Innovating Bureaucracy for a More Capable GovernmenT Contents Acknowledgments    v Executive Summary    ix Introduction    1 Who Are the Bureaucrats?    5 Public employment    5 Public sector compensation    10 The gender pay gap     13 Wages and selection    16 What Practices Shape the Work of Bureaucrats?    17 Merit-based selection, allocation, and promotion     17 The quality of management     19 Digital technology and bureaucracy     22 What Are the Attitudes and Behaviors of Bureaucrats?    25 Job satisfaction    26 Work motivation    29 Public service motivation    30 Trust    32 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T i How Productive Are Bureaucrats?    33 Conclusion    37 Annex: The Datasets Used in the Report    41 References    45 ii I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T FIGURES Figure 1.1 A Conceptual Framework for Government Capability . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Figure 2.1 The Public Sector Is a Large Employer Globally, Particularly of Skilled Workers. . . . . . . . 6 Figure 2.2 The Public Sector Is the Dominant Employer in Some Regions. . . . . . . . . . . . . . . . . . . . 6 Figure 2.3 Public Sector Employment Varies Considerably Cross-Nationally . . . . . . . . . . . . . . . . . . 7 Figure 2.4 The Wage Bill Has a Large Fiscal Footprint, but with Considerable Cross-Country Variation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Figure 2.5 The Wage Bill Does Not Crowd Out Capital Spending. . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure 2.6 The Public Sector Is a More Gender Equal Employer. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Figure 2.7 The Share of Women in the Public Sector Increases with Country Income Levels. . . . . 9 Figure 2.8  Public Sector Workers Are Older. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Figure 2.9 And Have Higher Academic Qualifications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Figure 2.10 Bureaucrats Spend Much of Their Time on ad hoc Tasks That Are Not Related to Their Core Competencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Figure 2.11 Public Sector Workers Receive a Wage Premium When Compared to Similar Private Sector Workers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Figure 2.12 The Premium Is Much Lower When the Public Sector Is Compared to the Formal Private Sector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Figure 2.13 Premiums Are Lower for More Skilled Occupations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Figure 2.14 Public Sector Workers Receive More Benefits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 Figure 2.15 Women Earn a Higher Wage Premium in the Public Sector. . . . . . . . . . . . . . . . . . . . . . . 13 Figure 2.16 The Gender Pay Gap Is Lower in the Public Sector Than in the Private Sector. . . . . . . . . 14 Figure 2.17 The Gender Pay Gap Is Lower in the Public Sector Than in the Private Sector. . . . . . . . . 14 Figure 2.18 The Public Sector Gender Pay Gap Has Declined over Time . . . . . . . . . . . . . . . . . . . . . . 14 Figure 2.19 Women Are Underrepresented in Senior Positions and Overrepresented In Clerical Occupations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Figure 2.20 Public Sector Occupational Segregation Is Persistent across Country Income Levels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Figure 2.21 Perceptions of the Quality of Recruits between BR and Non-BR Agencies in Indonesia. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Figure 3.1 Merit Is Generally the Main Criterion for Selection but Less So for Job Allocation. . . . . 18 Figure 3.2 Merit as the Main Criterion for Selection Varies Considerably Across Organizations. . . . 18 Figure 3.3 The Quality of Management Varies Considerably across Government Organizations within Countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Figure 3.4 Philippines: Perceptions of the Effect of Performance-Related Pay (PRP) on Goal Setting, Monitoring of Targets, Teamwork, and Performance Appraisals (percent). . . . . . . . . . . 21 Figure 3.5 Percentage of Staff Reporting That Politicians Had Improperly Tried to Influence Projects or Personnel Decisions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Figure 3.6 Developing Countries Have Invested Heavily in Digital Technologies. . . . . . . . . . . . . . . 22 Figure 3.7 Staff Believe That IT Has Improved Management and Productivity in Pakistan SRB. . . 24 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T iii Figure 4.1 Work Satisfaction Varies Considerably across Government Organizations within Countries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Figure 4.2 Bureaucrats Are Generally Unsatisfied with Their Pay Levels . . . . . . . . . . . . . . . . . . . . 26 Figure 4.3 There Is Considerable Pay Dispersion for Similar Workers in Rio de Janeiro Municipality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Figure 4.4 The Public Sector Wage Distribution Is Narrower than That in the Private Sector . . . . . 28 Figure 4.5 Work Satisfaction Is Corelated with the Quality of Management in Organizations in Nigeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Figure 4.6 Motivation Decreases over Time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Figure 4.7 Satisfaction with Wages Has No Association with Staff Motivation. . . . . . . . . . . . . . . . 29 Figure 4.8 Career Ambitions and Job Security Are Prominent Self-Reported Reasons for Joining the Public Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Figure 4.9 Percentage of Staff Reporting That They Trust Their Colleagues. . . . . . . . . . . . . . . . . . . 32 Figure 5.1 The Quality of Infrastructure Delivery Varies Considerably across Government Organizations in Nigeria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure 5.2 The Quality of Task Completion Varies Considerably across Government Organizations in Ghana . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 TABLES Table A.1 The List of World Bank Surveys of Bureaucrats. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Table A.2 Modules used in the Surveys. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Acknowledgments This report was produced by a core team consisting of Zahid Hasnain, Dan Rogger, Danny Walker, Kerenssa Kay, and Rong Shi. It draws on datasets compiled with the help of Elsa Araya, Erwin Ariadharma, Sheheryar Banuri, Pablo Becerra, Vanessa Cheng, Erika Deserranno, Verena Fritz, Tim Gindling, Emmanuel Kapee, Phil Keefer, Claire Le Barbenchon, Jennifer Ljungqvist, Massimo Mastruzzi, Aisha Nansamba, Saa David Nayuma Jr., David Newhouse, Daniel Ortega, Camilo Penuela, Vincent Pons, Imran Rasul, Ravi Somani, Maria Tambunan, and Martin Williams. The team is grateful to government officials from Brazil, Ethiopia, Ghana, Indonesia, Liberia, Nigeria Pakistan, and the Philippines for their cooperation in sharing data and for sharing their views in the surveys. The surveys were requested by our government counter- parts in each country and were conducted under their guidance and with their support. We thank the Office of the Head of Civil Service of Ghana; the Ministry of Public Service and Human Resource Development of Ethiopia; the Ministry of Administrative Reforms in Indonesia; the Liberian Civil Service Agency; the Office of the Head of Civil Service of the Federation of Nigeria; the Sindh Revenue Board in Pakistan; and the Department of Budget and Management in the Philippines. The report was produced under the overall guidance of Deborah Wetzel, James Brumby, Asmeen Khan, William Maloney, Shantayanan Devarajan, Asli Demirguc-Kunt, and Arianna Legovini. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T v ABBREVIATIONS DAI Digital Adoption Index GDP gross domestic product I2D2 International Income Distribution Database ISCO International Standard Classification of Occupations IT information technology PPP purchasing power parity SRB Sindh Revenue Board (Pakistan) WMS World Management Survey WWBI Worldwide Bureaucracy Indicators I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T vii Executive Summary The ability, motivation, and productivity of bureaucrats are key deter- minants of governments’ ability to implement policies and deliver infrastructure and services to citizens effectively. Despite their importance, the role of the administrators and managers who are responsible for regulating, financing, and monitoring the work ” largely due to lack of service providers has been a “black box, of good data. Drawing on new datasets compiled by the World Bank—a cross-country dataset on public employment and wages covering 114 countries, surveys of 20,000 civil servants in 7 countries, and microlevel administrative datasets in 2 countries—this report presents empirically driven findings on five questions: 1. What are the main features of the public sector labor market in terms of employment and compensation? 2. What are bureaucrats’ attitudes toward their jobs and their behav- iors toward each other? 3. How well are they managed? 4. Are they using digital technologies to innovate? 5. And finally, how can we measure whether they are productive? The public sector is often a country’s predominant employer and pays a wage premium, with significant implications for the overall labor market. Globally, public employment is 30 percent of wage employ- ment and 38 percent of formal sector employment, and accounts for the majority of formal sector employment in South Asia and Sub-Saharan Africa. Thirty-nine percent of all employees with tertiary education work in the public sector, and the figure is as high as 60 per- cent in several countries. Public sector workers also on average earn higher wages than private sector workers of similar education and age; the public sector wage premium is approximately 16 percent across I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T ix 72 countries and does not vary with country Public sector wages are helping recruit good qual- income. The size of the wage premium, however, ity staff, and recruitment is generally perceived as is sensitive to the choice of the private sector meritocratic by bureaucrats. In Indonesia, staff in comparator and differs across occupations. The higher-paid ministries were more likely to state average premium decreases to 9 percent if pub- that their ministry could recruit high-quality candi- lic sector workers are compared only to formal dates than staff in lower-paid ministries. A signifi- sector workers, and public sector pays relatively cant majority of civil servants in Ghana, Indonesia, lower wages than the private sector to senior offi- and Pakistan believed that meritocratic processes cials and professionals. These wage differentials, were followed in recruitment and the best can- however, underestimate the public sector pre- didates were selected. Merit, however, was per- mium for total compensation, as a much higher ceived to be less of a factor in allocating jobs to proportion of public sector workers receive bureaucrats after they were selected, reflecting benefits like health insurance or pensions than the greater managerial discretion in matching private sector workers. Given its large footprint individuals to jobs. on the labor market, the challenge of more and The relatively high public sector compensation, better jobs in developing countries hinges on however, is not resulting in more motivated a better understanding of the impact of public bureaucrats. There are two aspects to this weak sector employment and compensation practices association between wages and motivation. on the overall labor market. First, surveys of civil servants reveal that only 40 percent of respondents across seven coun- The public sector employs a higher proportion of tries surveyed were satisfied with their pay lev- women than the private sector and pays them els, despite a significant wage premium in each a fairer wage, although gender equality remains of these countries. A possible explanation could a challenge. In most countries, the share of be the considerable pay dispersion and pay women working in the public sector is higher inequity in the public sector, and the possibility than the share of women working in the private that bureaucrats benchmark their pay against their sector. Women’s average wages are 88 percent peers rather than private sector workers. Second, of male wages in the public sector, as compared even bureaucrats who are satisfied with their to 81 percent of male wages in the private wages do not have higher levels of self-reported sector, and the gender pay gap is also lower in motivation. Bureaucrats’ motivation levels also the public sector in 42 of the 60 countries for decline over time. This finding suggests that which we have data. This greater wage equality the high public expenditures on wages are not for women likely reflects the more limited scope yielding a motivated and productive workforce. for wage discrimination in the public sector, given the highly regulated nature of the public Management practices are more important than sector labor market. Nevertheless, there is sig- compensation in influencing the attitudes and nificant occupation segregation by gender, with behaviors of bureaucrats, and the quality of man- women underrepresented in higher-paying man- agement varies considerably across organizations agerial occupations, and overrepresented in within countries. Some reasons for declining moti- lower-paying clerical ones. This occupation seg- vation levels are structural, such as limited opportu- regation persists across country income levels. nities for promotion and limited movement across x I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T organizations. A typical civil servant spends 10 fundamentally, many bureaucracies lack the and 16 years in the same organization in Ethiopia necessary “analog complements” of sound man- and Nigeria, respectively. In the Sindh Revenue agement, incentives, and flexible work practices Board in Pakistan, despite receiving higher wages to make effective use of digital technologies. than civil servants in the rest of the government, 21 percent of bureaucrats would like to change Measuring the productivity of bureaucracy is their jobs in the next two years due to limited pro- difficult given the dearth of quantifiable outputs, motion opportunities. Day-to-day management on but the speed and quality of task completion can setting organizational goals and communicating provide a useful proxy measure. In Ghana, there these to staff, regular monitoring, task distribu- is considerable variation across organizations in tion, and the quality of performance assessments the rates of completion of tasks that organiza- is likely to affect staff attitudes and behaviors more tions committed to in their performance agree- than relatively infrequent events like promotions ments. The speed and quality of responsiveness and movement. Across our surveyed countries, of ministries to requests from the center of managers are failing to regularly involve staff in government can be another measure of produc- setting and monitoring goals and in problem solv- tivity, and there is also substantial variation in ing. These country averages, however, mask the these across Ghanaian organizations. considerable variation in the quality of manage- These findings also open several additional ment across organizations in Ethiopia, Ghana, and avenues for future research. A better understand- Nigeria, as well as across units within organiza- ing of the effects of the public sector labor mar- tions. This dispersion implies that the experience of ket on the overall labor market requires more being a bureaucrat, despite a common regulatory accurate comparisons between public and pri- framework, is highly dependent on local context. vate sector workers, controlling for unobservable Governments have invested heavily in digital characteristics like worker motivation. One prom- technologies, but bureaucracies are lagging in ising approach is to use panel data to focus on indi- digital innovation. There are many examples of viduals who transition between the two sectors digital technologies improving frontline agen- to more accurately estimate wage differentials cies: digital identification programs have reduced and other pecuniary and nonpecuniary aspects leakages in government welfare programs; elec- of compensation that influence workers’ employ- tronic filing of taxes has reduced the cost of tax ment decisions. The reasons for occupational compliance for businesses; and digitally enabled segregation by gender in the public sector are one-stop shops have reduced the time it takes also not well understood. Better understanding for citizens to receive services and lowered the determinants of bureaucrats’ motivation, and corruption. The impact on bureaucrats’ daily work, the role of different compensation policies and however, is more limited, in part because of lack of management practices, can be key to improving sufficient information technology skills and poor government capability. Finally, the personnel and infrastructure. In Ethiopia, in 40 percent of the management linkages between the upstream organizations surveyed, fewer than half of the bureaucracy and the downstream citizen-facing staff in the unit could use a computer for writ- agencies, and a more precise determination of ing a memo. Local administrators in Nigeria had which activities of administrators most effect ser- Internet access on only 3 percent of days. More vice providers, should be a priority for the future. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T xi Introduction Improving government capability is one of the main challenges of economic development. There is consensus around the core policies needed for developing countries to achieve equitable growth and reduce extreme poverty. But government capability—its ability to effectively implement these polices and efficiently achieve the desired outputs in regulation, infrastructure provision, and service delivery—varies considerably across countries and across policy domains within countries.1 The ability, motivation, and productivity of the personnel who popu- late government bureaucracies are key determinants of government capability. Capable organizations are those that can select high-ability personnel, provide them with the necessary resources, and motivate them to work toward the organization’s objectives and to serve the public. In Russia, 60 percent of the price variation in standard pro- curement contracts is due to the ability of individual bureaucrats and the quality of the organizations in which they work (Best, Hjort, and Szakonyi 2017). If the worst-performing 20 percent of bureaucrats can be made as effective as the median bureaucrat, the Russian gov- ernment would save 10 percent of its procurement costs. In Nigeria, there is substantial variation in the quality of organizational manage- ment across the federal government, and a one standard deviation increase in the quality of management would lead to a 32 percent increase in project completion rates (Rasul and Rogger 2017). Public sector compensation and employment practices also have significant implications for the competitiveness of the overall labor market, and on fiscal sustainability. Governments face important 1 This definition of capability is based on Andrews, Pritchett, and Woolcock (2017). I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 1 choices relating to the size of the public sector level indicators are derived primarily from over and the compensation of its workers. Low public 13 million household-level observations from sector wages can result in difficulties in recruit- the International Income Distribution Database ing and retaining qualified workers; but large (I2D2), the World Bank’s repository of harmonized wage premiums for public sector workers can household (labor force and welfare) surveys. discourage private sector jobs and lead to search Information on human resource management unemployment. A rising wage bill is also often practices and the attitudes and experiences of associated with problems of fiscal sustainability. bureaucrats draw on surveys of approximately 20,000 civil servants in seven countries: Ethiopia, Despite its importance, there is little rigorous Ghana, Indonesia, Liberia, Nigeria, Pakistan, and empirically based knowledge on the personnel the Philippines. These surveys cover either cen- aspects of bureaucracies that can be the basis tral and local governments (Ethiopia, Ghana, and for evidence-based policy making. There is an Nigeria); only central governments (Indonesia, enormous body of empirical research that shows Liberia, and the Philippines); or select agencies that the skills, incentives, and accountabilities of (the federal and subnational tax authorities in “street-level” bureaucrats like doctors, police Pakistan). Some of these surveys have been con- officers, and teachers are the main determinants ducted exclusively by the World Bank, and others of service delivery outcomes. The role of admin- have been done in partnership with academics. istrators and managers—the bureaucrats that The report also references public employment and are the focus of this report—who are respon­ wage bill analysis in Bosnia and Herzegovina and sible for policy-making, regulating, financing, and some subnational jurisdictions in Brazil based monitoring the work of the frontline service pro- on microlevel human resource and payroll data ” A major viders, however, remains a “black box. that the respective governments shared with the reason for this is lack of good data, with most World Bank. More details on these datasets are existing cross-national sources of information on given in the appendix. bureaucrats limited to expert perception-based approaches, such as the Bertelsmann Trans- Our conceptual framework for government capa- formation Index or the Worldwide Governance bility is a production function in which inputs, Indicators. The title of this report is a deliberate the “technology” of combining these inputs, play on words to underline the innovation of and a variety of exogenous factors determine an empirical approach to understanding public bureaucrats’ attitudes and behaviors, outputs, bureaucracies, and of filling this knowledge gap. and eventually outcomes (figure 1.1). Govern- ment capability is the conversion of policies, The report draws on new datasets compiled by assumed to be generated by politicians, to the World Bank and on analyses of existing data outputs, which in turn involves two interrelated sources. The main new cross-country dataset on drivers: the quality of (i) bureaucracy and (ii) front- public employment and wages is the Worldwide line activities in service delivery, infrastructure Bureaucracy Indicators (WWBI), which covers provision, and revenue collection. The frontline 114 countries from 2000 to 2016.2 The country- is the point of contact between the government worker (the teacher, doctor, police officer, or tax 2 The details on the variable definitions and methodology used are given in “The Worldwide Bureaucracy Indicators: Explanatory Note on the Dataset” (WWBI). official) and the client (the business, student, 2 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T ■ ◾ ▪ Figure 1.1  A conceptual framework for government capability Exogenous factors (political environment, socioeconomic factors) Bureaucracy Frontline agency Inputs “technology” Personnel Management practices Attitudes and Service delivery Policies Goods + IT systems behaviors of bureaucrats Revenue collection Outcomes Capital Business processes Infrastructure provision Source: Based on Fiszbein, Ringold, and Rogers 2011. patient, or crime victim). Bureaucracy is the set production function, and center around four of ministries and agencies that are upstream questions: from service delivery. While there is overlap between bureaucracy and frontline agencies 1. First, who are the government personnel? in tasks performed, the distinction is important What are the main features of public sector conceptually because the outputs of frontline labor market in terms of public employment agencies are more measurable than those of and the level and structure of compensation, bureaucracies, and bureaucrats are also more a key factor for the selection and incentives of likely to have multiple principals and multi­ bureaucrats? The report provides some basic dimensional tasks than service providers. facts. 2. Second, what are the core practices that The attitudes and behaviors of bureaucrats are shape the work of bureaucrats? These include an indirect measure of the productivity of the management practices around recruitment, bureaucracy. An extensive academic literature has promotion, and performance orientation, and shown that job satisfaction, work motivation, the extent to which bureaucracies effectively public service motivation, professional norms, use digital technologies to become more effi- and integrity have an important bearing on the cient and to innovate. productivity of bureaucrats. These attitudes and 3. Third, what do we know about the attitudes behaviors are influenced by both the quality and behaviors of bureaucrats, in terms of their of inputs, such as skills and type of personnel motivation, professional norms, and commit- selected, and the technology to combine these ment to serve the public? How do the inputs inputs, which in turn is determined by work prac- of the production function—public sector com- tices, business processes, and the use of infor- pensation policies and work practices—affect mation technology for managing staff, receiving the selection of high ability staff and correlate customer feedback, and automating tasks. The with bureaucrats’ levels of motivation? Given main findings of the report cover the personnel the difficulties in measuring the actual effort subset of these elements of the government and outputs of individual bureaucrats, these I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 3 attitudes are a proxy indicator of the produc- across organizations within countries despite tivity of personnel. a common de jure human resource manage- 4. And finally, how do we measure the produc- ment regime, underlining the need to have an tivity of the bureaucracy, given that many of empirically granular approach that can capture its outputs are intermediate and difficult to this within-country heterogeneity and measure quantify? the core components of bureaucratic capability from a representative sample of public officials. We examine the de facto situation in each of these areas rather than the de jure structures The report focuses primarily on the supply side of specified in regulations. One reason for this governance and does not delve into the political choice is pragmatic, as a discussion of the legal economy of public administration, for both con- framework across numerous countries would ceptual and methodological reasons. The domain be tedious. But substantively, and as is well of citizen engagement is largely at the point of known, the gap between law and actual prac- service delivery or revenue collection, and not at tice is large across many policy domains in the upstream administrative tier. It is unlikely that developing countries. And this divide is espe- bureaucrats have regular contact with citizens, cially apparent in personnel management given and any citizen voice would need to be transmit- the complexity of the regulatory regime that ted via “the long route of accountability”—from governs the different categories of public citizen to politician and then from politician to sector employees; the considerable discretion bureaucrat (World Bank 2003). Asking bureau- that organizations can have in setting pay and crats about their interactions with politicians employment for their staff; and the coordina- through surveys, however, is a difficult and sen- tion and collective action among thousands of sitive topic, and one that has been broached only bureaucrats and service providers necessary to cautiously in our work to date. Methodologically, convert policy actions to outputs. As evidenced it requires more experimental approaches, which in the following material, the attitudes and adds complexity to the surveys, and is an ambition experiences of bureaucrats vary considerably for future work. 4 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T Who Are the Bureaucrats? Public employment The public sector is a very large employer globally, particularly of formal sector and skilled workers. Overall, the public sector is responsible for 16 percent of total employment, 30 percent of wage employment, and 38 percent of formal sector wage employment (figure 2.1, panel a).3 Globally, on average 39 percent of all employees with tertiary edu- cation work in the public sector, with in many countries this number as high as 60 to 70 percent (figure 2.1, panel b). The public sector accounts for the bulk of formal sector employment and the share of employed workers with tertiary education in the Middle East and North Africa, East Asia, and Sub-Saharan Africa, and has a significant, albeit much smaller formal sector footprint in Latin America (figure 2.2). Apart from underlining the influence of public sector compensation policies on the broader labor market, the large size of the public sector also has important implications for the selection and motivation of public employees. We can reasonably assume that where the public sector is the largest formal sector employer, it is attracting individuals with a variety of motivations, both intrinsic and extrinsic, with implications for public sector productivity. Cross-nationally, the size of the public sector as a share of total employ- ment increases with a country’s level of economic development. The share rises from below 10 percent in low-income countries to above 3 Total employed individuals are defined as those workers, age 15 and older, who in the household surveys responded that they had a job in the prior week. Wage employees are those whose basic remuneration is not directly dependent on the revenue of the unit they work for and are instead paid in wages and salaries, piece work, or in-kind compensation, and therefore exclude self-employed workers. Formal sector wage employees are those who also have an employment contract, have health insurance, belong to a union, or are enrolled in a pension program. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 5 ■ ◾ ▪ Figure 2.1  The public sector is a large employer globally, particularly of skilled workers b. Proportion of total employees with tertiary education a. Public sector employment as working in the public sector 80 Formal employment 38.1% MRT LKA TJK VNM 60 ETH MWI TZA KSV BWA ALB KHM MOZ BFA MDV Percentage TCD RWA UKR Paid employment 30.2% LBR GHANGA MNG Global average=39% CHN 40 TGO PAK SRBTHA SLV KAZ PAN PRY GEO ECUCRI BRA URY ARG ZAR BOL MEX MDG GTM BGD PHL DOM PER 20 Total employment 15.2% PNG CHL COL 0 6 8 10 12 Log of GDP per capita Source: WWBI. Note: Panel a is based on data from 114 countries; panel b has 65 countries. 20 percent in high-income countries (figure 2.3, sector grows along with private formal sector panel a).This positive relationship, called Wag- wage employment (figure 2.3, panel b). There ner’s Law, reflects the increasing role of the state is considerable heterogeneity within these in providing social services as incomes rise (Dia- regional and income categories, revealing that mond 1977). There is however, no discernible countries also make choices in the numbers relationship between country income levels and of their government personnel. Public sector public sector employment as a share of salaried shares of total employment range from less employment, which suggests that the public than 2 percent to over 40 percent, and of paid ■ ◾ ▪ Figure 2.2  The public sector is the dominant employer in some regions b. Proportion of total employees with tertiary a. Public sector as a share of overall employment (%) education working in the public sector East Asia and Pacific East Asia and Pacific Eastern Europe and Central Asia Eastern Europe and Central Asia Latin America and Caribbean Latin America and Caribbean Middle East and North Africa South Asia Middle East and North Africa Sub-Saharan Africa South Asia 0 10 20 30 40 50 Sub-Saharan Africa Total employment Paid employment Formal employment 0 10 20 30 40 50 Source: WWBI. 6 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T ■ ◾ ▪ Figure 2.3  Public sector employment varies considerably cross-nationally a. Public employment as share of total employment b. Public employment as share of paid employment 50 80 SRB RUS 40 SYC UKR SRB KAZ 60 MDV MRT MDV TUN 30 MDA KSV FIN KAZ Percentage Percentage TJK GNB COM MDA KSV RUSSYC TUNBWA POL EST IRL PNG TCD UKR GBR 40 NERGIN STP ZMB SLETJK MNG MNG FRADEU ZARLBR ETH GEO ALB BWA FIN 20 ALB GRC CANUSA LKA POL GRC IRL SWZ BFA TZA VNM SWZ EST GBR COM STP LKA ARG AFGBEN DEU PAN MOZ GHA BOL CHN PAN FRA CANUSA MRT GEO CHN MUS TUR URY ITA LUX TGO LSO CMR PAKMAR AGO PRY THA ARG TUR CRI BRA MNE 20 MDG DOM PER URY AFGLSO DOM MWI KHM ECU LBN BRA ITA 10 VNM ZMBNGA BOL PRY ECU MEX LBN ROM THA CHL SLV MNE PHL CRI ROM MUS MEX LUX LBR KHM PAKMAR SLV PER AGO PHL RWA BGD CHL ZAR MWI GHA GTM GTM COL MOZGNB ETH TGO GIN TCD BFASLE CMR TZA BGD COL NER MDGRW BEN APNG 0 0 6 8 10 12 6 8 10 12 Log of GDP per capita Log of GDP per capita Source: WWBI. employment from 10 percent to 70 percent, roughly a quarter of general government expen- with four- to fivefold variations in these shares ditures, albeit with considerable variation across at any given income level. countries (figure 2.4). While the wage bill as a share of GDP rises with country income, as per Public employment also has a large fiscal foot- Wagner’s Law, it tends to fall as a share of expen- print. Cross-nationally, the general government ditures due to the higher revenue mobilization of wage bill averages approximately 9 to 10 percent richer countries. These wage bill numbers under- of gross domestic product (GDP), and represents estimate, often dramatically, the full fiscal costs ■ ◾ ▪ Figure 2.4  The wage bill has a large fiscal footprint, but with considerable cross-country variation a. General government wage bill as a share of GDP b. General government wage bill as a share of expenditures 20 LSO 60 AFG MDV 50 PRY Wage bill as % of expenditure 15 FIN Wage bill as % of GDP SWZ SLV SWZ TUN AFG MAR PRY BWA FRA CAN MAR MDV TUN UKR ARG GRC IRL 40 MDG CRI GHA POL RUS EST AGO SRB LSO GHA 10 MOZ LBR ECU BRA ITA GBRUSA ETH SLE ZMB THAMUS MWI ZMBMDA SLVKSV LBN KSV LBN MUS LUX ZAR MOZCOM TZA ARG COM BENSTP MRT TURROM DEU 30 LBR BEN CRI THA BFA EST CAN BFA TJK MNG SYC GNB