1%    0.2%    1.0%    1.6%    2.1%    1.9%

     PACK SLOW            -0.5%    -0.6%      -0.6%   -0.7%   -0.7%   -0.7%   -0.6%   -0.7%   -0.6%   -0.6%

      PACK FAST           -0.5%    -0.5%      -0.5%   -0.5%   -0.5%   -0.4%   0.1%    0.5%    0.9%    0.5%

  PACK SLOW NOEQ          -0.2%    -0.1%      -0.2%   0.0%    0.0%    0.0%    0.2%    -0.6%   -0.5%   -0.6%

   PACK FAST NOEQ         -0.2%    -0.2%      -0.2%   -0.2%   -0.1%   -0.1%   0.5%    0.8%    1.1%    0.8%

Source: World Bank PROMIS model for Croatia



Figure 9.23 Policy packages SLOW, FAST – share of old age pensioners choosing combined pension




                                                      160
Source: World Bank PROMIS model for Croatia

An important question that came out of the simulation results is if the government should
waive the transfer from returned accumulations and cover the residual deficit instead. Two
straightforward answers can be offered. First, the higher adequacy which is shown above. If
the equal-choice transfer is not implemented, net deficits would be smaller than in the
baseline but with significantly lower adequacy due to many individuals not having the
opportunity for reaping the benefits of the multipillar system. Second, the implementation
of policy of justful choice, which should have been implemented from the beginning of
allowing the choice. Revenues from the return of accumulations by individuals who chose the
PAYG only pensions have not been earned on equal-choice grounds. It can be argued that the
imbalance is eliminated by the EQCHOICE and the policy packages incorporating it.
Accordingly, the state is waiving the revenue earned on unfair grounds and channeling it into
higher adequacy in the long run.



 Policy packages including PAYG-raising adequacy, such as more generous indexation,
 equal-choice transfer, higher pension supplement and minimum pension, improve
 pension adequacy but also strongly raise the pension expenditures. They could be
 implemented in a fiscally sustainable manner only with strong retirement age increase
 which simultaneously generates PAYG savings and higher adequacy. The stronger the
 retirement age increase, the wider the space for considering how to structure the PAYG
 measures. Implementing the equal-choice transfer sets the choice on equal-footing and
 raises pension adequacy but requires the state to waive most of the extraordinary
 revenue emerging from individuals choosing the PAYG pension and handing their
 savings to the state. Faster increase of retirement age, together with indexation 85%-
 15%, pension supplement raised to 30% and minimum pension to 106% of APV for each
 service year, and equal choice transfer, is the policy package recommended by the
 World Bank in addition to other individual policy measures identified in Section 7.


                                              161
10. Conclusion

Pension reforms in Croatia in the past three decades have not succeeded in reversing the
negative impact of demographic trends on pension adequacy. Keeping the overall
expenditures and contribution rate as pension policy anchors left the pension adequacy
endogenous in balancing the system in the medium run. Pension adequacy in Croatia has
been gradually eroding and poverty risk for elderly has been increasing.
Sources of adequacy deterioration have been both intrinsic, embedded in system design, and
external. Pension formula with below-wage valorization and indexation, mechanically
yielding replacement rate and benefit ratios’ decline; short service periods due to numerous
early retirement opportunities; reluctance to raise retirement age; underreporting of income;
large share of individuals on minimum income; underutilized second fully funded pillar with
large opt-back to PAYG have all contributed to low adequacy of the Croatian pension system.
The time has come for the Government of Croatia and pension system stakeholders to
address the issue of pension adequacy systemically and decisively.
The findings of this Report suggest that Croatian multipillar system can continue serving
Croatian population better than either of the monopillar systems (only PAYG or only second
funded). Reforms overhauling pension system’s structure, such as abolishing either of the
three pension pillars, would not improve pension system’s outcomes. However, the current
system needs a face-lift of all its three pillars - PAYG, the second mandatory funded and third
voluntary funded. Measures recommended in this Report are based on the analysis of the
system with individual pension model and microsimulation pension model established on
actual data of Croatian contributors and beneficiaries. The recommended measures include:
Measures for improvement of pension adequacy and/or fiscal sustainability in both
mandatory pillars

