88486 Pay Flexibility and Government Performance A Multicountry Study June 2014 © Copyright 2014 International Bank for Reconstruction and Development/THE WORLD BANK Contents Acknowledgmentsiii Executive Summary v Introduction1 The Theoretical Framework 5 Performance-related pay 5 Differentiation6 Hypotheses7 Context matters 8 Assessing the Evidence: Review of Literature on Performance-Related Pay 13 The Case Studies: Pay Flexibility Schemes 20 Performance-related pay 21 Differentiation26 Assessing the Evidence from the Case Studies: Performance-Related Pay 30 Direct pay flexibility levers 30 Indirect pay flexibility levers 36 Assessing the Evidence from the Case Studies: Pay Differentiation 40 Direct pay flexibility levers 40 Indirect pay flexibility levers 42 Summary of Empirical Analysis 44 Fiscal impact 44 The degree of delegated authority 44 Conclusions and Policy Recommendations 47 How did the hypotheses hold up? 47 Implications for pay policy in developing countries 49 Notes53 References55 i ii PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Boxes 1 Arguments for Long-Term Career-Based Incentives Rather Than Short-Term Performance‑Based Incentives for Public Sector Staff 7 2 The “Effort Bargain” 8 3 Active Management and Managerial Effort 9 Figures 1 The Link between Pay Flexibility and Performance 9 2 Aggregate Findings on PRP 14 3 Findings by Research Quality (High Internal and External Validity Studies Only) 14 4 Findings for Craft and Coping Jobs 15 5 Numbers of Studies and Their Findings 18 6 PRP for Selected Careers in Brazil’s Federal Government 22 7 Minas Gerais PRP Bonus Size Increases over Time Based on Consistency of Performance 23 8 PRP as a Percentage of Total Pay in the Philippines 24 9 Pay Differentiation in Indonesia 27 10 Pay Differentiation in Brazil 28 11 Pay Differentiation in the Russian Federation 29 12 Perceptions of Effort in the BR Agencies in Indonesia 31 13 Results of the Performance Evaluation of Permanent Staff in Minas Gerais 32 14 Perceptions of Attendance in BR and Non-BR Agencies in Indonesia 33 15 Perceptions of PRP and Individual Performance Appraisals, Effort, and Motivation in Coping Jobs in the Philippines 34 16 Overall View among Civil Servants of the PRP Scheme in Coping Jobs in the Philippines 35 17 Perceptions of the Effect of PRP on Goal Setting, Monitoring of Targets, Teamwork, and Performance Appraisals in the Philippines 39 18 Perceptions of Effort, Morale, and Quality of Recruits between BR and Non-BR Agencies in Indonesia41 19 Perceptions of Pay Inequity across Ministries and Agencies and within BR Ministries in Indonesia42 20 Staff Perceptions of Management and Willingness to Accept Restructuring in Indonesia 43 21 The Link between Pay Flexibility and Performance 47 Tables 1 Hypotheses on the Impact of Pay Flexibility on Civil Service Performance 10 2 Five Key Design Elements of PRP Schemes 10 3 James Q. Wilson’s Classification of Job Types 11 4 The Studies Reviewed 13 5 Key Design Features of PRP in Case Study Countries 25 6 A Sample of Performance Indicators for the Group-Based Bonus Scheme in the Philippines 35 7 Summary of Evidence about the Impact of Pay Flexibility on Civil Service Performance 45 Acknowledgments This report was prepared by a multicountry Philippines by Zahid Hasnain and Sheheryar team of World Bank staff and staff from other Banuri, and a literature review on performance- international and national institutions. The related pay by Zahid Hasnain, Nick Manning, lead authors are Zahid Hasnain and Nick Man- and Jan Pierskalla (Ohio State University). ning (World Bank), drawing on case studies The team is grateful to government officials from Chile by Mariano Lafuente (Inter-Amer- from Korea, Indonesia, the Philippines, Malay- ican Development Bank); Thailand and Malay- sia, Thailand, Chile, Brazil, and Russia for sia by Willy McCourt (World Bank); Brazil by their cooperation in sharing their views and Barbara Nunberg (Columbia University) and providing data for the study. Renato Brizzi (consultant); the Russian Fed- The team is also grateful to a number of eration by Maya Gusarova (World Bank) and World Bank colleagues who helped facilitate Nikolay Klishch (National Research University data collection and commented on and pro- Higher School of Economics); the Republic of vided guidance for this work. They include Korea by Pan Suk Kim (Yonsei University); Indo- Erwin Ariadharma, Soren Davidsen, Pablo nesia by Zahid Hasnain; and the Philippines by Fajnzylber, Frederico Gil-Sander, Kai Kaiser, Zahid Hasnain and Sheheryar Banuri (World Alma Mariano, Yasuhiko Matsuda, Shabih Bank). It also reflects insights from surveys of Mohib, Anne Sevilla, Geoffrey Shephard, Staf- government officials in Indonesia by Phil Keefer fan Synnerstrom, Maria Tambunan, Rogier (World Bank) and Sheheryar Banuri and the van den Brink, and Tony Verheijen. iii iv Executive Summary This study examines whether financial incen- diverse jobs is difficult, it is often necessary Pay flexibility can tives through pay flexibility can improve the to use proxies that reflect the “engagement” performance of staff in government bureaucra- of staff (their commitment to the organiza- potentially improve cies. Its main messages for pay policy are: tion’s mission and observed willingness to performance directly • Pay flexibility can improve performance. use flexibility in working methods to help through financial • Pay f lexibility works most strikingly in deliver a product or service) or “organiza- changing managerial behavior. tional citizenship” (their willingness to pro- incentives and • Improving public sector performance does not vide extra effort to achieve goals). indirectly through need to wait for systematic pay rationalization or pay simplification throughout government. The study aims to provide policy advice to the improved management • Pay flexibility can work with rather than range of governments that are considering instead of long-term career incentives. these reforms. In Organisation for Economic • The strategy and implementation of pay Co-operation and Development (OECD) flexibility reforms must take into account countries, pay flexibility is a departure from the extent of fragmentation and complexity the archetypical single pay scale, which has of the existing public sector pay structure in been found to be too rigid in an increasingly the country. dynamic labor market and unable to meet per- formance objectives under fiscal constraints. About the study By contrast, in emerging market and middle- Pay flexibility is defined in the study as: income countries pay flexibility was either part • Performance-related pay (PRP): Enabling pay of a move to liberalize a rigid, centralized pay to differ for workers doing the same job by model or introduced on top of an already com- linking a portion of pay to the achievement plex pay regime in an attempt to formalize a of performance targets. haphazard structure that had evolved with little • Differentiation: Differences in pay between regard to the larger fiscal or incentive impact. apparently similar workers across agencies, Drawing on the large theoretical literature career groups, and geographical locations on PRP, and on the much more limited one on that reflect the need to compete for specific differentiation, the study identifies two ways skills in the labor market. that pay flexibility can potentially improve performance: And performance is measured by: • A direct effect of the financial incentive on • Improvements in the quality of staff in a staff behavior and quality, which we call the government ministry or agency through direct pay flexibility lever. recruiting and retaining more highly skilled • An indirect effect through more effective and motivated personnel. management that in turn improves staff • Improvements in individual line employees’ behavior, which we call the indirect pay flex- “effort.” Because measuring effort across ibility lever. v vi PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E The primary question We used the literature to distill specific performance are rarely available, and informa- hypotheses on the possible impacts of PRP and tion on inputs, behaviors, and processes can in this study is of differentiation: usually be indirectly imputed only through whether pay flexibility • PRP hypotheses: staff perceptions. The primary question in this • Hypothesis 1. PRP as a direct pay flexibil- study is whether pay flexibility contributes to contributes to these ity lever: these changed methods of working and rela- changed methods • Hypothesis 1a. PRP can have a direct tionships to the job in ways that have a positive incentive effect by improving individ- impact on outputs and outcomes. of working and ual engagement and organizational Each of these hypotheses is hotly debated in relationships to the citizenship and by inducing staff to the literature, with several arguments against exert more effort to achieve the out- both PRP and differentiation. The main job in ways that have puts and outcomes linked to the points are that PRP can cause many perverse a positive impact on incentive. consequences when performance cannot be outputs and outcomes • Hypothesis 1b. PRP can have a direct accurately measured, that PRP can crowd out incentive effect on staff quality, result- intrinsic motivation, and that the pay inequity ing in the recruitment and retention that both PRP and differentiation create hurts of more qualified staff who are likely staff morale, effort, and teamwork. The study to do well under the PRP scheme explores whether and in what contexts these (sorting). hypotheses hold. Two key contextual variables • Hypothesis 2. PRP can act as an indi- are the design features of the PRP plan and the rect lever by providing incentives for nature of the public sector job in which PRP more effective management, resulting is implemented. The relevant characteristics of in improvements in the performance the job (based on James Q. Wilson’s typology) dialogue with staff—organizational goal are whether the job outputs are easily observ- setting, and teamwork toward achieving able and measurable to managers and exter- organizational goals and linking indi- nal agents (craft jobs, such as teaching, health vidual performance appraisals to those care, and revenue administration) or not (cop- organizational goals. ing jobs, such as finance, planning, and other • Differentiation-based hypotheses: core policy functions). The five PRP design ele- • Hypothesis 3. Differentiation can act as a ments are whether the incentive is individual direct lever through the recruitment and or group based, the time horizon of the incen- retention of better-quality staff for prior- tive, the nature of the performance evaluation, ity activities (sorting). the size of the incentive, and the probability of • Hypothesis 4. Differentiation can act receiving the incentive. indirectly by providing incentives for The study evaluates existing evidence greater effort by managers, pushing through a comprehensive literature review of them to improve the performance dia- PRP. It then examines pay flexibility for the logue with staff and to increase efforts by core public administration in emerging market line staff. countries through case studies of PRP and dif- The evidence used to examine these ferentiation in Brazil, Chile, Indonesia, Korea, hypotheses is for the most part self-reported Malaysia, the Philippines, Russia, and Thai- perceptions of changed behaviors and falls land. Data in the case studies were collected short of the ideal of actual measures of out- through structured interviews with experts puts and outcomes. Wherever data on outputs and, in Indonesia and the Philippines, large (such as revenue collection, fines for nonfilers representative surveys of government officials. of taxes, and teacher attendance) and even outcomes (such as student test scores) were Findings from the literature review available, they were used in the study. How- on PRP ever, this report focuses on the core adminis- The literature review, unlike other reviews of tration, for which comprehensive measures of PRP, disaggregates the evidence by the quality PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  vii of the empirical study, the different types of frontline managers became aware of the per- The literature review public sector jobs based on the craft and cop- verse incentives that a myopic focus on only ing classification, and the country context. The revenue collection could generate and later and the case studies main findings: 93 of the 153 studies show some included taxpayer facilitation to ensure sus- find that PRP can form of positive effect of PRP, and 65 of the 110 tainability of revenue collection over the long increase productivity studies of craft and coping jobs find positive run. There was also some evidence of a positive effects. Limiting the analysis to high-quality sorting effect in Minas Gerais. In the World in teaching, health studies of craft and coping jobs, 37 of 53 report Bank survey of government officials in Indone- care, and revenue positive results. (The evidence is overwhelm- sia, the revenue agency scored the highest on ingly for craft jobs, since there are only three questions gauging effort and staff engagement. administration high-quality studies of coping jobs, and there In Chile, PRP was viewed quite favorably in are no high-quality studies of coping jobs in the revenue authority and the civil registry. In developing countries.) Most of the literature Malaysia the only positive reports of PRP were has explored the effect of PRP on staff effort, from the revenue authority. with only a few studies examining the impact By contrast, there was little support for on staff quality through sorting, and even Hypothesis 1 for coping jobs, and in some fewer exploring the impact of indirect pay flex- cases evidence of negative effects. By defini- ibility levers. tion, these are jobs for which outputs can- Overall, the literature review finds sup- not be easily measured, so the performance port for Hypothesis 1a for jobs such as teach- evaluations that form the basis of the finan- ing, health care, revenue administration, and cial incentive are either based on subjective job placement, which have more measurable evaluations by supervisors and review panels outputs and outcomes. There is more limited or on quantitative input or process measures. support for Hypothesis 1b of PRP improv- Subjective appraisals put the onus on manag- ing sorting. Context is important, with more ers to credibly distinguish among staff, diffi- positive findings in developing countries, cult in most bureaucracies. The tendency in especially in teaching. Several studies identi- Brazil, Chile, and Thailand was for the vast fied problems of unintended consequences, majority of staff to be given a best or next or “gaming” of the incentive program. There best performance rating. As a result, the per- is not enough evidence for Hypothesis 1 and formance bonus was given out with close to Hypothesis 2—the direct incentive effect and a probability of 1 and therefore became a de the indirect incentive effect through improved facto salary supplement that could not have management—in the core civil service in non- a direct incentive effect. Where quantitative OECD countries. measures were used, they tended to be heavily process-oriented, increasing the risks of gam- Findings from the case studies: PRP ing behavior. Brazil, Chile, Korea, Malaysia, the Philippines, However, evidence from Brazil and Indo- Thailand, and, to a limited extent, Indonesia nesia revealed that PRP had an impact on the have implemented PRP. These PRP plans dis- extremes of the performance distribution by play a high degree of diversity in the five main reducing staff absenteeism and helping disci- design features, and all these countries have pline blatantly incompetent staff. Outlier staff plans that apply to both craft and coping jobs. members were fairly easy to identify, so manag- The case studies reinforce support for ers could credibly sanction them. Hypothesis 1 for craft jobs. In Brazil, especially Korea, Malaysia, and the Philippines tried in the state of Minas Gerais, the PRP scheme to counter the tendency toward uniformly led to higher productivity in the police and high performance ratings and an equal dis- the revenue authority, as measured by weap- tribution of the performance bonus by man- ons seizures, number of police operations, and dating a forced distribution of performance revenue collection. Notably, the performance ratings. This risky policy can harm staff targets for the revenue authority evolved as morale, and its efficacy depends on the level v i i i PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E In Indonesia of trust in an organization and the legiti- Findings from the case studies: macy of performance appraisals among staff. Differentiation differentiation helped In Korea the plan has been generally well The case studies found strong support for gain buy-in from staff received. For example, when given the choice Hypothesis 3. Evidence from Brazil, Chile, between individually differentiated incentives Indonesia, and, to a lesser extent, Russia—all for restructuring and and equal pay for all within a group, 32 of 44 of which have significant differentiation— pushed management to government agencies opted for the former reveals that the higher paid agencies or groups and none for the latter (with the others choos- of staff were better able to attract and retain improve recruitment ing some form of mixed option). By contrast, high-quality staff. This increase in staff qual- procedures and in the Philippines, which has a combination ity did not appear to come at the expense of of individual- and group-based incentives demotivation through greater pay inequi- performance appraisals from a mandated ranking of both working ties for other groups. As the Indonesia survey units and individuals within those units, the showed, pay inequity has been the norm in survey revealed little direct effect on staff many of these countries, and the added ineq- effort. Instead, the survey found a divergence uity from differentiation may be only margin- of views, depending on the performance rank- ally more demotivating for staff in the less ing of the respondent, the effectiveness of the privileged agencies. Differentiation needs to individual performance appraisal process, the be limited, however, to a few priority groups of transparency of individual performance rat- staff. Without these constraints, as the recent ings, and the impact of the incentive on staff experience of Brazil suggests, differentiation morale. can create disruptive competition among agen- The case studies found support for Hypoth- cies over salary increases and discourage inter- esis 2: PRP did improve performance through agency cooperation. the indirect route of better management, There was also some support for Hypoth- specifically through setting better goals and esis 4. Data from Indonesia suggest that dif- encouraging greater teamwork. In the Phil- ferentiation can spur other organizational ippines the World Bank survey showed that reforms, where it helped gain buy-in from staff staff believed that the performance bonus for restructuring and pushed management to scheme had motivated management to focus improve recruitment procedures and perfor- more on organizational target setting, moni- mance appraisals. toring the accomplishment of those targets, engaging staff in this process, and fostering Conclusions and policy greater teamwork and collaboration among recommendations staff. Overall, staff were positive about the Overall, six messages can be drawn more gen- PRP scheme, despite the lack of credibility of erally for pay policy, particularly for the more the individual performance appraisal process, challenging low-income country contexts suggesting that the direct and indirect pay in Africa, South Asia, and elsewhere. First, flexibility levers were having opposite effects pay flexibility can improve performance, an on individual productivity and organizational important finding given the general skepti- citizenship. In Minas Gerais, staff believed that cism in the public administration literature on PRP was integral to the performance agenda this topic. These reforms are not a silver bul- in the state and helped clarify expectations let, and they involve tradeoffs and risks. Poorly and individual goals and targets, broadly com- designed pay flexibility plans can cause consid- plemented results-based management reforms, erable harm by encouraging perverse behav- and encouraged greater delegation of human ior and unintended consequences, and