Using Microdata for Strategic Human Resource Management and Fiscal Planning in the Public Sector Rafael Alves de Albuquerque Tavares1 , Daniel Ortega Nieto2 , and Eleanor Florence Woodhouse3 1 ao Paulo University of S˜ 2 World Bank 3 Department of Political Science, University College London December 14, 2021 Abstract We present a microdata-based approach for governments to improve their strategic human resource management and fiscal planning. Using a series of examples from Latin American countries, we demonstrate how some basic statistics created using payroll and human resource management data can help policy-makers gain insight into the current and future state of their government’s wage bill. We argue that this constitutes an important first step towards tapping the potential of existing bodies of payroll and human resource microdata that are currently underused. We believe that our approach can help policy-makers make difficult decisions by breaking down the causes of problems and putting numbers to the ways in which certain policy choices will translate into longer term consequences. ∗ Authors’ names are listed alphabetically. This article is based on technical support provided to several governments across Latin America. The team was led by Daniel Ortega Nieto. Our thanks go to Vivian Amorim, Paulo Antonacci, Francisco Lima Filho, Sara Brolhato de Oliveira, Alison Farias, and Raphael Bruce for their part in the work presented here. Practitioner Points • Data collection practices: We recommend that where possible governments centralize their human resources (HR) data collection systems and render such data accessible to insights teams. If such data do not exist, even in a disparate fashion, we strongly advise governments to start to collect, in a centralized manner, payroll and human resource management information system (HRMIS) microdata. We advise governments to make these data public where possible (anonymizing the data, naturally) to improve transparency. • Fiscal planning: We advocate for the analysis of HR data to be better integrated with fiscal planning. To be able to leverage payroll and HRMIS microdata, govern- ments must encourage civil servants from the treasury and HR department(s) to collaborate more closely. This could be achieved through allocating dedicated por- tions of civil servant workload to the task of sharing and analysing data or creating dedicated interdepartmental roles to push forward and undertake the collection and analysis of payroll microdata for strategic human resource management (SHRM). • Service delivery: By better integrating HR data and wage bill planning, policy- makers can improve service delivery to citizens. For example, projections of which categories of public servants will retire or transfer across posts allows managers to identify where additional resources will be required to ensure a continuity of service provision. This logic can be extended to the integration of personnel data with the wider dataverse available to the policy-maker. For example, using data on demographic changes amongst citizens allows for predictions of changes in service demands. The interaction of such analytics on the demand and supply sides of service delivery allows policy-makers to use their resources intelligently. • Insulation of SHRM and fiscal planning: Political considerations can impede the implementation of successful SHRM and fiscal planning. We recommend that governments insulate certain aspects of planning offices’ work from the ebb and flow of politics. This could come hand-in-hand with our second practitioner point, 3 to carve out explicit portfolios or roles dedicated to collecting and analysing HR microdata, by ensuring that such work is undertaken by public servants reporting to a independent agency, rather than to a minister. 4 1 Introduction In this chapter, we offer policy-makers ways to use human resources (HR) microdata to improve strategic human resource management (SHRM) and fiscal planning. More specifically, we present practical examples of how to use human resource management information and payroll system (HRMIS) data to strengthen wage bill projections, gain better insights into the dynamics of the public sector labor market, and strengthen evidence-based personnel policy. Our approach offers ways to tap the potential of HRMIS data that are widely available but underused. The types of analysis that we propose can support public sector managers in their decision-making by rendering explicit some of the consequences of key human resource management (HRM) choices and simulating what distinct scenarios might look like. The approach we describe is the use of administrative data related to individual employment and compensation to model the dynamics of the public sector workforce and its associated costs. By taking an analytical lens to the administrative data the public sector holds on its employees, that data becomes a means of better understanding characteristics of public administration. This includes determining simple statistics such as the ratio of pay and allowances across distinct groups of employees, the different job placements and training opportunities secured by officials across time and the institutional environment, and extrapolations of core variables such as the wage bill under current laws and regulations. With these generally straightforward statistics - that any government with a HRMIS should have access to1 - significant improvements can be made to how to address potential HRM shortcomings and related fiscal issues: strategic workforce planning, replacement rates, salary inequalities within and/or across government agencies, distribution of pay for performance benefits, retirement of personnel, and projections of payroll costs, among others. Data analytics based on personnel data have proliferated in recent years and enable 1 The IMF, in fact, estimates that over 130 countries report comprehensive government finance statistics and that, on average, countries have about 25 years of data at their disposal (Gupta et al., 2016, p.11). 5 organizations to understand analytics across the HRM cycle (Davenport, 2019) – from the attractiveness of distinct positions advertised by the organization (measured by the number of applicants) to diversity and inclusion (measured by, for instance, ethnic diversity in different ranks of hierarchy in the organization), to name just a few. Yet, as outlined in chapter X (=HRMIS chapter), many government organizations lack HRMIS systems with which to register this data. In this chapter, we thus limit the HRMIS analysis to personnel data which is often more readily available and registered by governments – such as age (by registering date of birth) and gender – while acknowledging that this only presents a small fraction of HRMIS data analytics which can be analyzed with more widely available data. 1.1 Common Sources of Microdata Two key terms that we will be using throughout the chapter are the government wage bill and payroll and HRMIS microdata. It is important to have a clear idea of what they are moving forward. The government wage bill is the sum of wages and salaries paid to civilian central government and the armed forces. Wages and salaries consist of all payments in cash (not in kind) to employees in return for services rendered, before the deduction of withholding taxes and employee pension contributions. Monetary allowances (e.g., for housing, transportation) are included in the wage bill, whilst generally pensions are not (World Bank, 2021c). Payroll and HRMIS microdata, instead, are two separate data sources that we leverage in our analyzes. Both capture individual-level information about employees and should be easily available to most governments. Payroll data include information pertaining to one’s contract (job position, contract type etc.), one’s personal details (date of birth, national insurance number, address etc.), salary and tax details (amounts and dates of payments, tax codes etc.), and leave, holidays and benefits. These data are generally collected by the HR or finance department that administers the salaries of public employees. As such, these data are automatically updated as they must reflect promotions or changes in role or leave allocations. 6 HRMIS data, instead, can be used to enrich payroll data in that they also capture information such as an employee’s gender or educational qualifications or prior professional experience. However, they tend to not be updated in the same way as payroll data because they are usually taken as a snapshot at the recruitment stage and, thus, capture an employee only at a single point in time (i.e., if the employee earns a degree after having started a position, this will not necessarily be reflected in the HRMIS data). When we refer to HR data more broadly, we refer to this combination of payroll and HRMIS data for active (or currently employed) civil servants2 . Whilst many governments have access to data collected via their HRMIS, they sometimes struggle to extract them and use them to their full potential. This can be due to a number of issues, including outdated HRMIS or analytical capacity, HR departments being decentralized, meaning that central government administrators only have access to partial data, or lack of long- term strategic HR planning. In Section 3, we offer some insights into how to get such data into shape for analysis and, in Section 4, we present how to undertake simple and effective analyzes using said data to improve SHRM. 1.2 Capitalizing on Government Microdata We focus on SHRM and the government wage bill for a number of reasons. First, the wage bill has considerable fiscal impact as it represents a significant portion of government spending, around one fifth of total spending according to the International Monetary Fund (IMF) (Gupta et al., 2016, p.2) or around 9 to 10 percent of GDP and roughly a quarter of general government expenditures according to the World Bank (World Bank, 2021b, p.8). Second, it is likely that, across the globe, in the coming years and decades pressures on wage spending will increase as “[a]dvanced economies are facing fiscal challenges associated with ageing populations while also needing to reduce high public debt levels [and] [e]merging markets and low-income countries have pressures to expand public service coverage in the context of revenue and financing constraints and the need for higher public 2 Some governments also gather and record data on pensioners and survivors. Having this additional set of data can useful, especially to improve the government’s understanding of retirees’ profiles and the overall fiscal impact of pensions. Given that this subject opens a whole set of new analyzes, we do not comprehensively discuss the use of pensions data in this chapter. 