   •   Fast increase of statutory retirement ages for old age pension by 2030, followed by
       increasing it with life expectancy,
   •   Fast increase in early retirement age with narrowing the early retirement period to 3
       years in relation to statutory retirement age,
   •   Increase of age and insurance period for long-term insured,
   •   Increase of penalty for early retirement pension,


Measures for improvement of pension adequacy in PAYG only

   •   Indexation based on 85%:15% wage-inflation rotational formula,
   •   Increase of disability and survivor pensions by extending added period,
   •   Extending the period for calculation of added period increase for disability and
       survivor’s pensions,
   •   Abandoning the requirement of 35 years of insurance period for later retirement
       bonus,
   •   Increase of pension supplement from the current 4%-27% to 7%-30%,

                                             162
   •   Increase of minimum pension from 103% to 106% of APV per year of insurance,


Measures for improvement of the proportionality between contributions and benefits

   •   Limiting the application of the initial factor for long-term insured to maximum of 60
       months,
   •   Equalizing the contribution rate for other income with the general contribution rate,


Measures for improvement of second pillar and/or combined pensions

   •   Increase of the 2nd pillar contribution when the social and fiscal conditions are
       appropriate and in balance,
   •   Introducing choice-neutral transfer scheme to PAYG pensions,
   •   Implementing Incremental Allocation System (IAS),
   •   Expanding the eligibility for cash withdrawal from second pillar at retirement,
   •   Enriching the information sheet for new retirees,
   •   Further liberalization and simplification of the investment limits,
   •   Implementing capital markets development program, including listings of SOEs,
   •   Issuing securitized financial instruments for financing public investment,


Measures for improvement of voluntary pension savings

   •   Application of non-linear cascading incentive scheme,
   •   Indexing thresholds for government incentives with inflation,
   •   Increase of the exit age from third pillar and conditioning early withdrawal option with
       returning of government incentives,
   •   Increase of existing cash withdrawal option but not earlier than statutory retirement
       age,
   •   Liberalization of the choice of payout method irrespective of amount of pension
       savings,
   •   Further liberalization of the investment limits,
   •   Abolishing fiscal incentives for people who pay contributions to the third pillar after
       statutory retirement age,
   •   Introducing auto-enrollment covering all insured individuals with participation of the
       employers and the state.
According to the analyzes conducted in this Report, simultaneous implementation of these
measures, especially the strongest and mutually balancing ones would i) reverse the trend of
eroding pension adequacy, ii) set the Croatian pension system on a sustainable fiscal path,
and iii) assure intergenerational sustainability by keeping the total pension contribution rate
unchanged at 20%.




                                             163
Annexes

Annex 1: Pension Savings, Credit Ratings and Economic Development

Figure 0.1 00Sovereign ratings vs. total assets in private and funded pension plans in 2021

                                          14


                                          12
         Sovereign credit ratings scale




                                                                                                                  R² = 0,2886
                                          10


                                          8


                                          6


                                          4


                                          2


                                          0
                                               0,0              50,0            100,0           150,0           200,0              250,0
                                                     Assets under management in privately funded pension plans as % of GDP as of
                                                                                       2021

Methodological notes: (1) Sample includes 38 OECD and 24 non-OECD countries including Croatia, altogether 62 countries.
Countries from Africa, micro-states and Russia and Ukraine are excluded from the sample. Kosovo is excluded due to lack of
ratings data. Data are for 2021. Data on accumulated assets was missing for four countries, so 2019 values were used
(Belgium, France, Dominican Republic, and India). (2) S&P sovereign credit ratings table was translated to numeric scale from
1 to 14, where any C is 1, B and BB- is 2, BB is 3, BB+ is 4, BBB- is 5, BBB is 6, BBB+ is 7, A- is 8, A is 9, A+ is 10, AA- is 11, AA is
12, AA+ is 13, and AAA is 14. Outlook is not considered. (See Standard and Poor's Methodology „How We Rate Sovereigns“
(2019)).
Source: OECD Pension Markets in Focus 2022, Statistical Annex, Table A.B.3. for assets in private and funded pension plans;
tradingeconomics.com credit ratings list for sovereign credit ratings.




