SIMLAB: A Tool to Analyze Distributional Impact of Structural Reforms The Case of Kosovo Monica Robayo-Abril Ana Maria Oviedo The World Bank Poverty and Equity Global Practice Pristina, January 23th, 2020 Outline 1. Motivation 2. Overview of SIMLAB 3. Model 4. Baseline Results 5. Counterfactual Policy Simulations 6. Simulation Tools ▪ Comments 1. Motivation Motivation Stylized Facts • Existence of a sizable informal sector in Kosovo, regardless of definition used • Persistently high unemployment and long unemployment durations, especially among young • Poverty reduction limited by lack of access to job opportunities Motivation Stylized Facts • Unemployment rates in Kosovo are the highest in the Europe and Central Asia (ECA) region, and among the highest in the world, particularly among the young. Youth vs. Overall Unemployment Rate (% Labor Force), ECA Countries, circa 2018 35 Unemployment, total (% of total labor force) (national 30 XKX 25 20 MKD estimate) GRC 15 MNE BIH ESP ALB SRB GEO TUR ITA 10 FRA HRV UKR CYP LVA FIN PRT LTU IRLSVK KGZ BELSWE 5 SVN DNK EST BGRLUX CHEAUT BLR RUS ROU DEU GBR NLD NORPOL HUN ISLMDA CZE 0 0 10 20 30 40 50 60 Unemployment, youth total (% of total labor force ages 15-24) (national estimate) Motivation Stylized Facts • As such, workers in Kosovo waste approximately half of their productive life, the worst performer in the ECA region. Average years of employment potentially lost, 2016 Source: World Bank (2018b). Note: Methodology based on Arias et al. (2014) Motivation Stylized Facts • Expected years of schooling, measured as the number of years of education a child born today can expect to achieve at the age of 18, are relatively low (scope for Human capital policies – school levels). Expected years of school 16 12 Expected years of school 8 4 0 TJK SRB BIH ROU EST ESP CHE HRV IRL PRT FRA LUX GRC BGR LVA CYP RUS GBR AZE TUR GEO NLD DEU ALB DNK ITA SWE KGZ XKX SVK HUN UKR POL KAZ ISL BEL SVN FIN NOR AUT CZE MKD MDA MNE LTU Source: WB Human Capital Project 2018 • • 100 200 300 400 500 600 0 Estonia Netherlands Poland Switzerland Denmark Motivation Stylized Facts Slovenia Belgium Finland Sweden United Kingdom Norway Germany Ireland Czech Republic Austria Latvia France Iceland Portugal Russian Federation Italy Slovak Republic Luxembourg Spain (scope for Human capital policies – productivity). Lithuania Hungary PISA mean math score, 2018 Belarus Croatia Turkey Ukraine Greece Cyprus Low quality of education (Math PISA Scores -the lowest in ECA). Serbia Albania Bulgaria Romania Montenegro Kazakhstan Moldova BiH Georgia N. Macedonia Kosovo Students lose approximately 5.1 years of schooling because of low quality of education Motivation Stylized Facts • Low access to quality education translates into lower productivity: a child born today will lose almost 44 percent of its lifetime productivity Human Capital Index 0.9 0.8 0.7 0.6 0.56 0.5 0.4 HCI 0.3 0.2 0.1 0.0 LUX LVA ROU GEO HRV EST FRA PRT TJK BIH BGR GRC ESP SRB CHE GBR DEU NLD RUS CYP AZE TUR SWE IRL ALB KGZ KAZ ITA AUT UKR SVK POL BEL CZE SVN MKD XKX MDA MNE HUN LTU NOR FIN DNK ISL Source: WB Human Capital Project 2018 Motivation Structural barriers affecting labor market performance • Previous evidence suggests the following structural barriers to employability: • Lack of skills and skill mismatch (World Bank, STEP Report and Kosovo Jobs Diagnostic (2017)) • Information asymmetries and frictions → Informal networks and connections matter • Lack of dynamic formal sector job creation from costly regulatory burden and inefficient and extensive inspection systems. Source: WB Human Capital Project 2018 Motivation The Role of Information Asymmetries and Frictions • Resolving structural barriers requires a combination of policies including: • Human Capital policies (Education quality and completion) • Matching efficiency policies • Formalization policies • This implies: • Different time horizons and costs • Results may affect different population groups differently • Policymakers need to understand the trade-offs and cost-benefit of different policies, as well as potential distributional impacts Motivation Policy Questions • What are the main drivers behind high unemployment in Kosovo? • What human capital policies (i.e. education vs. productivity) should be implemented to reduce unemployment and what are the distributional impacts? Which population groups (low-skilled vs. high-skilled) should be targeted, and in which sectors (formal/informal)? • How important is formalization to reduce unemployment and what are the distributional impacts? • How important