7 investment” (Gupta et al., 2016, p.2). Thus, in order to ensure that they are able to continue to deliver essential public services when facing increasing financial constraints, governments must invest in fiscal planning and SHRM. The approach we propose allows government organisations to leverage their HR data better and make use of evidence for decision-making. Such a strategic use of HR data can also have a significant fiscal impact, helping to avoid short-termism and here-and-now pressures that may cast a long shadow over government organisations’ ability to undertake their work and offer the best services to citizens under the budget constraints faced by government. A good example to illustrate the potential of using payroll microdata to provide empirical evidence for the pros and cons of policy decisions is a case in the Brazilian state of Alagoas. Here, the estimates of a decreasing pupil-per-teacher ratio helped to inform the government’s decision to recruit fewer teachers whilst maintaining the quality of the public education delivered to its citizens by opening up fiscal space to better provide other, more needed public services. One of the major advantages of taking an analytical lens to SHRM and wage bill data is that it supports governments to jointly improve workforce and fiscal planning in a coordinated way. These two aspects of government work should occur in tandem, but in practice this is rarely the case. Workforce planning “is a core HRM process that helps to identify, develop and sustain the necessary workforce skills [...] [it] ensures that the organisation has the right number of people with the right skills in the right place at the right time to deliver short and long-term organisational objectives” (Melchor, 2013, p.7). The ultimate goal of public sector workforce planning is to optimize the number and type of staff employed and the budget of the department or government in question. By the ‘type’ of staff, we mean their professional skill set: are they able to contribute to completing the mission of the organisation they serve? The identification of the needs of a specific organisation and the human resources required to achieve its goals is at the heart of strategic workforce planning (Kiyonaga, 2004; Selden, 2009; Jacobson, 2009): “[a] goal of workforce planning is to identify the gap between those needs and the available labor supply for government to continue providing quality services and fulfill its mission” 8 (Goodman et al., 2015, p.137). Fiscal planning, instead, refers to the way in which governments use their spending and taxation to influence the economy. As such, it can be improved by developing a better understanding of when certain groups of employees are going to be hired or retire, for example, allowing for more accurate revenue forecasting which influences the budget approved by the government. One area that the IMF has identified as important for improving fiscal planning is precisely the strengthening of links between wage bill management - specifically wage determination processes - and fiscal frameworks (Gupta et al., 2016, p.2). 1.3 Strengthening Traditional Approaches One additional application of our microdata-driven approach is to help to bridge the gap between macro analysis and traditional functional reviews, two common approaches used to date for the analysis of the government wage bill and the distribution of work functions across the civil service, respectively. The former relies on macro-level analysis that leverages the use of indicators such as the wage bill as a share of GDP and government employment per capita to gauge the appropriate size and cost of civil service. By relying on macro indicators, these analyzes have often led to simplistic policy prescriptions in the context of fiscal crises. The latter strain of analysis relies on functional reviews. Using mostly legal documents, regulations, and interviews, these reviews scrutinize the goals, tasks, and resources of units inside the government to improve efficiency and effectiveness. Functional reviews, thus, have multiple goals, but generally aim to assess how work is distributed across the civil service and to identify potential duplication of work through the functions performed by different departments. The analysis may produce results that are not integrated with an overarching strategy of reforming the civil service based on fiscal constraints. By undertaking microdata analyzes, one can complement functional reviews by not only looking at government functions, but also gaining greater insight into other relevant dimensions of government organisation, such as staffing and competencies. To illustrate, 9 if one undertakes a functional review and discovers that two departments perform similar functions, a parallel microdata-powered analysis can identify the distribution of compe- tencies across the two departments. Perhaps one department has a natural advantage in taking full responsibility for the function given the stronger staff strength they have. Or perhaps there needs to be a redistribution of staff to more effectively distinguish the roles and activities of the two departments. Micro-level analysis can be used to help to reconcile and to complement the fiscal- oriented nature of macro analysis and the flexible and detailed aspect of functional reviews. This can be done through the use of some simple descriptive statistics, such as the drivers of payroll growth (variation in total payroll, wages, and number of employees), the distribution of the workforce according to levels in the career ladder, progressions and promotions over time and how much they cost, amongst others, and via a model-based simulation of the wage bill, with the fiscal impacts of policies that improve and consolidate wage bill spending. One contribution that our chapter makes is to demonstrate some of the potential uses of and synergies between payroll and HRMIS data. By breaking down data silos, governments can start to better leverage data that is already at their disposal to gain insights into how to manage certain processes, such as adjusting the wage bill and improving fiscal planning. In short, our chapter aims to lay out a practical, practitioner-friendly approach to the government wage bill that can improve SHRM and fiscal planning with relatively little technical expertise and data that should be accessible (with a relatively low cost of extraction) to any government with a HRMIS. It offers significant advantages in terms of using the untapped potential of lakes of payroll and HRMIS microdata and helping governments to use evidence in order to navigate difficult policy decisions. 10 2 Strategic Human Resource Management and Fiscal Planning The public administration literature on SHRM focuses on identifying how it is used across different levels of government (Choudhury, 2007; Jacobson, 2010; Goodman et al., 2015), evaluating the effectiveness of different types of SHRM (Selden and Jacobson, 2009; Selden, 2009), and which factors influence the successful implementation of SHRM strategies (Pynes, 2004; Goodman et al., 2015). However, it is widely recognized that there is a paucity of empirical research on public sector SHRM (Choudhury, 2007; Goodman et al., 2015; Reitano, 2019), with much of the existing literature being normative in nature, relying on small samples, or being significantly dated. Moreover, the extant literature has a strong US focus, with little to no evidence from the rest of the world.3 Broadly, SHRM and wage bill data is underused as a source of analytics data for better understanding the characteristics and nature of the public administration and public service. One central finding of the literature that does exist is that many local governments do not have workforce plans in action (Jacobson, 2010). In their survey of the largest US municipal governments, Goodman et al. (2015) find that “[v]ery few local governments make use of comprehensive, formal workforce plans” [p.147].4 This is confirmed by other studies focusing on specific geographical regions, such as Jacobson (2010) or Frank and Zhao (2009). Local governments have been shown to demonstrate a lack of the technical know-how and resources required to undertake SHRM (Choudhury, 2007; Jacobson, 2010; Melchor, 2013). Small local governments, in particular, often lack the fiscal, professional and technical expertise to be able to innovate successfully (French and Folz, 2004). As such, local governments may shy away from more complex econometric approaches to processes such as budget forecasting because they lack the know-how to do so (Frank and Zhao, 2009; Kavanagh et al., 2016). This is precisely where our approach comes into its 3 If not for limited analyzes looking at very specific issues, often from the healthcare sector. With the notable exception of Colley and Price (2010) who examine the case of Queensland public service. 4 47% of the human resource management professionals they surveyed reported engaging in little or no workforce planning for their municipalities and only 11% reported that their municipalities had a centralized, formal workforce plan (Goodman et al., 2015, p.148). 