some resource management authority from the cen- task reallocation and gaming behavior should tral personnel office to line ministries and be expected. With proper monitoring, these agencies. Similar positive management effects tradeoffs can be managed. At a minimum, were noted in Korea and, to a more limited the study suggests that pay flexibility plans be extent, in Malaysia. implemented in craft jobs with a monitoring PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  ix regime in place to detect and respond to gam- of the country’s existing public sector pay Pay flexibility works ing behavior. structure. Although pay simplification is not Second, pay flexibility works most strik- a necessary prior step before introducing pay most strikingly in ingly in changing managerial behavior, which flexibility, the extent of “messiness” of the pay changing managerial has several implications. The PRP scheme regime has implications for the pay flexibility behavior, which designed for coping jobs should encourage strategy. In simpler systems, ambitious, across- these management changes. A large group- the-board pay flexibility reforms pose less risk has implications for based bonus PRP scheme is preferable, despite if there is explicit recognition of possible per- the design of pay the difficulty in establishing unit-level perfor- verse behavior and unintended consequences mance targets, because it bypasses the problem and if experimentation and learning-as-you- flexibility schemes of distinguishing between individuals’ perfor- go are built into the plan. Ideally, even in mance and puts the spotlight on management these systems pay flexibility would be initially improvements as the key enabler linking PRP restricted to a few organizations to limit the to better performance. Pay flexibility does not administrative burden of the validation and need to be introduced across the board—it monitoring systems necessary for effective pay should be introduced asymmetrically where flexibility. In complex systems there is the risk there is some basic minimum level of manage- that pay flexibility could degenerate into yet rial competence and should complement per- another element of the messy pay regime with formance budgeting and other management few productivity gains and a further weaken- reforms. In terms of sequencing, PRP should ing of central fiscal control and management come after initiation of results-based manage- coherence. This risk can be mitigated by limit- ment so that it has some managerial resources ing pay flexibility to a select few high-priority to build on, such as performance indicators organizational “islands,” which are chosen and progress review mechanisms, and provides either because they are the highest priority incentives to frontline managers to better use or because they are managed relatively well. these resources. These are the staff whose productivity improve- Third, improving public sector perfor- ments are considered to be the most important mance does not need to wait for systematic pay for government performance. rationalization or pay simplification through- And sixth, many questions remain, and out government. “Purposeful” pay complexity much more research is needed. The evidence through pay flexibility can improve perfor- this study has drawn on primarily concerns mance at the margin even when layered on self-reported perceptions of processes and top of a chaotic pay regime and poor human behavior change and not improvements in resource management. This is an encouraging actual outputs and outcomes, and so causal finding given the technical and political chal- claims cannot be made. The impact of pay lenges of comprehensive pay rationalization flexibility reforms varies considerably based and the poor track record of such reforms. on contextual factors that go beyond the two Fourth, pay flexibility can work with rather factors—the type of public sector job and the than instead of long-term career incentives. design features of the pay flexibility scheme— The PRP scheme in Minas Gerais is a good that this study has looked at. How pay flexibil- example of complementarity between short- ity interacts with existing formal and informal and long-term incentives, since the size of rules, culture, and the institutional arrange- the incentive increased with sustained good ments and capacity within government to col- performance. lect and validate data on performance and Fifth, the strategy and implementation of to coordinate pay flexibility across the pub- pay flexibility reforms must take into account lic sector are all key issues that require more the extent of fragmentation and complexity investigation. x Introduction An effective and efficient government is a public administration excluding key service The center sets the central objective for countries at all levels of delivery staff such as teachers, medical person- economic development. Government bureau- nel, and police. These limitations are in part to rules of the game cracies are very different from their private sec- keep the analytics manageable and in part to for the wider public tor counterparts because they are bound by a home in on one of the more dynamic areas of sector as well as for legal framework and subject to influences from human resource management in both OECD external actors that create a set of incentives in and developing countries over the past two the private sector which a performance orientation, in the sense decades. of managers and staff working purposefully to This focus on the civil service is important achieve an outcome, may be a minor objective. for three reasons. First, the core administra- In some low- and middle-income countries tion is the conduit of political authority— the public sector is seen more as an avenue public servants in central agencies can claim, for political patronage and employment than explicitly or implicitly, a reform mandate as the executor of government policies aimed from those politicians with whom they work at growth, poverty reduction, and delivery of day to day. Second, civil service pay is often essential services. Yet there are few alternatives the implicit standard for pay across the public to government’s extensive role in these areas sector. And third, the center sets the rules of and public sector reform remains a key devel- the game for the wider public sector as well as opment challenge. for the private sector. For example, the pub- This study examines whether financial lic expenditure and financial accountability incentives through pay reform, specifically arrangements created by the center (budget, pay flexibility, can improve the performance accounting, and audit) determine how public of public bureaucracies. It asks this question and private schools and hospitals providing with the objective of providing policy advice services on behalf of the government are held to the range of governments, particularly in accountable. The rules and procedures set emerging market and low- and middle-income by the center determine how publicly owned countries, that have introduced, or aim to banks, harbors, and airports, bodies that are introduce, these reforms. It is therefore a delib- outside general government, are regulated erately narrow, but focused and empirically and how their contingent liabilities are to be detailed, analysis of one factor in the set of managed. They thus shape the procurement variables in human resource management— arrangements that determine how many ser- recruitment, promotion, training, and man- vices are obtained. agement practices—that impact the incentives of individuals in organizations. Within public What is “performance” and how bureaucracies, the study focuses largely on should it be measured? the “civil service” or “core public administra- The study takes the simple view that organiza- tion,” which is normally defined as the civilian tional performance can be assessed through 1 2 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E The study assesses the following behavioral changes among the are devised by the political leadership (Weber staff of an organization: 1978). The accompanying employment and sal- performance primarily • Improvements in the quality of employees ary systems, intended to allow bureaucrats to through behavioral in a government ministry or agency through fulfill their role with minimal political interfer- the recruitment and retention of more ence, have often featured a work relationship changes among staff skilled personnel. governed by public not civil law; common and in organizations • Improvements in employees’ “effort.” compressed salary scales based on grades and Because measuring effort across diverse seniority not task; standardized and test-based jobs is difficult, it is often necessary to use recruitment; secure tenure; and generous ben- proxies concerning the “engagement” of efits and regular across-the-board salary raises. staff (their commitment to the organiza- While large private sector bureaucracies have tion’s mission and observed willingness to historically developed similar characteristics, use flexibility in working methods to help more recent assessments of pay arrangements deliver a product or an output) or “organi- in private organizations characterize them as zational citizenship” (their willingness to having higher pay dispersion, higher average provide extra effort to achieve goals).1 pay for skilled employees, less secure employ- The evidence for these behavioral changes ment, more explicit performance incentives, in the study is largely based on self-reported and greater sensitivity of the individual wage perceptions of staff and falls short of the ideal to current supply and demand in the labor of actual measures of outputs and outcomes. market (Eldridge and Palmer 2009; Perry, Eng- Wherever data on outputs (for example, rev- bers, and Jun 2009). enue collection, fines for nonfilers of taxes, In Organisation for Economic Co-operation and teacher attendance) and even outcomes and Development (OECD) countries, moves (for example, student test scores) are avail- toward pay flexibility in all economic sectors able they are used. However, since the focus arose in the 1980s in the context of attempts to of this report is the core administration, such liberalize labor markets in response to increas- comprehensive measures of performance are ing international economic competition. Pay rarely available, and information on inputs, flexibility was usually complemented with flex- behaviors, and processes can often only be ibility in other aspects of personnel manage- indirectly imputed through staff perceptions. ment, such as numerical flexibility (the ability The primary question addressed in this study of employers to adjust the number of workers is whether pay flexibility contributes to these or hours worked to changes in demand); func- changed methods of working and relationships tional flexibility (the ability of employers to to the job in ways that are considered in the reorganize the competencies associated with current literature to have a positive impact on jobs); and distancing (displacement of employ- outputs and outcomes. ment contracts by contracting out noncore tasks). In the public sector, traditional pay What is pay flexibility? arrangements were, it was argued, unable to Pay flexibility as an aspect of public sector ensure that performance objectives were met pay policy is defined in this study as a depar- within fiscal constraints, and many OECD ture from the traditional civil service pay countries began to move from the archetypi- model that emerged in the process of nation- cal single pay scale toward more flexible pay state formation and the creation of modern arrangements that have “pay for performance” bureaucracies in the nineteenth and early and “pay for skills,” and not “pay for seniority,” twentieth centuries (Ketelaar, Manning, and as their defining feature. Turkisch 2007). As captured in (although not Pay flexibility takes place within constraints. prescribed by) the writings of Max Weber, a The level and structure of pay needs to be set modern bureaucracy acts, in principle, as a so that it maintains long-term fiscal sustain- lasting, impartial, rule-abiding, and nonpar- ability and ensures that the wage bill does tisan executor of laws and regulations, which not crowd out other essential government PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  3 expenditures. Affordability constrains the resource specialists, peer panels, or out- In developing extent to which compensation can be used as a side agencies. The financial incentive can lever to improve performance. be a combination of base pay and one-off countries, in practice Although developing countries have not bonuses or merit increases of base pay, and a somewhat ad hoc made substantial formal moves toward pay it can be awarded on an individual, small flexibility has emerged f lexibility, in practice a somewhat ad hoc team, or larger departmental basis. flexibility has emerged. In many countries • Differentiation: Differences in pay between the civil service pay structure is far from the apparently similar workers across agencies, Weberian model, encompassing a variety of career groups, and locations that are pri- allowances and salary supplements. Pay as a marily a function of the specific skills that result varies considerably between individuals the agency competes for in the labor mar- performing similar tasks based on individual ket and of labor costs and the cost of living circumstances. This ad hoc pay flexibility is in the localities where agencies operate. in part a cumulative result of uncoordinated Differentiation can affect all staff paid by measures taken over time to improve the per- the organization, with the result that there formance of, or to respond to lobbying by, par- are agency-specific pay scales; particular ticular groups of staff—for example, in salary occupational groups or cadres; specific indi- increases through special allowances for ser- viduals with scarce but vital skills; and spe- vice delivery staff or revenue officials. In these cific locations entailing particular hardship. cases, pay flexibility is in part an outcome of Pay flexibility reforms are always framed political circumstance, such as bargaining in terms of improving performance, but they between different employee unions and cen- can be an element of two quite distinct strate- tral authorities. Attempts to instill a mission gies: to liberalize a rigid and centrally driven orientation in these public administrations pay model, or to formalize an already complex have not been very successful. One approach, haphazard or asymmetric structure that has often favored by the World Bank and other evolved with little regard for the larger fiscal or development partners, of trying to rational- incentive impact. The first strategy introduces ize pay through the traditional single pay scale asymmetry and complexity within the pay sys- or a limited number of pay scales has a poor tem in contrast to the apparent simplicity and track record (Independent Evaluation Group order of standardized, centrally determined 2008). Many developing countries, seeking to pay scales. The second emphasizes “purposeful emulate OECD countries but also trying to complexity” to bring order to an unmanaged improve the quality and motivation of staff, approach to pay policy. Both pay flexibility have been experimenting with linking pay strategies aim to provide incentives for perfor- to performance and providing higher pay to mance improvements for a target group of civil select groups of staff that are deemed particu- servants or to make pay more competitive for larly important. that group. The first is usually done through This study defines pay flexibility as compris- an ex post assessment of whether the target ing two key design elements, which can be pres- group of staff delivered on contracted outputs ent in varying degrees. These elements are: or objectives, the second through an ex ante • Performance-related pay (PRP): 2 Enabling assessment of the relevant labor market in pay to differ for workers doing the same which the target group competes. job by linking a portion of their pay to the achievement of performance targets. How Data sources for the study performance is measured, who measures The study draws on two sets of data. First is a it, and how it is linked to salary can all vary comprehensive review of 153 studies on PRP considerably. Performance can be based that, unlike other reviews, disaggregates the on qualitative assessments or quantitative available evidence by the quality of the empiri- measures of inputs, outputs, or outcomes, cal study; by differences in public sector con- and assessed by direct supervisors, human texts, particularly the different types of public 4 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Pay reforms are sector jobs; and by country context (developing This empirical analysis is more an assess- country or OECD settings). This disaggrega- ment of the general plausibility of these not a silver bullet tion allows for more nuanced conclusions on hypotheses than a rigorous empirical test. to enhance public the impact of performance pay in the public The strength of this study is the breadth and sector. richness of its contextual coverage, which sector performance Second is a set of case studies of OECD and comes at the expense of empirical depth emerging market countries (Brazil, Chile, and its ability to support causal statements. Indonesia, Korea, Malaysia, the Philippines, The evidence primarily concerns percep- Russia, and Thailand) that have undertaken tions of processes and behavior changes either one or both of these pay flexibility and not improvements in actual outputs and reforms. Data in the case studies were collected outcomes. Much of the evidence from the through structured interviews with country case studies is anecdotal, based on conversa- experts and, in Indonesia and the Philippines, tions with government officials and different large perception surveys of government offi- “experts.” Even the large perception surveys cials.3 These case studies, unlike the literature of government officials in Indonesia and Phil- review, focus mainly on pay flexibility in the ippines, which are in many ways state of the core civil service. The case studies exclude key art in core public administration research, service delivery staff such as teachers and doc- provide little evidence to establish causality. tors, since the literature review analyzes a num- The study is therefore more empirically tenu- ber of high-quality studies of PRP for teaching ous than the recent impact evaluation litera- and medical staff that provide strong evidence ture on performance incentives in teaching of causal relationships. There are few high- and health care, but it is more rigorous than quality studies of PRP in the core civil service, much of the work to date on pay reform in the and therefore the case studies focus on this core public administration. area. There is also no significant literature on It is important to emphasize also that this pay differentiation, which also is covered in the study is not premised on the notion that pay case studies. reforms are a silver bullet to enhance pub- lic sector performance. It is improbable that Organization of the study such silver bullets exist. Any reform is about The next section outlines the theoretical tradeoffs, and whether these tradeoffs are framework, summarizing the main arguments worthwhile is highly context specific. Improv- for and against PRP. It also reviews the much ing the performance of large bureaucracies smaller literature on pay differentiation. This is extremely difficult, and the key is to capi- discussion is then used to distill some hypoth- talize on small openings to achieve marginal eses on the impact of pay flexibility on public improvements in productivity (World Bank sector performance through direct levers, which 2012). Can pay f lexibility provide such an change the incentives of staff, and through opening, and are the potential tradeoffs, indirect levers, which change the incentives of which are discussed in the next section, worth frontline managers, in both cases by acting it in that particular country? The study also on the incentives of staff to behave differ- reviews the empirical evidence from research ently in their jobs and by changing incentives that is limited to pay flexibility, assuming that about joining, or remaining in, the public ser- all else is constant; this limitation is both to vice. These hypotheses are then examined in keep the task manageable and to examine a the main