GTM PHL PER RUS ETH SLE TGO PER MEX CHL PAN MWI TGO AGO BWA CHL POL GRC FIN USA 5 ZAR MDG TZA NER GNB CMR PHL ALBCOL URY GEO TJK CMR KHM MDA UKR SRB MRT DOMBRA MEX ROM FRA GBR GIN TCD KHM ECU TURPAN RWA GTM DOM 20 TCD ALBCOL SYC ITAIRL LUX KAZ GEO MNG DEU NERGIN STP URY BGD RWA KAZ BGD 0 10 6 8 10 12 6 8 10 12 Log of GDP per capita Log of GDP per capita Source: World Bank staff calculations based on IMF data. Note: PPP = purchasing power parity; I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 7 of public sector workers, given the generous Wage expenditures have a built-in momentum pensions benefits that they enjoy. In Brazil, for resulting from automatic pay increases linked example, the wage bill is 13 percent of GDP, and to seniority, limiting governments’ flexibility in public sector pension expenditures are another 4 responding to fiscal constraints. The magnitude percent of GDP (World Bank 2017). of this “natural rate of wage growth” is condi- tional on the size of the pay increments in the pub- Higher wage expenditures are associated with lic sector salary scales. The example of Brazil is greater fiscal imbalances, but not with crowd- indicative, where microlevel data allows for mod- ing out of other expenditures. While there is no eling scenarios that can decompose the effects cross-national relationship between the level of wage bill increases due to additional hiring of the wage bill and fiscal deficits, increases from those due to staff moving up the pay scale in wage expenditures over time are correlated with increasing years of service. In Rio de Janeiro with increasing fiscal deficits (IMF 2016 ). How- municipality, for example, the wage bill increases ever, surprisingly, there is a positive relationship by 2.1 percent in real terms annually, even when between wage expenditures and capital expen- staffing levels are fixed and there is a one-to-one ditures, suggesting that concerns that higher replacement of relatively higher paid retirees with spending on wages automatically reduces spend- relatively lower paid new hires, and annual nom- ing on growth-enhancing infrastructure are exag- inal increases to the salary scale are limited to gerated (figure 2.5). This finding does not detract inflation. Brazil is an extreme case of high seniority- from the importance of ensuring that the con- based pay increases, but it is representative of a siderable resources spent on personnel yield general phenomenon, as almost all public sector commensurate results in government outputs. salary scales have seniority-based pay increments that result in a natural rate of wage bill growth. ■ ◾ ▪ Figure 2.5  The wage bill does not Women are disproportionately represented in the crowd out capital spending public sector, although they are still outnumbered Wage bill vs. Capital spending (GDP) by men. Figure 2.6 shows the share of women MRT PRY in the public sector and in private sector wage 10 ETH employment for the countries in the Worldwide Capital Expenditures as % of GDP 8 Bureaucracy Indicators (WWBI), with the vertical AFG BFA MDA and horizontal lines marking gender equality in TUN 6 GNB ALB PER NAM employment in the two sectors, respectively. MOZ CHL BGR 4 KEN GAB In several countries (those in the right half), GIN ARM BEN MEX IDN UGA MKD CMR POL the majority of public sector workers are women, PAK HTI UKR CRI 2 BDI URY while there are fewer countries, mostly high SEN NER GTM TZA MLI DOM ARG LBR BRA SLE income, in which women are a majority of private 0 SLV AGO 0 5 10 15 sector workers (in the top half). Most countries, Wage bill as % of GDP however, are in the bottom left quadrant, which reg coef: 0.349; p-value: 0.00 shows that women have a lower share of employ- Source: World Bank staff calculations based on BOOST data. BOOST is a WB tool that collects and ment than men in both the public and private sec- compiles detailed data on public expenditures from national treasury systems and presents it in a simple user-friendly format. tors. Women’s share of public sector employment 8 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T ■ ◾ ▪ Figure 2.6  The public sector is a more ■ ◾ ▪ Figure 2.7  The share of women in the gender equal employer public sector increases with country income levels Proportion of female as share of each sector Proportion of female in the public sector 0.8 80 FIN GBREST RUS KAZ UKR 0.6 CANDEU FRAIRL 60 MNG BRA USA CHL MWI GEO CRI ARG PHL ITA MWI URY PAN SRB BOL MDV COL Private sector NAM PNG PRY ALBDOM IRL DEU Percentage PER BW A MEX KHM THA LUX PNG DOMURY RUS MDG VNM GTM SLVLKA ECUTHA EST FRA GRC CANUSA LUX GRC CHN ZAF BWACOL CHL GEO ARG BGRBRA MNG CHN ITA GBR 0.4 NGA PER UKR 40 SRB KAZ FIN TGO HTI RWA MDG ETH VNM SLV PAN MEX UZBCRI PHL BFA TZA GHA RWA ETH NGA TZA GAB LKA BOL ZWEECU HND MDV TJK KSV KENBFA ALB GTM NIC MOZ NPL GMB SLB GHA BGD ZAR KHM MOZKSV ZAR MRT 0.2 BGD TJK TCD LBR JOR 20 LBR TGO WBMRT G CAF PAK PAK EGY 0 0 0 0.2 0.4 0.6 0.8 6 8 10 12 Public sector Log of GDP per capita Source: WB staff calculations based on WWBI. Note: The vertical and horizontal dashed lines in figure 2.6 depict gender equality in employment shares in the public and private sectors, respectively. increases with country incomes, although there are disproportionately represented in these occu- is significant variation in these shares at any given pations. This issue of gender segregation by occu- income level, and women remain underrepre- pation is discussed in more detail later. sented even in high-income countries (figure 2.7). One reason for this increase in the share of women Public sector workers are also older and more could be linked to Wagner’s Law; as countries educated than private sector wage workers. In develop, a greater proportion of the public sector figure 2.8 and figure 2.9, the 45-degree line indi- workers are in the service sectors, and women cates equal values for countries for the categories ■ ◾ ▪ Figure 2.8  Public sector workers ■ ◾ ▪ Figure 2.9  And have higher are older academic qualifications Age in each sector Education in each sector 50 Percent of population with secondary education or above ITA 1.0 UKR RUS GEO MNG SRB UZB 45 IRL GRC SRB EST DEU FIN BGR PER LKA USA CHL GBR FRA 0.8 KAZ MDV ZAF TJK ARG URY KSV PAN JOR PHLCOL CAN KAZ GHA LUX GEO ZAR GABPRY MEX Private sector 40 CHL RUS GMB DOM NAM VNM UKR ALB Private sector BGR MDG ECU NGA ZAR LBR 0.6 LBRNGA PNGCAF BRA SLB CRI ZWE URY ARG THA EGY THA ZAF WB G CHN PAN HND NIC GMB SLB MNG ECU BRACRIMRT PER KSV COL HTI CHN PAK 35 JOR UZBHTITCD GAB GHA BGD TJK CAF MEX MDG SLVDOM SLV KHMBGD PNG ZWE NPL BWA NAM PHL TGO 0.4 TCD ETH GTM MOZ TGO MOZ GTM MWI BOL BWA TZA NPL MDVWB G VNM NIC PRY BFAPAK HND KEN EGY TZA KEN RWA MRT 30 ETH 0.2 BFA RWA KHM MWI 25 0 25 30 35 40 45 50 0 0.2 0.4 0.6 0.8 1.0 Public sector Public sector Source: WWBI. Note: The 45 degree line depicts equal values on the two axes in both figure 2.8 and figure 2.9. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 9 depicted in the two axes. In both figures, coun- ■ ◾ ▪ Figure 2.10 Bureaucrats spend tries largely cluster below the 45-degree line, much of their time on indicating higher shares for public sector work- ad hoc tasks that are ers for the category. The data reveal that the not related to their core competencies public sector has older workers (figure 2.8) and a higher proportion of workers with secondary or Percentage of time spent on ad hoc tasks 60 higher education (figure 2.9). These systematic differences between public and private sector workers have implications for any comparative 40 analysis between the two labor markets, espe- Percent cially public-private wage differentials. 20 These high academic qualifications do not nec- essarily imply high skills matched to functions. Academic credentials may not be matched to 0 Ethiopia Pakistan (SRB) the competencies needed in jobs, nor do they necessarily imply sufficient learning and technical deficiencies underline the urgency of regular tech- knowledge for the job. An emphasis on academic nical and management training of bureaucrats. credentials in public sector employment can also But even in organizations that make training a trigger the proliferation of low quality tertiary edu- priority, as in the Sindh Revenue Board (SRB) in cation establishments with little impact on out- Pakistan, there is a tendency for the training to comes, as has been documented in Indonesia for inadequately reflect the needs of different staff teachers (Ree et al. 2017). In Ethiopia, discussions groups. Seventy-four percent of SRB staff under- with key informants suggest that quality of new went training in the past year, but only 15 percent hires has declined because of the very rapid expan- sion in public employment (World Bank survey). of them found the training useful for their job, pri- A significant share of staff whose qualifications marily because the training was too generic and have been tested through competency assess- did not meet the specific needs of their jobs. ments failed to exhibit sufficient basic knowledge. Public sector compensation Many bureaucrats lack basic IT skills and spend much of their time working on ad hoc tasks that Public sector wages are an important determi- are not related to their main job. In Ethiopia, in nant of personnel quality and motivation and, 64 percent of the organizations surveyed, fewer given the large size of the public sector, have an than half of the staff in the unit could create an important influence on the broader labor market. Excel spreadsheet or a Power Point presentation, The cross-country heterogeneity in the size of the and in 40 percent of organizations, fewer than half wage bill reflects the policy choices that countries of all staff were able to use a computer for writing make on employment and compensation. If pub- a memo. In Ethiopia and Pakistan, staff can spend lic sector workers are paid significantly less than as much as 60 percent of their daily time at work similar workers in the private sector, the resulting on noncore tasks, which points to weaknesses difficulty in recruiting and retaining qualified work- in management practices (figure 2.10). These ers will adversely affect the quality of publicly 10 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T provided goods and services. On the other hand, assumes that workers with the same personal large wage premiums for public sector workers characteristics should be paid the same wage may indicate that public sector workers are a priv- irrespective of the sector of employment, and ileged group who are protected from competition therefore a public sector wage premium rep- with other workers. Large premiums could also resents an economic rent (Moulton 1990). How- encourage youth to queue for public sector jobs, ever, wages are also dependent on the type of leading to high rates of youth unemployment. occupation, as some jobs entail more responsibil- ities and therefore pay more for observably simi- In general, the public sector pays a wage premium lar workers. It is possible that the public sector for “similar” workers compared to the private sec- has a different distribution of jobs—larger pro- tor. The standard approach in the labor economics portions of workers in managerial, professional, literature is to estimate an earnings regression and clerical occupations, and fewer in sales and where wages are a function of observable worker laborer occupations—and that these jobs require characteristics, such as education, age (a proxy for additional responsibilities that are not accounted work experience), gender, location, and the sector for in premium estimates that control only for of employment (public sector or private sector). workers’ observable characteristics. Figure 2.12 We estimate these regressions using the WWBI.4 shows the premium when public sector workers Figure 2.11 shows the premium when the public are compared to formal private sector workers sector is compared to all private sector salaried who are likely to work in jobs that have equiv- employees, irrespective of the type of job and alent levels of responsibility. The public sector controlling only for worker characteristics. The premium reduces to 9 percent globally, with average public sector wage premium is 16 per- 29 of the 44 countries in the sample having a cent across the 72 countries in the data, with 55 statistically significant public sector earnings of the 72 countries having a premium. There is premium. considerable heterogeneity in the size of that pre- mium across countries, varying from a penalty of The public sector wage premium is lower for more 20 percent to a premium of 60 percent. The size of skilled occupations. We estimated the wage pre- the premium does not vary with country incomes, mium for the main occupational categories as an a surprising finding, as academic studies based alternative approach to account for both job and on a smaller sample of countries have reported worker characteristics and the possibility that the higher premiums for developing countries. public sector wage premium compensates for the additional requirements necessary for these The public sector premium is lower when public occupations.5 Using the International Standard workers are compared to private workers doing Classification of Occupations (ISCO) of the Inter- similar jobs. The analysis just noted implicitly national Labor Organization, the data reveal that the public sector has wage penalties for senior officials, professionals, and technicians and pays 4 The data includes all wage and salaried employees, so that the estimates are the public sector earnings premium relative to private wage and salaried employees. premiums for clerks and workers in elementary The analysis excludes self-employed workers, as many do not report any wages. The wage data in the I2D2 does not include bonuses and other payments to occupations (figure 2.13). employees, which is likely to underestimate the public sector premiums given the generally higher proportion of pay given in the form of allowances in the public sector. It also does not include in-kind payments, which will likely also underesti- 5 Our dataset has information only on these main occupational categories, and we mate the public sector premium when total compensation is taken into account. cannot further disaggregate into the three-digit occupations within these categories. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 11 ■ ◾ ▪ Figure 2.11  Public sector workers ■ ◾ ▪ Figure 2.12  The premium is much receive a wage premium lower when the public when compared to similar sector is compared to the private sector workers formal private sector Public sector wage premium Public sector wage premium versus all private employee versus formal employee 0.6 0.6 BWA PAK BWA COL PNG SLV CRI 0.4 LBR 0.4 TGO PHL ECU LUX LBR COL ECU GTM Wage premium Wage premium MOZ RWA TZA GHA TZA PAK SLV MEX CRI TCD GHA MDV PAN URY DEU 0.2 MDG URY 0.2 MDG BFA BOL PRY ARGGRC CAN BRA MWI MOZ BGD NGA IRL PRY BRA MEX CHL ETH CHL ITA TCD GTM GEO KSVPER KAZ BGD ARG GEO 0 BFA BOL KSV TJK 29 out of 44 countries>0 RWA ALBCHN PAN 0 ALB EST GBR 55 out of 72 countries>0 DOM VNM UKR CHN USA DOM MWI FIN PER RUS ZAR MNG RUS −0.2 TJK −0.2 SRB ZAR 6 8 10 12 6 7 8 9 10 Log of GDP per capita Log of GDP per capita Source: WWBI. Note: The premium controls for observable worker characteristics (education, age, and gender) in figure 2.11 is based on data from 72 countries. Figure 2.12 is based on data from 44 countries. The public sector premium is higher when con- workers have either health insurance or social secu- sidering total compensation and not just wages. rity, the two most important and widely provided The premium estimates just noted are for basic benefits, than private sector workers in all coun- wages only and exclude monetary allowances and tries in the sample. Combined with the public sec- benefits. As figure 2.14 shows, the public sector tor wage premiums, the data suggest significantly provides disproportionately more benefits than the higher average total compensation in the public private sector. A higher proportion of public sector sector compared to the private sector. These higher ■ ◾ ▪ Figure 2.13 Premiums are ■ ◾ ▪ Figure 2.14 Public sector lower for more skilled workers receive occupations more benefits Benefits in each sector Senior officials Percent of employees with health insurance or social security 1.0 CHN Professionals CHL GEOURY BGR 0.8 BRA Technicians CRI Private sector DOM 0.6 ARGCOL ZWE ALB ECU Clerks ZAF PRY LKA SLV 0.4 PER MEX MDG MOZ BWA EGY NIC GTM BOL Elementary occupations 0.2 LBR GHA TZA RUS BFA MWI KSV HND BGD −0.1 −0.05 0 0.05 0.1 TCD NPL HTI KHM 0 Wage premium 0 0.2 0.4 0.6 0.8 1.0 vs formal employee Public sector Source: WWBI. Source: WWBI. 12 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T public sector benefits persist when controlling for ■ ◾ ▪ Figure 2.15  Women earn a higher wage worker demographics. The premiums are likely premium in the public sector to be even higher when in-kind benefits, such as Overall housing, and nonpecuniary benefits, such as job Sub-Saharan Africa security, are factored into total compensation. South Asia The public sector also pays a higher wage premium Latin America and Caribbean to women. Across all countries, women receive Eastern Europe and Central Asia a wage premium of 27 percent, as compared to East Asia and Pacific 14 percent for men, and the premium is higher in –0.1 0 0.1 0.2 0.3 all regions of the world (figure 2.15). The reasons For women For men for these gender differences could be due to both Source: WWBI. employment and wage factors. Women could be disproportionately represented in public sector occu- What is less well known is that the gender pay pations, such as teaching and nursing, which have a gap is lower in the public sector. Figure 2.16 higher wage premium. Wage discrimination against shows that across all regions in the world, and women may also be lower in the public sector across all country income categories, the ratio given that salary scales are determined by regula- of female-to-male average wages is higher tions and may leave less discretion for managers to in the public sector than in the private sector. differentiate pay based on gender for similar occu- Globally, in our sample of countries, women’s pations and workers. We explore these issue in the average wages are 88 percent of male wages in following sections. the public sector, as compared to 81 percent of male wages in the private sector. The public sec- tor wage gap is substantially less than the private The gender pay gap sector wage gap in the Middle East and North Africa as well as in South Asia, and is roughly It is well known that women globally earn signifi- equivalent in Europe and Central Asia. Surpris- cantly less in the private sector than men for doing ingly, the difference between the public and the same work. In the United States, women private sector pay gaps increases with country famously earn “79 cents to the dollar a man income levels. Across countries, the pay gap is earns” for the same hours worked, and similar lower in the public sector in 42 out of the 60 coun- disparities hold across developed and developing tries in our sample (See figure 2.17. The countries countries for wage labor.6 In the developing world, to the left of the 45 degree line are those in which the biggest source of income gaps are lower rates the female-to-male mean wage ratio is higher in of labor force participation of women. As a result the public sector). This lower pay gap could be in urban Ghana women only earn about two-thirds a reason, in addition to the higher public sector of what men earn, in Colombia they earn half, and wage premium for women, why females gener- in Sri Lanka they earn less than a third of their ally have a higher share of employment in the male counterparts (Sandefur 2018). public sector than in the private sector. 