11 own. With very few, simple statistics that any public organisation with a HRMIS should have access to, local and national HR departments can make a marked improvement to the use of their SHRM data. Although the issue of a lack of capacity for SHRM seems to be most acute at the local level, it has also been documented in national governments. The Organisation for Economic Co-operation and Development (OECD) describes how its member states have “experienced problems with developing the necessary institutional capacity to engage in workforce planning both at the level of the central HRM body and the budget authority, and at the level of HR departments, professionals and front line managers” (Melchor, 2013, p.15). Strategic human capital management was identified by the US General Accounting Office (GAO) in 2001 as a government-wide high-risk area, given that many agencies were experiencing “serious human capital challenges” and the combined effect of these challenges placed “at risk the ability of agencies to efficiently, economically, and effectively accomplish their missions, manage critical programs, and adequately serve the American people both now and in the future” (GAO, 2001a). Strategic human capital management remains ‘high risk’ to this day (GAO, 2021) and is proving difficult to improve upon, with “[s]kills gaps [...] identified in government-wide occupations in fields such as science, technology, engineering, mathematics, cybersecurity, and acquisitions [...] [and] emerging workforce needs in the wake of the COVID-19 pandemic” (GAO, 2021). As such, simple timely ways to improve upon SHRM - such as the approach that we propose - are urgently needed. Another important obstacle to successful SHRM and fiscal planning highlighted by the existing literature are political considerations. Successful SHRM requires support and planning from top management, given that data have to be systematically collected and analyzed over long periods of time. If elected figures are more interested in satisfying concerns in the ‘here and now’ and are unwilling to invest in longer-term HRM and fiscal strategies, this can pose a significant challenge. This is especially true in smaller, local governments where leadership tends to be more centralized and informal and where frequently no separate personnel departments exist (Choudhury, 2007, p.265). Thus, local 12 governments appear to be more susceptible to a lack of long-term planning in that they are more likely to lack the technical know-how to be able to do so or to face direct political pressures (Wong, 1995; Kong, 2007). It seems to be especially important, then, to take into consideration the nature and size of government when examining SHRM (Reitano, 2019). As Choudhury (2007) notes, “the conditions of effective human resource management at the federal, state, or large urban levels often are not a good fit for smaller jurisdictions” [p.265]. That being said, we believe that our approach can cut across different levels and sizes of government given that it relies on data that should be widely available to small and large governments alike. The extant literature has also paid a significant amount of attention to what Goodman et al. (2015) refer to as the ‘perfect storm’ of “human capital crisis that looms for local governments due to the number of employees who will be eligible for retirement or early retirement in the near future offers significant opportunity for the use of workforce planning to help with forecasting the labor pool and fine tuning recruitment efforts” [p.147]. Such a storm is still brewing in many countries around the world, both at the local and national levels. A significant number of studies explore the issue, which already in the early 2000s was becoming evident with predictions that over 50% of (US) government senior management would retire as the baby boomer generation came to retirement age (GAO, 2001b; Dychtwald et al., 2004; Kiyonaga, 2004; Wilkerson, 2007; Jacobson, 2010; Pynes, 2009). Nowadays, the issue of retirement and the subsequent talent shortage due to a smaller pool of younger public officials being available to replace retiring officials is aggravated by significant budget constraints in the public sector. Agencies are “freezing recruitment and not replacing employees who retire. The problem [being] that countries continue cutting budgets without scaling back agencies’ and ministries’ missions, compromising the ability to serve the public” (Melchor, 2013, p.15). This makes SHRM all the more important as governments need to use their available resources as wisely as possible, to continue to deliver essential services to the public. Another obstacle to successful SHRM that has been identified by the existing literature is a lack of adequate data (Anderson, 2004). For example, in the empirical context 13 of Queensland, Australia, Colley and Price (2010) argue that, in this case, there was “inadequate workforce data to support workforce planning and thereby identify and mitigate workforce risks” [p.203]. Several other studies echo this finding that public organisations in many countries find it difficult to obtain an accurate picture of the composition of their workforces (Rogers and Naeve, 1989; Pynes, 2004; OECD, 2007). Colley and Price (2010) note that “[t]here is general agreement in the public service HR literature that the ideal is a centralised whole-of-service database to meet the common workforce planning needs of agencies [...] However, establishing such databases is time consuming and costly, which limits its appeal to an incumbent government focused on short term budget and election cycles” [p.204]. Again, then, we see that political short-termism can obstruct successful SHRM before, even, one considers the lack of technical expertise or time/capacity that HR professionals may suffer from (as we saw earlier in this section). Our proposed approach speaks to this obstacle to SHRM in that it requires only very few, basic statistics with which to better leverage HR data. In addition to the direct challenges of enacting SHRM, there are also a series of important ways in which SHRM and fiscal planning interact. In order to enact more effective and sustainable fiscal planning there are numerous ways in which the management of government wages can be improved upon and can better take into consideration fiscal concerns. For example, the IMF notes that wage bill increases have been shown to be associated with worsening fiscal balances: “rather than crowding out other items in the budget, increases in the wage bill have on average been associated with increases in other government spending and with a deterioration of the overall balance” (Gupta et al., 2016, p.14). As such, policy-makers should be especially wary of increasing the wage bill when the budget is tight. Furthermore, if SHRM is not undertaken so as to employ the right type and amount of workers, this can have a negative fiscal impact. If there is a wage premium in the public sector, this can “increase private production costs, including wage costs, as well as result in additional ‘deadweight losses’ associated with distortionary taxation” (Gupta et al., 2016, p.15). In fact, wage penalties can also have detrimental fiscal effects in that 14 difficulty recruiting and retaining qualified workers will adversely affect the quality of publicly provided goods and services and could also contribute to corruption (World Bank, 2021b, p.8). As such, attention should be paid to ensuring that public sector salaries are calibrated with those of the private sector for comparable jobs and that they are adjusted according to broader changes in the population, society and the economy at large (Somani, 2021). Indeed, advanced economies have been found to struggle to adjust employment levels in response to demographic changes - such as the decline in school-aged children, which had led to an oversupply of teachers (Gupta et al., 2016, p.20) - which can lead to significant fiscal concerns that could be avoided with a more forward-thinking HRM strategy. 3 Payroll and HRMIS Microdata and Related Challenges Before delving into what analysis can be done with payroll and HRMIS microdata, it is important to provide further discussion of the kind of data we are talking about and the type of variables one can extract from such data sources. First, we describe payroll microdata before turning to HRMIS microdata. Payroll microdata are drawn from the administrative datasets that governments use to follow and register the monthly compensation of civil servants and their underlying items. It usually covers most of the government’s contracts with its employees and sometimes contains some demographic characteristics of civil servants and their occupational information (e.g., the department or unit where the civil servant is located, type of contract, date of entry in the civil service, etc.). In some contexts, these sets of information are collected independently by different teams. HRMIS data, instead, as anticipated in the introduction, are additional data often collected by recruitment units that can enrich payroll data with information about employees’ gender, education level, and professional sector, for example. To undertake our analyzes, we combine these two types of microdata. In Table A1, we present an example of a hypothetical combined payroll-HRMIS micro 15 dataset with the main variables (columns) and observations (lines) needed for the type of analysis we propose in this chapter. This table represents the minimum required data to be able to undertake the analyzes that we propose. Each line represents an individual and their respective contract with the government, and each of the columns points to some relevant variable for the analysis, such as the unit where the civil servant is located, age, gender, date of entry in the civil service, type of contract, etc.. An individual might have more than one contract with the government, for example, a teacher with two part-time job positions. Ideally, the database should have information about the government’s employees for the last 10 years, so that one can retrieve some variables of interest based on historical data (for example, the average number of years of service before retirement). Ideally, governments should have the aforementioned information of all their public employees readily available, but based on our experience working with several