empirical sections of the study, the variable that has a particular relationship with review of the empirical literature on PRP, and other aspects of pay reform. It cannot reach a the review of the findings on pay flexibility conclusion that when pay flexibility “works” it from the case studies. The final section sum- is more important than other variables. These marizes the main findings and offers some are important issues that merit a different policy directions. study. The Theoretical Framework Performance-related pay observable results. Critics have pointed out PRP is a topic of The case for and against PRP has strong the- that these conditions are rare in the public oretical underpinnings in neoclassical eco- sector. Results are hard to define, let alone considerable debate in nomics, organization theory, and behavioral measure, particularly in policy or administra- the academic literature economics. In the microeconomic principal- tive, as opposed to service delivery, organiza- agent model, the problem that needs to be tions. Civil servants often work in large teams addressed is that the principal—the manager, under the supervision of multiple managers, the school principal, the hospital administra- complicating the attribution of performance tor—has to induce effort from his or her staff and responsibility of evaluation. A necessary but cannot easily monitor their work. Under condition for PRP may also be the presence of these conditions a fixed pay contract gives the high levels of trust and transparency between employer little leverage to influence employee employees and management to avoid arbitrary effort after hiring decisions have been made, a implementation and worker dissatisfaction problem that becomes worse if employees are (Kellough and Lu 1993). Implementing PRP hard to fire. PRP is a means of addressing this in the absence of objective measures of results, problem of moral hazard by tying observable lack of attribution, and lack of trust can breed outputs, which are presumably correlated with resentment among staff and demotivate them, unobservable effort, to pay. thereby resulting in lower effort. The principal-agent model suggests that A related problem is one of perverse incen- PRP can potentially ameliorate another prob- tives or unintended consequences, which come lem, that of adverse selection. The agent has in several variants, the most extensively stud- access to private and valuable information at ied of which is the multitasking problem (Hol- the time of contract signing and low- and high- mstrom and Milgrom 1991). When multiple skill applicants are hard to distinguish based tasks are performed, giving incentives for only on public information. Hiring agencies need to a subset of those that are observable and con- offer contracts that induce high-quality appli- tractible will not necessarily improve overall cants to apply and deter low-quality applicants outcomes. Instead, employees may shift effort from misrepresenting their qualifications. from noncontracted to contracted tasks, which PRP can alleviate this sorting problem, since under some circumstances can lead to worse higher-quality personnel who expect to per- outcomes. For example, the task of teaching form better under this pay system will be more can involve both instruction based on sound likely to apply than low-quality applicants. curricula and coaching on test-taking strate- The model also offers strong arguments gies, and poorly designed incentive schemes against PRP. Such incentive schemes require can encourage teachers to reallocate effort to the ability to measure some relevant results the latter and away from the former (known to which pay is linked, and their design as “teaching to the test”) to the detriment should tightly link the agent’s actions to these of human capital accumulation. Similarly, 5 6 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E PRP can provide financial incentives for the provision of school Q. Wilson (1989) and Jean Tirole (1994) have meals to children to reduce malnutrition can recognized that the multiple demands placed frontline managers result in reduced teaching and worse student on a public organization, and the multiple with incentives to learning outcomes. interests or principals it needs to serve, make Closely linked to the multitasking problem it difficult to define a “goal” for the organiza- increase awareness is “gaming,” or manipulation of the incentive tion that can orient the staff. The introduction among staff of system, which comes in two forms: manipula- of PRP can start the process of goal or mis- tion of the data used to measure the perfor- sion orientation by triggering improvements organizational goals mance incentive output, or manipulation in management practices. PRP can provide by defining explicit of the output itself. Typical examples of the frontline managers with incentives to increase former are the setting of easy-to-reach per- awareness among staff of organizational goals performance standards formance targets and other forms of manipu- by defining explicit performance standards, lating data to demonstrate that the targets encourage frontline managers to focus more have been achieved. The latter includes “cream on working with their staff toward achieving skimming” or “cherry picking”—the deliber- these organizational goals and tracking them ate selection of beneficiaries to improve pro- regularly, and increase the link between indi- gram effects (Heckman, Heinrich, and Smith vidual and organizational goals in individual 1997)—or other forms of manipulation such as performance assessments, thereby inculcat- the provision of high-calorie food to students ing a focus on results within the organization during test days (Figlio and Winicki 2005) and (Marsden 2004, 2009; OECD 2005). In sum, ratchet effects under which managers reduce PRP works indirectly by institutionalizing regu- their output increases to a modest increment lar discussions of performance between man- so that expectations and future targets will be agement and staff, thereby altering the “effort set at a low level. bargain,” as elaborated in boxes 2 and 3. Finally, an argument coming from organiza- tional theory and behavioral economics is that Differentiation PRP can reduce the intrinsic motivation that The theoretical literature on pay differentiation people have in their jobs (Frey and Osterloh is much more limited. The main argument for 1999). These schemes cause workers to change it is that differentiation enables an agency to set their perception about organizational goals pay at a level that is appropriate for a given task and values—for example, that the organiza- in a specific labor market. This is deemed nec- tion’s goals are not about public service but essary given the considerable pay differentia- more about private profitability—leading to tion in the private sector in which interindustry an overall reduction of effort. This crowding- wage differentials exist even after controlling out of intrinsic motivation can be especially for differences in human capital of employees salient if performance pay is introduced using (Dickens and Katz 1987; Krueger and Sum- antagonistic framing and can stifle creativity mers 1988; Groshen 1991). Differentiation can and collaboration. also reflect other factors, such as differences These problems have prompted many to in union power and wage bargaining arrange- argue that the approach to performance in ments, sectoral monopoly or oligopoly profits, the public sector should be based on long- industry-specific technology shocks and innova- term career-based incentives rather than PRP tions, and managerial approaches. Advocates (box 1). of pay differentiation in the public sector stress The pros and cons of PRP have focused functional similarities and competition with the largely on the direct effects of financial rewards private sector to motivate mirroring reforms. on individual incentives. Organizational theo- Overall, pay differentiation across agencies is rists have also pointed to possible indirect meant to better reflect the heterogeneity of pub- effects of PRP even in light of the difficulties lic services and signal commitment to particular associated with measuring outputs amid a mul- organizational goals to employees and outside tiplicity of tasks. Scholars as diverse as James observers (Bender and Elliott 2003). PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  7 Arguments for Long-Term Career-Based Incentives Rather Than Short-Term Box Performance-Based Incentives for Public Sector Staff 1 Many argue that in complex public sector environments, with (Burgess and Metcalfe 1999). Competitive promotions are considered complex and occasionally contradictory objectives and multiple more reliable measures of ability than short-term performance principals, incentives for performance should rely on information assessments since “[s]enior civil servants are likely to be as that is hard to game over the longer term (Burgess and Metcalfe motivated by [promotions] as they are financial rewards [since] the 1999). There are three sets of long-term career-based incentives: incentives to game to achieve reputational rewards are somewhat opportunities for long-term enhancement of rewards and status, lower than the incentives to game in relation to financial rewards. competitive promotions, and deferred compensation. This is for the simple reason that if the reward is reputation, a reputation for gaming amongst professional peers undermines the The central idea behind long-term career benefits is that workers reward itself” (Ketelaar, Manning, and Turkisch 2007, 16). However, exert effort in order to influence actual or potential employers’ early promotions risk distorting the employer’s perceptions of fast- beliefs about their talent, and that while performance information rising staff, tending to promote them more automatically. There is can be gamed in the short term, over time real performance also some evidence of diminished cooperation between staff who becomes evident. Even when employees are paid a fixed wage, are in the same pool of candidates for promotion (Lazear 1989).1 they are thus motivated by the effect their effort will have on future wages (Holmstrom 1982). However, long-term career-based The final long-term career incentive widely used in the public incentives require some cofactors: employees cannot signal talent sector is deferred compensation, where upward-sloping wage and effort to employers that fail to look afresh at effort on a profiles can be structured to reflect experience and expertise. regular basis or that are more interested in nonmerit based The argument is that deferred compensation provides incentives signals, or if the breadth and complexity of employee tasks to workers to exert effort early in their careers in order to be are such that it cannot be clear where and whether they are promoted or not to be fired and hence lose a pay-off later in succeeding (Dewatripont, Jewitt, and Tirole 1999a, 1999b). their tenure (Lazear 1981). However, if there is no serious risk of losing the long-term compensation gains, the rewards are simply Competitive promotions have been extensively reviewed within the provided in exchange for length of tenure or seniority. However, framework of tournament theory with promotions seen as prizes if rewards for seniority are provided together with a credible allocated to the workers who rank higher than all others over a threat of nonadvancement or dismissal for poor performance, given period. Some evidence suggests that successive rounds of seniority is simply an easy-to-measure proxy for experience and competition for jobs can reveal otherwise hidden performance traits serves to attract risk-averse but talented workers. Note 1. See also the emphasis given to employment security and recruitment in the seven human resource management practices identified by Pfeffer (1998a, 1998b) as key to organizational effectiveness. These have been validated more widely, though an empirical review of the impact of these practices found no direct relationship between employment insecurity and organizational performance. The review did note that insecurity seemed to hinder development of other useful human resource management practices with a stronger link to performance (Ahmad and Schroeder 2003). In developing countries the argument for Western bureaucracies in the nineteenth and differentiation, also called asymmetric pay early twentieth centuries precisely to combat reform or agency-level pay reform, is primar- problems of patronage and corruption (Odden ily pragmatic. The experience of comprehen- and Kelley 2002). Pay differentiation across sive whole-of-government pay reforms in these agencies can also reduce cooperation and com- countries has been disappointing, primarily plicate coordination across sectors or regions because they lacked the political support of (Rexed and others 2007). elites and were resisted by key stakeholders such as employee unions (Lindauer and Nun- Hypotheses berg 1994). An agency-based approach may Performance improvements are characterized therefore be a more feasible alternative for in this study as increased effort by line staff, incremental productivity improvements, with proxied as staff engagement or organizational successful implementation creating a demon- citizenship, and better staff quality. The review stration effect across government (Nunberg of the literature suggests a number of hypoth- and Taliercio 2012). eses on how pay flexibility can cause these The main argument against pay differentia- improvements (also illustrated in figure 1): tion is that the departure from centralized and • PRP hypotheses: uniform salary scales reduces transparency and • Hypothesis 1. PRP can act as a direct equity in civil service remuneration, thereby lever by: breeding resentment and demotivating staff. • Hypothesis 1a. Having a direct Uniform salary schedules were introduced in incentive effect through increased 8 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Box The “Effort Bargain” 2 “[E]very employment contract … consists of two elements: (1) … accessible to higher management only through the eyes an agreement on the wage rate … i.e. a wage-rate bargain; and of firstline managers” (Marsden 2004, 352), to move some of (2) an agreement on the work to be done, i.e. an effort bargain” the goalposts concerning work procedures and performance (Behrend 1957, 505). Early work on productivity assumed that expectations. Marsden argued that while research showed that the relationship between these two elements was fixed in some PRP for jobs within the core civil service had little impact on mechanical way—without a change in the wage-rate bargain, measures of individual motivation, the accompanying regular there would be no change in the effort bargain. Some flexibility institutionalized performance discussions improved productivity could be found to adapt that employment contract, but only in the agencies as a whole. Empirical work has highlighted “within certain limits” (Coase 1937, 391). the significant motivational effect of regular communication between frontline managers and their staff (Ahmed and others Marsden, in examining why U.K. efforts to introduce PRP in 2010). the core civil service during 1980–2000 had failed to improve psychological motivation for staff, argued that the effort bargain A more recent line of research concerning staff engagement could be changed in nuanced ways. “Less visible [than pay has reached a similar conclusion. There is a close link between adjustments], but just as important for management, is its ability high levels of employee engagement and positive discretionary to revise job boundaries, and redefine the nature and standards behavior or effort. The empirical literature tends to come from of performance that it requires from employees. These standards, the private sector and is largely from developed countries, which may include qualitative aspects of performance, are but there is an emerging literature concerning middle-income usually the subject of a tacit understanding between staff and countries (Kular and others 2008). Engagement improves affective management” (Marsden 2004, 352). commitment (“I like this place”) and normative commitment (“I should stay”), which are both positively behavior changing.1 In effect, frontline managers were using the informality that Purcell and others (2003a) found that such commitment was is always present in complex bureaucratic tasks, where “[i] affected by how large-scale objectives were explained and t is common for jobs to deviate considerably from their formal interpreted day to day: “[T]he vital ingredient in linking people job descriptions” and their inside knowledge of what workers management to business performance … is primarily the task of were actually doing, since “[t]he features of a given job are line managers” (Purcell and others 2003b, 72). Note 1. Rafferty and others (2005) review the literature and find empirical associations between commitment and increased job satisfaction (Vandenberg and Lance 1992); increased job performance (Mathieu and Zajac 1990); higher sales (Barber, Hayday, and Bevan 1999); lower employee turnover (Cohen 1991); less intention to leave (Cohen 1993; Balfour and Wechsler 1996); and lower absenteeism (Cohen 1993; Barber, Hayday, and Bevan 1999). individual engagement and organi- recruitment and retention of better-qual- zational citizenship and by induc- ity staff for priority activities (sorting). ing staff to exert more effort toward • Hypothesis 4. Differentiation can act achieving outputs and outcomes indirectly by providing incentives for linked to the incentive. greater effort by managers, putting the • Hypothesis 1b. Having a direct incen- spotlight on management to improve the tive effect on staff quality, resulting in performance dialogue with staff. This the recruitment of higher quality staff improved performance dialogue in turn who are likely to do well under the results in greater effort by line staff. scheme (sorting). These potential direct incentive effects • Hypothesis 2. PRP can act as an indirect on productivity, and indirect effects through lever by providing incentives for greater changed management practices, are hotly effort by managers on the performance debated. Table 1 summarizes the arguments dialogue, resulting in better organiza- for and against them. tional goal setting, teamwork toward achieving organizational goals, and link- Context matters age of individual performance apprais- The main objective of the empirical analysis is to als to those organizational goals. This examine whether and under what contexts these improved performance dialogue in turn hypotheses hold. This study considers three results in greater effort by line staff. contextual variables in particular: the nature of • Differentiation-based hypotheses: the PRP scheme; the nature of the job for which the • Hypothesis 3. Differentiation can act scheme is being implemented; and the degree of as a direct lever by resulting in the delegated authority that agencies have to manage PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  9 Box Active Management and Managerial Effort 3 PRP encourages frontline managers to institutionalize performance where they can exercise the greatest amount of discretion, discussions with staff, thereby renegotiating the “effort bargain.” “bringing human resource policies to life” (Hutchinson and Purcell The logical next questions are: What flexibilities can managers 2003). use to propose to staff in this renegotiation? And why should frontline managers themselves be motivated to undertake such a In sum, line managers play a crucial role in providing positive renegotiation, which is, after all, an additional effort for them? feedback and recognition and in finding opportunities for staff to develop and the scope to exercise responsibility (Armstrong 2010). In considering the first question, it is important to underscore They thus play a major part in influencing employee attitudes that these are likely to be informal, “between the rules” changes toward the organization and their jobs (Purcell and Hutchinson to working practices. The emphasis on managerialism within the 2007). Where this shows in practical terms is the frontline public sector in the managerial and academic literatures since the manager’s role in modest modifications to a staff member’s work 1970s has highlighted that the real tradeoff is between increased allocation, responding as well as possible to the skills of the accountability for managers and reduced structural restrictions staff member and demonstrating a