6 See Blau and Kahn (2000) for a review of US studies, and the World Development The public sector gender pay gap has also Report 2012: Gender Equality and Development (World Bank 2012) for a comprehensive discussion of the global evidence. declined over time. The female-to-male ratio of I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 13 ■ ◾ ▪ Figure 2.16  The gender pay gap is lower in the public sector than in the private sector a. Female-male pay ratio b. Female-male pay ratio (Most recent year available) (Most recent year available) East Asia and Pacific Low income Eastern Europe and Central Asia Lower middle income Latin America and Caribbean Middle East and North Africa Upper middle income South Asia High income Sub-Saharan Africa 0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 Private sector Public sector Private sector Public sector Source: WB staff calculations based on WWBI. Note: The female-to-male ratio is the mean female wages divided by the mean male wages for all employees in the public and private sector respectively average wages increased from 0.85 in 2000–7 to sector mirrors the decline that has been observed 0.9 in 2008–2016, rising in 27 of the 35 countries in the academic literature for the private sector. for which we have data in each time period (See figure 2.18. The countries to right of the 45 degree There are many possible explanations for the line are those for which the female-to-male wage gender pay gap. In theory, the it could be due ratio was higher in 2008–16 compared to 2000–7). to the systematic differences in worker char- This decline in the gender pay gap in the public acteristics (for example, age, education, and ■ ◾ ▪ Figure 2.17  The gender pay gap is ■ ◾ ▪ Figure 2.18  the public sector gender lower in the public sector pay gap has declined than in the private sector over time Gender pay gap Gender pay gap over time Public sector only Female-to-male wage ratio in the public sector 1.5 1.2 Female-male wage ratio average 2000–7 DOM SLV PHL SLB MOZ SLV RWAWBG BWA COLJOR 1.0 1.0 PAK GHA LBRTHA KEN CRI PAN HND LKA JOR NAM ZAF LKA URY MDGALBMNG VNM KSV SRB ECU GEO BFA PNG ZAR GHA ZAR ARG MEX PER GTM GTM CRI COL HND CHNPRY URY BOL MDV EGY GABCHL RUS HTI KAZ PER MEXMNG PAK PAN NPL UKR BGR UZB BRA PRY MDV CHL ECU ARG ETH CAF 0.8 ETH MDGALB NIC MRT BOL KAZ TJK MRT GMB BRA 0.5 RUS MDA 0.6 0 0.4 0 0.5 1.0 1.5 0.4 0.6 0.8 1.0 1.2 Female-to-male wage ratio in the private sector Female-male wage ratio average 2008–16 Source: WB staff calculations based on WWBI. Note: The female-male ratio is the mean female wages divided by the mean male wages for all employees in the public and private sector respectively. 14 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T experience) between men and women; to wage major occupational groups. Women make up discrimination; or to employment segregation by roughly 30 percent of “senior officials” (managers gender, which means differences in the types and executives) in both the public and private of jobs that women and men do. There are few sectors and are approximately 50 percent and significant differences in education qualifications 45 percent of the “professionals” and “techni- by gender in the public sector, although women ” respectively, in the public sector, much cians, may have less work experience because of inter- higher than in the private sector. This relatively ruptions in their career due to childcare. There is high female representation in these two catego- also less scope for wage discrimination—paying ries is not surprising as these two occupational less to women working in the same job and with groups include teachers and nurses, jobs that tra- the same education and experience as men—in ditionally viewed as being female ones. Women the public sector given that wage levels are set in also make up 55 percent of clerical occupations. legislation and managers have limited discretion in setting pay. This legal equality between men Occupation segregation persists across country and women is a major nonpecuniary benefit of income levels. The proportion of women who are the public sector. senior officials is below 20 percent in low-income countries and increases with country incomes, The pay gap in the public sector is driven largely but is below parity with men even in the upper- by occupation segregation by gender. Women middle-income countries. By contrast, the propor- are underrepresented in higher-paying mana- tion of women in clerical occupations increases gerial occupations and overrepresented in lower- with country income; women make up a surprising paying clerical ones. Figure 2.19 shows the 80 percent of clerical public sector occupations proportion of women globally in some of the in high income countries (figure 2.20). ■ ◾ ▪ Figure 2.19 Women are ■ ◾ ▪ Figure 2.20  Public sector occupational underrepresented in senior segregation is persistent positions and overrepresented across country income in clerical occupations levels Proportion of female by occupation Proportion of female by occupation (Public sector only) Senior officials Low income Professionals Lower middle income Technicians Upper middle income Clerks Elementary occupations High income 0 0.1 0.2 0.3 0.4 0.5 0 0.2 0.4 0.6 0.8 Private sector Public sector Senior officials Professionals Clerks Source: WB staff calculations based on WWBI. Note: These are the major ISCO occupational groups where an occupation is defined as “a set of jobs whose main tasks and duties are characterized by a high degree of similarity.” I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 15 ■ ◾ ▪ Figure 2.21  Perceptions of the quality of recruits between BR and non-BR agencies in Indonesia b. “Graduates from elite universities consider a career at your a. “Your ministry/agency easily recruits high quality staff” ministry/agency as best possible public sector option” BR agencies 11 12 77 BR agencies 8 14 78 BR status of agency BR status of agency Non-BR Non-BR 24 23 53 18 27 55 agencies agencies Disagree or strongly disagree Neither Agree or strongly agree Disagree or strongly disagree Neither Agree or strongly agree Note: BR = strategically important Bureaucracy Reform agencies. The reasons for employment segregation in the were able to probe whether higher-paying orga- public sector are not well understood. Presum- nizations within a country were perceived to ably, some of the findings from studies of the be attracting a better pool of candidates than private sector are relevant and include gender lower-paying organizations in that country. differences in time use, particularly household In Indonesia, staff in the higher paid ministries work and childcare, that constrain women’s occu- were more likely to state that their ministry could pational choices, and informational market failures recruit high-quality candidates than staff in the that result in low female participation in certain lower-paid ministries. At the time of the survey occupations, perpetuating women’s lack of access in 2014, there were two- to fourfold variations to the social networks of other females necessary in staff pay levels between the few central min- to gain access to these jobs (World Bank 2012). istries and agencies that were deemed to be strategically important and undergoing internal Wages and selection reforms, termed Bureaucracy Reform (BR) agen- cies, and others. Our survey asked respondents Do the high public sector wages encourage in the BR and non-BR agencies their views on high-ability candidates to seek government agency prestige and quality of new recruits. employment? While common sense would Seventy-seven percent of survey respondents ” there has been assume the answer is “yes, in the BR agencies, as compared to 53 percent surprisingly little empirical exploration of this of respondents in the non-BR agencies, either question. One recent experimental study, in agreed or strongly agreed with the statement which researchers could exogenously vary the that their agency could recruit high-quality staff wage offers for the same advertised position, (figure 2.21). There was a similar difference in showed that higher wages indeed attracted responses between the BR and non-BR agencies higher quality candidates (Dal Bo, Finan, and to the question on whether graduates from elite Rossie 2013). While such randomized allotment universities considered a career in their agency of wage offers is not feasible at scale, our surveys as the best possible public sector option. 16 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T What Practices Shape the Work of Bureaucrats? Management practices, information technology (IT), and business processes are the main elements of the government production function that shape the work of bureaucrats and influence on the attitudes and behaviors of bureaucrats. Both the public administra- tion and the labor and organizational economics literatures stress the importance of management, although emphasizing different aspects of it. Merit, instead of political appointments, as the main criterion for selection and promotion is the hallmark of the Weberian bureaucracy and can help create strong professional norms that drive performance. Principal-agent approaches underline the cen- trality of monitoring and incentives to motivate staff to perform. The World Management Survey (WMS), a rigorous method of quantify- ing managerial and organizational practices, has revealed that the quality of management is the main driver of innovation and produc- tivity in firms across the world (Bloom and Van Reenen 2007; Cirera and Maloney 2017). Our surveys of civil servants used an adapted instrument to measure the quality of management of public sector organizations. Merit-based selection, allocation, and promotion Merit is generally used as the main criterion for selection and promotion of staff. Except for Liberia, a significant majority of respondents in the surveyed countries believed that “the selec- ” or selection tion process identifies the best people for the job, was based on interviews and written examinations (figure 3.1, left panel). A similar pattern of responses was found for questions I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 17 ■ ◾ ▪ Figure 3.1  Merit is generally the main criterion for selection but less so for job allocation a. Merit is the main criterion for selection b. Merit is the main criterion for job allocation 87% 97% 84% 79% 75% 59% 60% 60% 45% 42% All Ghana Indonesia Pakistan (SRB) Liberia All Ghana Indonesia Nigeria Pakistan (SRB) on merit in promotion, which suggests that for the civil service, he or she must be allocated patronage and politicization is less prevalent a job, and potentially a series of jobs as the per- than commonly believed in developing-country son moves through the service. Merit is viewed bureaucracies. Most civil servant jobs are as less of a factor in job allocation, with the dif- advertised, and examinations are a usual pro- ferences between selection and allocation sub- cedure to screen applicants, particularly at the stantial in Ghana and Indonesia (figure 3.1, panel entry level, and our data suggest that these b). This divergence is not surprising, as matching formal mechanisms work well. individuals to jobs requires much more manage- rial discretion and cannot be easily specified in Merit, however, is less of a factor for determining procedures. Allocation is clearly just as import- job allocation. Once an officer has been selected ant as recruitment, as getting the right people into the right job requires that they are actually ■ ◾ ▪ Figure 3.2  Merit as the main criterion for posted there, and not just that they are in ser- selection varies considerably vice. Understanding these dynamics of internal across organizations labor markets within the public service is an area 1 that requires further research. .8 These national averages mask considerable within-country variation across organizations .6 in the use of merit. Figure 3.2 charts organi- .4 zational averages of stated beliefs in merito- cratic recruitment for each of the organizations .2 Ghana Indonesia in Ghana, Indonesia, and Pakistan. Each marker 0 Pakistan on the graph represents the proportion of staff 0 .2 .4 .6 .8 1 in an organization that believe merit is the main Ranking of organisation method for recruitment. Overall, these data are Note: Binary variable constructed with “Agree” or “Strongly Agree” coded as 1 and “Strongly disagree,” “Disagree,” and “Neither agree nor disagree” coded as 0. consistent with other findings in the literature 18 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T that most bureaucracies are “neither islands of to carry out their tasks; and the regularity and excellences nor basket cases. ” (Meyer-Sahling, robustness of performance evaluations. In our Schuster, and Mikkelsen 2018). surveyed countries, and following the practice of the WMS, these questions on management Merit matters for staff motivation; organiza- practices were aggregated into an overall “man- tions where selection and promotion are more agement quality” index. In Ethiopia, Ghana, meritocratic are likely to have more motivated and Nigeria, for example, the index varied con- staff. In Pakistan, organizations that were siderably across departments, ministries, and self-assessed by civil servants as being more local administrations. In Ghana, management meritocratic in promotions also had staff that quality not only varies across organizations but were more satisfied with their experience in the also the units within organizations (figure 3.3). service. Our findings are consistent with those This dispersion in management quality implies of Meyer-Sahling, Schuster, and Mikkelsen that the experience of being a civil servant, (2018), who conducted the only other compa- despite a common regulatory framework, is rable cross-national survey of bureaucrats. They highly dependent on the organization that the find that civil servants who reported that they individual is employed in and underlines the were hired through political connections were importance of local context on government less motivated to work hard and serve the pub- capability. lic or were less satisfied with their jobs. Other Goal setting and monitoring are the strongest cross-national studies also find that merit-based aspects of management across the countries recruitment is associated with lower levels of surveyed. Setting goals and targets, and dis- corruption (Rauch and Evans 2000). aggregating them for units and individual staff members, can provide a sense of direction and The quality of management purpose to employees. In terms of targeting, The quality of management varies considerably across public sector organizations within coun- ■ ◾ ▪ Figure 3.3  The quality of management varies tries, underlining the need for a granular under- considerably across government organizations within countries standing of public administration. The WMS conceptualizes good management as regular Diversity in management scores across divisions in Ghana’s civil service 1 monitoring and communication of organization goals and activities, efficient personnel and .5 policy management, and the judicious use of Management score incentives. In our surveys, the WMS is modi- 0 fied for relevance to public sector bureaucra- −.5 cies and asks respondents questions on goal setting and how these are communicated to −1 Organisation average staff; the extent of monitoring of the achieve- Division average −1.5 ment of these goals; how managers distrib- 0 10 20 30 40 50 ute tasks across employees, involve staff in Ranking of organisation problem solving, and give staff the autonomy Source: Rasul, Rogger, and Williams 2017. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 19 overall performance is “middling” on average, contribute toward decisions) and flexibility (the with cross-country scores of around 3 out of ability of staff to adjust to new demands and a maximum of 5. This score indicates that minis- ways of working) are low across our sample of tries and departments in bureaucracies assign countries, which indicates that because there is targets to their organizations, which are then no system for identifying and solving problems, broken down to managerial and employee-level staff are not involved in providing solutions or targets. These are generally well understood adjusting their tasks to meet these new require- by mid-level staff, although they are not nec- ments. Suggestions may be taken from staff, essarily communicated in a clear and concise but these occur on a rare and ad hoc basis, and manner to lower level staff, which suggests it is nonmanagerial staff do not actively contribute not always clear how the targets contribute to in staff meetings and rarely provide any kind of their organization’s goals. Importantly, the tasks feedback. New ideas or practices are adopted, assigned to staff on a day-to-day basis are not but in an informal or isolated manner. Gener- always related to those targets. The average ally, organizations are slow to integrate new score for monitoring is similar to that of target- practices into their operations. ing in most countries surveyed, which indicates that organizations generally track a limited num- How can managers be incentivized to take ber of performance indicators somewhat reg- management seriously? Management requires ularly, which are seen and reviewed by senior effort and motivation on the part of managers. management only. Although some of this data While performance pay schemes have signifi- may be shared with relevant staff members, cant limitations when applied to bureaucrats, there is limited communication about perfor- as opposed to frontline staff, due to the prob- mance and how is it is monitored and reviewed. lems of multitasking and measurement, they This suggests that although some monitoring is may have a role in incentivizing managers. Per- taking place at the directorate level, it is done formance pay can potentially act as an indi- inconsistently. rect lever by providing incentives for improved management, resulting in improvements in the By contrast, performance evaluations, incen- performance dialogue with staff: organizational tives, staff involvement, and flexibility are the goal setting, teamwork toward achieving orga- weakest aspects of management. Perhaps nizational goals, and linking individual perfor- unsurprisingly, given our findings on percep- mance appraisals to those organizational goals tions of meritocracy in the bureaucracy, average (Marsden 2009). scores for performance monitoring and incen- tive systems are low, indicating that overall poor Evidence from the survey in the Philippines sug- performance is addressed inconsistently and gests that performance incentives can improve on an ad hoc basis. Performance may be eval- management. Staff perceptions are depicted in uated through a formal system and rewarded figure 3.4 and reveal positive responses to ques- (financially or nonfinancially), but there is no tions that cut across individual performance rank- system or clear criteria for doing so. The aver- ings. Interviewees noted that the performance age scores for staff involvement (the ability bonus scheme had motivated management to be of staff to become involved in operations and more focused on target setting and monitoring 20 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T ■ ◾ ▪ Figure 3.4  Philippines: perceptions of the effect of performance-related pay (PRP) on goal setting, monitoring of targets, teamwork, and performance appraisals (percent) Performance a. "Because of PRP, management is more focused Performance b. "As a result of PRP, management is more diligent ranking of on working with staff to serve the public's interest" ranking of in setting performance indicators and targets" respondent respondent Top Top 11 21 78 9 15 76 performers performers Second best Second best 10 21 69 9 17 74 category category Third best Third best 18 22 60 13 23 64 category category Disagree or strongly disagree Neither Agree or strongly agree Disagree or strongly disagree Neither Agree or strongly agree Performance c. "As a result of PRP, staff in the unit are working Performance d. "Because of PRP, the performance appraisal ranking of better together to achieve targets" ranking of process has significantly improved" respondent respondent Top Top 7 14 79 21 27 52 performers performers Second best Second best 8 14 78 23 30 47 category category Third best Third best 13 18 69 29 33 38 category category Disagree or strongly disagree Neither Agree or strongly agree Disagree or strongly disagree Neither Agree or strongly agree and in engaging staff in the process. Staff across or strongly agreeing). Taken together, the evi- the performance spectrum strongly believed dence suggests that performance pay had incul- that management was more focused on working cated improvements in management that can with staff to serve the public’s interest (between over time help create a performance culture in 60 percent and 78 percent agreeing or strongly the bureaucracy. agreeing) and more diligent in goal setting and in monitoring accomplishment against goals The politician-bureaucrat nexus is central to the (between 64 percent and 76 percent agreeing functioning of the public sector labor market, or strongly agreeing). They were similarly clear particularly the selection and motivation of pub- in their views that teamwork in achieving depart- lic employees. Senior bureaucrats need to be mental performance targets had improved due accountable to elected officials, but this respon- to performance-related pay (between 69 percent siveness needs to be balanced with appropri- and 79 percent agreeing or strongly agreeing). ate rules and protections for civil servants that Staff were also of the view, although less clearly, regulate the pressure on bureaucrats from polit- that performance pay had triggered improve- ical actors. Informal interactions between poli- ments in the performance appraisal process ticians and civil servants may be a means for (between 38 percent and 52 percent agreeing politicians to circumvent these protections. Our I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 21 surveys explored these informal engagements Report 2016, is a measure of the strength of in various ways, such as the extent to which these information systems. The DAI has three politicians had influenced the selection of proj- components: digital identification systems as ects or procurement decisions (Philippines); core platforms for interoperability; core admin- whether politicians had tried to influence hiring istrative systems to automate and streamline decisions or promotions (Ghana); or the propor- government activities; and on-line services for tion of projects in which there had been political businesses and citizens. The Index rises in value interference (Nigeria). The relatively low levels with country incomes, but there are numerous of self-reported interactions of bureaucrats developing countries, such as India and Rwanda, with politicians in Ghana and the Philippines that score much higher than their predicted are in stark contrast to the highly politicized values, pointing to the considerable investments bureaucracy of Nigeria, where 77 percent of made in digital technologies (figure 3.6). staff reported that the recent projects they had worked on were subject to some degree of There is increasing empirical evidence that political interference (figure 3.5). digital technologies are improving the function- ing of “street-level” bureaucrats. Digital iden- tification programs, such as India’s Aadhaar, Digital technology enable poor countries to leapfrog the decades- and bureaucracy long processes that developed countries went Governments have invested heavily in digital through to build their traditional, paper-based technologies to automate core administrative civil registration systems. These programs are tasks, improve the delivery of public services, a platform for a variety of services, and have and promote transparency and accountability. The Digital Adoption Index (DAI), developed ■ ◾ ▪ Figure 3.6 Developing countries by the World Bank for the World Development have invested heavily in digital technologies ■ ◾ ▪ Figure 3.5  Percentage of staff 1 KOR SGP reporting that politicians JPN CHL MYS URY PRT ITAARE DEU had improperly tried SRB BRA KAZ RUS LTU EST ESP MLT ISRAUT .8 IND NLD to influence projects or AZE ALB COL MEX TUR BHR BEL FRA LUX ARG USA NOR EGY ZAF ROU LVA personnel decisions RWA NGA ARM GEO GTM CHN SWE CHE MARPRY MUSHUN CZESVN BRN FIN MNG BIH ECU PER THA POLSVK CRI OMN .6 KEN LKA TUN BOL DOM HRV SAU CAN GBRAUS QAT PAK MDA JOR Index score UGA TZA BGD AFG GHA SLV IRN MKD BGR MNE PAN TTO NZL DNK KGZ ZMB UZB IDN VNM GRDVENSYC Philippines 25% NPL HND BTN JAM LBN NIC DMA BRB KWT IRL CMR UKR AGO PHL FJI BWA SEN VCT GRCCYP BDI SLE LCA ISL .4 COD GMB ETH MRT GNB KHM CPV DZA PNG BLR IRQ MLIYEM MOZGIN ZWE COG TCD BLZ SUR BHS MWI MDG COM LSO TLS HTI TJK SDN WSM TON NAM GAB MDV Nigeria 77% LBR GUY CUB BFA LAOSWZ VUT ATG KNA TGO MMR SLB .2 NER BEN KIR CAF TKM Ghana 13% TUV MHL GNQ 0 1,000 5,000 25,000 All 38% GNI per capita Source: DAI. 22 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T been shown, for example, to reduce leakages studies that complementary organizational and in government welfare programs (Muralidha- institutional changes are necessary to reap the ran, Niehaus, and Sukhtankar 2014). Electronic benefits of IT, both by firms and governments filing of taxes has reduced the administrative (Brynjolfsson and Hitt 2000; Cirera and Maloney burden to businesses and citizens by mini- 2017; Garicano and Heaton 2010). Bureaucracies mizing interactions with tax officials, reducing can have a disincentive to reorganize their work opportunities for rent seeking, and lowering to take full advantage of digital technologies, as the time it takes businesses to file tax returns, the concomitant efficiency improvements can make payments, and receive refunds (Kochan- lead to a reduction in the agency’s budget and ova, Hasnain, and Larsen 2017). Citizen service staff. The opportunities offered by digital tech- centers and one-stop shops that provide citizen nologies for better collaboration and integration and business services—such as registration, across government also bump into bureaucratic licensing, records, and bill payments—in a sin- structures and budget and legislative processes that reinforce vertical stovepipes. There is lit- gle physical space or web portal are increas- tle evidence, even in high-income countries, ingly common in developing countries and have that digital technologies have fundamentally greatly reduced the time and hassle for citizens changed the way bureaucracies are managed to receive services. And digitally enabled moni- (Fountain 2001). toring of service providers has reduced teacher and doctor absenteeism in India, Niger, Paki- Our surveys are also suggestive of this imbal- stan, and Uganda.7 ance between heavy investments in IT, on the one hand, and the effective usage of these sys- The impact of information systems on core tems, on the other. In some countries there is administration, however, is less clear. The empiri- a more basic constraint of lack of IT skills (as cal evidence is very limited, with one study show- discussed earlier), irregular electricity, and inad- ing that e-procurement systems improve the equate funding for maintenance of IT systems. competitiveness of public procurement (Lewis- Local government officials in Ethiopia and Nige- Faupel et al. 2016). However, many large digital ria stated that they had Internet access on only technology projects fail, as evidenced by the 21 percent and 3 percent of days, respectively, low self-evaluated success ratings of the World and several local governments had no access at Bank-funded digital technology projects. Only all. In Ethiopia, 79 percent of respondents men- 35  percent of the 530 information technology tioned that all professional staff have access to projects funded by the Bank from 1995 to 2005 computers, but 47 percent of respondents dis- were rated as satisfactory or above, compared agree that maintenance and management of IT to 56 percent of all projects (World Bank 2016). equipment was sufficient. At 5 of the 18 local governments surveyed in Nigeria, managers Digital technologies require “analog comple- stated that they never had access to electricity, ments” to have impact. This was the main mes- and half the organizations only had power for sage of the World Development Report 2016, half the day on average. and it encapsulates the findings of numerous Conditional on sound management and ade- 7 See WDR 2016 for further examples. quate skills, digital technologies are making a I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 23 ■ ◾ ▪ Figure 3.7  Staff believe that IT has difference in bureaucracies. A case in point is the improved management and Pakistan SRB, a meritocratic and well-managed productivity in Pakistan SRB organization, where the staff strongly agreed Has the Management Information System... that the management information system has led to staff working harder, both because they Made some staff redundant 1% have more access to information and because they are being more closely monitored. It has Led to staff exerting more effort because of greater monitoring 77% provided more information on staff performance that is used in the annual performance appraisal Led to staff exerting more effort 85% process. And it has not led to any staff redun- because they had better information dancies (figure 3.7). These findings exemplify the Provided better information for the complementarity between technology and man- annual performance review process 85% agement for productivity improvements. 