governments from Latin America and the Caribbean (LAC) countries, we know that governments face challenges when it comes to their wage bill microdata. Theses challenges can be organized in two dimensions. First, governments may not be able to collect information about all of their employees, leading aggregate figures to be potentially wrong or biased. This can happen if wage bill microdata collection is not centralized and the information of some units/departments is missing in the data. Using the example from Table A1, this would be reflected in fewer observations (lines) in the data than in the actual government bureaucracy. A second dimension relates to the number of different aspects that are being collected to describe the bureaucracy. In Table A1, these are captured in the number of columns in the dataset. For example, in a recent analysis that was undertaken in the context of project with a LAC country, the wage bill data did not have information about when public employees started their careers in the civil service, making it difficult to figure how the experience in a position, measured by the years of service, are related to wage levels and, as a consequence, the total cost of hiring a new civil servant for that position. With theses issues in mind, practitioners should be cautious about what the available wage bill microdata can tell them about the current situation of bureaucracy in aggregate terms and which aspects can be explored to provide insightful ideas for governments to 16 better manage their SHRM and fiscal planning. In Figure 1, we propose a simple wage bill microdata “quality ladder” to help practi- tioners separate good data from bad data. We organize the ladder in 5 levels, with the first level being the one with lowest quality and fifth level with the highest quality. In Level 0, there is a missed opportunity regarding HRMIS data analysis, as the minimum required data are not available (see Table A1 for a reference). This is because the information for public employees is scarce, inaccurate, inconsistent, and scattered across government units or career lines, such that almost any indicator or statistic based on such data would be wrong or biased. Statistically, it is impossible to draw inferences from incomplete data especially where there are worries that the missingness is correlated with relevant features of the underlying values of the variables in the data. To see this, you need only think of a couple of the reasons for which a government agency may not report HR microdata: because they lack the capacity or manpower to do so (in such a case, only the agencies with greater capacity would present their data offering a skewed vision of the performance of the government at large) or because they are not mandated to do so and, as such, will not spend precious resources reporting HR data (again, here drawing inferences from such data would give a misleading impression of the government at large as only the agencies with reporting mandates would provide their microdata for analysis). In level 1, some analysis can be performed for the units or careers for which there are data available. However, for the reasons outlined above, such analyzes must be applied only to the units or career lines for which data are available and very careful considerations must be made about why and how such missingness in the data is occurring. A good example of this is a situation where the wage bill data gathering is decentralized, and some government units collect data whereas others do not. For instance, if only the education and health departments can fill Table A1 with information about their employees, the analysis should be restricted to these units, and the government should start collecting data from other units of the government. In level 2, not only the basic information shown in Table A1 is readily available, but one is also able to connect these data with additional data sources and explore specific 17 features of job contracts. Using the above example, it could be the case where the wage bill data for teachers can be connected to students’ performance in standardized tests, allowing for the analysis of teachers’ productivity in the public sector. Level 3 illustrates a situation in which the information outlined in Table A1 is collected for a large part of the bureaucracy in such a way that one can undertake an aggregate analysis of the wage bill expenditures based on the microdata. In Section 5, we present an example of such an aggregate analysis, with a projection of the wage bill for future years based on the data from the Brazilian Federal Government. We would like to note that levels 2 and 3 of the quality ladder can be ranked differently depending on the objectives of the analyzes to be performed. For example, when analyzing the impact or value-added of teachers on students’ performance, having a productivity measure in the wage bill data for teachers can be especially useful. Given the fiscal nature of the analyzes undertaken in this chapter, having a wage bill data set that allows the analyst to create aggregate figures is particularly important. Because of that, we decided to rank a comprehensive data set for all civil servants without productivity measures above a data set with partial productivity measures in our quality ranking. In level 4, one can not only undertake the analysis described in level 3, but can also merge in other data sources that might be available, and connect them with the overall fiscal landscape of the government. Building on the same example used to describe level 2, one could assess both the fiscal impacts and the productivity impacts of a policy that adds a pay-for-performance scheme to teachers’ compensation based on the performance of students on standardized tests. 18 Level 4 Level 3 + additional data for all units and career lines Level 3 Minimum data required available for all units and career lines Level 2 Level 1 + additional data for some units and career lines Level 1 Minimum data required available for some units and career lines Level 0 Some information available for a group of public employees Figure 1: HR Microdata Quality Ladder Building a HRMIS that climbs the ladder described in Figure 1 can be politically costly and requires sustained investment in the technical skills that underlie data management. The benefits are the improved understanding of the public sector that such an effort provides. The next section outlines the basic analytics such databases provide the foundation for. Without the qualities outlined in Figure 1, such analytics are undermined and can be distortionary. But with a sound foundation of quality and comprehensive data collection, these descriptives can support substantial fiscal efficiencies and improved service delivery. In the country cases on which section 4 builds, these investments have paid off many times over. 4 Descriptive Statistics In this section, we present some descriptive statistics that can help policy-makers gain insight into the current and future state of their government’s wage bill. Along with each insight, we present examples from the wage bill analyzes that we undertook in different LAC countries. As aforementioned, the data that are required for these analyzes should be available to any government that has a HRMIS. That being said, we recognize that there are sometimes significant challenges to obtaining such data - especially in contexts where these datasets are not held centrally - and to getting them organized in such a way as to undertake these analyzes. We posit that there is great untapped potential in 19 the payroll and HRMIS data that governments collect and propose a way to start using these data lakes, where they exist. Where they do not exist, we recommend starting to centralize HR microdata to be able to undertake these types of analyzes. We present our proposed descriptive statistics in three groups. The first provides a general overview of the wage bill and HR management systems to give the reader a sense of how HR practices can impact upon the wage bill. The second addresses how these HR microdata can be used to identify inequalities in terms of representation within the public sector. The third then proposes a way to address some of these inequalities by adopting a forward-looking perspective to apply changes to fiscal policy to avoid such inequalities or inefficiencies occurring again in the future. 4.1 General Overview of the Wage Bill and HR Management We first address how HR management practices can impact the wage bill and offer some examples of the insights that can be gained by better exploiting payroll and HRMIS microdata. Drivers of payroll growth. Changes in wage bill expenditures can be attributed to changes in employment levels and changes in the average wages of civil servants. A wage bill increase as a result of increased employee hiring is usually accompanied by an expansion in the coverage of public services. Wage increases do not have an immediate impact on the provision of public services, but may have a medium and long-term impact on the attraction, retention, and motivation of civil servants that could enhance the productivity of the public service and lead to better service provision. Figure 2 presents a simple way of analyzing what is driving wage bill variation. By setting the starting year as a baseline, we can see in this example from the Brazilian federal government’s wage bill that most of the increase in the wage bill expenditures came from increases in civil servants’ compensation. In fact, between 2008 and 2017, spending on Brazilian federal executive personnel rose by 2.9% per year in real terms. This growth was made up of a 1.8% increase in average salaries and a 1.2% increase in the number of public servants. This kind of figure can also be applied to analyze specific sectors and career lines in the 20 government, such as the education sector and, within that, teachers. Besides that, making sector-specific or career-specific analysis is a way of providing insights with partial data, as one should be cautious when making aggregate claims from microdata if not all wage bill data is available. 