link between the tasks and on their use of inputs (money, resources, and people) (Knies and organizational goals and how they fit with the contribution of Leisink 2013). But loosening those formal structural constraints others (Purcell and Hutchinson 2007). is difficult: “Overall public sector management reforms are inducing more entrepreneur-like behavior patterns and attitudes On the second question, there are three likely reasons why on the part of public managers … (but) entrepreneurialism frontline managers might want to expend additional effort to find can only be introduced at the expense of the more traditional small, “between the rules” informal ways to better motivate their input-oriented type of governing with accountable scrutiny” (Koch staff. First, it seems reasonable to assume that putting money 1996, 39–40). More likely, it is the behavior of managers, rather on the table for staff performance attracts senior management than their formal power, that leads to successful implementation attention and so making frontline managers sensitive to their at the unit level (Norrgren and Schaller 1999). This view is unit’s performance regardless of whether they are themselves confirmed by the research in the United Kingdom by Purcel on PRP. Second, managers will have a heightened awareness of and others (2003a), who note that high levels of organizational the organization’s objectives since they are now spending much performance are not achieved simply by having well-conceived of their managerial time translating them into work targets for human resources policies and practices in place: what makes the their staff. Third, performance discussions with their staff are, difference is how these policies and practices are implemented. inevitably, performance feedback sessions for themselves, likely How frontline managers show leadership, deal with employees, bringing options for these informal “between the rules” issues to and exercise control are major issues—and it in these areas their attention. Figure The Link between Pay Flexibility and Performance 1 Direct pay Indirect pay Functional exibility policy exibility policy improvements levers levers Greater effort by managers Greater effort by H2 staff (higher H1a engagement and PRP organizational citizenship) H1b H4 Better-quality staff Differentiation (sorting) H3 their personnel on their own discretion rather PRP schemes can be distinguished based than by centrally managed rules. Other factors on five design elements: whether the bonus in the case studies, not treated as systematically is based on individual or group performance, but nonetheless important, include the nature the time horizon of the financial incentive, the of the political context and potential synergies nature of the performance evaluation, the size and complementarities with other reforms. of the bonus, and the probability of receiving 10 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Table Hypotheses on the Impact of Pay Flexibility on Civil Service Performance 1 Arguments for Arguments against PRP Hypothesis 1. Direct pay flexibility levers Hypothesis 1a. Direct incentive effect on effort (PRP can directly affect individual engagement and organizational citizenship) • Induces staff to exert more effort toward achievement of the outputs • No effect on effort: Difficult to measure outputs in the public sector and outcomes linked to the incentive given lack of objective performance measures • Negative effect on effort: Given the difficulties in measuring outputs, the perceived unjustified pay inequity breeds resentment • Negative effect through unintended consequences: “Gaming behavior” • Behavioral economics: Crowding out of intrinsic motivation, which reduces effort Hypothesis 1b. Direct incentive effect on staff quality (PRP can have a direct effect in improving the recruitment and retention of better-quality staff) • Attracts higher-quality staff who are likely to do well under the scheme (sorting) Hypothesis 2. Indirect pay flexibility levers (PRP can act indirectly by providing incentives for improved management) • Positive effect through better management of the “effort bargain,” in • Negative effect: Pay inequity results in harmful competition, hurting particular improvements in: teamwork and reducing staff trust in management • Organizational goal setting • Teamwork task allocation toward achieving organizational goals • Linking individual performance appraisals to organizational goals Differentiation Hypothesis 3. Direct pay flexibility levers (Differentiation can have a direct effect in improving the recruitment and retention of better-quality staff) • Direct incentive effect on staff quality: More targeted recruitment • Induces staff in the lower-paying segments of the government to of high-quality staff for priority activities within governing fiscal exert less effort; reduces pay transparency; and increases pay constraints (sorting) inequity, thereby breeding resentment Hypothesis 4. Indirect pay flexibility levers (Differentiation can act indirectly by providing incentives for improved management) • Puts the spotlight on management to improve the performance • Hurts interagency cooperation dialogue the bonus (table 2). Rewarding team perfor- incentive effect, while large bonuses can fur- mance can have certain advantages, ranging ther encourage gaming and under extreme cir- from reduced evaluation costs to avoiding cumstances result in “choking under pressure” harmful competition between employees. How- and therefore hurt performance (a phenome- ever, basing rewards on team outputs can also non known as the Yerkes-Dodson Law). Finally, lead to free-riding, where some team members if the probability of receiving the performance reduce their effort and rely instead on the work bonus is either close to 0 or to 1, the incentive of others. The awards can be one-off bonuses will have no impact (Bruns, Filmer, and Patri- or merit increments to salary that are perma- nos 2011). nent and cumulative, or can be based on the On nature of the job, this study relies on past year’s performance or on multiple years’ the fundamental insight of James Q. Wilson performance. Performance evaluations can (1989) that government bureaucracies vary be based on quantitative performance targets along two dimensions: whether the tasks per- or subjective assessments against individual formed by the individuals in the organization, results agreements. Small bonuses have little or the inputs of labor, are easily observable and Table Five Key Design Elements of PRP Schemes 2 Individual- or Time horizon Nature of the Size of Probability of group‑based awards of the incentive performance evaluation the award receiving the award Individual awards can breed Short-term annual bonuses Quantitative performance Small awards have If the probability of harmful competition, but or longer-term merit targets may be more limited effects, but very receiving the performance team awards can encourage increments. Single- or objective but are rare for large awards can further bonus is either close to 0 free-riding multi-year performance public sector jobs encourage gaming and or 1, the incentive will have appraisals perverse incentives no impact Source: World Bank staff. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  11 measurable to managers and external agents; services like child immunization, are also more Coping jobs and whether the outputs or outcomes of these measurable. Other examples include tax collec- jobs are easily measurable. Inputs may be hard tion, job placement services, and auditing. present the most to monitor because they may be highly techni- In the bottom row are “procedural jobs” and challenging functional cal and esoteric in nature (for example, a doc- “coping jobs.” Both are characterized by diffi- contexts for PRP tor performing a diagnosis) or because these cult-to-measure outputs, but again they differ actions cannot be readily observed since staff in the observability of the production process are in faraway locations or act out of view of to an outsider. Procedural jobs like the mili- managers (for example, forest rangers or rural tary have clearly defined and highly routinized teachers). Outputs may be hard to measure and monitored inputs, which can be specified because indicators may be difficult to find, or, in standard operating procedures that prolif- more fundamentally, to define what the goal of erate and that regulate every detail of work. a particular agency is. Administrative jobs in general policy units of Table 3 provides a classification of job types, the central government neither produce easily with the simplifying assumption that jobs with measurable outputs nor have transparent pro- multiple dimensions are located within the duction processes. These coping jobs present cell that represents the most complex of those the most challenging functional contexts for dimensions.4 The top left box describes “pro- PRP and more generally for inducing a mission duction jobs,” in which outputs are easily mea- orientation in the organization. surable; the production process consists of For this study we use the distinction between repeatable, mechanical tasks that are observ- jobs with observable and unobservable produc- able to an outside monitor; and controllability tion processes as a rough proxy for distinguish- is likely to be high. Typical examples are manu- ing between jobs that tend to be found in the facturing factory-floor jobs, the postal service private or public sectors, respectively. Within the (where letter sorters can be observed and the latter, we use the distinction between those pub- speed of mail delivery measured), and municipal lic sector jobs where outputs are measurable and services like garbage collection. If the produc- those where they are not as a more precise proxy tion process is not directly observable but out- for distinguishing between broad public service puts remain measurable, such jobs are termed jobs and those in the core administration. We “craft jobs.” With recent advances in measuring take those studies of PRP in coping jobs as likely learning outcomes, teaching can be classified as the best measure of the impact of PRP in the pri- a job in which the exact process of production marily policy jobs in the core civil service. is hard to ascertain but desired outputs are to The study also looks at organizational some degree quantifiable. Similarly, some of the autonomy, or delegation, in the use of the outputs of health care, particularly in preventive factors of production—the extent to which Table James Q. Wilson’s Classification of Job Types 3 Actions or internal production process of the job Observable Not observable Production jobs: Simple repetitive stable tasks, Craft jobs: Application of general sets of skills to unique specialized skills tasks, but with stable, similar outcomes Relatively easily Examples: Manufacturing, sales, simple clerical tasks, Examples: Auditing, revenue collection, teaching, medical measurable postal service, garbage collection practice, job placement work Outputs from the job Procedural jobs: Specialized skills; stable tasks that can Coping jobs: Application of generic skills to unique tasks be routinized through standard operating procedures, but whose outcomes cannot be evaluated in the absence of Not easily unique outcomes alternatives measurable Examples: Military Examples: Policy formulation, administration; managerial jobs in large private sector organizations Source: Adapted from Wilson (1989). 12 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E agencies have the authority to manage their better able to make these decisions. Delegated personnel on their own discretion rather than authority allows agencies to complement new by centrally managed rules. This autonomy organizational goals with adjustments in the can include the ability to hire and fire staff, workforce, hiring and firing staff members, establish agency hiring standards and salaries, reorganizing teams, and providing financial and make outsourcing decisions while retain- incentives (Rexed and others 2007). Civil ser- ing central control only over the agency’s total vice agencies that have little control over per- wage bill.5 sonnel planning or salary structures have few Delegation can potentially impact the effi- tools to reorganize the production of services, cacy of both PRP and differentiation. For PRP since their main input is human capital. Joint to work, managers in an agency need some authority over service delivery, input alloca- degree of freedom to define the agency and tion, and overall costs increases the responsi- individual goals and to evaluate staff perfor- bility and accountability of senior management mance, the assumption being that these man- as the overseeing authority and thus allows agers have more information about outputs better assessment of performance and service and inputs in these jobs and are therefore quality. Assessing the Evidence: Review of Literature on Performance- Related Pay The first piece of evidence on the impact of pay been exclusively on craft jobs and production flexibility is based on a comprehensive review jobs, with none to date on coping jobs. of the existing literature on PRP that disag- These reviewed studies were grouped into gregates the studies by type of job, using the three categories to capture the effect of PRP: classification discussed in the previous section, positive if their findings provide evidence for and study quality. In total, 153 empirical stud- the effectiveness of incentive schemes;6 neutral ies of PRP were considered in this review (see if the study is largely descriptive or finds con- Hasnain, Manning, and Pierskalla [2012]) for tradictory evidence; and failed if the evidence the full list), of which 110 were of craft and cop- indicates no effect or a negative effect for ing jobs and 17 were of coping jobs specifically PRP. Figure 2 shows the overall frequency of (table 4). The research to date on the subject results. Most of studies (93 of the 153) present has largely focused on advanced countries—in evidence for the effectiveness of performance the review 127 studies are in OECD countries, pay schemes, with experimental studies show- and only 26 are in developing countries. The ing more positive findings than observational empirical literature also employs a range of ones. methodologies, from earlier observational and In drawing conclusions, however, it is often qualitative studies to the more recent important to distinguish the findings by study field randomized control trials (RCTs) and lab- research quality. Study quality was ranked oratory experiments explicitly aimed at teasing in two different ways. First, each study was out causality. These experimental studies have assessed for its internal validity, or the strength Table The Studies Reviewed 4 Types of jobs Country and methodology Production Procedural Coping Craft Unclassified Total OECD study 27 0 16 71 13 127 Observational 14 0 16 58 13 101 Field RCT 7 0 0 13 0 20 Lab experiment 6 0 0 0 0 6 Developing country study 1 0 1 22 2 26 Observational 0 0 1 15 2 18 Field RCT 0 0 0 6 0 6 Lab experiment 1 0 17 1 0 2 Total 28 0 17 93 15 153 Source: Hasnain, Manning, and Pierskalla 2012. 13 14 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Figure Aggregate Findings on PRP 2 Aggregate findings Number of studies Findings by study type Number of studies Failed Neutral 100 80 Positive 75 60 50 40 25 20 0 0 Failed Neutral Positive Observational Field Lab experiment experiment Source: Hasnain, Manning, and Pierskalla 2012. of the causal arguments being made, using a Findings by Research Quality (High Internal and External Validity Studies five-point ranking (from weak to strong): Figure Only) 1. No empirical study or faulty research design.7 3 Number of studies 60 2. Descriptive; small sample size.8 3. Secondary data analysis or descriptive data analysis; small sample size; some statistical analysis.9 40 4. Quasi-experimental design; reasonable sample size; conclusions based on statistical analysis.10 20 5. Laboratory experiments; RCTs; large sam- ple size; strong statistical analysis; strong conclusions. 0 Second, studies were also evaluated on the Failed Neutral Positive dimension of external validity, or the extent Note: Seventy-two studies. to which causal connections drawn in the spe- Source: Hasnain, Manning, and Pierskalla 2012. cific context of the study would remain valid if replicated in other contexts. For example, lab Figure 3 applies these two quality filters to experiments and RCTs offer strong evidence these 153 studies, resulting in 72 high-quality of causality (high internal validity) but in a spe- studies (ranked 4 or 5 on the internal valid- cific context—they tell us the average impact of ity scale, and “high” on the external validity a particular intervention in a particular location scale);11 54 of these high-quality studies show with a particular sample at a particular point positive results. in time. They are often accused of being low The effect of PRP in craft and coping jobs on external validity because the study subjects is particularly interesting, since these jobs (usually college students in the case of labora- most closely resemble public sector organiza- tory experiments) are not representative of the tions. Figure 4 presents the evidence for all general population, or in this case the popula- the reviewed studies for craft and coping jobs tion of interest (civil servants), and the require- (110 studies) and for the high-quality studies ments of the experiment imply conditions that of craft and coping jobs (53 studies). The over- may not approximate real world settings. all evidence is generally quite positive, though PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  15 Figure Findings for Craft and Coping Jobs 4 Number of studies by country context, craft and coping jobs only (110 total) Findings for high-quality (high internal and external validity) relevant studies only (53 total) Number of studies Failed Number of studies Failed 50 Neutral 50 Neutral Positive Positive 40 40 30 30 20 20 10 10 0 0 OECD Developing country Craft Coping Source: Hasnain, Manning, and Pierskalla 2012. the evidence is almost exclusively for craft jobs, treatment assignment (Lavy 2008, 2009). and mostly in OECD settings. These studies by Lavy also show significant There is an extensive and growing litera- gains in student achievement. They also iden- ture on performance pay for teachers. In the tify changes in teaching methods, enhanced United States most observational studies after-school teaching, and increased teacher have primarily examined the impact of per- responsiveness as the key mechanisms for these formance incentives on student test scores, improvements. though a few studies, such as Clotfelter, Diaz A number of field experiments have evalu- and others (2004) and Clotfelter, Glennie, and ated the impact of performance pay for teach- others (2008), show, using detailed data from ers on reducing absenteeism and improving North Carolina’s schools, that accountability learning outcomes. The findings are gener- and performance pay systems can help retain ally mixed but are, interestingly, more positive quality teachers. The evidence is mixed with for developing countries. Duflo, Hanna, and regard to student test scores. Cross-sectional Ryany (2010) show that monitoring teacher studies from the American National Edu- attendance through tamper-proof cameras cational Longitudinal Survey show positive linked to financial incentives in rural India results (Figlio and Kenny 2007; Winters and led to a strong reduction in teacher absen- others 2009). A number of studies identify teeism and increased students’ test scores. By problems of gaming, such as outright cheating contrast, Kremer and Chen (2001) found that (Jacob and Levitt 2003; Jacob 2005) or, more in Kenya subjective monitoring arrangements subtly, the adjustment of the caloric content by an individual in the institutional hierarchy of school lunches to improve cognitive ability (such as a school’s headmaster) may not work on test days (Jacob and Levitt 2003; Figlio and because the monitor might shirk, try to avoid Winicki 2005; Jacob 2005). confrontation, or collude with the workers. Outside the United States an analysis by These studies suggest that impersonal, exter- Atkinson and others (2004) finds clear positive nal monitoring through technology coupled effects of performance pay for British schools. with a clear, credible, and automatic threat of A set of observational studies uses data from punishment and promise of reward was the key an Israeli policy experiment with tournament- design