are frictions and information asymmetries? Can policies affecting matching efficiency reduce unemployment and what are the distributional impacts? • What is the impact of these set on policies on GDP per capita growth and poverty? ▪ Comments 2. Overview of SIMLAB Overview What is SIMLAB? • SIMLAB (Simulation of Policies in Labor Economics) structural model of the labor market with search and matching frictions capable of simulating a rich array of labor market and other relevant macroeconomic variables. • SIMLAB captures both the micro-distributional as well as macroeconomic effects of selected policy reforms, when the main channel of transmission is the labor market. • KEY OBJECTIVE: simulate ex-ante effects of selected policy reforms, including human capital policies, formalization, labor market policies and policies affecting matching efficiency • USES: SIMLAB can be used as a standalone to inform policy and operations or to complement other country diagnostics (SCDs, Job Diagnostics and PSIAs) Overview How does SIMLAB compare with other modeling and simulation approaches? • It allows for ex-ante analysis • Requires the simulation of a counterfactual sample, which represents the population characteristics of interest. • Analysis is marginal and behavioral • Agent response to policy reforms are considered when defining the counterfactuals • New lens to assess distributional impact of reforms through a structural model • Search models are well established in the academic literature. • Search frictions are critical in the labor market • Partial equilibrium vs Equilibrium search models : Canonical DMP Overview How does SIMLAB compare with other modeling and simulation approaches? simulation approaches? Search Models: The Flow Perspective on Labor Markets Overview How does SIMLAB compare with other modeling and simulation approaches? • Idiosyncratic component and wage dispersion: Not all models of search are designed to answer distributional questions (core of micro-search literature) • SIMLAB model is designed to address distributional questions • A user-friendly microsimulation tool with these features is extremely useful for policymakers to simulate ex-ante distributional impacts impact of policy reforms before they are implemented Overview Structural Models vs Reduced-Form for Policy Evaluation Advantages of structural approach for policy design • Define how outcomes relate to preferences • Describe the mechanism through which effects operate • Provide the framework for understanding how a particular policy may translate in different environments • Designed to analyzed counterfactual policies, quantifying impacts of specific outcomes in the short and longer run • Not limited to analyze past historical events • Equilibrium approach: understand how both firms and workers are affected by policy reform (equilibrium model) Overview Structural Models vs Reduced-Form for Policy Evaluation Disadvantages • Relies on simplifying or identifying assumptions, and some economic choices must be left out • Important to understand the basics behind it • Entry cost: Learning to solve them is high (numerical methods) • For this reason we develop a user-friendly web-based tool that users can use to simulate in real time State of the art literature combines randomized experiments and structural modeling for policy evaluation Use experimental evidence to validate structural model • Estimate structural model based on data from control group • Predict the impact of the experiment in the model See Low and Meghir (JEP 2017) for a full discussion Overview Methodology 1. Build a search and matching friction model of a labor market with formal and informal sector 2. Solve the model numerically 3. Simulate the model and calibrate/estimate the structural parameters using micro data (i.e. standard labor force survey) 4. Run counterfactual policy experiments • Quantify the ex-ante aggregate, compositional and distributional impact of set of policies on the labor markets 5. Model Extensions: impacts on GDP per capita and poverty 6. Tool Development ▪ Comments 3. Model Model Ex-ante Heterogeneity: 8 observed worker types We want distributional impacts so we group young (15-29) and old (30-69) workers into 4 educational categories, for a total of 8 worker types/ groups • Lower Secondary or below: includes no school, elementary education, and 8/9 years lower secondary • Upper secondary -Vocational: includes 2/3 years or 4/5 years of vocational school • Gymnasium/Complete Secondary: includes upper secondary - general (gymnasium) and high school • Complete Tertiary