130 125,9 % (December 2018 price level) 120 114,1 110 110,5 100 90 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Wage Bill Average Wages Civil Servants Figure 2: Drivers of the wage bill variation over time, Brazilian Federal Government Breakdown of the wage bill by sector. The idea of breaking down the change in the overall wage bill expenditures in changes in the number of civil servants and in average wages can also be applied to understanding how civil servants and wage bill expenditures are distributed among priority areas. Extending the analysis to the sector level can shed light on the needs and targets of the government in areas such as education, health and security. For example, in the case of the Brazilian state of Rio Grande do Norte (see Figure 3), 86% of the civil servants are distributed in priority areas, while the wage bill expenditures for these same sectors amount to 82% of the total wage bill spending. In particular, the education sector employs 41% of the public servants and accounts for 34% of the total wage bill. 21 13.7% 18.3% 34.0% 41.1% 21.6% 26.9% 23.6% 20.8% Education Health Safety Other Education Health Safety Other (a) Number of Public Employees by Sector (b) Wage Bill Expenditures by Sector Figure 3: Wage Bill Breakdown by Sector, Brazilian State of Rio Grande do Norte, 2018 Distribution of civil servants by career ladder levels. Progressions and promo- tions along career ladders are a common path for governments to connect higher labor productivity to wage increases. Based on this link between productivity and wages, we can expect that a longer tenure in the civil service reflects a knowledge gain that should equip employees with better tools with which to deliver public services. By analyzing how civil servants are distributed along career ladder levels, policy-makers can assess if the ladder structure of civil service careers reflects these increases in productivity. Ideally, we should expect to see a smooth distribution of civil servants across the different levels. In Figure 4, we use as an example the career of tax auditors in the Brazilian federal government. We can see that more than 80% of the public employees are in the final step of their career, which suggests that there may be margin for improving the design of the career structure and the requirements for progression or promotion to better reflect labor productivity gains. 22 9th level 84.0 8th level 0.1 7th level 0.9 6th level 6.1 5th level 0.1 4th level 0.1 3rd level 2.5 2nd level 2.9 1st level 3.4 0 20 40 60 80 100 % Figure 4: Distribution of civil servants by career ladder levels, Brazilian Federal Government Strict progression rules and high turnover of civil servants. On the other hand, situations where the rules for career progressions and promotions are too strict may end up leading to difficulties in retaining public employees, along with their acquired knowledge and expertise. To illustrate such a situation, we can examine the case of on ) Uruguay’s central administration, where civil servants are assigned to one scale (escalaf´ and ministry. As movements across ministries and scales are rare and can only take place with special authorization, grade progression is the only career path available for civil servants. As a result, this limited room for vertical promotions may end up hindering productivity and motivation, as well as increasing turnover. In Figure 5, we can see the share of employees that were promoted in 2019 (Figure 5a), and the turnover of employees (Figure 5b) by Ministry in Uruguay’s Central Administration. Less than 5% of employees were promoted to a higher grade in almost all ministries, while 7% of employees entered the Central Administration in 2019 and 6% exited in that same year. In some ministries 23 the exit rate was even higher than the entry rate. This high turnover can be interpreted as a sign of the challenges in retaining civil servants and it also represents a hidden cost for the government due to loss of expertise and the cost of training new staff. (a) Grade Progressions by Ministry (b) Turnover by Ministry Figure 5: Grade Progressions and Turnover, Uruguay Central Government Distribution of pay-for-performance allowances. Pay-for-performance is a useful tool to stimulate productivity in the civil service. In theory, it rewards high-performing public employees and inspires low-performers to perform better. However, there is much debate regarding the extent to which performance pay succeeds in improving civil service performance (cf. Hasnain et al. (2014)). We posit that our approach can help the policy- maker understand whether pay-for-performance is working in the context in which they work. For example, one problem that can arise is when all employees receive performance payment. In Figure 6, using data from the Brazilian federal government, we display on the x axis all careers next to each other (each vertical line represents a specific career track), and on the y axis the percentage of that profession that received a performance bonus. We show that in 2017 at least 90% of employees received performance-related payment in 164 of the 187 careers that offer such schemes. This can indicate that the pay-for-performance scheme in question is not successful in differentiating between good and bad performers. 24 100 80 60 % 40 20 0 Careers Figure 6: Distribution of pay-for-performance allowances, Brazilian Federal Government 4.2 Inequality in the Public Sector Wage Bill Having given a general overview of key features of the public service, we turn to the use of HRMIS data to understanding inequalities in the service. Such inequalities may come in different forms and have correspondingly different impacts on the efficiency or qualities of the state. Representativeness. Many government’s strive to recruit officials in a way that ensures the administration as a whole is broadly representative of the population that it serves. For example, by having personnel from across the country’s regions in rough proportion with the distribution of population across those regions. Normatively, such con- siderations are important to make given that, in a democratic setting, bureaucracies should represent the populations they serve. Moreover, empirically it has been demonstrated that more representative bureaucracies - dependent on the policy domain - can affect important phenomena such as citizens’ trust in the government and willingness to cooperate with 25 the state (see, for example, Theobald and Haider-Markel (2009); Riccucci et al. (2014); Van Ryzin et al. (2017)). Though there may be good reason for this principle to not hold strictly, HRMIS data allows the degree of representativeness of the administration to be acurately articulated and to act as the foundation of an evidence-based debate on the matter. Pay inequity. Inequality in payments in the public sector can reflect underlying differences in responsibilities or can be a sign that inconsistent compensation rules are being applied. For example, we expect the government to reward managers and high performing employees with better compensation than entry-level civil servants, but we do not expect it to award significantly different levels of compensation for employees with same attributes, same jobs, and same tenure, following the generally observed principle of equal pay for equal jobs. For the case of the Brazilian Federal Government Tax Auditors (see Figure 7b), we can see that there is a huge wage dispersion for similar workers. Gross pay can vary five-fold for workers with similar levels of experience, which is largely a result of nonperformance-related payments and is not related to base salary either. Related to this is the need for governments to devise pay schedules that incentivize officials to want to keep exerting effort to rise up the career ladder, while also being aware that equity in pay is a key issue for some official’s motivation. To measure inequality due to differences in responsibilities and career level, we can analyze the pay scale compression5 of the government’s units. A higher wage compression (a smaller wage gap between management level and entry level) is associated with greater difficulty in motivating personnel to progress up through the public service, as the increased responsibility is not adequately compensated. For example, in the case of Uruguay (see Figure 7a), the wage compression in the Central Administration is low by international standards, but varies greatly across ministries. Having a low wage compression by international standards is good for equity, but the implications for civil servants’ productivity and motivation are unclear. Low pay compression can generate positive attitudes across civil servants 5 Wage compression is generally defined as the ratio between high-earners and low-earners in a specific organization. In this chapter, we define the wage compression as the ratio between the 90th percentile and the 10th percentile of the wage distribution of the organization. 26 if responsibilities are also spread accordingly across the civil service; but it might also indicate that the salary structure is not sufficiently able to incentivize and reward workers’ efforts or workers with additional responsibilities. 