24 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T What Are the Attitude and Behaviors of Bureaucrats? The attitudes and behaviors of bureaucrats—concerning job satisfaction, work motivation, public service motivation, and trust— are key for government capability and performance. These attitudes are especially important given that monitoring and incentives, the standard toolkit for motivating workers in the principal-agent frame- work, are less applicable in bureaucracies due to the nature of admin- istrative jobs. Measurability of inputs, outputs, or outcomes is the necessary condition for performance incentives to be effective in the public sector, but these schemes are by nature largely confined to jobs in service delivery or revenue collection. For example, experimental studies have shown that financial incentives can improve the effort of tax collectors, school teachers, and health facility staff (Khan, Khwaja, and Olken 2014; Muralidharan and Sundararaman 2011; Basinga et al. 2010; Gertler and Vermeersch 2012). Most jobs in bureaucracies lack the measurable performance indicators necessary to condition such incentives, underlining the importance of selecting the right workers with high public service motivation and of norms and values based on professionalization (Dixit 2002; Perry and Wise 1990; Grant 2008; Wilson 1989). Intrinsic motivation and prosocial motivation are particularly important determinants of bureaucrats’ productivity. Intrinsic motivation is the desire to work hard for the enjoyment of the task itself rather than for the rewards, such as higher pay, promotion, or recognition. Public service, or prosocial, motivation is the desire to work and exert effort for the benefit of others and not for oneself. A large psychology and public administration literature shows that it is the combination of intrinsic and prosocial motivation that drives effort and work of public sector workers (Grant 2008; Perry and Hondeghem 2008), a finding I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 25 that is corroborated by more recent experi- Indonesian civil servants. Nigeria has the highest mental studies that show that more prosocial overall level of satisfaction in our surveys, with 89 doctors and nurses perform better in their jobs percent of civil servants neutral or satisfied with (Callen et al. 2015; Deserranno 2017). their job overall. We do not see major differences in terms of gender or on a range of other demo- Our surveys explore job satisfaction, work moti- graphics. However, there is substantial variation vation, and public service motivation in a variety across government organizations. Figure 4.1 of ways and analyze how compensation and plots, for each organization in our samples, the management practices are associated with proportion of civil servants in that organization motivation. These include general questions that are neutral or satisfied with their job. These about job satisfaction, questions about overall proportions are plotted against the percentile of motivation levels and how they have changed average satisfaction at an organization within the over time, and questions about why bureaucrats country. Ghana is a relative outlier in the extent of joined the public sector and the reasons why variation its organizations exhibit in average satis- they continue to work there. We also explore the faction. The other surveys fluctuate between 60 extent to which bureaucrats are satisfied with their wages, whether managers are effectively percent and 100 percent of staff satisfied with screening for candidates with prosocial motiva- their jobs overall. tion, and whether sound management is associ- Despite their relatively high overall compensation, ated with higher motivation. bureaucrats are generally unsatisfied with their pay. On average, only 40 percent of the survey Job satisfaction respondents in the seven countries were satisfied General satisfaction with jobs varies considerably with their pay levels, despite a high public sector across countries and across government organi- wage premium in each of these countries (fig- zations within countries. Fifty-three percent of ure 4.2). These wage satisfaction numbers range Ghanaian civil servants are neutral or satisfied with their jobs, as compared to 85 percent of ■ ◾ ▪ Figure 4.2 Bureaucrats are generally unsatisfied with ■ ◾ ▪ Figure 4.1  Work satisfaction varies their pay levels considerably across government 57% 54% organizations within countries 49% 46% 100 40% 28% Proportion satisfied (percent) 26% 80 18% 60 40 All Ghana Philippines Indonesia Nigeria Pakistan (SRB) Liberia Ethiopia Ghana 20 Indonesia Nigeria 0 Pakistan 0 20 40 60 80 100 Note: Percentage of respondents reporting that they are satisfied with Ranking of organisation (percent) their pay. 26 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T from a low of 18 percent in Ghana to a high of 57 ■ ◾ ▪ Figure 4.3  There is considerable pay percent in the Philippines. Other surveys confirm dispersion for similar workers in these findings; a cross-national survey of 23,000 Rio de Janeiro municipality civil servants in 10 countries also found that only a. Admin and professional staff 37 percent of civil servants were satisfied with their 0 20 40 60 80 salary (Meyer-Sahling, Schuster, and This incon- Monthly gross salary (R$ thousands) gruity between actual wages and perceptions of wages is puzzling and a concern, as it suggests that the high fiscal outlays on wages, a significant 0 5 10 15 20 25 cost to taxpayers, may not be yielding a motivated b. Support staff public workforce. One reason for this dissatisfac- 0 10 20 30 40 tion could be a result of bias, with bureaucrats benchmarking their pay against that offered by a small segment of high-paying private sector com- panies, often multinationals. Unfortunately, this 0 5 10 15 20 25 bias is reinforced by standard “pay benchmark- Years of service ing” studies conducted by human resource con- Source: Calculations based on micro payroll data. sulting firms that usually restrict their sample to Note: Each dot is data for an individual employee. large, formal sector firms and often find a public sector wage penalty. These studies also usually do Brazilian states, and in Bosnia and Herzegovina, not consider total compensation, in-kind as well and is likely reflective of the complex regulatory as monetary, or the considerable nonpecuniary regimes that govern individual compensation, benefits of public sector jobs. which results in nontransparent and highly varied wages, depending on the unique employment Another reason for dissatisfaction with wages history of an individual. could be the considerable dispersion and inequity in pay in the public sector compared to similar Yet another reason for pay dissatisfaction could workers in similar occupations. Studies have be pay compression or relatively flat pay levels shown that relative wages can have a significant over the course of an employee’s career. While impact of job satisfaction and worker productivity the I2D2 surveys are cross-sectional and cannot (Card et al. 2012; Breza et al. 2018). Figure 4.3 track individuals over time, examining the dis- presents data from Rio de Janeiro municipality of tribution of wages provides a rough measure monthly gross wages of employees in two of the of pay progression. A standard measure is the major occupational groups by years of service. pay compression ratio, defined as the ratio of Each point in the graphs represents the data of an the 90th percentile wage to the 10th percentile individual employee and reveals that wages vary wage. The public sector compression ratio aver- five- to tenfold for staff with similar experience in ages approximately 7 across the 63 countries similar occupational groups. There is significant in the sample, with no correlation with country pay dispersion even within specific occupations, income levels (figure 4.4, panel a). This num- and even for the same occupation within organi- ber, however, is driven by a few outliers, and in zations. A similar pattern is observed in several most countries the ratio is approximately 4. Pay I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 27 ■ ◾ ▪ Figure 4.4  The public sector wage distribution is narrower than that in the private sector a. Public sector pay compression ratio b. Pay compression ratio in each sector 20 20 15 15 Compression ratio Public sector 10 10 5 5 0 0 6 7 8 9 10 0 5 10 15 20 Log of GDP per capita Private sector Source: Calculations based on the WWBI. Note: Pay compression ratio is the ratio of the 90th percentile wage to the 10th percentile wage. Data is for 63 countries. compression in the public sector is lower than Management practices matter for job satisfac- in the private sector in 45 of the 63 countries tion. In Nigeria, a standard deviation improve- (figure 4.4, panel b). This suggests that even ment in management related to staffing and though the public sector pays a wage premium, flexibility is correlated with a 7 percent increase public sector workers will be dissatisfied with in the likelihood that bureaucrats state they their pay if they only compare their wages to the are satisfied with their job. Figure 4.5 displays top of the distribution of private sector wages. this relationship, with each point in the scatter Anecdotally, it does seem that bureaucrats are graph an organization. It shows a positive slop- selective in their benchmarking. ing upward trend in the relationship between the quality of management (increasing on the y axis) and the proportion of public officials ■ ◾ ▪ Figure 4.5  Work Satisfaction Is Corelated who state they are satisfied with their job. Inter- with the Quality of Management in Organizations in Nigeria estingly, no such relationship exists between a management index related to monitoring 50 and incentives and motivation. In Ethiopia, we find that better management practices improve 0 satisfaction by a very similar magnitude to in Management score Nigeria, and that giving staff substantive roles –50 in the policy-making process is the dominant positive driver of this. –100 As an aside, we find little evidence in any of these countries that time in service substan- –150 50 60 70 80 90 100 tially impacts on satisfaction and public ser- Proportion of staff satisfied with job vice motivation, indicating that the incentive 28 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T environments are critical determinants of pub- ■ ◾ ▪ Figure 4.6  Motivation decreases over time lic sector motivation and satisfaction. 50 40 Work motivation The extent of motivation among bureaucrats 30 Percent varies both cross-nationally also across organi- 20 zations within countries. Our surveys measure employee motivation in several ways, ranging 10 from general questions of respondents’ satis- Ethiopia faction with their jobs, reasons for joining the 0 Pakistan public sector, and assessment of their current 0 50 100 150 200 motivation levels compared to when they joined Staff motivation the public sector. The Ghana, Indonesia, and Pakistan, surveys ask about the experience of self-reported satisfaction with their public working in the public sector relative to the pri- sector career is surprisingly low. Figure 4.7 vate sector and find substantial cross-national shows the data from Ethiopia, where bureau- differences, with 53 percent of Ghanaian civil crats who report higher levels of wage satisfac- servants neutral or satisfied with their jobs tion are no more likely to report higher levels of relative to the private sector, as compared to motivation. These correlations are also low in 85 percent of Indonesian civil servants who are Ghana (0.20), Indonesia (0.18), Nigeria (0.24), satisfied. Within Ghana, satisfaction levels vary and Pakistan (0.05). These data suggest that considerably across organizations, with between while pay may be a component of the motiva- 60 percent and 100 percent of staff satisfied tion of public officials, other factors are likely with their jobs overall. Civil servants’ motivation levels also decrease ■ ◾ ▪ Figure 4.7  Satisfaction with wages has no over time. One of the questions in our surveys association with staff motivation asks respondents to imagine that when they 200 entered the service their motivation was 100 per- cent; they are then asked what their motivation 150 is now relative to then. In Ethiopia, 61 percent of Staff motivation bureaucrats feel less motivated now than when 100 they first entered the service, 31 percent feel just as motivated, and 8 percent feel more moti- 50 vated (figure 4.6). In Pakistan (SRB) the decline in motivation is smaller. 0 Compensation is weakly associated with motiva- 1 2 3 4 Wage satisfaction tion. Across the countries, the correlation between Note: Each data point represents a survey respondent score on wage satisfaction (from 1 to 4, the level of self-reported wage satisfaction and with 4 the most satisfied) and motivation (from 0 to 200). I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 29 more important. This finding is consistent respondents consider merit to be the single with the literature and supports the view that most important criteria for promotion, and bureaucrats are largely motivated by the non- 81 percent are confident they will receive a pro- monetary rewards of working within the ser- motion if they perform well. However, 21 percent vice (Perry and Hondegheim 2008). of respondents would like to change jobs in the next two years, the top reason being limited Part of the reason for low motivation may be promotion opportunities where they currently the low career mobility across government work. This view reflects the structural constraints organizations. A typical civil servant spends of the SRB, as a there is no clear career trajec- 16 years in the same organization in Nigeria, and tory for staff to transition into the broader civil 10 years in Ethiopia. In Indonesia, the average service once they reach the upper levels of the bureaucrat spends 95 percent of his or her time organization. in the service at the organization at which they are surveyed. In Nigeria, only a quarter of offi- There is no clear relationship between aggregate cials surveyed were satisfied with the number management quality and work motivation. Intu- of transfers they have had, and almost half itively, we would expect to find a positive rela- wished to be transferred more. The reasons tionship between the two. However, our survey why there is such little horizontal mobility in results are unclear. One reason for this weak bureaucracies needs to be better understood, association is that the quality of management but such long tenures within an organization is assessed using only managers’ responses, underline the importance of management to and the level of motivation is measured for non­ keep staff motivated. They also point to the managerial staff only. Therefore there could be need to institutionalize whole-of-government significant differences in experience not cap- management cadres, such as the Senior Exec- tured by these two variables. Another reason is utive Service in Chile, Korea, and some other that there may be other factors, such as a lack of countries of the Organization for Economic resources or general dissatisfaction with wages, Co-operation and Development, which are open that could be affecting motivation regardless of to all staff through a competitive process and can the quality of management in their organiza- be an aspiration for all bureaucrats. tions. And finally, and perhaps most important, The lack of career mobility is a particularly aggregating management practices into a single motivation-dampening factor in specialized, index underestimates the complementarities “islands of excellence” organizations. The between the different dimensions of manage- Pakistan SRB is a case in point. It is a small, ment and the need to have all of them in place semi-autonomous revenue authority that pays to impact motivation. higher wages than the rest of the civil service and has strong meritocratic and competitive processes in place for recruitment and promo- Public service motivation tion, as evident from the survey responses. Public service motivation can be defined as Ninety-five percent of SRB staff were satisfied the desire to serve the public interest and is with their wages, and 64 percent were satis- empirically measured using the Perry scale. The fied with their benefits. Eighty-four percent of norms of serving the public and of practicing 30 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T ■ ◾ ▪ Figure 4.8  Career ambitions and job security are prominent self-reported reasons for joining the public sector a. Philippines b. Pakistan (SRB) 75 100 Future career ambitions 28.4% Fits my educational background 23.6% Percent naming reason Public service 16.9% Security 10.1% Salary 50 7.4% Working conditions 6.8% Status 25 3.8% Other 2.0% Location 1.4% 0 od ty Be ry on ible ts nc e w urs t p ad rso ocia tial isf s M n n r he u 0 10 20 30 tio io fi ri la l s tat lo ne ho S en cu iss Ot sa ac em ork na l s se ot Percent b at Go Jo x en Re Fle l ab as Pe va Ad self-sacrifice are integral to the ethic of bureau- of public service and mission (figure 4.8). In cracy, a conception that goes back to the work Nigeria, “The chance to serve Nigeria” was the of the classical scholars of bureaucracy. Sur- 8 most popular choice (37 percent), followed vey-based measures of public service moti- by “I was interested in the type of work” vation generally follow the approach of Perry (29 percent). These responses contrast with (1996), with questions along four dimensions: some studies that find that more intrinsically (i) attraction to public policy making, (ii) com- motivated citizens enter the public sector, but mitment to civic duty and the public interest, they are not surprising given that the high total (iii) compassion, and (iv) self-sacrifice. Our sur- compensation and considerable nonpecuniary veys simplify the Perry measure and focus on rewards of public sector employment are likely bureaucrats’ reasons for joining the public sec- to incentivize more materially oriented individ- tor and their desire to serve the public relative uals to seek public employment (Banuri and to concerns for personal career growth. Keefer 2016). This finding also underlines the importance of recruitment strategies to influ- Civil servants join the public sector for both ence the type of individuals who apply for gov- material and intrinsic reasons. The ideal of the ernment posts.9 Weberian bureaucrat who joins the public ser- vice to serve his or her country is only partially Better management is associated with greater met in our surveyed countries. Bureaucrats in public service motivation. In Ghana, the manage- the Philippines and Pakistan (SRB) listed job ment practices related to greater staff involve- security (79 percent) and future career ambi- ment in decision making increases bureaucrats’ tions (28 percent) as the two main reasons for score on the Perry public service motivation joining the government, respectively, ahead 9 See Dal Bo, Finan, and Rossie (2013) and Ashraf et al. (2014) for the importance of emphasizing public service in job advertisements to encourage more intrinsically 8 See Perry (1996) for a discussion. motivated individuals to apply I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 31 scale, in particular the score on the compassion in that survey in Pakistan and the Philippines, and public interest subindices. only 22 percent and a mere 3 percent of respon- dents, respectively, agreed that most people who they knew personally, or who were in Trust their neighborhood, could be trusted. Bureau- Bureaucrats are generally trusting of their crats report higher levels of trust for employees immediate colleagues but less so of those in in