                                                                                        164
Figure 0.2 Sovereign ratings vs. GDP per capita at PPP in 2021

                                          14


                                          12                                                   R² = 0,7323

                                          10
         Sovereign credit ratings scale




                                          8


                                          6


                                          4


                                          2


                                          0
                                               0   20000   40000     60000      80000      100000     120000   140000
                                          -2
                                                                   GDP at PPP as of 2021

Source: OECD Pension Markets in Focus 2022, Statistical Annex, Table A.B.3. for assets in private and funded pension plans;
IMF WEO for GDP per capita at PPP in USD.




                                                                         165
Annex 2: Alternative determination of the basic pension factor


The implicit contribution rate gives an indication of the implicit (neutral) value of the basic
pension factor. Basic pension factor is an important part of the pension formula determining
the basic PAYG pension. Currently, it is set at 0.75 (=15/20), determined by the Law. However,
if the 15% of the gross wage paid to the first pillar as pension contribution is not sufficient for
payment of all insurance-based pensions from the first pillar, then the rest should be covered
by the budget. The budget is financed by general taxation. Tax obligation is the same for those
who receive full-PAYG pension and those who receive combined pension. Therefore, those
with the combined pensions should have the same right to receive government transfers
given to the public pension system as those who take full PAYG pension. Their equal
participation in financing the budget should be reflected in their basic pensions. One way of
doing that is by re-calculation of the basic pension factor.
Implicit basic pension factor can be calculated as an appropriate and fair value of the factor
for the calculation of basic pension of those who decide to take the combined pension.
Implicit basic pension factor is the ratio between implicit contribution rate for the first pillar
and total implicit contribution rate. Figure O.3 shows that it was around 0.8 in the years up to
2020 and then declined a bit. The results vary with the economic cycle and pension policy
measures. One may expect that implicit contribution rate for the first pillar, and implicit basic
pension factor in turn, will increase back to around 0.80 due to an increased deficit of the
pension system for insurance-based pensions in years to come, as suggested by the Draft
Government Budget for 2024. The reasons behind this are implementation of the new model
of subsistence pensions, higher minimum pensions, and the delay in spending effects of
regular pension indexation.
Figure 0.3 Implicit contribution rate to finance total pension expenditure and implicit basic pension
factor
30.0                                                                                                     0.81

25.0                                                                                                     0.80

20.0                                                                                                     0.79

15.0                                                                                                     0.78

10.0                                                                                                     0.77

 5.0                                                                                                     0.76

  -                                                                                                      0.75
        2012   2013   2014      2015      2016      2017      2018     2019      2020      2021   2022
                             Total implicit contribution rate*
                             Implicit contribution rate for Pillar I - multi-pillar system**
                             Implicit basic pension factor - multi-pillar system (RHS)***




                                                        166
Note: *Implicit total contribution rate is estimated at the level that expected contribution revenues are sufficient to cover
total pension expenditures. It is expressed in % of gross wage (gross income subject to contribution payment) by scaling up
current rate of 20% to get revenues sufficient to cover total pension expenditure. Total pension expenditure is calculated as
actual pension expenditures arising from the general pension legislation recorded by HZMO without one-off transfers (covid,
energy and inflation benefits) and without pension additions based on special regulation (war veterans, military and policy
pensions, etc.). Pension expenditures made by the second pillar (without one-off expenditure of 15%) are included in total
pension expenditure.
**Implicit contribution rate for Pillar I is total implicit contribution rate minus contribution rate for the Pillar II of 5%.
*** Implicit basic pension factor is ratio between implicit contribution rate for the Pillar I and total implicit contribution rate.