feature for program success. based teacher competition for bonuses and a A field experiment in 50 Kenyan schools regression discontinuity and difference-in- linking teacher salaries to student test scores difference design to approximate random failed to find lasting effects (Glewwe, Ilias, 16 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E There is a large and and Kremer 2010): teacher attendance did treatment and control groups. Similarly, a RCT not improve, and teachers did not adjust their implemented by Grady, Lemkau, and Cad- growing empirical teaching methods or conduct more prepara- dell (1997) found no clear effects of financial literature on PRP tion sessions. Students in treated schools did incentives on mammography referrals by pri- perform better during the program duration, mary care physicians. in education that but these gains did not carry beyond the study By contrast, a set of studies, also focusing surprisingly finds more period. However, a large-scale field experiment on pediatric immunizations, found that per- in a representative sample of 300 government- formance incentives increased immunization positive results in run rural primary schools in India found that rates by several percentage points compared developing countries bonus pay linked to the mean improvement of with the control group (Fairbrother, Hanson, student test scores in an independent learn- and others 1999; Fairbrother, Siegel, and oth- ing assessment led to a statistically significant ers 2001). An RCT at the clinic level found that and substantively meaningful improvement of financial incentives improved treatment of student outcomes (Glewwe, Ilias, and Kremer smoking cessation outcomes (Roski and oth- 2010). ers 2003). Work on performance pay for cogni- RCTs in the United States have been quite tive services interventions by pharmacists also negative on the effect of teacher incentives found positive effects (Christensen and others on student outcomes. A field experiment con- 2000). ducted in New York City public schools also To our knowledge, the only two RCTs on failed to find statistically significant effects performance pay in health care in a low- of team incentives for teachers on student income country are a study by Basinga and outcomes (Fryer 2011). A related study that others (2010) in Rwanda and a study by Singh also assessed the effects of the New York City (2010) in India. Basinga and others used an group incentive program on classroom activi- RCT design to evaluate performance pay in ties and teacher turnover and qualification, Rwandan primary health care centers. The apart from test scores and teacher effort, simi- authors took advantage of a sequenced roll- larly found no effects (Goodman and Turner out of the scheme across Rwandan health care 2010). A three-year experimental evaluation of facilities, collecting data on child preventive the Project on Incentives in Teaching in metro- care and prenatal delivery. To isolate the per- politan Nashville schools also found no signifi- formance-pay effect from a general increase cant effects of bonus incentives on student test in resources, comparison facilities received an scores (Springer and others 2010). equivalent increase in their budgets. The study, In the health sector a number of RCTs have using information from 166 facilities and 2,158 been implemented to determine the role of households, found large effects on all central performance pay on health worker produc- outcome measures, but with particularly strik- tivity, patient treatment, and outcomes. The ing effects for services with the highest payoffs majority of studies assess these questions in the and smallest necessary staff effort. context of OECD health care systems. Kouides Singh (2010) treated three groups of moth- and others (1998) implemented an RCT offer- ers and staff providing child care and nutri- ing financial incentives to a randomly selected tional advice to them in Chandigarh, India. In set of primary care physicians based on influ- one group the workers received performance enza immunization rates of the elderly as part pay; in a second group the workers had no per- of a Medicare demonstration project. Doc- formance pay but the women that they worked tors in the treatment group performed more with were separately given factual information immunizations. However, Hillman and others about nutrition; and the third group received (1998, 1999) used two RCT designs to provide both treatments. The study found that chil- incentives for cancer screenings for women dren’s weights improved only in the third of age 50 and above and pediatric immuniza- group compared with the control group. tions, respectively. In both studies the authors Note that nearly all the studies in the health document no significant difference between care sector focus on fairly narrow types of PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  17 performance pay and specific, single-outcome of lower-tier workers was used as an outcome There are too few measures in preventive care rather than on measure. The study found evidence of both overall multidimensional patient treatments incentive and sorting effects—that is, manag- high-quality academic and outcomes. ers support their high-productivity workers studies of coping Revenue authorities and job placement and fire the least qualified employees. jobs to draw any agencies also provide examples of craft jobs To date there have been few high-quality where, though work methods are hard to studies of coping jobs. Of the three studies firm conclusions observe, the outputs—number of audits con- reviewed, two were of performance pay for ducted, tax fines collected, job-seekers finding managerial positions in the private sector, and employment within a specified time period— only one was for core administrative jobs in the are more easily measurable. Kahn, De Silva, public sector. All these studies showed posi- and Ziliak (2001) examined a 1998 incentive tive effects of the performance incentive. For scheme in Brazil and found that it resulted example, Hochberg and Lindsey (2010) found in a 75  percent increase in fines per inspec- a positive impact of stock options for company tion. Burgess and others (2010) used an RCT rank-and-file on firm profits. Dowling and to examine the impact of a pilot team–based Richardson (1997) used an employee percep- incentive scheme based on revenue collec- tion survey to examine PRP for managers in tion and audit targets introduced in 2002 in the U.K. National Health Service and found a a U.K. indirect tax assessment and collection modest positive effect of the incentive on man- agency. The tax yield increased for the treat- ager motivation and effort. No high-quality ment team relative to the control group, with studies were found of this type of job within the increases due to more time spent on audits, developing countries. which resulted in recovery of greater tax reve- To summarize, when we winnow the pool nue. By contrast, Bertelli (2006) found that, in of studies to identify the subsets that are most the U.S. Internal Revenue Service, the incen- relevant to the tasks facing senior administra- tive scheme “crowded in” intrinsic motivation tors within the core civil service in non-OECD at the lowest pay levels and crowded it out at settings, the number of studies from which the highest levels. Similarly, a set of studies we might draw policy lessons becomes quite of performance incentives for agencies with small (figure 5). Therefore, no conclusions can responsibility for training and recruitment be drawn from the existing literature on the found considerable evidence of gaming among effects of PRP in these organizational contexts. agency staff in the choice of termination date The review does enable us to conclude that of the training for the participants (Asch 1990; the incentive theory prediction—that PRP has Heckman, Heinrich, and Smith 1997; Courty a role to play in craft jobs where outputs are and Marschke 2004). readily observable, such as teaching, health Only a few studies have attempted to evalu- care, and revenue administration jobs—holds ate the sorting effect of PRP. An experiment true, even though the day-to-day actions of with 115 Australian students that tried to dis- staff are unobservable. This apparently con- tinguish the potential incentive and sorting founds, at least in the short term, the concern effect of performance pay found evidence for in behavioral economics about crowding out both hypotheses (Cadsby, Song, and Tapon intrinsic incentives. This conclusion is sup- 2007). In addition, it found that low-produc- ported by observational and experimental tivity subjects were less likely to sort into pay- studies in developing countries, where the evi- for-performance jobs and that subjects with dence is generally more positive than in OECD higher levels of risk aversion avoided PRP, settings. suggesting important unintended side effects. At the same time, several observational stud- In a field experiment in the private sector ies identify problems with unintended conse- (Bandiera, Barankay, and Rasul 2006), some quences, generically subsumed under gaming managers were treated with the introduction the incentive scheme, which can subvert the of a performance-pay system, and productivity original intentions of the reforms. With the 18 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Figure Numbers of Studies and Their Findings 5 Actions or internal production process of the job Observable Not observable Production job Craft job Outputs from the job Relatively easily measurable Not easily Procedural job Coping job measurable Total sample: 153 Positive for PRP: 61% Public sector equivalent sample: 110 Positive for PRP: 60% High-quality public sector equivalent sample: 53 Positive for PRP: 70% High-quality core civil service equivalent sample: 3 Positive for PRP: 100% High-quality core civil service equivalent non-OECD sample: 0 Positive for PRP: n/a current evidence though, it remains unclear performance incentives can work, but the stud- whether incidents of gaming have a net nega- ies employ easily measurable performance tive effect in the presence of increased pro- indicators and use fairly unrepresentative sub- ductivity. Furthermore, while explicit incentive ject pools. Both concerns should caution policy schemes certainly increase the opportunity makers against accepting the results indepen- for gaming, standard civil service arrange- dently of other research. However, similar ments have their own unintended incentive results have been found across a varied set of effects, with employees engaging in behavior experimental settings, test locations, and sub- that increases the chances of easy work assign- ject pools, and the overall findings do resonate ments or promotions. It is simply unknown with the observational literature, improving whether existing forms of gaming are worse overall credibility and external validity. than similar behavior under performance pay. The strongest form of evidence comes from In addition, there might exist important cul- field experimental studies for craft jobs that tural differences in the prevalence of gaming neatly address concerns of internal and exter- performance standards in the public sector nal validity. Here, evidence is somewhat more between developed and developing countries. mixed. Several studies of teacher incentive pro- While to our knowledge no explicit research grams found no or transient effects of bonus on this question exists, work on the prevalence pay systems in U.S. schools, but in the devel- of corruption, behavioral norms, and the effec- oping world evidence has been more positive. tiveness of anticorruption efforts suggests that The discrepancy between teacher incentives gaming in highly corrupt bureaucracies might in the developed and developing world could be more problematic. stem from the relative magnitude of incentives For RCTs, the evidence again speaks in favor compared with normal salary, or from higher of the potential utility of performance pay marginal effects in the education production for craft jobs. Comparing various laboratory function in developing countries. Many fac- experiments, the results suggest that explicit tors affect the education process, all of which PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  19 are likely lacking in many developing country In sum, the evidence in relation to the schools. Improving one input aspect, such as hypotheses set out earlier is primarily around teacher presence and effort, could have con- Hypothesis 1: that PRP has a direct impact on ceivably larger marginal effects than the same staff effort and on the type of staff recruited input improvement in a developed country or retained. The literature shows no relevant school. support for Hypothesis 1 for coping jobs but For coping jobs in developing countries, the reasonable evidence concerning craft jobs evidence has to draw more on the case studies, (both within and outside OECD settings) even which will not meet the rigorous standards of though gaming is a persistent problem. Some an RCT but can nonetheless provide insights sorting effect is noted, primarily in OECD on the tradeoffs involved. country settings. The Case Studies: Pay Flexibility Schemes Fiscal sustainability Brazil, Chile, Indonesia, Korea, Malaysia, the limited to one six-year term in which to deliver Philippines, Russia, and Thailand represent on his promises, needed to quickly infuse was a trigger for the an interesting diversity in their pay flexibility some dynamism into a notoriously sluggish introduction of PRP schemes. It’s worth noting that performance bureaucracy. improvement was usually only one reason for Beyond these fiscal and political motiva- in Brazil and Korea, introduction of these plans, with fiscal and tions, it is not apparent from the case studies echoing moves in political reasons also important. Fiscal con- that these pay systems had achieved a mini- straints meant that the wage bill had to be mum threshold of functionality before intro- the OECD countries brought under control, and a different mecha- duction of second-generation pay flexibility a decade earlier nism for salary increases had to be found or reforms. With the exception of Thailand and, to be more explicitly linked to productivity surprisingly, the Philippines, these countries improvements. Fiscal sustainability was in par- had highly fragmented pay regimes. In gen- ticular a trigger for the introduction of PRP in eral, it was the norm for pay to vary between Brazil and Korea, echoing moves in the OECD workers doing similar jobs based on their countries a decade earlier. Korea introduced agency, occupational group, or geographic PRP in the aftermath of the Asian financial location, or because of idiosyncratic personal crisis in 1997, and Brazil similarly moved in characteristics. While in some cases, these this direction in the wake of the fiscal crises differences reflected official policies as speci- affecting Brazilian states in the early 2000s as fied in laws or regulations, more often they part of a package of budget and management were uncoordinated outcomes of ad hoc policy reforms to increase the efficiency of the public changes or political power, and they often pro- sector. vided incentives for unproductive behavior. Political crisis or a change in regime was As an example, in Indonesia base pay another motivation. Politically, PRP has often accounts for less than 20  percent of total been viewed as a means to enforce responsive- compensation for mid- to senior-level staff, ness, particularly among senior bureaucrats, to and there are hundreds of different types of the political leadership, or as a politically more allowances and supplementary payments for acceptable way of increasing public sector different categories of staff. Among these are wages. In Thailand the main driver for reform honoraria, which are additional payments to was former prime minister Thaksin Shinawa- staff for taking on tasks and responsibilities in tra, who had a degree of executive power that addition to their regular duties, such as attend- was unprecedented in Thai politics and who, ing workshops or participating in meetings given his business background, sought to use that are called by other agencies, or working PRP to infuse private sector responsiveness on a team on a special project. These hono- into the civil service. In the Philippines, PRP raria are paid in cash and are not included in was the initiative of the newly elected reform- the personnel budget, and they can amount minded president Benigno Aquino, who, to a third of total cash compensation. It is not 20 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  21 hard to see that honoraria can induce unpro- supplement) by mandating a distribution of PRP is ubiquitous in ductive behavior, since they are paid in a performance rankings of organizations or highly nontransparent manner and encourage staff. the Chilean public staff to maximize the number of project teams PRP is ubiquitous in the Chilean pub- administration, with they belong to and the number of meetings lic administration, with schemes that apply schemes that apply and workshops they participate in. to most central administration institutions In almost all cases, pay flexibility measures (across-the-board schemes) and others that to most central added another layer to an already complex pay are specific to particular institutions, such as administration regime, but one that was, at least rhetorically, the revenue authority. The across-the-board motivated by the desire to enhance individual schemes are an annual institutional-level institutions (across- and agency productivity. The question then is bonus based on the achievement of ministe- the-board schemes) whether the tradeoffs in terms of increased rial/agency targets and a team-level bonus inequity and possible lack of transparency and defined internally in each agency. Until 2011, and others that are lower intrinsic motivation were offset by the the agency-level targets consisted of achieve- specific to particular direct and indirect productivity-enhancing ment of outputs and improvements in pro- effects of these changes. cesses such as basic human resource practices, institutions planning and management control, procure- Performance-related pay ment, and audit. Each agency defined its action In the case countries, Brazil, Chile, Korea, plan in agreement with the Budget Office and Malaysia, the Philippines, Thailand, and, to a self-evaluated its accomplishments of the pre- limited extent, Indonesia have implemented vious year, with a group of experts validating PRP. These PRP schemes display a high degree the report. This procedure changed in 2011, of diversity in the five main design features with the parent ministry having a much stron- (summarized in table 5 below). The schemes ger role in discussing the agency’s targets and in Brazil, Chile, the Philippines, and, to some a stronger focus on results rather than on extent, Korea have both individual- and group- outputs and processes. For the working unit based incentives, either through two separate scheme, the concerned line minister or head schemes or through a formula that combines of agency divides the institution by groups and assessments of group and individual perfor- defines annual goals and performance indica- mance. Chile and Korea also have distinctive tors for each (collective performance agree- PRP schemes for senior civil servants. The ments), and the internal audit unit evaluates award can be paid either as an annual bonus, the level of accomplishment annually. The size as in Chile and the Philippines, or a merit of the award ranges from 4 percent to approxi- increment, as in Korea and Thailand. In Brazil mately 8  percent of base pay for both the the bonus increases based on the number of agency-level and working unit–level schemes, good successive performance evaluations staff implying that staff in a high-performing unit have. The performance assessment mecha- in a high-performing agency (usually over nism usually uses measures of organizational 95 percent of public sector staff) could receive and individual performance, but with notable an annual bonus of as much as 16 percent of variations in the use of quantitative standard- base pay. ized indicators versus individually specific A government-wide individual PRP scheme results agreements and short- versus long-term was introduced in 1998 but abandoned in measures. The size of the awards varies consid- 2003, as supervisors faced with the challenge of erably, from a low of 3 percent of basic pay in forcing a normal curve—when traditionally