75 BRL Thousands, December 2018 price level 65 55 45 35 25 15 0 10 20 30 40 50 Years of service Individual total compensation Regression line (a) Wage compression by Ministry, Government (b) Wage dispersion and tenure, Brazilian Fed- of Uruguay eral Government Figure 7: Measuring Pay Inequity Inequity of Pay Based on Increasing Wage Components. A good compensation system should allow the government to select high quality candidates and offer incentives to align each public servant’s interests with those of society. Some desirable payment system characteristics include the ability to link wage gains with skills and performance and the transparency of the wage components. Having a large number of salary components can hinder transparency and generate inequalities. For example, in the case of Uruguay’s Central Administration, there are 297 different salary components, of which 53 are “basic” and 244 are “personal”.6 Each entity has some discretion to define the compensation its employees receive, thus reducing transparency and potentially creating payment inequalities. From Figure 8, we can see that this discretion is reflected in the distribution of “personal” payments (see Figure 8b), which, differently from the distribution of “basic” payments (see Figure 8a), follows a non-standard distribution. The non-standard distribution of “personal” payments suggests both a lack of transparency and an unequal 6 The salary structure in the public administration consists of multiple salary components, grouped in “basic” and “personal” components. “Basic” payments are determined based on the specific position (plaza), which represents the set of tasks, responsibilities and working conditions associated to each civil servant, and they include sueldos al grado and compensaciones al cargo. All civil servants also receive “personal” payments, which are specific to each individual employee 27 pay structure, based on the increase of payment line items. (a) Distribution of Basic Payments, Government (b) Distribution of Personal Payments, Govern- of Uruguay ment of Uruguay Figure 8: Inequity of Pay in Wage Components Wage Inequality by Gender. Gender equality is a key indicator of progress towards making the public sector workforce more diverse, representative, innovative, and able to provide public services that better reflect citizens’ needs. According to the OECD (2019), women are over-represented in the public sector workforce of OECD countries. However, this is not true across the globe; in fact, the Worldwide Bureaucracy Indicators show that public sector gender equity is correlated with country income (World Bank, 2021a). Part of the issue lies in providing similar levels of compensation for women and men, where some systems discriminate against women. In some cases, the wage gap can discourage women from entering the civil service or applying for higher positions in the organization. In that sense, identifying potential gender wage gaps in the public sector is important to foster the diversity of public employees. In Figure 9, we analyze the gender wage gap in Uruguay’s public sector workforce. The results suggest that overall, after controlling for working hours, age, type of contract, grade, tenure and occupation, there is not a statistically significant gender gap in wages, but this varies across ministries. 28 Note: The graph shows regression coefficients and 95% confidence intervals for the interaction between the female dummy and for Ministry Fixed Effect. Each point represents the average salary difference with respect to Ministerio de Salud Publica, after controlling for worker’s characteristics. Figure 9: Gender Gap Across Ministries, Government of Uruguay While there are many other margins of potential inequality in the service, and be- tween the public service and the rest of society, these examples showcase the power of government microdata in identifying the extent and distribution of inequities across public administrations. 4.3 Fiscal Analysis Having considered what the wage bill is, how HR management can affect it, and how HR practices can affect important questions of the character of and equity within the bureaucracy, we now turn our attention to how such practices can affect the fiscal health of a polity. Setting compensation schemes such as initial wages and wage increases related to progressions and promotions are some of the key tools we have at our disposal to attract, retain, and motivate civil servants. On the other hand, they can be a cause of long-term 29 fiscal imbalance, as public sector employees usually work for more than 15 years.7 For example, careers with high starting salaries may attract qualified candidates, but when combined with slow or small wage increases related to progressions, this can lead to demotivated public employees. In such a situation, a reform that keeps starting salary levels high and increases the additional pay related to progressions/promotions may cause a situation of fiscal unsustainability of the wage bill. By understanding the fiscal impact of current career compensation schemes and of potential reforms of them, policy-makers can better manage the public sector’s HR in the long term. In Figure 10, we present examples of how these compensation features can be visualized. In the case of the Brazilian state of Mato Grosso (see Figure 10b), we find that for some of the careers the first three progressions more than double public employees’ salaries. Police chief 11 11 11 Police soldier 13 22 0 Support staff (Education) 30 4 5 Clerical (University) 4.8 Temporary teacher 20 25 0 PGCPE/PCC Careers 6.5 Environmental analyst 27 28 30 Defense (civil) 6.9 Superior level (Health) 46 20 27 Others 7.8 Technical level (Health) 25 30 40 Sector-End-Area 8.0 Support staff (education) 56 18 24 Permanent teacher 56 18 24 Physician (University) 8.5 Administrative analyst 35 35 31 Elementary Professor 9.0 Analyst (Dev. Econ. Social) 35 35 31 Higher education Professor 10.3 Technical level (Dev. Econ. Soci 30 33 40 Police 11.1 Police scrivener 42 32 34 Autarchy 11.2 Police investigator 42 32 34 Prison guard 48 35 35 Regulatory Agencies 13.4 College professor 95 30 5 Diplomacy 15.2 Control 17.6 0 50 100 150 Planning and management 20.7 Progressions (%) Justice 24.1 First Second Third 0 5 10 15 20 25 BRL Thousands, December 2018 price level (b) Wage increases related to progressions and (a) Starting Wages by Groups of Careers, Brazil- promotions, Brazilian State of Mato Grosso Gov- ian Federal Government ernment Figure 10: Career Types and Wages Besides starting salaries and wage increases, another important piece of information for policy-makers implementing strategic workforce planning pertains to when public officials retire. Getting a clearer picture of when public employees retire is of critical importance for strategic workforce planning and fiscal planning. One needs to understand who will retire and when in order to plan successfully for incoming cohorts of civil servants in terms of both their numbers and the competencies they will need to have. When 7 For example, a Brazilian Federal Government employee works for an average of 30 years before retiring. 30 large numbers of public servants are all due to retire at the same time, this can offer a window of opportunity for policy reform. For example, in the case of the Brazilian federal administration, the World Bank’s projection for retirement using 2017 data was that 22% of public servants would have retired by 2022 and 40% were expected to retire by 2030 (see Figure 11). This situation presented an opportunity for administrative reform, to restructure career systems and rationalize the number of existing civil servants in order to better plan ahead both in terms of the workforce and in fiscal terms. The use of HR microdata to undertake such analysis helped to inform the debate about a civil service reform.8 40 Number of civil servants (in thousands) 30 20 10 0 2019 2024 2029 2034 2039 2044 2049 2054 Figure 11: Retirement Projections, Brazilian Federal Government 8 encia de uma reforma administrativa”, Valor Econˆ See: “Banco Mundial aponta urgˆ omico. October th 10 , 2019 (in Portuguese). 31 5 Projections of the Wage bill In this section, we present a HR microdata-based model based on the building blocks presented in Section ??. With information about initial wages, wage increases related to career progressions, and expected date of retirement, the policy-maker can project the expected fiscal impact of civil service reforms, the design of new careers, and fiscal consolidation policies. Using counterfactual scenarios can also help governments in promoting diversity and reducing inequalities in the civil service, fostering policies and services that better reflect citizens’ needs. Payroll and HRMIS microdata represent an important tool for the analysis of HR and fiscal policies. They can help policy-makers lay out the trade-offs among competing ao the government was policy objectives. For example, in the Brazilian state of Maranh˜ seeking to understand the fiscal impacts of wage increases for teachers along with increased recruitment of police personnel. By representing graphically the relevant statistics and comparing, first, the decreasing trend of the pupils-per-teacher ratio and its effect on the demand for new teachers and, second, the levels of violence in the state when compared with peers and the ratio of policemen-per-inhabitant, led decision-makers to obtain a more realistic picture of the available employment policies. In this section, we use some of the figures from Section ?? to lay out the building blocks of a policy-oriented model for projecting wage bill expenditures. The model can help policy-makers make difficult choices more transparent by showing the real costs and benefits of potential civil service reforms. In practice, this is how we make the projections. First, we set up the HR microdata in a structure similar to that we described in Section 3 and reported in Table A1. Ideally, the database should contain payroll and HR information for the last ten years. If monthly data is not available, it is possible to use a representative month of the year.9 The wage bill data from previous years will then be used to estimate some of the parameters of the model, and the most recent month/year data will be used as a starting point for the 9 The ‘representative month’ should allow for extrapolating the monthly wage bill expenditures and number of civil servants for the whole year. 32 projections. Second, with the microdata set up, we group civil servants according to similarities in job position and/or common legal framework. The inputs of government’s HR managers is critical to this first part of the model, because setting the number of groups should both reflect the bulk of the civil service careers and allow for more fine-grained policy options. In that sense, there is no “magic number” of groups, but a context-based number. In practice, we tend to cluster civil servants in a range between 