                                                               167
Annex 3: How many first pillar pensions in Croatia are earnings-related?

Although the general pension formula in the first pillar is earnings-related, more than half
of pensions paid have no relation to earnings. This finding is based on crude estimation of
non-earnings-related pensions using comparable data over the last ten years. Non-earnings-
related pensions are: i) pension under special regulations, ii) minimum pensions approved by
post-1998 legislation, iii) minimum pensions and pensions with poverty protection
supplement approved before 1999, and iv) survivors’ pensions. In 2012 were 49.8% of these
pensions and their share gradually increased to 54.2% in September 2023 (Figure O.4).
The share of earnings-related pensions is declining mostly due to a sizable increase in the
proportion of minimum pension beneficiaries. In September 2023, around 25% of all
pensions were minimum pensions, compared to around 18% in 2012. New legislation of
minimum pensions in power since 1999, including two recent increases in the minimum
pension point value (2019 and 2023), resulted in continuously increasing number of minimum
pension beneficiaries. There is no official data on the proportion of new old-age pensions
approved as minimum pensions, but our estimate is that it could be more than 40% in 2023.
The number of minimum pension beneficiaries under previous legislation (before 1999) is
naturally declining. The proportion of pensions under special regulation is on the gradual rise
since 2018. The main reason is the increasing number of war veterans that qualify for special
kind of minimum pension.
The share of survivors’ pensions is declining due to shrinking gender employment
differential and wage gaps. In 2023, a further decline in the number of survivors’ pensions
was caused by policy measures. Increase in the value of survivors’ pension led to a higher
number of those who find survivor’s pension to be higher than their own pension. However,
now it is possible to continue receiving their own pension and take up to 27% of the survivors’
pension. That option motivated some of the former survivor pension beneficiaries to return
to their own pension and take additional survivor’s benefit. In our estimation of non -earnings-
related pension in Figure O.4, pension consisting of own personal pension plus a part of
survivors’ pension is not counted as non-earnings-related pensions, although it might be.
Namely, the principle of proportionality between earnings and subsequent pensions is
seriously compromised in case of mixed pension (own plus a part of survivors). In September
2023, there were 73.5 thousand of such pensions (6% of total). That number will increase
because not all claims for such a pension among existing pensioners are posted and
processed. It is likely that there will be more than 100 thousand of such pensions in 2024. This
type of pension might also be popular among new pensioners. Although treated as earnings-
related pensions in our structure of pensions, the appearance of this type of benefit




                                              168
contributes to observation that the importance of own earnings for one's own pension is on
the decline.159
Figure 0.4 Non-earnings related PAYG pensions (in % of total pensions)

 60.0%

                                                                                             2.1%        2.0%
 50.0%                                                2.9%          2.6%        2.3%
                                           3.1%
                 4.9%

 40.0%                                                18.8%         19.5%       20.2%       20.6%       22.9%
                13.4%                      18.3%

 30.0%
                15.6%                      14.2%      14.4%         14.5%       14.7%       15.1%
 20.0%                                                                                                  15.1%


 10.0%
                15.8%                      15.6%      15.5%         15.3%       15.3%       15.2%       14.2%

      0.0%
                2012                       2018       2019          2020        2021         2022     SEP2023

             Survivors (without special regulation)            Pensions under special regulations
             Minimum pensions                                  Minimum and protection supplement (pre-1999)

Notes: Pensions under special regulation include war veterans and military personal pensions, but also pensions approved
by general regulation in the amount that is determined by special regulation. Minimum pension refers to benefits
determined by post-1998 regulation, while minimum pensions and protection supplements refer to benefits determined by
pre-1999 regulation. Both categories of minimum pensions are shown without survivors' pension to avoid double counting.
Survivors’ pensions are without pensions under special regulation. In 2023, a pension consisting of own pension plus a part
of survivors’ pension is treated as earnings-related pension.
Sources: HZMO Statistical information for 2018-2023 and HZMO working tables for 2012.