and above: includes tertiary university, post university / master and doctorate Model Proxy of Informality based on firm size and occupation • Firm Size and Occupation definition of Informality A worker is considered informal if the following two criteria are met: • Paid salary employees working in small firms • Firm size<10 (observed range) since cannot identify those with 5 or less • Employees in non-professional occupations (clerks, service and market sales workers, skilled agricultural, craft workers, machine operators, elementary occupations) • Social protection (SP) definition of Informality: cannot be used since very few workers report being covered by health and pension in Kosovo, since public health insurance is not mandatory Model Human Capital Policies Policies that enhance human capital accumulation along several dimensions Aggregate, How? What? compositional and Compulsory Accumulation distributional effects schooling and distribution Education legislation, of schooling enhancing accessibility (could be biased Labor Market Impacts policies and affordability toward women, of schools young, old, vocational, gymnasium) Poverty Impacts On-the-job training, post- schooling executive Productivity training, Macro Impacts enhancing experience- Accumulation policies enhancing and distribution (internships), of productivity training (same vouchers, schooling) continuing education Model Formalization Policies Policies that aim to reduce size of informal sector or increase the size of the formal sector How? What? Aggregate, compositional and Reduce regulatory burden distributional effects Increase among formal-sector firms: vacancies in • offering grant payments the formal Policies to sector Labor Market or tax benefits for promote Impacts companies hiring formal ‘formally’, Reduce job employment destruction • Facilitate starting a rates in the business, get credit, or formal Poverty Impacts deal with permits if firms sector do comply with labor market regulations Increase Macro Impacts cost of Policies to posting a Enforce compliance with reduce vacancy (or regulations that leads to compliance informal informal-sector firms cost) in the employment internalize some costs of informal hiring informally. sector Model Matching Efficiency Policies Policies that aim to increase the job finding rate, or reduce the time it takes for a worker to find a job Aggregate, How? compositional and What? distributional effects • Efficiency of PES • Efficiency of private Labor Market sector employment Increase Impacts providers matching efficiency Policies • Improving information (higher job that asymmetries (Labor finding rate Poverty Impacts promote Market Observatories, for same matching LMIS, informational number of efficiency campaigns) vacancies) • Tackling barriers to Macro Impacts geographic labor mobility • ALMP Model Want aggregate, compositional, and distributional effects of policies Want impact of human capital, formalization, and matching efficiency policies on: • Labor market • Division of labor force into unemployment, informal and formal (aggregate effects) • Mix of worker types in two sectors (compositional effects) • Distribution of productivity and wages across sectors (distributional effects) • Disaggregated LM indicators by age and educational attainment (distributional effects) • Poverty rates • Important potential impacts on income distribution and poverty via their effects on the level and distribution of wages and employment • Macro- GDP per capita growth using Shapley decomposition • Employment and productivity from search model • UN population projections + LFP target growth Model Extensions : Impact on GDP per capita and Poverty GDP per Capita Shapley Decomposition Method- Decompose output per capita as follows: = Product Empl Rate LFP Dependency Ratio Where Y= Total output Model Outputs N= Total population E= Employment L= Labor force A= Working-age population Poverty Imputing household per capita consumption from the 2017 Household Budget Survey (official survey to measure poverty) into the counterfactual sample simulated by the model ▪ Comments 4. Baseline Results Baseline Results Model Performance: Model fits well selected data moments Variable Model Data AGGREGATE EMPLOYMENT AND UNEMPLOYMENT RATES u 0.44 0.43 ni 0.11 0.11 nf 0.45 0.46 LM TIGHTNESS AND INFORMAL SECTOR SIZE θ 0.64 φ 0.24 MEASURES OF WAGE DISPERSION µ(lnw)f - µ(lnw)i 0.40 0.41 σ(lnw)f / σ(lnw)i 1.39 1.36 σ(lnW) 0.38 0.39 MEAN EMPLOYMENT DURATION (years) µ(t)I 5.33 5.34 µ(t)f 11.05 11.03 Baseline Results How tight is the Kosovo Labor Market? Labor