5 to 20 groups. For example, in the case of some Brazilian states, we defined seven main groups: teachers, military police, investigative police, physicians, education support staff, health support staff, and others. These groups were defined to reflect the main public services Brazilian states are responsible for: public security, secondary education, and mid to high complexity health care. In another example, for the Brazilian Federal government we defined 15 career groups, which included university professors as Brazilian public universities are mostly federal. Third, after setting the clusters of careers, we estimate some basic parameters for these groups using the microdata from previous years: the number of retirees by year for the following years, average tenure when retiring, initial wages, years between progres- sions/promotions, real increases in salaries related to progressions/promotions legislation, real increases in salaries not related to progressions/promotions legislation, and attrition rate, which is the ratio of new hires to leavers. Some of these parameters were shown in Section ??. For example, Figure 11 shows estimates for the number of retirees by year for the Brazilian Federal Government with data from 2008 to 2018. Fourth, we use the most recent month of the wage bill database and our estimated parameters to track the career path of current employees until their retirement and that of new civil servants that will replace retiring civil servants. Because of the fiscal nature of the model, the wage bill estimates tend to be less accurate for long term projections. Based on experiences with LAC governments, we recommend using at most a 10-year span for projections. Using the estimated parameters, we come up with a baseline projection, the trajectory of the wage bill expenditures with ‘business as usual’ as extrapolated from 33 the data on the past years. In other words, the expected wage bill spending if we take the same wage increases from the last years, the same expected tenure before retirement, and the same replacement rate of new civil servants per retiring employee. Finally, after having a baseline projection of the wage bill, we are able to simulate reforms that implement changes to some the estimated parameters. For example, if the government wants to analyze the fiscal impacts of a reform that increases the recruitment of teachers, we simply change the rate of replacement of the career group of teachers. In another example, if the government wants to consolidate wage bill expenditures by freezing wages for the next 2 years, we can change the parameter for salary increases that are not related to progressions or promotions. The list of potential policy scenarios includes hiring freezes or targeted pay increases for specific classes of employees. The model is meant to be flexible to adapt to the government’s needs, so policy-makers can test different reform options and hypotheses. 5.1 Example from the Brazilian Federal Government To exemplify the use of the model, in this section we present the wage bill projections of the Brazilian federal government for the 2019-2030 period that were undertaken using HR microdata from 2008 to 2018. For example, Figures 10a and 11 from Section ?? are graphical representations of the starting wages and the number of retirees by year, respectively. Figure 12 presents the baseline projection of the wage bill , and Figure 13 provides a decomposition of the wage bill projection across current and new employees. Brazil is something of an outlier amongst LAC countries in that it has very high quality administrative data meaning that it makes for a good example of the more advanced types of analyzes one can undertake with HR microdata once a comprehensive, centralized data collection system has been put in place. 34 150 140 BRL Billions (2017 price levels) 130 120 110 100 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 Wage Bill data Baseline Figure 12: Baseline Wage Bill Projection 100 0 1 6 11 14 16 18 20 21 22 23 23 23 80 60 % 100 99 94 89 86 40 84 82 80 79 78 77 77 77 20 0 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 Hired before 2018 Hired after 2018 35 Figure 13: Decomposition of Wage Bill Projection Between Current and New Employees, Brazilian Federal Government With the projection of a baseline scenario for the wage bill expenditures in the coming decade, we are able to compare it to different policy scenarios. To better organize reform options, we can separate them into pay-related and employment-related reforms. In the context of Brazil, the federal government’s main objective was to simulate reforms that could lead to fiscal savings. We presented nine policy options, two of them related to employment reforms, and the other seven related to pay policies. Based on these specific policies, we projected scenarios each with a set of pay-related and employment-related policies: • Scenario A: Replacement of 100% of the retiring employees and no real increases in salaries for 10 years. • Scenario B: Replacement of 90% of the retiring employees and no nominal increases in salaries for the first 3 years. • Scenario C: Replacement of 80% of the retiring employees and no nominal increases in salaries for the first 3 years, after that no real increases in salaries for the next 7 years. Figure 14 provides a graphical presentation of the baseline projection, along with the three outlined reform scenarios. In Scenario A, a policy of no real wage increases is implemented starting in 2019. Since the y-axis measures wage bill expenditures in real prices for 2017, the policy of correcting salaries only for inflation leads to an almost steady line in the chart. Scenarios B and C implement tighter policies, with a nominal freeze in salaries for the first three years starting in 2019, along with fewer hires of new employees to replace retiring civil servants. The bulk of the difference in savings between scenarios B and C come from the years after 2022, in which scenario B goes back to the baseline wage bill expenditures, whereas in scenario C salaries are corrected for inflation. 36 150 140 BRL Billions (2017 price levels) 130 120 110 100 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 Wage Bill data Baseline Scenario A Scenario B Scenario C Figure 14: Wage Bill Projection and Policy Scenarios To put these different scenarios in perspective and compare their effectiveness in providing fiscal savings, we show in Figure 15 the fiscal savings accumulated throughout the years with each reform scenario. In 2018, the wage bill expenditures in the Brazilian federal civil service amounted to a total of BRL 131 billion. Based on the projections of the model used in this analysis, in 2026 scenario A saves approximately 12% of the 2018 wage bill expenditures, scenario B saves 19%, and scenario C saves 24%. Besides these differences in total savings, in scenarios B and C the government achieves larger savings in the short term, while compensating with smaller savings after a few years, whereas in scenario A the total savings are spread out over the years. 37 40 35 30 BRL Billions (2017 price levels) 25 20 15 10 5 0 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 Scenario A Scenario B Scenario C Figure 15: Cumulative Fiscal Savings from Policy Scenarios Experimenting with combinations of policies before their implementation so as to understand their fiscal impact has the potential to save governments significant proportions of their wage bill. Similarly, such extrapolations can be extended to the descriptive analysis outlined in previous sections so that governments can better understand how personnel policy reform will impact the character of the public service. With enough good quality data, governments can leverage their SHRM and wage bill data to provide evidence-based planning of their fiscal expenditures and personnel dynamics into the future. 6 Conclusions We have presented a microdata-based approach for governments to improve their SHRM and to develop realistic civil service compensation and employment strategies. We have also demonstrated how such strategies can allow policy-makers to make better fiscal choices. We have used a series of examples from LAC countries to demonstrate how 38 the use of relatively basic payroll and HRMIS statistics can help policy-makers gain insight into the current and future state of their government’s wage bill. We posit that this constitutes an important first step towards tapping the potential of existing bodies of payroll and HRMIS microdata that are currently underused. We believe that our approach can help policy-makers make difficult decisions by breaking down the causes of problems and putting numbers to the ways in which certain policy choices will translate into longer term consequences. On the basis of our experience in using HR microdata for such analyzes, we have a series of practical recommendations to make. Our first recommendation pertains to the collection of the data required to undertake the analyzes we propose. Although, in theory, any government with a HRMIS should have access to these data, we know from our experience working with governments that extracting and cleaning such data can be a difficult task. As such, we recommend that, where possible, governments should centralize their HR data collection systems and render such data accessible to insights teams. If such data do not exist, even in a disparate fashion, we strongly advise governments to start to collect, in a centralized manner, payroll and HRMIS microdata. If governments are able to break down existing inter- and intra-departmental data silos and embed data analytics into their institution culture, they stand to gain a much clearer idea of - amongst many other phenomena - the composition of their workforce, how to use this workforce more effectively, and how to plan, budget, and staff for future challenges. This is a central recommendation from our experience working with these microdata. As we laid out in Section 3, the quality and coverage of the data that you have at your disposal affects the usefulness of the analyzes that you can