159
    In this simple division of first pillar pensions between those that are related to earning and those that are
not, one must note that pensions treated as earnings-related are still subject to rules that influence the degree
of their dependence on earnings. This refers to disability pensions, pension for those with long service or those
working on arduous and hazardous jobs, pensions including added service period of mothers (parents), etc. Also,
most of the pensions taken some 20 and more years ago are mix of original benefit and various supplements
that were implemented in the past. Nevertheless, the presented division of pensions in two parts, earnings- and
non-earnings related, helps us to confirm weakening of proportionality between contributions paid and
pensions received in the first pillar.

                                                              169
Annex 4: Net pension benefits and net replacement rates projected by the TRR model


The theoretical replacement rate (TRR) model is applied to inform the discussion in Section 4 and
Section 7, with the results presented in gross terms, whereas here, the selected results are
presented in net terms. The first gross pension benefit is divided by the last gross wage to get the
gross replacement rate. However, for the policymakers, net replacement rates may be relevant for
communication reasons. The trends are basically the same in gross and net terms due to the
assumptions used in simulations of taxation in the future.

In simulations of wage and pension taxation, we start with the tax rules in 2024, and for future years
all the nominal brackets are adjusted to increase in line with the assumed gross wage growth.
Personal tax allowance, the upper bracket for application of the second (the higher of two) tax rate,
and the brackets for application of reduced pension contribution rate for the first pillar are all
increasing in line with the gross wage growth. The tax rates of 20% and 30% are applied in all years as
we assume that these rates are close to the average for Croatia. Currently, each local government unit
decides on the tax rate subject to certain limitations set by legislation. Other projection features and
assumptions of the TRR model are the same as explained in Section 4.1.



Figure 0.5 Net TRR for a male average wage earner in cases of old-age and early retirement (M65/40
and M60/35)




Note: Person is assumed to be passive pension fund member and experiences flat wage profile in the career.
Source: WB team projections.


Figure 0.6 P2/P1 in net terms for three generations of new old-age pensioners and different wage levels
(M65/40)


                                                          170
Note: Person is assumed to be passive pension fund member and experiences flat wage profile in the career.
Source: WB team projections.

Figure 0.7 Theoretical net replacement rates for female average wage earner with flat wage profile
having two children in cases of old-age and early retirement (F65/40 and F60/35)




Source: WB team projections.




                                                          171
Table 0.1 Determination of the net pension benefit for the full-PAYG pension and the combined pension
depending on the average value point (female average wage earner having two children in case of early
retirement F60/35), retirement at January 1, 2024 (in EUR)
                                                    Average VP=0.60   Average VP=1.00    Average VP=2.00
                                                  Full-PAYG Combined Full-PAYG Combined Full-PAYG Combined
                                                   pension pension pension pension pension pension
First pillar
Gross pension benefit by point formula for the
effective insurance period                           235.83     198.77      393.06     331.29     786.11      662.58
Gross pension benefit by point formula for the
added insurance period due to parenthood                6.74       5.05      11.23        8.42      22.46      16.85
Pension supplement (27%)
                                                      65.49       55.03     109.16      91.72     218.31      183.44
Added benefit to reaching the minimum
pension                                              108.75       91.38           -          -           -          -
Gross pension benefit from the first pillar
                                                     416.82     350.24      513.44     431.43 1,026.89        862.87
Second pillar
Gross pension benefit from the second pillar
(annuity)                                                  -      44.13           -     73.55            -    147.11
Total

Total gross pension benefit
                                                     416.82     394.37      513.44     504.99 1,026.89 1,009.98
Personal income tax
                                                        0.00       0.00       0.00        0.00      48.87      47.10
Net pension benefit
                                                     416.82     394.37      513.44     504.99     978.01      962.87
Source: WB Team.

Figure 0.8 Net TRR for old-age pension of average wage earner (M65/40) depending on wage profile
and investment behavior




Note: A passive member follows default ABC fund category distribution, while an active member chooses the most active
ABC fund category allowed (pre- and post-2024 legal changes).
Source: WB team projections.