market tightness, or fraction of vacancies to the unemployed job seekers calibrated at 0.64, within the usual range of estimates for other countries US Colombia Ranging between 0.2 and 0.9 Ranging between 0.3 and 0.65 Source: St Louis Fed Source: Morales, Lobo 2017 Baseline Results Model matches formal/informal employment rates by age and education Informal-sector employment rate drops significantly with education, especially among tertiary educated workers Informal-Employment Rate by Educational Level and Age 30% 20% Baseline 10% 0% Gymnasium/Complete Gymnasium/Complete Upper secondary -Vocational Upper secondary -Vocational Lower Secondary or below Complete Tertiary and above Lower Secondary or below Complete Tertiary and above Data - Baseline Model Secondary Secondary 15-29 30-64 Age and Education Level Source: Own estimates based on 2017 Kosovo LFS Note: Informal and formal sector employment rate expressed as percentage of labor force Baseline Results Model matches unemployment rate by age and education Unemployment rate decreasing with education for old workers, but staggering high for young Mean Unemployment Rate by Educational Level and Age 80% 70% 60% 50% 40% 30% 20% 10% 0% Baseline Gymnasium/Complete Gymnasium/Complete Upper secondary -Vocational Upper secondary -Vocational Lower Secondary or below Lower Secondary or below Complete Tertiary and above Complete Tertiary and above Data - Baseline Model Secondary Secondary 15-29 30-64 Age and Education Level Source: World Bank estimates based on 2017 Kosovo LFS Baseline Results Model approximates well the distribution of wages by sector Kernel Density of Log-Wages, By Sector- Model (Simulated) vs Data 1.5 1 .5 0 0 1 2 3 4 x Informal-Data Informal-Model Simulated Formal-Data Formal-Model Simulated Baseline Results Low-Wage incidence-Defining the thresholds Poverty-Level Wage is defined as the minimum wage per hour needed to lift a family out of certain size out of poverty, assuming one person working full-time (8 hours a day) Equivalent Poverty Line Per Household per day Poverty Line per Average capita per day Average Household Size (2011 PPP Family of Family of Household among the Poor Dollars) Family of 5 Four three Size (5.1) (5.8) 1.9 9.5 7.6 5.7 9.8 11.0 3.2 16 12.8 9.6 16.4 18.5 5.5 27.5 22 16.5 28.3 31.7 Poverty Level Wages (2011 PPP dollars) Poverty Line per Average capita per day Average Household Size (2011 PPP Family of Family of Household among the Poor Dollars) Family of 5* Four three Size (5.1) (5.8) 1.9 1.2 1.0 0.7 1.2 1.4 3.2 2.0 1.6 1.2 2.1 2.3 5.5 3.4 2.8 2.1 3.5 4.0 Baseline Results Model approximates well low wage incidence Low Wage Incidence: This measure represents the percentage of individuals with labor earnings below poverty level-wages. Calibration: Data-based vs. Simulated Statistics Variable* Model Data Low wage incidence 4.0$ per hour Poverty -Level Wage is Pf 0.11 0.13 defined as the minimum Pi 0.46 0.43 wage per hour needed to P 0.18 0.18 lift a family out of certain Low wage incidence 2.3$ per hour size out of poverty, Pf 0.00 0.00 assuming one person Pi 0.00 0.00 working full-time (8 P 0.00 0.00 hours a day) Low wage incidence 1.4$ per hour Pf 0.00 0.00 Pi 0.00 0.00 P 0.00 0.00 Low wage incidence 2/3 median Pf 0.10 0.11 Pi 0.44 0.39 P 0.17 0.16 Baseline Results The model matches reasonably well the consumption poverty rates using an auxiliary regression 2017 Poverty Rates (%), by Poverty Line – Model vs Data 25 21.6 20 18.9 15 Percent 10 5 2.7 3.1 0 $1.90 $3.20 $5.50 Poverty Line (USD Dollars a Day at 2011 PPP) Data Model Baseline Results Why is overall unemployment so high? Job destruction rates are similar to other countries in the region… Quarterly Job Destruction Rates 0.025 0.020 Jobs per quarter 0.015 0.010 0.005 0.000 Source: For Kosovo and North Macedonia, WB estimates based on SIMLAB model. For OECD Countries, Hobijn & Sahin, 2007 NY Fed Baseline Results Job Destruction Rates But some groups experienced significantly larger job destruction rates… • In the informal sector, job destruction rates among old increase sharply with education • Older and highly educated workers tend to retain their informal-sector jobs for relatively short time Quarterly Job Destruction Rates, old (15-64) 0.5 Jobs per Quarter 0.4 0.3 0.2 0.1 0.0 Lower Secondary or 2-3 y. of Upper Complete Upper Complete Tertiary and below secondary Secondary (4+ y) above Informal Formal Baseline Results Job Destruction Rates • Similar pattern among the young but even more pronounced. • Less educated young workers tend do retain their formal-sector jobs for a long- time, but