undertake and, consequently, the power of the insights you can gain. Our second recommendation is that the analysis of HR data be better integrated with fiscal planning. Our approach can help to bridge and can complement functional reviews and macro analyzes and, as such, can reconcile the fiscal-oriented nature of macro analysis with the detail of functional reviews. For this to be effective, however, governments must encourage civil servants from the treasury and HR department(s) to collaborate more closely. This could be achieved through allocating dedicated portions of 39 civil servant workload (from both the Treasury and the HR department) to the task of sharing and analysing data in collaboration or creating dedicated interdepartmental roles to push forward and undertake the collection and analysis of HR microdata for SHRM. By better integrating HR data and wage bill planning, policy-makers can also improve the services that are delivered to citizens. Thinking back to the example we mentioned in the introduction, by incorporating demographic changes into their projections regarding how many teachers to hire (given the falling pupil-per-teacher ratio caused by lower fertility rates) policy-makers in Alagoas were able to identify an area in which to make substantial savings and to better target their HR strategy to hire different categories of civil servants that were not oversupplied. As such, the state was able to provide better quality services to its citizens by hiring civil servants in areas where greater personnel was needed, rather than in the education sector where there was an excess of teachers. Our third recommendation relates to how political considerations can impede the implementation of successful SHRM and fiscal planning. We recommend that governments, in addition to centralising HR data collection systems, seek to insulate certain aspects of planning offices’ work from the ebb and flow of politics. This could come hand-in-hand with our second recommendation, to carve out explicit portfolios or roles dedicated to collecting and analysing HR microdata, by ensuring that such work is undertaken by public servants reporting to a independent agency, rather than to a minister. All three of our recommendations pertain to how governments can better institutionalize SHRM and improve their analytical capabilities with data that should be relatively easy to collect and use. By developing a culture of centralizing and sharing such data - always anonymized and stored and shared in full respect of employees’ privacy and rights - governments can improve their ability to identify and resolve issues pertaining to the workforce and fiscal planning alike, as we have laid out. Moreover, such analyzes are simple to undertake meaning that without hiring significant numbers of data specialists, governments can leverage such data through existing staff with even minimal levels of data literacy. We hope to have illustrated the benefits of combining HRMIS and payroll data to inform SHRM and fiscal planning, and to have inspired practitioners to exploit 40 these data’s potential for more and better evidence-based policy-making. 41 References M. W. Anderson. The metrics of workforce planning. Public Personnel Management, 33 (4):363–378, 2004. E. H. Choudhury. Workforce planning in small local governments. Review of Public Personnel Administration, 27(3):264–280, 2007. L. Colley and R. Price. Where have all the workers gone? Exploring public sector workforce planning. Australian Journal of Public Administration, 69(2):202–213, 2010. T. Davenport. Is hr the most analytics-driven function? https://hbr.org/2019/04/is- hr-the-most-analytics-driven-function, 2019. K. Dychtwald, T. Erickson, and B. Morison. It’s time to retire retirement. Public Policy and Aging Report, 14(3):1–28, 2004. H. A. Frank and Y. Zhao. Determinants of local government revenue forecasting practice: empirical evidence from Florida. Journal of public budgeting, accounting & financial management, 2009. P. E. French and D. H. Folz. Executive behavior and decision making in small US cities. The American Review of Public Administration, 34(1):52–66, 2004. GAO. High-risk series: an update. https://www.gao.gov/assets/gao-01-263.pdf, 2001a. GAO. Federal employee retirements: expected increase over the next 5 years illustrates need for workforce planning. 2001b. GAO. High-risk series: dedicated leadership needed to address limited progress in most high-risk areas. https://www.gao.gov/products/gao-21-119sp, 2021. D. Goodman, P. E. French, and R. P. Battaglio Jr. Determinants of local government workforce planning. The American Review of Public Administration, 45(2):135–152, 2015. 42 S. Gupta, D. Coady, M. Fouad, R. Hughes, M. Garcia-Escribano, T. Curristine, C. Ab- dallah, K. Dybczak, Y. Endegnanew, M. Francese, et al. Managing government compensation and employment-institutions, policies, and reform challenges. Policy Papers, 47, 2016. Z. Hasnain, N. Manning, and J. H. Pierskalla. The promise of performance pay? Reasons for caution in policy prescriptions in the core civil service. The World Bank Research Observer, 29(2):235–264, 2014. W. S. Jacobson. Planning for today and tomorrow: Workforce planning. Public human resource management: Problems and prospects, 5:179–202, 2009. W. S. Jacobson. Preparing for tomorrow: a case study of workforce planning in North Carolina municipal governments. Public Personnel Management, 39(4):353–377, 2010. S. C. Kavanagh, D. W. Williams, et al. Informed Decision-making Through Forecasting: A Practitioner’s Guide to Government Revenue Analysis. Government Finance Officers Association, 2016. N. B. Kiyonaga. Today is the tomorrow you worried about yesterday: meeting the challenges of a changing workforce. Public Personnel Management, 33(4):357–361, 2004. D. Kong. Local government revenue forecasting: the California county experience. Journal of Public Budgeting, Accounting & Financial Management, 2007. O. H. Melchor. The government workforce of the future: Innovation in strategic workforce planning in OECD countries. 2013. OECD. Ageing and the public service: Human resource challenges. OECD Publishing, 2007. OECD. Gender equality in public sector employment. OECD Publishing, 2019. J. E. Pynes. The implementation of workforce and succession planning in the public sector. Public Personnel Management, 33(4):389–404, 2004. 43 J. E. Pynes. Strategic human resources management. In S. W. Hays, R. C. Kearney, and J. D. Coggburn, editors, Public human resource management: problems and prospects, pages 95–106. Longman, New York, NY, 2009. V. Reitano. Government and nonprofit personnel forecasting. The Palgrave Handbook of Government Budget Forecasting, pages 361–376, 2019. N. M. Riccucci, G. G. Van Ryzin, and C. F. Lavena. Representative bureaucracy in policing: Does it increase perceived legitimacy? Journal of Public Administration Research and Theory, 24(3):537–551, 2014. J. Rogers and C. Naeve. One small step towards better government. Public Utilities Fortnightly, 3:9–12, 1989. S. C. Selden. Human capital: tools and strategies for the public sector. Sage, 2009. S. C. Selden and W. Jacobson. Government’s largest investment: human resource management in states, counties, and cities. In P. W. Ingraham, editor, Public human resource management: problems and prospects, pages 82–116. Johns Hopkins University Press, Baltimore, MD, 2009. R. Somani. The returns to higher education and public employment. World Development, 144:105471, 2021. N. A. Theobald and D. P. Haider-Markel. Race, bureaucracy, and symbolic representation: Interactions between citizens and police. Journal of Public Administration Research and Theory, 19(2):409–426, 2009. G. G. Van Ryzin, N. M. Riccucci, and H. Li. Representative bureaucracy and its symbolic effect on citizens: A conceptual replication. Public Management Review, 19(9):1365– 1379, 2017. B. Wilkerson. Effective succession planning in the public sector. Watson Wyatt Worldwide, pages 1–7, 2007. 44 J. D. Wong. Local government revenue forecasting: Using regression and econometric revenue forecasting in a medium-sized city. Journal of Public Budgeting, Accounting & Financial Management, 1995. World Bank. Five facts on gender equity in the public sector. https://blogs.worldbank. org/governance/five-facts-gender-equity-public-sector, 2021a. World Bank. Innovating bureaucracy for improved government performance: Main findings of the bureaucracy lab. World Bank Papers, 2021b. World Bank. Measuring government employment and wages, 2021c. 45 Table A1: Payroll + HR Microdata Example − Minimum data required March 2020 Individual Job Job Weekly Career Date of last Pension Year Month ID ID Date of birth Gender Education Date of entry Type of contract Area position working hours level progression Base salary Allowance 1 Allowance 2 Allowance 3 Vacation contribution Gross wage Net wage 2020 March 100001 1 1987-03-05 Female Secondary 2015-01-01 Statutory Education 20 III 2016-03-01 3500 0 0 0 0 440 3500 3060 2020 March 100001 2 1987-03-05 Female Secondary 2010-11-10 Statutory Health 20 IV 2013-03-01 1000 0 100 0 0 110 1100 990 2020 March 100004 1 1980-06-04 Female Superior 2008-03-02 Temporary Safety 30 VI 2020-03-05 4000 0 0 0 0 440 4000 3560 2020 March 100005 1 1985-02-03 Female No schooling 2009-05-03 Political appointee Other 40 III 2020-03-31 2500 200 0 0 0 275 2700 2425 46 March 2021 Individual Job Job Weekly Career Date of last Pension Year Month ID ID Date of birth Gender Education Date of entry Type of contract Area position working hours level progression Base salary Allowance 1 Allowance 2 Allowance 3 Vacation contribution Gross wage Net wage 2021 March 100001 1 1987-03-05 Female Secondary 2015-01-01 Statutory Education 30 III 2016-03-01 3500 0 0 0 0 440 3500 3060 2021 March 100002 1 1980-06-05 Male Primary 2010-11-10 Statutory Health 40 IV 2013-03-01 1000 0 100 0 0 110 1100 990 2021 March 100004 1 1980-06-04 Female Superior 2008-03-02 Temporary Safety 30 VI 2020-03-05 4000 0 0 0 0 440 4000 3560 2021 March 100005 1 1985-02-03 Female No schooling 2009-05-03 Political appointee Other 40 III 2020-03-31 2500 200 0 0 0 275 2700 2425