their own teams or units than employees in other organizations in the government. Given other organizations in government. In Pakistan the importance of teamwork, the extent of reg- SRB, for example, staff trust the general citizenry ular interactions and trust among bureaucrats more than bureaucrats from other ministries, is likely to be important for motivation and capa- which is surprising and suggestive of limited bility. Overall, the data suggest high levels of cross-agency interaction. trust, with on average 76 percent of staff across The determinants of trust within organiza- the five countries agreeing or strongly agreeing tions are not clear. In Ghana we find a negative with the question, “How much would you say impact of management quality on levels of trust. people can be trusted?” (figure 4.9). In Indo- Increasing the quality of management under nesia, even the organization with the lowest which an official works significantly reduces the levels of trust had 40 percent of officials stating likelihood that they will state that their ministers that they trust their colleagues. These trust lev- trust them. We find similar results in Nigeria, els are strikingly high when compared to levels where we asked, “To what extent would you of trust reported in the country in the World say employees of your organization trust each Values Survey (2010-2014 round). For example, other?” The likelihood that respondents stated that “All employees trust each other to under- take the commitments they make” is negatively ■ ◾ ▪ Figure 4.9  Percentage of staff correlated with management practices that reporting that they improve productivity and satisfaction. A stan- trust their colleagues dard deviation improvement in management Philippines 72% related to staffing and flexibility is correlated with a 37 percent decrease in the likelihood Pakistan (SRB) 80% that officials state they trust one another. Giv- Pakistan (FBR) 56% ing civil servants autonomy to work in the way that best suits them may damage the cohesion Liberia 98% of the organization. This is in contrast to the Indonesia 66% fact that in many settings officials state that Ethiopia 82% they trust each other more than their family or the general public. The tight rules of the public All 76% service may bind officials together but distort Note: Pakistan (FBR) is the Federal Bureau of Revenue other aspects of their work. 32 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T How Productive Are Bureaucrats? The need for a better understanding of bureaucracy is fundamen- tally about the need to increase the productivity of government, but measuring productivity is difficult. The preceding pages have explored the various elements of the government production function that we expect to impact productivity, but ultimately, we would like to measure productivity directly and explore more explicitly the links between these elements and productivity. Productivity is defined in economics as the ratio of outputs to inputs, with total factor produc- tivity, or the ratio of output to all factors of production, being a key determinant of long-run economic growth. Measuring the outputs of the public administration, however, is complicated. Beyond the prob- lem of defining these outputs, no market prices exist for many of its products (such as regulations); these products frequently have broad externalities (such as the public health benefits of immunization); or they have contingency value (such as planning for disasters); and they rely on a system of government units for their creation, making attribution of benefits or production complicated. There has been little effort to confront these challenges, although recently a series of studies has argued for proxying public sector outputs by a func- tion of their administrative costs.10 This provides an improvement over existing approaches, but it does not confront the concerns just mentioned nor decompose productivity into its constituent parts. Budget execution or audit data can provide a proxy measure for the quantity and quality of outputs. While expenditures do not imply social benefit, the short-run rate of financial disbursement can be an indicator of the long-run health of a project. In Bangladesh, we found For work in this direction, see, among others, Atkinson (2005), Dunleavy and Carrera (2014), and Gemmell et al (2017). 10 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 33 that disbursements halfway through imple- others completing literally nothing (figure 5.2). mentation were significantly higher for proj- These organizational averages mask substantial ects that would eventually be completed to a division-level variation. Walking across the corri- satisfactory quality compared with those that dor in a Ghanaian ministry may mean passage would never be completed. In Nigeria, audits through a transformation in productivity. of infrastructure projects by independent (non- The speed and quality of responsiveness to intra- government) engineers in teams with civil governmental requests can be another measure society organizations showed that many orga- of productivity. In Ghana, we measured how nizations never produce anything at all, while long it took line agencies to respond to requests other public agencies complete all their projects from the center of government, with the quality to a satisfactory quality (figure 5.1). of these responses assessed by civil servants, retired or still active, in terms of their adherence Moving beyond infrastructure, activity audits to both government guidelines as well as broader examining the completion rates of tasks that benchmarks of quality. We find that roughly a organizations committed to in their performance quarter of agencies never respond, and a third agreements with the central government are do so late, some many weeks after deadlines. another potentially valuable approach to mea- This approach enables us to identify those suring productivity. In Ghana, independent audi- organizations that are a bottleneck to the wider tors assessed the completion rates and quality productivity of the public service. of these tasks and found considerable variation across organizations, with some agencies com- Measuring productivity and its organizational pleting almost everything they commit to and determinants is an agenda for future work on ■ ◾ ▪ Figure 5.1  The quality of infrastructure ■ ◾ ▪ Figure 5.2  The quality of task delivery varies considerably completion varies considerably across government across government organizations in Nigeria organizations in Ghana Diversity in productivity across organizations in Nigeria’s civil service Diversity in productivity across divisions in Ghana’s civil service 1 1 Proportion of projects completed .8 .8 Proportion of projects .6 .6 .4 .4 .2 .2 0 0 0 20 40 60 0 10 20 30 Organization ranking by proportion of projects initiated Ranking of organization Proportion started Proportion completed Division average Oganisation average Source: Rasul and Rogger 2017; Rasul, Rogger and Williams 2017. 34 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T bureaucracy. Combining detailed measures improvements in public service (Best, Hjort, of the characteristics of public organizations and Szakonyi 2017; Rasul, Rogger, and Williams, with productivity data can provide quantitative 2017). Analysis of the Ghanaian management estimates of the benefits of civil service reforms. and productivity data, for example, implies that And the potential impact of these reforms is enor- a single standard deviation increase in the qual- mous, given, as studies have shown, that small ity of public management would increase GDP improvements in quality can lead to substantial by 8 percentage points. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 35 Conclusion This report is centered around a set of core questions on bureau­ cracy and its effects on government capability. What are the main features of the public sector labor market in terms of employment and compensation? What are bureaucrats’ attitudes toward their jobs and their behaviors with each other and the public? How well are they managed? Are they using digital technologies to innovate? And finally, how can we measure whether they are productive? The report used original, microlevel data in attempting to answer these ques­ tions, with often surprising findings, which open up potentially fruitful avenues for future research. We find that the public sector labor market is distinctive in both the characteristics of public employees and in the levels and structure of compensation. Bureaucrats are older, have higher academic qualifications, and are more likely to be women than private sector wage employees. These are the observable and measurable differ­ ences between workers in the two sectors. Public sector workers are also likely to be different on unobservable dimensions, particularly in their motivation for employment, in part because of the different compensation package offered by the public sector. The public sector pays a wage premium, particularly for lower skilled occupations, pro­ vides a higher proportion of compensation in the form of pecuniary and nonpecuniary benefits, and is fairer to women. These distinctive features of public sector compensation likely impact the types of individuals who seek public employment, and their motivation and commitment to serve the public. Our survey findings suggest that public sector workers are both intrinsically and extrinsically motivated, but it is still likely that the distribution of workers on this motivation dimension is different from that in the private sector. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 37 Our finding that the public sector pays a wage are 85 percent of male wages in the public premium is tentative, because it does not effec­ sector, as compared to 79 percent of male tively account for these differences in types of wages in the private sector, and this pay gap individuals who select to enter the public sector. has declined over time. Women are under­ As shown, the scale of the premium depends on represented in higher-paying managerial occu­ the choice of the private sector comparator, but pations, and overrepresented in lower-paying this is an incomplete finding. The cross-sectional clerical ones, and this occupation segregation wage regressions that are used to calculate the is the main driver of wage inequality and per­ premiums control for only the observable worker sists across country income levels. The rea­ and job characteristics and, therefore, cannot sons for employment segregation in the public account for the differentially motivated workers sector are not well understood and are an area between the two sectors (Disney 2017). One requiring further research. method to control for these selection effects is to use panel data and focus on the individuals The relatively high and more gender-equal pub­ who move between the public and private sector lic sector wages are helping select qualified and estimate the differences in their compensa­ candidates but are not sufficiently motivating tion levels. Since we would see the same indi­ bureaucrats after they have joined the service. vidual in both sectors, we can control for both There are two aspects to this weak association the observable and unobservable characteristics between wages and motivation, which suggests in estimating the wage differentials. This analysis relatively low public sector productivity. First, is planned for Brazil. bureaucrats are dissatisfied with their pay even though they generally receive a compensation The public sector can have significant distortion­ premium. The reasons for this low wage satis­ ary effects on the broader labor market, given faction could be the considerable wage disper­ its large size and generous compensation. These sion and pay inequity in public sector, and the include skewing individual employment prefer­ fact that bureaucrats may be benchmarking their ences toward the public sector and away from the pay against their peers rather than with workers formal private sector; raising reservation wages in the private sector. It may also reflect the rela­ for private sector jobs and contributing to volun­ tively flat pay progression given limited promo­ tary unemployment; and possibly even impacting tion opportunities in the public sector. Second, individuals’ education choices away from science, even bureaucrats who are satisfied with their technology, engineering, and mathematics fields, wages do not have higher levels of motivation. given that many public sector jobs seek more This finding contributes to the debate in the aca­ generalist skills. This effect of the public sector demic literature on whether pay can incentivize on the broader labor market and competitiveness of the economy is an area of future research, one performance and suggests that the existing is planned for Bosnia and Herzegovina through an civil service compensation schemes—pay that innovative labor force survey. is largely determined by seniority and only weakly tied to performance—are not motivating staff. There is gender inequality in pay and employ­ Whether alternative wage policies can motivate ment, although the gender wage gap is lower bureaucrats is a potentially rich avenue for further in the public sector. Women’s average wages research. 38 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T Management practices influence the attitudes how it is changing organizational practices, and and behaviors of bureaucrats and their productiv­ what impact it is having on bureaucrats’ atti­ ity. While some reasons for bureaucrats’ declin­ tudes and behaviors. One obvious area where ing motivation levels are structural, such as the digital technologies can have big impact is on limited opportunities for upward mobility in gov­ public employment and pay practices, given that ernment careers, others are likely due to poor many developing countries have invested in inte­ management. While merit is, surprisingly, the grated human resource and payroll systems. main criteria for recruitment across our sample These systems provide a rich source of data, of countries, it is less of a factor in job allocation. yet they are underutilized as an analytical tool, Day-to-day management is more likely to impact even in relatively advanced countries like Brazil. staff attitudes than infrequent life-cycle events A potentially valuable area for future World Bank like recruitment, job allocation, and promotion, technical assistance is helping countries use and the quality of these management practices, these systems to prepare regular reports that as measured by the World Management Survey, help improve wage bill management and public vary considerably across organizations. Manag­ employment policy. ers are failing to regularly involve their staffs in setting and monitoring goals for their organiza­ Finally, this report has suggested some new tion and in problem solving. Narrowing down approaches to measuring the productivity of which management practices particularly impact bureaucracy, but much more work is needed bureaucrats’ attitudes and behaviors, and how to in specifying the mechanisms through which incentivize managers to take management more bureaucrats interact with frontline agencies to seriously, are also areas requiring more research. deliver outputs to citizens. Task completion is a useful proxy for productivity in administration, but The intersection between management prac­ is not very meaningful for citizens. The personnel tices and digital technologies has only been and management linkages between the upstream lightly addressed in this report and in our data­ bureaucracy and the downstream citizen-facing sets. While lack of information technology (IT) agencies—how the daily work of bureaucrats skills and infrastructure constraints are limiting impacts the attitudes, behaviors, and productivity the effective usage of digital technologies in of teachers, doctors, and engineers—is an area low-income countries, we have not explored in that is underexplored and should be a priority detail how management and staff are using IT, for future research. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 39 ANNEX The Datasets Used in the Report Worldwide Bureaucracy Indicators A better understanding of public bureaucracies requires, at a mini- mum, country-level, comparable, cross-national data on the person- nel dimension of the state. Existing measures are almost exclusively based on expert perceptions, which have several limitations, such as measurement inconsistencies across countries and over time. To fill this gap, the Governance GP in collaboration with the Poverty GP and the DEC Data Group created a new dataset—the Worldwide Bureaucracy Indicators (WWBI)—on public sector employment and wages, and public-private wage and employment comparisons. This dataset draws on the following primary sources: 77 The International Income Distribution Database (I2D2), a module of the Global Monitoring Database: The I2D2 harmonizes nationally representative household surveys—both welfare and labor force surveys—from around the world, presenting data using the same variables and coding in each country and survey. 77 The Luxembourg Income Study: This source similarly harmonizes household surveys from several, mostly high-income countries. 