among more highly educated workers the pattern is reversed. Quarterly Job Destruction Rates, young (15-24) 0.5 0.4 0.413 Jobs per Quarter 0.4 0.3 0.312 0.3 0.2 0.2 0.1 0.1 0.0 Lower Secondary 2-3 y. of Upper Complete Upper Complete Tertiary or below secondary Secondary (4+ y) and above Informal Formal Baseline Results Why is overall unemployment so high? The high levels of unemployment observed are explained by three factors: i) among low educated workers, the high levels of job destruction in the formal sector, especially among young ii) among highly educated workers, the high levels of job destruction in the informal sector, both for young and old workers iii) the low levels of job creation in the formal sector at all educational levels. Baseline Results Job Finding Rates In the baseline model, contact/job finding rates rates are extremely low, suggesting the frictions and asymmetries played an important role explaining poor labor market outcomes. Job Finding Rate m(θ) • Estimate rate of arrival of 0.06 jobs per quarter (extremely low job finding rate). • Since the unit of time is a quarter, it takes 16 quarters (4 years) for a worker to make a contact that can lead to a job. Low Matching Efficiency 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 Kosovo 0.06 Italy Macedonia 0.10 Belgium Portugal Job Finding Rates Ireland Baseline Results Spain Netherlands Greece Slovak Republic Hungary France Germany Poland Czech Republic For OECD Countries, Hobijn & Sahin, 2007 NY Fed Luxembourg Denmark Quarterly Job Finding Rates United Kingdom Switzerland Finland Austria Australia Source: For Kosovo and North Macedonia, WB estimates based on SIMLAB model. Japan New Zealand Sweden Canada Iceland Norway United States Baseline Results Estimated Parameters: Job Creation • Most young workers accept almost all formal or informal job opportunities that arrive, regardless of their education level. • Older workers with high schooling levels are less likely to accept informal-sector offers Probability of Accepting a Job Offer 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 Lower Secondary 2-3 y. of Upper Complete Upper Complete Tertiary Lower Secondary 2-3 y. of Upper Complete Upper Complete Tertiary or below secondary Secondary (4+ y) and above or below secondary Secondary (4+ y) and above 15-29 30-64 Informal Formal Baseline Results Key determinants of high unemployment This evidence suggests that: • Lack of vacancies and low contact/ job finding rates, rather than high reservation wages are critical components behind the high unemployment • An exception are the older and educated workers, who seem to have higher reservation wages and are less willing to take informal-sector opportunities • Job destruction rates among uneducated young are unusually high; they cannot retain their formal-sector jobs for long ▪ Comments 5. Counterfactual Policy Simulations Illustration: the Case of Kosovo Shapley Decomposition: Macro Targets and Projections • Time Horizon: 13 years to get to new steady state (2017-2030) • Projected Working Age/Population Cumulative Growth: 0.24% Source: UN Population Projections • Target Cumulative Growth in LFP: 40.19% • Target LFP in 2030: 60 percent • Current LFP in 2017: 42.8 percent • Target Annual GDP Growth Rate: 4% • Target Annual GDP per capita Growth Rate: 3.83% ▪ Comments Policy Experiment 2 Increase educational attainment for all: • Half of lower secondary move to upper secondary • Half of upper secondary move to tertiary Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64) Experiment 2 Variable Baseline (Increase education among all workers) Aggregate employment and unemployment rates u 0.44 0.43 ni 0.11 0.02 nf 0.45 0.55 Labor market tightness and Informal sector size θ 0.64 0.53 φ 0.24 0.05 Measures of wage dispersion µ(lnw)f - µ(lnw)i 0.40 0.43 σ(lnw)f / σ(lnw)i 1.39 1.40 σ(lnw) 0.38 0.36 mean employment duration µ(t)i 5.33 5.26 µ(t)f 11.05 11.96 Cumulative growth Productivity 11.62 Employment Rate (Employment /Labor Force) 1.50 Labor Force/Working-Age 40.19 Working-Age/Total Population 0.24 Simulated GDP per capita 53.54 Target GDP per capita 63.00 Compound annual growth Simulated GDP per capita 2.71 Target GDP per capita 3.83 Error 1.12 Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64) Aggregate Effects • This human capital policy induces some ‘formalization’: employment shifts from the informal to the formal sector, with a small decrease in unemployment • Unemployment decrease by 1 pp • Informal Sector size decrease by 9 pp • Formal Sector size increase by 10 pp Having more education means that, on average, workers can find formal-sector jobs easier, so they become pickier when it comes to accept prospective matches, and as a result, informal-sector firms are less willing to post vacancies. Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64 years) Macro Effects • The labor markets are less tight •  decreases, since there are less vacancies overall relative to job seekers. • The productivity gains generated from this policy (primarily from high productivity in both sectors but also from the movement of workers to a more productive sector) are large • Cumulative 11.6 percent over the 13 year period • A small employment rate growth and a large productivity growth lead to moderate GDP per capita growth. • 2.7 percent vs target 3.8 percent Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64) Compositional Effects More workers with complete tertiary sort into the formal sector; therefore, formal-sector workers tend to be more educated on average and more productive Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64) Compositional Effects While more workers with vocational school are absorbed in the informal sector Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64) Distributional Effects Higher wages in both sectors (mostly formal), on average, due to more workers with higher educational attainment and productivity Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64) Distributional Effects Incidence of low wages is significantly reduced in both sectors Low Wage Incidence, 4 $ USD per hour 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Pf Pi P Benchmark Experiment 2 Poverty rate (at each poverty line) is reduced significantly as a result of this human capital policy Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64 years) Distributional Effects A significant fraction of low educated workers (in particular, workers with lower secondary or below (old and young) become unemployed Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64 years) Distributional Effects Formalization happens along the distribution.. Counterfactual Policy Simulations Experiment 2: Increase educational attainment for all (15-64 years) Distributional Effects Informality drops, with larger decreases among uneducated old and young ▪ Comments Policy Experiment 8 Increase matching efficiency (job finding rates) by 20 percent Counterfactual Policy Simulations Experiment 8: Increase matching efficiency by 20 percent Experiment 8 (Increase matching Variable Baseline effiiciency A by 20%) Aggregate employment and unemployment rates u 0.44 0.36 ni 0.11 0.18 nf 0.45 0.45 Labor market tightness and Informal sector size θ 0.64 0.92 φ 0.24 0.36 Measures of wage dispersion µ(lnw)f - µ(lnw)i 0.40 0.39 σ(lnw)f / σ(lnw)i 1.39 1.39 σ(lnw) 0.38 0.37 mean employment duration µ(t)i 5.33 5.30 µ(t)f 11.05 11.13 Cumulative growth Productivity -2.73 Employment Rate (Employment /Labor Force) 13.36 Labor Force/Working-Age 40.19 Working-Age/Total Population 0.24 Simulated GDP per capita 51.06 Target GDP per capita 63.00 Compound annual growth Simulated GDP per capita 2.62 Target GDP per capita 3.83 Error 1.21 Counterfactual Policy Simulations Experiment 8: Increase matching efficiency by 20 percent Aggregate Effects • Reducing friction and mismatches lead to a large reduction in unemployment: large fraction of unemployed workers move to the informal sector • Unemployment decrease by 7 pp • Informal Sector size increase by 7 pp • Formal Sector size change very little Higher job finding rates means that, workers can find jobs easier, so they become pickier when it comes to accept prospective matches. More outside options increase reservation wages. Informal sector firms offer more vacancies to attract “pickier” workers. Counterfactual Policy Simulations Experiment 8: Increase matching efficiency by 20 percent Distributional Effects Higher formal and informal-sector wages, on average, due to more educated and productive workers in both sectors Counterfactual Policy Simulations Experiment 8: Increase matching efficiency by 20 percent Distributional Effects Incidence of low wages is significantly reduced in both sectors Low wage incidence 4.0$ per hour (2011 PPP Dollars) 50 40 30 Percent 20 10 0 Pf Pi P Baseline Experiment 8 Poverty rate (at each poverty line) is reduced significantly as a result of improved LM outcomes (employment and wages) at the bottom Counterfactual Policy Simulations Experiment 8: Increase matching efficiency by 20 percent Distributional Effects Unemployment is significantly reduced among all workers, but particularly among the uneducated