                                                        172
Figure 0.9 Net TRR depending on real wage growth and real rate of pension fund returns ((M65/40,
average wage earner)




Note: Baseline scenario assumes annualized real wage growth of 1.5% per annum and net real investment returns of 3.4%
in A, 2.2% in B and 1.0% in C. Pessimistic/optimistic scenarios assume +/- 1 pp in these rates.
Source: WB team projections.


Figure 0.10 The impact of indexation rules on the net TRR
             a. 100% wage indexation and 100% inflation valorization/indexation




                                                        173
              b. 70%:30% and 85%:15% wage/inflation valorization/indexation




Note: The results refer to the average wage earner, M65/40, passive member with flat wage profile.
Source: WB Team




Table 0.2 Net replacement rates in case of increasing contribution rate for the second pillar to 7% and
decreasing contribution rate for the first pillar to 13% since January 1, 2025 (old-age pension with 40
years of service for average wage earner, men)
                                                     2024               2026              2044       2064
Full PAYG pension                                      54.0%              53.5%             49.8%      45.6%
Combined pension – baseline (no-reform)                 53.5%             53.3%             52.7%     50.5%
Combined pensions – reform with
                                                        53.5%             53.3%             53.0%     52.3%
corrected basic pension factor
    Corrected basic pension factor                     0.7500            0.7457            0.7025    0.6525
Combined pensions – reform with
                                                        53.5%             53.4%             55.2%     56.2%
unchanged basic pension factor
Source: WB Team.




                                                          174
Annex 5: Description of the Croatia PROMIS microsimulation model

The Pension Reform Options Micro Simulation (PROMiS) tool is a general-purpose simulation
software for modeling complex social policy scenarios. It simulates individuals moving
through various states of life (e.g., employment, retirement, disability, survivorship). PROMiS
facilitates modeling process by providing: (i) a flexible scenario and output description
interface; and (ii) Monte-Carlo simulation framework, which allows tracking individuals across
various life events over time by applying probabilistic transition paths to individuals moving
between the life states. The Monte-Carlo simulation core and the model specification user
interface are distributed as a single Microsoft Windows executable file. The model
specification, in addition to the description of life events, normally includes the descriptions
of policy rules and outputs, and underlying aggregate data and micro datasets. The model
specifications are developed using Java Script Object Notation (JSON), while the underlying
data can be specified in various common data formats, including STATA data files, CSV, and
Excel.


Data foundation for the Croatia PROMIS

In addition to macroeconomic and demographic aggregates described in Section 8 of the
report, the Croatia PROMiS model uses individual micro data on contributors and existing
beneficiaries extracted from the HZMO and REGOS databases as of December 31, 2020. The
contributors’ data include individual demographic characteristics, such as date of birth and
sex (used in the model to simulate mortality), as well as the information allowing to compute
contributory wages, last contribution amount and its date, and the length of service and
individual points accumulated up to date. For the individuals participating in the second pillar
defined contribution scheme, the data contain individual account balances. The pensioners’
data in addition to demographics provide PAYG pension amounts and the indicator variables
defining whether it is a minimum or a maximum pension recipient and whether it is a full
PAYG pension or a basic PAYG pension of a second pillar participant.

For modeling purposes, the contributors’ data were split into two pools: (1) active
contributors comprised of individuals with at least one contribution in 2020; (2) dormant
contributors including individuals with no contributions in 2020. The contributors can retire
in one of the following pension categories (types): (1) regular old age pensioners defined as
individuals who reached statutory retirement age160 and with at least 15 years of service; (2)
early old age pensioners who can retire at most 5 years before the normal retirement age
(with pension reduced) if they have at least 35161 years of service; (3) long service pensioners
who can retire at 60 provided they have at least 41 years of service; (4) special pensioners
who can retire early without pension reduction if they worked in special occupations. In
addition, some contributors can retire as disabled or as international regular or early old age
pensioners with partial pension paid only for years worked in Croatia if they worked abroad
and fulfilled the overall length of service (LOS) vesting requirement through work history in
Croatia and abroad.