77 The International Comparisons Program (ICP) wage survey: The ICP is a global statistical partnership that calculates purchasing- power-parity prices based on detailed comparative price data, including government wages for a standard set of occupations. The WWBI also capture some data from secondary sources, namely the International Labor Organization’s (ILO’s) employment database (ILOSTAT), and a new dataset on the wage bill produced by the International Monetary Fund. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 41 The country coverage and time series in the data- their abilities draw on surveys of approximately set varies by indicator. The bulk of the dataset 20,000 civil servants in seven countries: Ethiopia, is based on the I2D2 and covers 90 countries Ghana, Indonesia, Liberia, Nigeria, Pakistan, and and the period 2000 to 2016. The variables in the the Philippines. The surveys aim to contribute WWBI include the public sector wage bill; pub- to the development of diagnostic tools that will lic and private sector shares in total, wage, and allow us to better understand the incentive envi- formal sector employment; demographic char- ronments that lead to different types of behavior acteristics of public and private sector work- and the determinants of service delivery in the ers; relative wages within government across civil service. What distinguishes civil servant a standardized set of occupations; the benefits surveys from other types of surveys is not only given to public and private sector workers; public- their scale and scope but the clear shift from private wage differentials (controlling for various form to function. Much of the previous work in worker characteristics); gender wage differen- civil service reform defaulted to optimal “forms” tials in the public and private sectors; and wage for the civil service motivated by theory. Through distributions of public sector workers. the creation of improved survey modules, the survey builds a foundation for the systematic collection of data directly from individuals with Surveys of bureaucrats first-hand experience. Some basic meta data Information on human resource management on these surveys is provided in table A.1. All the practices, the attitudes and experiences of surveys entailed face-to-face interviews with a bureaucrats, and the key restraints to bureau- representative sample of civil servants, with very crats performing their duties to the best of high response rates. *** ■ ◾ ▪ Table A.1  The list of World Bank surveys of bureaucrats Year Unit of Selected Response Country Survey name fielded Unit of analysis observation sample rate Sampling method Ethiopia Ethiopia Civil 2016 Organizations and Individual officials and 2197 99.4% Random sampling Servants Survey officials organizations across 3 tiers of government Indonesia Public Employeee 2012 Organizations and Individual officials 4000 95.0% Random sampling Survey of officials Bureaucracy Reform Liberia Liberia Civil 2016 Central government Individual officials and Census of Monrovia- Servants Survey ministries organizations based officials Pakistan (SRB) Pakistan Survey of 2017/18 Departments of SRB Individual officials and 148 100.0% Census of SRB officials Tax Officials organizations based in Karachi Philippines Philippines Public 2013 Government Individual officials 2573 100.0% Stratified samplinq Sector Survey departments Nigeria Nigeria Civil 2010 Federal,state and local Individual officials and 5432 99.8% Stratified sampling Servants Survey civiI service organisations organizations and officials Ghana Ghana CiviI 2015 Central civil service Individual officials and 3111 96.0% Census Servants Survey organizations and officials organizat ions 42 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T To obtain reliable information on public sector surveyed, questionnaires are split into three organizations, it is important to recognize that tracks: an employee-level track, a director-level protocol and language use in civil services are track, and a head-of-organization (political appoin- country specific. The questionnaires are there- tee) track. While some modules are covered in fore created in consultation with members of all three tracks, others were designed to be partner government organizations. Pilot inter- administered to specific tracks only, and not to views are conducted to ensure question viability, all sampled individuals. Some questions and as well as to ensure that question wording and sections are also split into sector-specific tracks. phrasing has remained relevant and suitable to the context of each country surveyed. Table A.2 presents a comprehensive list of all modules used in the survey questionnaires To ensure that the questions remain relevant to in Ghana, Indonesia, Ethiopia, Liberia, Nigeria, each broad level of staff within the civil services Pakistan, and the Philippines. ■ ◾ ▪ Table A.2  Modules used in the surveys Module Description Demographics and work history Includes the basic characteristics of the official to allow us to separate the experience of the service by different groups as well as investigate the career trajectory of the individual civil servant, and thus determine what the context of their experience of the service has been to date Management practices Investigates the nature of basic management practices in performance management, targeting, talent management, monitoring, autonomy, and incentives Turnover Investigates the structure and nature of turnover in the organization from a management perspective Recruitment and selection Aims to identify the criteria on which recruitment into the service/ organization is based Attitude and motivation Explores officers’ attitude to different aspects of their job and their organization to get a sense of officers’ motivation and the sources from which they derive it—includes questions aimed at identifying perceptions on career expectations, culture, and mission alignment Time use Explores how the official uses his/her time during a typical work day or week by incorporating experimental methods for the measurement of time use in order to help build the knowledge base of how best to measure this aspect of civil servant life Bottlenecks Investigates the challenges encountered by officials in undertaking daily tasks, as well as the barriers to policy implementation and obstacles in the delivery of service delivery Stakeholder engagement Explores who officials interact with in their day-to-day work with the aim of understanding their current relationships in the service, and how this affects their experience Information Aims to explicitly assess the level of information that officials have about the demographic conditions of their jurisdiction and the state of services in their respective sectors Information technology Investigates information systems in the organization, the type of information available, and how accessible it is Reforms Aims to give us a sense of how individual officials feel about government reforms, and how these have impacted operations on the ground in each of the organizations Corruption/Ethics Explores civil servants’ attitudes toward, and perceptions of, different forms of corruption in their organization through two methods: direct elicitation and vignettes Benchmarking Assesses the extent to which woreda (district) operations benchmark against indicators measuring institutional quality (Ethiopia) (continues on page 44) I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 43 ■ ◾ ▪ Table A.2  Modules used in the surveys (Continued ) Module Description Public service motivation and Uses the Perry (1996) public service motivation scale and Levenson’s Internality, Powerful Others, and Chance (IPC) scale locus of control (1981) to measure motivation and behavioral characteristics of civil servants through two self-administered questionnaires Capacity Experimental module testing the basic skills of civil servants in fields relevant to their work Service delivery Explores civil servants’ attitudes to different approaches to service delivery, and attempts to gauge their perceptions on prioritization, ethics, and engaging with the public Ease of doing business Investigates processes that take place at the divisional and organizational levels with a focus on time frame, monitoring, and barriers to implementation Internal labor market of the Assesses the staffing needs and labor gaps of the organization, with a focus on planning, prioritization of resources, and public sector labor matching. The module is broken down into three subsections: (i) labor demands by managers, (ii) labor supply by employees, and (iii) labor-matching. Skills and knowledge assessment Assesses officials’ knowledge of basic technical skills and basic laws and regulations relevant to their sector 44 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T References Andrews, M., L. Pritchett, and M. Woolcock. 2017. Building State Capability: Evidence, Analysis, Action. Oxford, UK: Oxford University Press. Ashraf, N., O. Bandiera, and K. Jack. 2014 “No Margin, No Mission? A Field ” Journal of Public Experiment on Incentives for Public Services Delivery. Economics 120: 1–17. Atkinson. A. 2005. Atkinson Review: Measurement of Government Output and Productivity for the National Accounts. Basingstoke: UK. Palgrave Macmillan. Bandiera, O., Nava, A., and Lee, Scott S. 2016. Do-gooders and go-getters: career incentives, selection, and performance in public service delivery. Working paper. Banuri, S., and P. Keefer. 2016. “Pro-social Motivation, Effort, and the Call to ” European Economic Review 83: 139–164. Public Service. Basinga, P., Paul J. Gertler, A. Binagwaho, Agnes L. B. Soucat, Jennifer R. ” Paying Primary Health Care Sturdy, Christel M. J. Vermeersch. 2010. ” World Bank Policy Research Working Centers for Performance in Rwanda. Paper 5190. Washington, DC. Best, M., J. Hjort, and D. Szakonyi. 2017. “Individuals and Organizations as ” Sources of State Effectiveness, and Consequences for Policy Design. National Bureau of Economics Working Paper 23350. Blau, F ” Journal of Economic ., and L. Kahn. 2000. “Gender Differences in Pay. Perspectives 14 (4): 75–99. Bloom, N., and John Van Reenen. 2007. “Measuring and Explaining Manage- ” Quarterly Journal of Economics ment Practices across Firms and Countries. 122 (4): 1351–1408. Breza, E., S. Kaur, and Y. Shamdasani. 2018. “The Morale Effects of Pay ” Quarterly Journal of Economics 133 (2): 611–663 Inequality. Brynjolfsson, E., and L. M. Hitt. 2000. “Beyond Computation: Information ” Technology, Organizational Transformation, and Business Performance. Journal of Economic Perspectives 14 (4): 23–48. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 45 Callen, M., S. Gulzar, A. Hasanain, and Y. Khan. 2014. ” Public Sector: Evidence from Police Departments. “The Political Economy of Public Employee Absence: Journal of Labor Economics 28 (1): 167–201. Experimental Evidence from Pakistan. ” National Gemmell, N., Nolan, P. and Scobie, G. 2017. “Public Bureau of Economics Working Paper 22340. sector productivity: Quality adjusting sector-level Card, D., Mas, A., Moretti, E. and Saez, E. 2012, ” New Zealand Pro- data on New Zealand schools. “Inequality at work: The effect of peer salaries ductivity Commission Working Paper 2017/2. ” American Economic Review on job satisfaction. Gertler, P., and C. Vermeersch. 2012. “Using Perfor- 102(6), 2981–3003. ” mance Incentives to Improve Health Outcomes. Cirera, X., and William F. Maloney. 2017. The Innova- World Bank Policy Research Working Paper 6100. tion Paradox: Developing-Country Capabilities and Washington, DC. the Unrealized Promise of Technological Catch-Up. Grant, A. M. 2008. “Does Intrinsic Motivation Fuel the A World Bank report. Washington, DC. Prosocial Fire? Motivational Synergy in Predicting Dal Bo, E., F. Finan, and Martin A. Rossie. 2013. ” Jour- Persistence, Performance, and Productivity. “Strengthening State Capabilities: The Role of nal of Applied Psychology 93 (1): 48–58. Financial Incentives in the Call to Public Service. ” IMF (International Monetary Fund). 2016. Managing Quarterly Journal of Economics 128 (3): 1169–1218. Government Compensation and Employment: Deserranno. E. 2017. “Financial Incentives as Signals: Institutions, Policies, and Reform Challenges. Experimental Evidence from the Recruitment of Washington, DC: IMF. Village Promoters in Uganda. ” Buffett Institute Khan, Adnan Q., Asim I. Khwaja, and Benjamin A. Global Poverty Research Lab Working Paper Series. Olken. 2014. “Tax Farming Redux: Experimental Diamond, Jack. 1977. “Wagner’s ‘Law’ and the Develop- ” Evidence on Performance Pay for Tax Collectors. ” Developing Economies 15 (1): 37:59 ing Countries. Quarterly Journal of Economics 131 (1): 219–71. Disney, R. 2017. https://worldbankgroup.sharepoint. Kochanova, A., Z. Hasnain, and B. Larson. 2016. com/portals/hub/_layouts/15/PointPublishing.aspx? “Does E-Government Improve Government Capac- app=video&p=p&chid=26a6dca1-c93b-4e51-8d4e- ity: Evidence from Tax Administration and Public 575a41cea731&vid =e9498ecd-af13-4904-aac5- ” World Bank Policy Research Working Procurement. 0cf538fb050b Paper 7657. Dixit, A. 2002. “Incentives and Organizations in the Lewis-Faupel, S., Y. Neggers, B. A. Olken, and R. Pande. ” Journal of Public Sector: An Interpretative Review. 2016. “Can Electronic Procurement Improve Infra- Human Resources 37 (4): 696–727. structure Provision? Evidence from Public Works Dunleavy, P., and L. Carrera. 2014. Growing the Pro- ” American Economic Journal: in India and Indonesia. ductivity of Government Services. Cheltenham, Economic Policy 8 (3): 258–83. UK: Edward Elgar. Marsden, D. 2009. The Paradox of Performance- Fiszbein, A., D. Ringold, and H. Rogers. 2011. Making Related Pay Systems: Why Do We Keep Adopting Services Work: Indicators, Assessments, and Them in the Face of Evidence That They Fail to Benchmarking of the Quality and Governance of Motivate? London: London School of Economics, Public Service Delivery in the Human Develop- Centre for Economic Performance. ment Sectors. World Bank Policy Research Work- Meyer-Sahling, J., C. Schuster, and K. Mikkelsen. 2018. ing Paper 5690. Washington, DC. Civil Service Management in Developing Countries: Fountain, Jane E. 2001. Building the Virtual State: What Works? Report for the UK Department for Information Technology and Institutional Change. International Development (DFID). Washington D.C: Brookings Institution Press. Moulton, B. 1990. “A Reexamination of the Federal- Garicano, L., and P. Heaton. 2010. “Information Tech- ” Private Wage Differential in the United States. nology, Organization, and Productivity in The Journal of Labor Economics 8 (2): 270–93. 46 I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T Muralidharan, K., P. Niehaus, and S. Sukhtankar. 2016. on an Unconditional Teacher Salary Increase in “Building State Capacity: Evidence from Biometric Indonesia. Quarterly Journal of Economics 133 (2): ” American Economic Review Smartcards in India. 993–1039. 106 (10): 2695–2929. Sandefur, J. 2018. “Chart of the Week: Gender Pay Muralidharan, K., and V. Sundararaman. 2011. “Teacher Gaps around the World Are Bigger Than You Think, ” Performance: Experimental Evidence from India. ” and Have Almost Nothing to Do with Girls Schooling. Journal of Political Economy 119 (1): 39–77. https://www.cgdev.org/blog/chart-week-gender- Perry. J.L. 1996. “Measuring Public Service Motiva- pay-gaps-around-world-are-bigger-you-think-and- tion: An Assessment of Construct Reliability and have-almost-nothing-do-girls ” Journal of Public Administration Research Validity. Wilson, J. Q. (1989), Bureaucracy: What Government and Theory 6(1): 5–24. Agencies Do and Why They Do It. New York, Basic Perry, J.L and A. Hondeghem. 2008. “Building Theory Books. and Empirical Evidence about Public Service Moti- World Bank. 2003. World Development Report 2004: ” International Public Management Journal vation. Making Services Work for Poor People. Washington, 11 (1): 3–12. DC: World Bank. Perry, J. L., and L. R. Wise. 1990. “The Motivational World Bank. 2012. World Development Report 2012: ” Public Administration Bases of Public Service. Gender Equality and Development. Washington, DC: Review. 50(3): 367–373. World Bank. Rasul, I., and D. Rogger. 2017. “Management of World Bank. 2016. Making Politics Work for Devel- Bureaucrats and Public Service Delivery: Evidence opment: Harnessing Transparency and Citizen ” Economic Journal from the Nigerian Civil Service. Engagement. Policy Research Report. Washington, 128: 413–446. DC: World Bank. Rasul, I., D. Rogger, and M. Williams. 2017. “Manage- World Bank. 2016. World Development Report 2016: ment and Bureaucratic Effectiveness: A Scientific Digital Dividends. Washington, DC: World Bank. ” Mimeo. Replication. World Bank. 2017. A Fair Adjustment: Efficiency and Rauch, J., and P. Evans. 2000. “Bureaucratic struc- Equity of Public Spending in Brazil. Washington, DC: ture and bureaucratic performance in less devel- World Bank. ” Journal of Public Economics 75: oped countries. WWBI Worldwide Business Indicators). Databank. 49–71. Washington, DC: World Bank. https://datacatalog. Ree, J., K. Muralidharan, M. Pradhan, and H. Rogers. worldbank.org/dataset/worldwide-bureaucracy- 2017. “Double for Nothing? Experimental Evidence indicators. I N N O V AT I N G B U R E A U C R A C Y F O R A M O R E C A PA B L E G O V E R N M E N T 47