young Counterfactual Policy Simulations Experiment 8: Increase matching efficiency by 20 percent Macro Effects • The labor markets are tighter •  increases, since there are more vacancies overall relative to job seekers. • The productivity losses generated from this policy are small • Cumulative -2.7 percent over the 13 year period • Despite a large employment rate increases, negative productivity growth and the projected population decline lead to small GDP per capita growth. • 2.6 percent vs target 3.8 percent Counterfactual Policy Simulations Summary • The results of the counterfactual policy experiments show that human capital policies induce some ‘formalization’: employment shifts from the informal to the formal sector, with a small decrease in unemployment. • Formalization policies, when implemented alone, are not sufficient to generate growth, since productivity losses can be large and in some cases, reducing the size of the informal sector may lead to higher unemployment • To this extent, formalization policies must be combined with human capital policies to achieve significant productivity gains, robust economic growth and substantial reduction of incidence of low wages Counterfactual Policy Simulations Summary • Policies affecting matching efficiency are fundamental to decrease unemployment, considering the high level of frictions and information asymmetries in the Kosovo labor market • With this policy, unemployment decrease substantially among all groups, especially among the uneducated young • Incidence of wages below poverty-level wages is also reduced • To decrease overall and youth unemployment, matching efficiency policies are critical ▪ Comments 5. Simulation Tools Simulation tools SIMLAB: Development of Excel Tool • USERS: Internal WB users who may require detailed and expansive set of policy options, but do not have modeling expertise or MATLAB installed in its computers. • INTERFACE: Excel as main interface; user discuss with SIMLAB team a set of pre-selected policy options. • MECHANICS: MATLAB codes with pre-selected policy choices run by SIMLAB team; outputs being automatically pulled back into Excel in standardized format and delivered to the WB users. • REPLICABILITY: SIMLAB team worked towards improving and systematizing an early pilot undertaken in the context of Kosovo, Ecuador and North Macedonia to ensure that this kind of product can be offered across countries. Simulation tools Development of Excel-based application Simulation tools Development of Online tool USERS: Internal WB users and External INTERFACE: Available through online portals where the user can input alternative policy choices with no programming or modeling expertise required and obtain results in real time MECHANICS: Simulations run in R in the background and return standardized results in the online portal. SECURITY: can be completely public or operate in a secure environment and do not require purchase of statistical software. Simulation Tools Other Labor Simulation tools OTHER LABOR SIMULATION MODELS Sorbonne and Paris University- Worksim: Model to simulate the French Labor Market with search, multi-jobs firms and bounded rationality http://worksim.lip6.fr/ Economic Policy Institute -Minimum Wage Simulation Model https://www.epi.org/publication/minimum-wage-simulation-model- technical-methodology/ Penn Wharton Budget Model (PWBM) – Dynamic OLM to study tax reforms https://budgetmodel.wharton.upenn.edu/our-model-0 Simulation Tools Development of interactive web app from R The online microsimulation tool for Kosovo is available in the internet using the link below: https://datanalytics.worldbank.org/content/389/ Future Work or Possible extensions • Improvement in data quality • Wages collected in brackets • Vacancy data • Auxiliary models that could be integrated into the core model to tailor the specific structural parameters to the policy intervention • Reduced-form approaches (correlations) • Elasticity scenarios based on historical data or previous studies Examples • Range of impacts of specific ALMP or policies tackling geographic mobility on matching efficiency/job finding rates • Impact of education investments on schooling • Impact of reduction in regulatory burden on formal-sector vacancies • Cost-benefit analysis of alternative policy options • Other extensions? ▪ Comments Thanks! For more information, visit our SIMLAB Website: https://worldbankgroup.sharepoint.com/sites/Poverty/Pages/ToolsSIMLAB-07162019- 230742.aspx (only available in the intranet) or contact us directly at: mrobayo@worldbank.org aoviedo@worldbank.org