160
      65 years currently for males and increasing to 65 years for females by 2030.
161
      35 years for males and increasing to 35 years for females by 2030.

                                                        175
Based on eligibility to retire in certain pension category, we further segmented the
contributory data as follows:

      Activity status and LOS requirements                        Retirement eligibility
 Active contributors with at least 15 years of     Eligible to retire as regular old age, early old age,
 service                                           disabled, long service, and special pensioners
 Dormant contributors with at least 15 years of    Eligible to retire as regular old age, early old age,
 service                                           disabled, long service, and special pensioners
 Dormant contributors with less than 15 years of   Only eligible to retire as international old age
 service                                           and international early old age

The retirement rates for different pension types were developed using base year data on
pensioners who retired in 2020. The retirement rates and eligibility determine the flow of
new pensioners by pension type.


Key building blocks of Croatia PROMIS


The PROMiS model for Croatia consists of the following modules:
   1. Projection of population structure aggregated by single-year age and gender.
   2. Projection of key macroeconomic indicators (GDP, inflation and economy-wide
      nominal wage) and pension indexation parameters (e.g., 70% of nominal wage growth
      and 30% of inflation indexation).
   3. Projection of aggregate counts of active and dormant contributors based on working
      age population counts (based on the projection of population structure in #1).
   4. Projection of active contributors at the individual level and their wages and
      contributions. Projection of dormant contributors at the individual level. These
      projections are based on initial 2020 stocks of active and dormant contributors and
      projected aggregate counts (from the projection in #3). The new contributors are
      created to fill the gap between the stock of contributors remaining from the previous
      year and projected aggregate counts. The age-gender structure of newly created
      contributors corresponds to the age-gender structure of new contributors in the base
      year.
   5. Projection of retirement flows from active and dormant contributors into various
      types of retirement (as described in “Data foundation for the Croatia PROMIS” section
      above).
   6. Calculation of retirement benefits for new pensioners based on retirement flows
      (from the projection in #5) and benefit formulae for different pension types.
   7. Projection of pensioners’ stock and the benefits of existing pensioners using initial
      stock of existing pensioners (as described in “Data foundation for the Croatia PROMIS”
      section above), and the retirement flows and benefits (from #6), and pension
      indexation parameters (from #2).



                                               176
   8. Aggregation of individual contributions to calculate contribution revenues.
      Aggregation of individual benefits to calculate pension expenditure, system deficits or
      surpluses, replacement rates, benefit ratios, etc.
   9. Descriptions of output indicators.


The calculations in 1 to 9 are performed according to the following flow chart:




                                             177
Population counts by     Macroeconomic indicators      Stock of pensioners,
age and sex              and indexation parameters     indexation of benefits,
                                                       calculation of
                                                       expenditures
                         Benefit formulae for
Working age population   calculation of new benefits

  Active contributors    New old age pensioners
  and calculation of
  wages,
                         New early pensioners
  contributions, and
  revenues
                          New disabled pensioners


                         New long service
                         pensioners
                          New special pensioners



  Dormant                New old age pensioners
  contributors with
  LOS>=15 years
                         New early pensioners


                          New disabled pensioners


                         New long service
                         pensioners
                         New long service
                         pensioners
  Dormant
  contributors with      New international old age
  LOS< 15 years          pensioners


                          New international early
                          pensioners




                         Deceased                              New survivors


                                 Aggregation


                               Output indicators


                                  178
References

Ahec-Šonje, Amina, Milan Deskar-Škrbić and Velimir Šonje (2018): Efficiency of public expenditure on
education: comparing Croatia with other NMS. MPRA paper 85152.

Bađun, Marijana, Milan Deskar-Škrbić and Vedrana Pribičević (2014): Government size and efficiency
as constraints to economic growth: comparing Croatia with other European countries. Post-
Communist         Economies,    26:3,     297-323,      DOI:     10.1080/14631377.2014.937089;
https://www.worldbank.org/en/events/2014/12/16/public-finance-review-restructuring-spending-
for-stability-and-growth

Egert, Balasz (2012): The Impact of Changes in Second Pension Pillars on Public Finances in Central and
Eastern Europe. CESifo Working Paper Series No. 3801.

European Commission, Economic Policy Committee - Social Protection Committee, 2020, Joint Paper
on Pensions 2019, https://economic-policy-committee.europa.eu/system/files/2020-01/Joint-Paper-
on-Pensions-2019.pdf

European Commission - Secretariat-General (European Commission), 2021, Green Paper on Ageing:
Fostering solidarity and responsibility between generations, COM(2021) 50 final of 27 January 2021,
https://op.europa.eu/en/publication-detail/-/publication/d918b520-63a9-11eb-aeb5-01aa75ed71a1

European Commission, Directorate-General for Employment, Social Affairs and Inclusion, 2021
Pension Adequacy Report – Current and future income adequacy in old age in the EU. Volume 1,
Publications Office, 2021, https://data.europa.eu/doi/10.2767/013455

European Commission, 2015, The 2015 Ageing Report: Economic and budgetary projections for the
28 EU Member States (2013-2060), The European Economy series 3/15,
https://ec.europa.eu/economy_finance/publications/european_economy/2015/ee3_en.htm

European Commission, 2018, The 2018 Ageing Report: Economic and Budgetary Projections for the
EU Member States (2016-2070), Institutional paper 079, doi:10.2765/615631 (online).

European Commission, 2021, The 2021 Ageing Report: Economic and Budgetary Projections for the
EU Member States (2019-2070), Institutional paper 148, doi:10.2765/84455 (online).

European Commission, 2023, The 2024 Ageing Report: Underlying assumptions and projection
methodologies, Institutional paper 257, doi:10.2765/960576 (online).

European Commission, 2024, The 2024 Ageing Report: Economic and Budgetary Projections for the
EU Member States (2022-2070), Institutional paper 279, doi:10.2765/022983 (online).

Government of the Republic of Croatia (2024): Program of the Government of the Republic of
Croatia 2024 – 2028, available at:
https://vlada.gov.hr/UserDocsImages/ZPPI/Dokumenti%20Vlada/2024/Program_16_Vlade_RH.pdf.

Grgurev, I., and I. Vukorepa, I. (2018): Flexible and new forms of employment in Croatia and their
pension entitlement aspects. Transnational, European, and National Labour Relations: Flexicurity and
New Economy, 241-262.

Nestić, D. and Tomić, I., 2012. Adequacy of Pensions in Croatia: What Can Tomorrow's Pensioners
Expect?. Privredna kretanja i ekonomska politika, 22(130), pp.61-100.


                                                 179
OECD (2020), "A framework for assessing the adequacy of retirement income", in OECD Pensions
Outlook 2020, OECD Publishing, Paris, https://doi.org/10.1787/760ca223-en.

OECD (2021), Pensions at a Glance 2021: OECD and G20 Indicators, OECD Publishing,
Paris, https://doi.org/10.1787/ca401ebd-en.

OECD (2023), Pensions at a Glance 2023: OECD and G20 Indicators, OECD Publishing,
Paris, https://doi.org/10.1787/678055dd-en.

Rudolph, H. P. (2016): Building Voluntary Pension Schemes in Emerging Market Economies. World
Bank Policy Research Working Paper 7779.

Sczepanski, M. and K. Kolodzieyczik (2021): Auto-enrollment Pension in Poland: Impact on HRM
Strategies and Pension Participation Rates. Journal of Eastern Europe Research in Business and
Economics, 1-13.

Sopek, Petar (2011): Efficiency of public expenditure on education in Croatia. Institute for Public
Finance Newsletter No. 61.

Vukorepa, I., I. Tomic, I., and P. Stubbs, (2017): ESPN thematic report on access to social protection of
people working as self-employed or on non-standard contracts.

The World Bank (2020): Public Expenditure Review in Science, Technology, and Innovation in Croatia.
The World Bank Brief.




                                                  180