all Thailand to a high of more than 100 percent employees had received the top qualification of basic pay in Chile (for the Senior Execu- in the performance appraisal system—rotated tive Service). Finally, some of the countries— the bonus between groups of employees so that Korea, Malaysia, and the Philippines—have each received an equal share over the medium tried to counter the usual tendency of equal term. Today, Chile has only an individual per- distribution of PRP (which renders it a salary formance bonus scheme for members of the 22 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Senior Executive Service (SES), which is based PRP for Selected Careers in Brazil’s Brazil, both at Figure Federal Government on evaluations by the immediate supervi- the federal level sor and uses targets established in individual 6 Brazilian reais (thousands) Bonus related to institutional performance and in the state of results agreements. The size of the maximum 15 Bonus related to bonus is determined by the Ministry of Finance individual performance Minas Gerais, has a Basic wage in consultation with the Civil Service Director- performance bonus ate and the concerned agency based on market conditions and an assessment of the level of 10 scheme that consists pay needed to attract and retain suitable can- of individual- and didates. The incentives can be high, with the bonus equal to as much as 100 percent of base 5 group-based awards pay if all targets are met in the case of the head of a hospital in a remote location, but are usu- ally a maximum of 30 percent of base pay for 0 other SES positions. Almost all SES members Water Agriculture Transport Mineral Of cer of Social receive the complete bonus, and the PRP is specialist specialist analyst resources Chancellor policy specialist Of ce analyst therefore a de facto salary supplement rather Source: Ministry of Planning. than a performance incentive. Brazil, both at the federal level and in the state of Minas Gerais, has a performance interesting variations and considerably more bonus scheme that consists of individual- and vigor in its application. The group-based group-based awards (this study covered only bonus, called the Productivity Premium, Minas Gerais at the state level). Remuneration began in 2003 and is based on the percentage practices are largely organized around career of institutional goals achieved, as identified categories that group professional positions on in the results agreements that agencies sign a common salary spine. In the federal govern- with the governor and that are cascaded down ment a range of careers have prescribed break- to working units. There is a time-in-service downs for performance allowances derived requirement, so that bonus size is also condi- from either individual or institutional crite- tional on the number of days that staff work. ria (figure 6). The performance assessment The individual bonus, adopted in 2007, applies allocates points based on the achievement of to all public servants recruited after 2003 and individual, working unit, and institutional is calculated on the basis of both institutional goals, with the individual component having goals reached (derived from the results agree- a weight of 20 percent and the group compo- ments for the whole agency) and the individual nent a weight of 80 percent. Each point has a performance appraisal of the specific public monetary value, the details of which vary by servant, with 30  percent weight given to the career group, and is at the discretion of the former and 70 percent to the latter. The value concerned agency. The institutional compo- obtained by measurement of these two catego- nent of the PRP is usually linked to goals set in ries is then multiplied by a factor that takes the multiyear plan for each government entity. into account the basic salary of the public ser- Interviews with government officials stressed vant and the percentage associated with the that there is no explicit system of results agree- number of satisfactory evaluation cycles in the ments across government units or even a pre- course of the civil servant’s career. This mul- cise set of indicators set out in these broad tiyear satisfactory evaluation requirement was umbrella plans. Nonetheless, some ministries designed to provide long-term incentives for have established institutional goals that are set performance, with the public servant seeing out in their own specific regulations, which are a substantial increase in bonus size over time. published annually online. For example, a civil servant with 5 satisfactory The Brazilian state of Minas Gerais has performance evaluations will receive a 10 per- a similar formula for PRP but with some cent bonus, while one with 15 will receive a PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  23 30  percent bonus, with even higher bonuses are explicitly stated in the policy guidelines: The Philippines until possible in the future (figure 7). improve agency performance as measured by The Philippines until recently had very little achievement of departmental performance recently had very pay flexibility, with public sector remunera- targets and key presidential priority programs; little pay flexibility, tion approximating the single pay spine across improve individual performance; and improve with public sector government departments and agencies and agency compliance with existing governance- allowances making up less than a quarter of related laws and regulations, such as transpar- remuneration total monetary compensation. In 2012, at the ency in procurement, financial management, approximating the initiative of President Aquino, the government and disclosure of information. The PRP introduced an individual- and group-based scheme provides an integrated group and indi- single pay spine performance bonus that applies to all central vidual award. across government government departments and agencies. The Departments and agencies can qualify for objective of this scheme, as stated in the presi- the performance bonus if they meet 90  per- departments dential order that promulgated it, is to help cent of their agreed performance targets and and agencies deliver on the Social Contract of the Aquino additional good governance criteria. Within administration. The Social Contract lays out departments, comparable working-level units a broad reform agenda to promote inclusive (policy bureaus, implementing units, and growth in the Philippines, with combating services) will be force-ranked into three cat- corruption and “professional, motivated, and egories, and staff within these units will also energized bureaucracies” as key elements for be force-ranked into three categories so that achieving the government priorities of higher 10  percent of units and 10  percent of staff revenue generation and improved outcomes in within those units are classified as best per- education, health, and other services. The PRP formers, 25 percent of units and staff are in the scheme is a deliberately ambitious, govern- second category, and 65 percent are in the bot- ment-wide reform initiative that is designed to tom category. While the government’s ambi- “shake the bureaucracy out of its lethargy” to tion is to have the working unit performance achieve some sustainable impact in the remain- targets cascaded to the individual performance ing three years of the Aquino administration. evaluations, the individual bonuses are cur- The scheme, as currently designed, is aimed rently based on the traditional, trait-based per- at achieving three objectives, not all of which formance evaluation tool that has long been in use in the civil service. Individual bonuses, Minas Gerais PRP Bonus Size which are flat sums and not a percentage of sal- Increases over Time Based on Figure Consistency of Performance ary, can vary from 35,000 pesos to 5,000 pesos, depending on where staff are in the three-by- 7 Performance bonus as a percent of monthly base pay three matrix of working unit and individual 80 performance, implying that as a percentage of pay the highest incentives can vary between 10 to 18 percent of total compensation for clerical 60 and junior technical staff to less than 5 percent for senior management (figure 8). 40 Almost all of Malaysia’s recent pay reforms have concentrated on PRP. PRP was introduced in 1992 as a merit increment scheme based on 20 a forced distribution of assessments: only 3 per- cent of employees could qualify for a single PRP 0 increment in addition to the “normal” incre- 3 5 10 15 20 25 30 35 ment, and only a further 2 percent could qualify Number of satisfactory for a double increment. (At the other end of the annual performance evaluations distribution, 5 percent of staff would not receive Source: Government of Minas Gerais. any increase, and would become subject to 24 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E PRP as a Percentage of Total Pay in In Korea, PRP has been one of several In Korea, PRP has been Figure the Philippines reforms, such as performance budgeting, one of several reforms 8 Percent of total pay Highest incentive Middle incentive aimed at strengthening performance man- aimed at strengthening 20 Lowest incentive agement, particularly in the wake of the Asian financial crisis. Historically the pay performance structure had been based on seniority, but 15 management, in 1999 the government introduced PRP to improve recruitment and create a more per- particularly in the 10 formance-oriented culture in the civil ser- wake of the Asian vice. These PRP schemes have spread widely in the public sector, with 71 percent of public financial crisis 5 bodies using PRP. Two types of PRP schemes are currently in use: an annual merit incre- 0 ment program for senior officials (members 1 5 10 15 20 25 30 33 of the Senior Civil Service [SCS]) and a per- Salary grade formance bonus program for middle- and Source: Department of Budget and Management. lower-level officials. For the SCS, PRP is based on individual disciplinary action.) This scheme was modified performance, organization-level performance substantially a decade later. A new “competence- based on the concerned staff’s position as based assessment” was introduced in response a manager (in terms of citizen satisfaction to complaints of manager favoritism by the civil with service delivery, citizen satisfaction on servants’ union, CUEPACS. In practice, this major public policies, and similar criteria), took the form of an assessment of employees’ and job-related abilities (core competencies, performance at a training workshop. Although customer-orientation, and so forth). The per- the competence assessment constituted only formance incentive is high, with a maximum 30 percent of the total assessment, in practice it of 15  percent of annual base pay increase to counted for more, since the competence assess- the top performers (as a merit increment, this ment was viewed as a better discriminator of is a permanent increase in base salary), and a performance than the managerial assessment. 10 percent and 5 percent increase for the next This competency-based assessment also eventu- two performance categories, respectively. The ally faced resistance from CUEPACS, this time nature of the performance agreement depends about the length of time that its members were on the type and task of the ministry. For exam- spending on assessment workshops, and the ple, policy-oriented departments have more indignity of having to go “back to school.” The qualitative targets, while the service delivery– system is being changed again, and the govern- related departments have more quantitative ment has set up a pay commission to review targets. overall pay arrangements. For junior staff, the minister in charge has Malaysia has performance evaluation at discretion in deciding what proportion of the agency as well as individual level. There PRP should be based on group or individual are National Key Result Areas; ministerial and performance. The most common is a fully department heads’ key performance indica- individual scheme, which provides for signifi- tors; “star ratings” of department performance cant variations in the size of the annual bonus conducted by the Malaysian Administrative across individual performance categories. Modernization and Management Planning This scheme is currently in operation in 30 Unit; reports by the auditor-general (which of 44 government agencies and is based on a are reported in the press); and ISO 9000 qual- forced distribution, with the top performers ity ratings and client’s charters. Government (the top 20 percent) getting an annual bonus departments are required to report their per- equal to 172.5 percent of basic monthly pay (or formance on all these measures. approximately 14 percent of annual pay), the PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  25 Table Key Design Features of PRP in Case Study Countries 5 Individual- and group-based awards Nature of the performance evaluation Time horizon Probability of Country Individual based Group based of the award Individual based Group based Size of the award receiving the award Chile • SES only • Achievement of • Annual bonus • Annual performance • Based on achievement • From 30% to 100% • Almost 100% of staff • Specific schemes for working unit only, paid in appraisal of SES of agency and working of base pay for SES, receive the individual certain agencies, goals (collective four installments members. Evaluation unit targets. Internal based on the job bonus (for SES) and such as the revenue performance during the year led by immediate audit unit evaluates • From 4% to 16% the institutional and authority agreements) supervisor and based achievement of working of base pay for working group bonuses • General across-the- • Achievement of on targets established unit targets the group awards board scheme was institutional goals in individual results (institutional and introduced but later (Management agreement (three years) working unit) for abandoned Enhancement all staff Program); applicable to almost all staff Brazil (federal • All staff. Based • Achievement of • Annual bonus • Details at the discretion • Based on goals • Specific legislation government) on productivity at working unit and only of the ministry but established at the for each career work, commitment, institutional goals based on a combination beginning of each identifies the knowledge, and • 80% weight in of self-assessment, evaluation cycle monetary value of compliance with rules overall bonus supervisor assessment, each point • 20% weight in and team member overall bonus assessment Brazil • Individual • Productivity • Annual bonus, • Based on a combination • Each institution defines • Productivity • 96% of staff receive (Minas Gerais) performance premium. Formula but the size of self-assessment, its action plan in premium: Maximum a satisfactory rating; allowance (ADE) based on working of the bonus supervisor assessment, agreement with the size of award is one 74% receive the • The ADE depends days without increases over and team member Secretary of Planning. month’s salary highest rating; and on both individual absenteeism and time with assessment Institution self-evaluates • ADE: Varies 22% the next highest performance achievement of successive accomplishment through with number rating assessment (70%) institutional results satisfactory a committee with at of successive and achievement of agreements performance least one member performance institutional goals evaluations of the Secretary of evaluations, ranging (30%) Planning from 5% to 70% of base pay Korea, Rep. • Two types of • For SCS, agency • Merit increment • For SCS: appraisal based • For SCS: Merit • Forced distribution schemes, one for performance is for senior staff; on agreement between increment as high for fully individual members of the one element of the annual bonus for agency head and heads as 15% of annual schemes: 20% of staff Senior Civil Service staff performance junior staff of working units and base pay get the highest bonus (SCS) and one for all evaluation includes individual • For other staff: • 30 government other staff • For all other staff, and organizational Bonus as high as agencies have fully the minister in performance- and job- 14% of annual individual bonus charge has discretion related abilities base pay schemes; none has a in deciding the full group bonus and mixture of group and equal individual bonus individual award scheme; and 14 have schemes combining individual- and group- based bonuses Malaysia • All staff from 1992 to • All staff have • Merit increment • Performance assessment • Forced distribution 2012. PRP suspended received one scheme was later with only 2% of staff at time of writing, month’s pay bonus modified to a receiving a double with possibility of in recent years competency assessment merit increment reintroduction for (Inland Revenue: 5 scheme, and then later senior staff only months, based on suspended performance) Thailand • All staff • None • Merit increment • Most employees • Most employees receive receive a pay something increase of less than 3% Philippines • All staff • Department- • Annual bonus • Based on traditional, • Departments/agencies • Varies from 1% to • Forced distribution: and unit-level “trait-based” individual need to meet 90% 18% of total pay 1% of staff receive the performance performance assessment of targets and 100% highest bonus of good governance conditions. Working units are then evaluated against their targets Indonesia • Bureaucracy Reform • Revenue authority: • Monthly • Fingerprint machines • Tax collections of the • A month’s BR • All staff in the agency (BR) allowance group award based allowances for are used to measure tax office allowance is cut get the BR allowance conditional on staff on tax collections individual BR attendance if staff fail to attendance pay; three bonus meet minimum payments for attendance the revenue requirements authority scheme Source: Urbanization study team. 26 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E In Chile key next category receiving 125 percent, the third Differentiation category receiving less than 85  percent, and In Brazil, Chile, Indonesia, and Russia total government agencies the bottom category (the lowest 10  percent) pay for a civil servant in a similar job varies departed from the getting no bonus. In departments where indi- across public entities based on the government vidual performance is hard to evaluate objec- agency, location, and the specifics of individ- former single pay scale tively, the minister can decide to distribute ual contracts. in the late 1970s the individual bonus equally, with the size of In Chile the “controlling institutions”— the bonus based purely on departmental per- among them the Superintendence of Pensions; formance. However, to date no government Superintendence of Health; the central public agency has opted for this variation. procurement office, ChileCompra; the Unit In Thailand, PRP was a major reform of for Financial Analysis; and the Defense of the the Thaksin government and mandated by the Competition Agency—and the “finance sector Civil Service Act 2008. Under the scheme, in institutions”—the Treasury, Budget Office, and each half-year cycle employees can receive a revenue authority—have separate pay scales salary increase of up to 6 percent. In theory, that are higher than those of other public enti- an employee could get 12  percent in a year, ties. These agencies departed from the former following the manager’s assessment of prog- single pay scale in the late 1970s, since they were ress against the key performance indicators viewed as key government agencies in charge of that employees are supposed to produce at the raising revenues, of regulating, and of exercis- start of every year. Because the PRP budget is ing control over the public sector. In addition, capped at 3 percent for each cycle (6 percent most ministries and agencies receive a “critical for the year), few employees in practice receive functions” allowance of up to 100  percent of 6 percent in a cycle, let alone 12 percent over their remuneration from the Budget Office to a year. In the first year of operation, most distribute among few staff, with agency heads employees received a pay increase of just below deciding who receives the benefit and how 3 percent. much selected staff receive. This allowance is The 2008 procedure was preceded by an normally used to increase the pay of key middle- elaborate job regrading and simplification management positions (for example, a Budget exercise, so that managers had a framework of and Planning Division chief for a line ministry) grades and job descriptions to use as the basis and a small number of staff. Moreover, approxi- for their PRP assessments. The public service’s mately 60 percent of public sector employees are 225 job classes were drastically simplified to contractual (contrata) staff with individual con- four broad job categories: executive, manage- tracts determined by the agency head in which ment, knowledge, and general. pay levels can be set at any point of the pay scale Indonesia does not have across-the-board within limits determined by the Budget Office. PRP in central ministries and agencies. How- Indonesia has a complex structure for ever, some agencies, particularly those that civil service pay with numerous allowances receive an additional allowance, have intro- that result in significant differentiation of duced fingerprint machines to record atten- pay between staff doing similar jobs. For staff dance and hours at the office, and these in central government ministries, cash com- additional allowances are cut if staff are pensation (excluding in-kind allowances for absent without permission or do not work the car, housing, and utilities) has three main required number of hours. A PRP scheme in elements: basic salary, allowances, and hono- the revenue agency gives a group bonus for raria. Basic salary and allowances are consid- achievement of property tax collection targets ered a “fixed” or guaranteed allocation that by relevant units. The bonus is given three staff receive regardless of the type or level times a year and can be between one and four of work that they are engaged in. Honoraria times monthly basic pay, depending on the are, at least on paper, given to staff for taking amount of tax revenues collected by the tax on activities and responsibilities above and office. beyond their regular duties. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  27 A major source of differentiation, and the In Brazil considerable pay differentia- As of 2011 Brazil’s focus of this study, is the significant variation tion across careers exists at both the federal in the levels and structure of pay between the and state levels. In recent years the number federal government few central ministries and agencies that are of careers has grown significantly, with each had 161 careers and deemed to be strategically important or under- ministry demanding its own set of identifiable similar professional going internal reforms. This reform exercise careers to validate its status in the bureaucratic is called Bureaucracy Reform (BR). The BR pantheon. As of 2011 the federal government structures, each agencies—as of 2012, 14 of 76 total central gov- had 161 careers and similar professional struc- with its own salary ernment ministries and agencies—include the tures, each with its own salary structures, usu- Ministry of Finance, the audit institutions, the ally varying in the size and terms of allowances structures, usually planning agency, and the personnel agency. and administrative arrangements, a framework varying in the size and They have additional, supplemental salary that is replicated at the state level. This differ- scales, though the BR pay scales have been entiation is instituted through legislation spe- terms of allowances largely determined in an ad hoc manner and cific to each position or career group, often and administrative differ across agencies, with seven separate BR through intensive bargaining with employee pay scales in place in BR agencies. unions. One noteworthy source of differentia- arrangements Average total monetary compensation in BR tion is “subsidy” remuneration for 32 priority agencies is between two and over four times careers, a unified compensation package that that of non-BR agencies, depending on staff guarantees salary increases for a specified seniority (figure 9). Pay is also more decom- period (and is thus not subject to periodic pressed in BR agencies, which means that pay negotiations) and makes all compensation is increases quite significantly for staff over their pensionable. career. Moreover, BR pay is based on a job The difference in pay between the federal evaluation–based parallel grading structure, and state levels is also significant. States pay so staff of similar seniority can be assigned to salaries significantly lower than their federal different grades, with the result that staff with equivalents, since they are constrained by the the same seniority can earn very different BR Fiscal Responsibility Law of 2000, which places a allowances within the agencies. For example, ceiling on state government wage expenditures. there can be a sixfold difference in the BR Figure 10 shows the variation in pay between allowance of the senior-most staff, leading to a careers in the federal government, variation significant variance in total pay. for a particular career between the federal Figure Pay Differentiation in Indonesia 9 Ratio of total monetary compensation in BR to non-BR agencies Average total pay of staff Ratio Pay compression ratio BR agencies Non-BR agencies 5 6 5 4 4 3 3 2 2 1 1 0 0 Increasing seniority of staff Increasing seniority of staff Note: The pay compression ratio is the salary of a particular staff as a multiple of the lowest paid staff. Source: World Bank staff calculations based on Ministry of Finance data. 28 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E government and the states, and variation in (figure  11). In federal ministries average pay pay across careers in the state of Minas Gerais. can be three times higher in ministries respon- Pay flexibility in Russia consists largely of sible for policy development and regulatory pay differentiation and significant use of PRP oversight compared with agencies undertaking in the service delivery sectors. For the core service delivery and supervision, and much of civil service, there is significant variation in this variation is at the senior manager level. pay levels across federal ministries, between Average compensation in territorial authori- federal civil servants working in Moscow ties is less than half the average compensation and those in the regions, across subnational in central authorities for similar jobs, with the administrations, and across staff in similar jobs gap larger in managerial jobs.12 Figure Pay Differentiation in Brazil 10 Pay ranges across subsidized careers in the federal government Variation in average pay for similar jobs across federal agencies Brazilian reais (thousands) Brazilian reais (thousands) Senior level 20 20 Entry level 15 15 10 10 5 5 0 0 Careers Transport Environment Mineral Supplementary Regulatory infrastructure production pension agencies Comparison of average wages for public policy specialist careers across states and the federal government Average wages across careers in Minas Gerais Brazilian reais (thousands) Brazilian reais (thousands) 20 30 25 15 20 10 15 10 5 5 0 0 Alagoas Minas Gerais São Paulo Goiás Espírito Santo Pernambuco Rio de Janeiro Mato Grosso Bahia Ceará Federal Teacher’s assistant Environmental specialist Public policies and government management specialist State prosecutor Tax auditor Police chief Sources: Ministry of Planning; Andrade 2011; Minas Gerais Civil Service Directorate. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  29 Figure Pay Differentiation in the Russian Federation 11 Rubles (thousands) Average monthly salary across federal ministries 100 75 50 25 0 Ministry/agency Average salaries across levels of government Rubles (thousands) Federal central 150 Territorial federal Regional Municipal Average norminal salary paid in private sector Average norminal salary paid in Moscow 100 50 0 All Managers Specialists Support specialists Source: Rosstat and HSE Institute for Civil Service Development data. Assessing the Evidence from the Case Studies: Performance-Related Pay In coping jobs the Direct pay flexibility levers in Brazil, both at the federal and state levels. The case studies reinforce the main finding The federal PRP scheme for tax collectors tendency has been for from the literature review that PRP can have was introduced in the late 1980s in the con- PRP to degenerate a direct incentive effect in craft jobs for which text of severe fiscal pressures and the need to performance, either individual or group, can raise tax revenues and was the subject of the into a de facto be measured with objective data. By contrast, empirical study by Kahn, De Silva, and Ziliak salary supplement in coping jobs, for which performance assess- (2001) reviewed earlier. This individual- and ments are more subjective, the vast majority of group-based scheme awarded a bonus to tax staff tend to get high ratings, with the result inspectors based on individual evaluations that PRP degenerates into a de facto salary sup- by supervisors as well as the collection per- plement. Some countries have tried to address formance of the local tax agency based on this problem by requiring mandatory distribu- fines collected and achievement of other tar- tion of performance ratings, but that solution gets (total tax collection, number of inspec- is not without implementation challenges and tions, collection of overdue taxes). The strong it depends heavily on the level of trust and rebound in revenues—significant increases in general functionality of human resource man- revenue collection and individual productiv- agement in the civil service. Our findings are ity, as measured by a 75  percent increase in that in low-trust environments mandatory dis- fines per inspection—was attributed in part tributions can have a negative impact on staff to PRP. Despite its apparent success, this fed- morale and performance. eral scheme was not politically sustainable, and under pressure from the employee unions was Craft jobs later abandoned and replaced by a high, fixed- Brazil provides several instances of productiv- pay system. ity improvements in the police and the revenue In Minas Gerais the Secretariat of Finance authority. In the state of Minas Gerais two of introduced an individual- and group-based the performance targets for the police were PRP scheme in 2005. The group award is for weapons seizures and police operations, in large part based on attainment of rev- both of which increased dramatically after the enue targets, with staff attendance required introduction of the performance incentive, to participate in the bonus. The individual the former by nearly 60 percent and the latter award was based on the achievement of goals by almost 50  percent. De Assis (2012) found that are linked to a cascade of results agree- that staff attributed this increased activity to ments. These results agreements have grown the introduction of agreed-upon institutional in sophistication over time in recognition of goals and the accompanying PRP financial the perverse incentives that a pure focus on incentives. revenue collection can engender. In this case, Revenue administration has been a priority staff had come to focus on increasing short- area for results-based management and PRP term revenues even though building enduring 30 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  31 relationships of trust with potential taxpayers Perceptions of Effort in the BR Figure Agencies in Indonesia In Chile, PRP was was a more effective way to broaden the tax base on a sustained basis. Individual evalua- 12 “Others of the same rank in the ministry contribute more than is expected of them” viewed quite favorably tions now increasingly focus on operational Percent of respondents Agree or in the revenue priorities (for example, concentrating efforts 80 strongly agree authority and the on large taxpayers) and also include taxpayer facilitation and relationship-building variables Civil Registry 60 such as frequency of site visits. In Chile, PRP was viewed quite favorably in the revenue authority and the Civil Registry. 40 The Civil Registry’s PRP scheme is based on net customer satisfaction as measured by a sur- 20 vey conducted by an independent source, nor- mally a university or consulting firm selected by competitive bid. The survey asks about wait 0 times and citizens’ experiences with online DG Tax DG Bappenas MenPAN Non-BR Treasury ministries services. Within the agency’s customer satisfac- BR ministries tion unit there is a perception that the bonus has helped improve staff performance. Source: World Bank survey. In Indonesia the revenue authority—the Directorate General of Taxation (DG Tax) of by the view of the authority’s own staff, who the Ministry of Finance—is the only govern- were generally neutral about PRP’s impact on ment agency that has introduced an additional staff effort. The revenue authority is also a rare performance bonus (on top of linking BR pay example of an agency that enjoys substantial to staff attendance, as in the BR agencies). This human resource management autonomy in bonus is based on achieving property tax col- Malaysia and where executives were motivated lection targets and goes to all staff of the tax and had the authority to create a new work cul- office. In the World Bank survey of govern- ture. PRP is thus part of a package of human ment officials, DG Tax ranked highest on the resource management reforms that has suc- staff effort question (“whether others in the ceeded in creating a performance culture in agency contribute more than is expected of the authority. them”) (figure 12). These differences among The case studies of PRP suggest that in craft the four BR agencies surveyed—DG Tax, the jobs a combination of group and individual Directorate General of Treasury (DG Trea- bonuses can be effective and that worries about sury) in the Ministry of Finance, Bappenas free-riding in group bonuses may be exagger- (the national planning agency), and MenPAN ated. The size of the bonus in many cases was (the national oversight agency on personnel significant, for example, in the case of Brazil- matters)—suggest that factors other than the ian tax officials. Another interesting finding BR allowance may have an impact, and the dif- is that managers, at least in some cases, were ferences between the two Ministry of Finance aware of the risks of gaming and had evolved agencies suggest that factors specific to DG performance assessments to minimize it, as Tax, such as its performance bonus, may be exemplified by the revenue authority in Minas important, a point that was also emphasized in Gerais. the expert interviews. In Malaysia the only positive accounts of Coping jobs PRP come from the revenue authority, where The case studies revealed several challenges in interviews with management revealed a per- implementing PRP for coping jobs that imply ception that staff effort is linked to the corpo- that the financial incentive had little direct posi- rate targets of increased revenue collections. tive effect, and in some cases possibly negative This sanguine management view is qualified effects, on staff effort. By definition, these are 32 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E In Brazil, PRP had jobs in which outputs cannot be easily measured The state government of Minas Gerais in and therefore the performance evaluations Brazil has also received considerable interna- some effect on that form the basis of the financial incentive tional attention for its performance orienta- the extremes of are either based on subjective evaluations by tion, but it has similarly found it difficult to use supervisors and review panels or on some quan- annual performance appraisals to meaning- the performance titative input or process measures. Given these fully distinguish among staff. The overwhelm- distribution by helping difficulties in performance assessment, the ing majority of civil servants receive satisfactory “steady state” in most bureaucracies is for the ratings (more than 70 points), though there is in disciplining blatantly vast majority of staff to be given a best or next some consideration to differentiating between incompetent staff best performance rating in performance evalu- the top performers (more than 90 points) and ations, with the result that the performance those who score above the threshold but are bonus is given with close to probability 1 and so not outstanding performers (between 80 and cannot have a direct incentive effect. Why most 90 points). This performance distribution has bureaucracies converge to this norm is a ques- not fundamentally changed since 2005, when tion that cannot be easily answered here. One the individual performance bonus scheme was reason may be the likelihood that staff will be introduced, with the top two performance rat- working with each other for a long period of ings given to almost 90 percent and 20 percent time, in contrast to the private sector, where low of staff, respectively (figure 13). Individual turnover encourages a nonconflictual attitude. PRP in Minas Gerais is therefore also largely a Chile, despite its reputation as having an salary supplement. efficient, performance-oriented public admin- Interviewees in Brazil did note that PRP had istration, abandoned a government-wide indi- some effect on the extremes of the performance vidual PRP scheme after five years because distribution by identifying the outlier staff and managers were rotating the award, with a third helping in disciplining and even dismissing of the staff getting the highest bonus every blatantly incompetent staff. PRP may also have year. Similarly, in the current PRP scheme reduced the number of leave-takers and dis- for senior managers almost all SES members couraged extended absences from service. receive the complete bonus, and PRP has Similarly, in Indonesia the World Bank therefore become a salary supplement rather survey of government officials found that the than a performance incentive. Interestingly, despite this high bonus, the turnover rate Results of the Performance Evaluation Figure of Permanent Staff in Minas Gerais among the SES cadre has averaged 50 percent over the past decade, reflecting the difficulty 13 Percent of staff Points More than 90 of attracting and retaining senior talent in a 100 80–90 highly dynamic labor market. 70–80 Less than 70 In Thailand most of the staff who were interviewed felt that PRP had only marginally 75 improved work effort because of the small size of the performance bonus (the bonus budget is 3–6 percent of the total salary budget), the fact that most staff got the bonus, and the influ- 50 ence of nonmonetary motivators such as sense of duty. Some staff suggested that the effect would be bigger if the amount was higher. 25 While staff have made the usual complaints about the weak links between pay and perfor- mance, overall they support the scheme. The government has recognized that the 6 percent 0 2005 2006 2007 2008 2009 2010 2011 ceiling on PRP increases is a constraint and is planning to increase it. Source: Minas Gerais Civil Service Directorate. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  33 linking the supplemental pay in the BR agen- that performance appraisals enjoy among staff. Without objective cies to staff attendance may have reduced the In Malaysia the forced distribution introduced proportion of staff leaving work early. In the in 1992 was abandoned 10 years later under performance BR agencies 11 percent of staff reported that pressure from the civil service union follow- measures, it is more than 20  percent of their coworkers in ing complaints of manager favoritism. It could very difficult for the agencies left work early or spent excessive in fact be argued that this policy did harm time on personal matters, far fewer than in staff morale, teamwork, and the relationship supervisors to credibly non-BR agencies, where 30 percent of the staff between management and staff. Similarly, in discriminate between reported that their coworkers were similarly the Philippines many of the staff interviewed delinquent (figure 14). This difference is quite complained about individual forced ranking as the achievement striking and cannot be fully explained by the being highly subjective and unfair and poten- of the majority of proliferation of fingerprint attendance record- tially hurting morale. Even in Korea, generally ing machines in ministries and agencies.13 regarded having as a high-performing civil the staff who are in To some extent, the use of PRP to sanction service, studies have noted significant differ- the “fat end” of the only the outlier staff makes sense, since they ences in perceptions of PRP between central can be easily identified and justified to most government and local government staff, with normal distribution staff as deserving special treatment. By con- PRP in the former being viewed as relatively of performance trast, without objective performance measures, well implemented and in the latter viewed gen- it is very difficult for supervisors to credibly erally negatively and not supported by staff discriminate between the achievement of the (Han 2010; Lee 2010). In less accomplished majority of the staff who are in the “fat end” of civil services, mandating performance distri- the normal distribution of performance, and butions can risk hurting individual and agency any significant variations in pay for this group performance. could easily be viewed as unfair and breed The World Bank survey of government resentment among staff. officials in the Philippines revealed that staff Some countries, such as Korea, Malaysia, in coping jobs clearly did not believe that the and the Philippines, have tried to counter this performance bonus had any positive effect tendency of uniformly high performance rat- on effort, as indicated by the unanimity in ings and equal distribution of the performance their disagreement with the statement that bonus by mandating forced distribution of per- the proportion of coworkers working late had formance ratings. This policy is risky, and its increased (with between 42 and 48  percent efficacy depends very much on the legitimacy disagreeing or strongly disagreeing) (fig- ure  15). Staff views diverged based on indi- Perceptions of Attendance in BR and Figure Non-BR Agencies in Indonesia vidual performance rankings, and therefore the size of the bonus, on the effectiveness of 14 “Percent of staff who leave work early or spend excessive time on personal matters” the individual performance appraisal process Percent of respondents More than 20 percent (between 37 and 35  percent disagreeing or 30 strongly disagreeing that it identified individu- als who did not contribute); the transparency of the individual performance rating (between 20 29 and 46 percent agreeing or strongly agree- ing that it was transparent); and the impact of the incentive on staff morale (between 28 and 40 percent agreeing or strongly agreeing 10 that it had demotivated their coworkers). The bottom-­ r anked performance category, which comprises the biggest group in the distribu- 0 tion, was quite clear that the rating process BR agencies Non-BR agencies was not transparent and that the performance Source: World Bank survey. bonus had demotivated their coworkers. 34 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Perceptions of PRP and Individual Performance Appraisals, Effort, and Motivation in Figure Coping Jobs in the Philippines 15 “As a result of PRP, the proportion of coworkers working late has increased” “The annual performance appraisal process identifies individuals that do not contribute” Performance ranking Disagree or strongly disagree Performance ranking Disagree or strongly disagree of respondent Neither Agree or strongly agree of respondent Neither Agree or strongly agree Top performers Top performers Second best Second best category category Third best Third best category category 0 25 50 75 100 0 25 50 75 100 Percent of respondents Percent of respondents “The rating process for the individual performance bonus is transparent” “PRP has demotivated your coworkers” Performance ranking Disagree or strongly disagree Performance ranking Disagree or strongly disagree of respondent Neither Agree or strongly agree of respondent Neither Agree or strongly agree Top performers Top performers Second best Second best category category Third best Third best category category 0 25 50 75 100 0 25 50 75 100 Percent of respondents Percent of respondents Source: World Bank survey. These perceptions suggest that the direct without its own complications. On paper at incentive effects of PRP in the Philippines are least, group-level bonuses can be more objec- weak and possibly negative. Yet the overall view tive and based on agency and working unit of the scheme among staff is positive, with even outputs and therefore potentially more accept- a majority of the bottom-ranked staff believ- able and more likely to have a direct incentive ing that it is a good idea (figure 16), suggesting effect. For policy units, as opposed to service that other factors might be at play, as explored delivery units, however, these performance below. indicators tend to be process oriented and For coping jobs, group-based PRP linked therefore more vulnerable to manipulation to working unit and institutional goals is not and gaming. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  35 Figure Overall View among Civil Servants of the PRP Scheme in Coping Jobs in the Philippines 16 Performance ranking “PRP is a good idea overall” Disagree or strongly disagree Neither Agree or strongly agree of respondent Top performers Second best category Third best category 0 25 50 75 100 Percent of respondents Source: World Bank survey. In the Philippines, for example, the desire indicator measures the ratio of authorized to to use quantitative output indicators (which received requests from departments for vari- in the Philippines are called major final out- ous budget and organizational change propos- puts) for all agencies has led to some subop- als. These indicators are vulnerable to gaming timal choices. The performance indicators (for example, splitting one advisory into two in for the Department of Budget and Manage- order to increase the output) and have unclear ment include the number of advisories issued links with performance. By contrast, the ser- (counting the number of documents) and vice delivery departments have more mea- the quality and timeliness of these advisories, surable and independently verifiable output in contrast to the output indicators of service indicators. delivery departments such as roads and educa- Gaming was a major concern in Chile’s tion (table 6). Similarly, another performance group-based bonus schemes, since indicators Table A Sample of Performance Indicators for the Group-Based Bonus Scheme in the Philippines 6 Department Performance indicators Policy and oversight departments Department of Budget and • Number of advisories and directives issued Management • Number of agencies whose spending capacity was evaluated • Release of funds for priority expenditures within five days after the receipt of approval from President Housing and Land Use Regulatory • Policies on housing, land use planning, and real estate development reviewed or conducted within the year Board • Proportion of local land use committee members trained • Percent of mediation conferences conducted within 60 days over total number of complaints filed Service delivery departments Department of Public Works and • Number of road maintenance projects Highways • Number of projects paving unpaved roads • Percent of projects completed in accord with plans and specifications Department of Education • Improvement in proportion of grade 6 pupils’ test scores • Improvements in school graduation rates • Proportion of private schools with permit to operate or acquired recognition Source: Data from the respective departments. 36 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E PRP can encourage were mainly process oriented, self-imposed, refuse to distinguish between staff in their and self-evaluated, resulting in the deliber- performance appraisals or targets are manipu- managers to foster ate setting of easy-to-meet targets. The Bud- lated so that most are achieved. The case stud- more teamwork get Office, which was charged with the review ies do not suggest that PRP caused a decline in process, could not feasibly control gaming intrinsic motivation or that a lack of pay trans- among staff and behavior in more than 200 institutions. More- parency caused problems, with the notable to improve the over, staff unions put considerable pressure exception of when it was implemented through on agency authorities to ensure that targets a forced distribution. In most cases the scheme dialogue between were achievable and no portion of the variable defaulted to a salary supplement, which is some managers and staff pay was at risk. Practically all Chilean public cases did have a positive effect on recruitment sector institutions have consistently achieved and retention (as in Minas Gerais) and in oth- their performance indicators, indicating that ers did not (Chile’s SES). Insistence on objec- PRP has served as a more politically viable way tive, measurable indicators for these jobs led to to grant pay increases rather than as a direct an abundance of input and process indicators, incentive for performance improvement. which resulted in some perverse consequences Similar problems of gaming were found and gaming behavior, though it is unclear what in the group-level schemes in Brazil. Some effect gaming had on agency productivity. One institutional goals received inordinate atten- can speculate that this plethora of indicators tion, drawing energy away from other essen- did cause some harm through performance tial tasks. Other organizations gamed goals measurement fatigue and skepticism of pub- by establishing low bars that the agency could lic employees and the creation of an army of not fail to meet—for example, the Ministry of employees in each agency to ensure that the Environment, which attained 300  percent of gaming was correctly executed and no staff’s some targets. pay was at risk. There is some evidence from the case stud- ies that PRP may have had a direct incentive Indirect pay flexibility levers effect of sorting. Inspired by Chile, the gov- PRP can theoretically improve individual and ernment of Minas Gerais instituted a Public agency productivity through another, more Entrepreneurs Program in 2007 that created indirect channel by encouraging gradual 90 positions, open to recruitment from the changes in management that in turn could private sector and within government, whose improve individual employee performance. job description was to help implement achieve- The three relevant management practices are ment of government goals. These public entre- goal setting, teamwork, and individual perfor- preneurs were charged with championing mance assessments. PRP can give managers statewide reforms across the public sector and incentives to shift an organization’s focus to were paid substantially on a performance basis. results rather than inputs through the disci- Half the public entrepreneurs came from the pline of setting organizational targets and reg- private sector, and recruits were generally ular monitoring of progress. It can encourage considered to be of high caliber. The meth- managers to foster more teamwork among staff ods of recruiting this cadre were also more and to improve the dialogue between manag- stringent than for the rest of the public service ers and staff. And it can improve the individ- and included an oversight committee and an ual performance appraisal process by better evaluation committee. As of 2009 the program linking individual results agreements to orga- was operating in 13 different ministries, and nizational goals. All of these practices help 25 percent of the positions were managerial or inculcate a “performance culture” or “mission other strategic positions. orientation” within the agency. In sum, the case studies show that PRP has The emphasis here is on gradual and evo- a limited direct incentive effect in coping jobs lutionary change. The usual path is for coping because the probability of receiving the incen- agencies in the first stages of PRP to have a tive is close to 1, either because managers surfeit of input and process indicators, which PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  37 leads to the problems mentioned. Over time, public expenditure reforms, and the perfor- In Minas Gerais however, these indicators can be refined and mance contracting through cascading results can approach, even in coping jobs, a degree of agreements, with the whole process coordi- results-based output focus—through, for example, the more nated by the center through the newly cre- management and PRP frequent use of perception surveys among ser- ated Secretariat for Human Resources and encouraged greater vice users (which are normally other govern- Management (SEPLAG) reporting to the gov- ment agencies). ernor. PRP was developed in the context of a delegation of human In Chile the government that took office in results-based management framework, and, resource management 2010 was well aware of the challenges of imple- as discussed, consisted of both individual- and menting PRP and made efforts to modify the group-based bonuses. authority to the performance management system through bet- In interviews, staff concluded that PRP was implementing agencies ter control from supervising ministries. This integral to the performance agenda in Minas agenda included setting and evaluating perfor- Gerais. Staff reported that PRP had helped mance targets for the agencies within their sec- clarify expectations and individual goals and tor, giving more weight to actual results than targets, and that these results were due to to processes, and encouraging agencies to set changes in the planning process, which led to higher targets by recognizing partial comple- a self-reported boost in morale and a perfor- tion of targets as opposed to recognizing full mance-oriented culture and better results on achievement or no achievement. the job. In Korea the general perception is that PRP Another notable aspect of the reforms in helped improve the quality of individual and Minas Gerais is that the package of results- organizational performance appraisals over based management and PRP—the data do time, complemented performance budget- not permit distinguishing between the two ing reforms, and has now achieved legitimacy on this point—encouraged greater delegation and support within the central government. of human resource management authority to In the initial years the system was viewed as the implementing agencies. The most signifi- very lenient, particularly for senior officials, cant feature of this autonomy was the ability who received uniformly high ratings. Pressure of agencies to change their organizational from the National Assembly led to the intro- structure and hence staff positions and func- duction of forced distribution of ratings in tions without prior approval from the central 2009. Forced distribution appears to have been finance and personnel agencies, provided implemented reasonably well, at least for senior these changes were budget neutral. This auton- civil servants in the central government, partic- omy was granted by SEPLAG on the condi- ularly after organizational performance mea- tion of performance: agencies could continue sures were linked to individual evaluations of to function autonomously if they performed senior civil servants. Government officials also satisfactorily and met at least 60  percent of perceived PRP as a tool to supplement their their targets. If they fell below this threshold, own salaries and to help managers manage autonomy would be withdrawn until their per- more effectively. The success of this scheme formance improved. This is a good example can be gauged by the fact that the majority of how results-based management, of which of government agencies (30 of 44) have opted PRP was an integral component, opened the for the individually differentiated forced rank- door for other human resource management ing scheme and none has opted for the purely reforms, with SEPLAG realizing that this group-based bonus scheme with equal individ- autonomy was necessary to make results-based ual payments for all staff within the group. management work. PRP in the state of Minas Gerais in Brazil In the Philippines there are signs that the was part of a package of reforms introduced PRP scheme is giving managers incentives to between 2003 and 2010 to institutionalize a take results-based management more seriously. results focus in the state administration. These The Philippines has had an agency perfor- reforms included administrative restructuring, mance framework called the Organizational 38 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E In Philippines, staff Performance Information Framework (OPIF) triggered improvements in the performance in place since 2007, but agencies have to date appraisal process (between 38 and 52 percent surveyed believed that viewed OPIF largely as a documentation exer- agreeing or strongly agreeing). Taken together, PRP had motivated cise with almost no budgetary implications. the evidence suggests that PRP has inculcated Group-based PRP, which uses OPIF indica- management improvements that can over management to tors, has stimulated efforts to improve OPIF time help create a performance culture in the increase its focus performance indicators and to create a system bureaucracy. for reviewing progress in achieving the targets, These positive findings from the survey may on target setting both within agencies and by the Department of explain why the performance bonus scheme is and monitoring Budget and Management. viewed generally favorably in the Philippines The World Bank survey of government despite the concerns about its impact on indi- and to engage staff officials in the Philippines asked questions to vidual morale discussed earlier. It suggests that in the process explore the hypothesis that PRP can improve the group bonus and the individual bonus in management practices in the three areas of the Philippines work in opposite directions: the goal setting and monitoring, better teamwork, individual bonus has a negative direct incen- and improvements in the individual perfor- tive effect on effort, while the group bonus has mance appraisal process. Staff perceptions a positive indirect effect through better man- reveal positive responses to each of these ques- agement. It is too early to tell what the overall tions that cut across individual performance effect on government performance will be. rankings (figure 17). Interviewees noted that In Malaysia, PRP has had a much more the performance bonus scheme has motivated problematic history, but the overall assessment management to increase its focus on target is that a culture of performance has gradually setting and monitoring and to engage staff come into being over and above the outcomes in the process. Staff across the performance of particular performance reforms and without spectrum strongly believed that manage- minimizing the vestiges of patronage that sur- ment was more focused on working with staff vive. In this analysis the enduring significance to serve the public interest (between 60 and of PRP and the other performance reforms is 78 percent agreeing or strongly agreeing) and that they have reinforced the message that the more diligent in goal setting and in monitor- government as an employer wants employees ing accomplishment against goals (between to raise their game. As in Minas Gerais, PRP 64 and 76 percent agreeing or strongly agree- in Malaysia has complemented results-based ing). They were similarly clear in their views management reforms that have received a fair that teamwork in achieving departmental per- amount of international attention. However, formance targets had improved as a result of the effect in Malaysia has been much weaker, PRP (between 69 and 79 percent agreeing or possibly because the use of the forced distri- strongly agreeing), a surprising finding since bution caused a severe reaction from the civil forced rankings might be expected to cre- service unions and may have further weakened ate harmful competition among staff. Staff managers’ already weak incentives to distin- believed, though less strongly, that PRP had guish between the performances of staff. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  39 Perceptions of the Effect of PRP on Goal Setting, Monitoring of Targets, Teamwork, and Figure Performance Appraisals in the Philippines 17 “Because of PRP, management is more focused on working with staff to serve the public’s interest” “As a result of PRP, management is more diligent in setting performance indicators and targets” Performance ranking Disagree or strongly disagree Performance ranking Disagree or strongly disagree of respondent Neither Agree or strongly agree of respondent Neither Agree or strongly agree Top performers Top performers Second best Second best category category Third best Third best category category 0 25 50 75 100 0 25 50 75 100 Percent of respondents Percent of respondents “As a result of PRP, staff in the unit are “Because of PRP, the performance appraisal working better together to achieve targets” process has significantly improved” Performance ranking Disagree or strongly disagree Performance ranking Disagree or strongly disagree of respondent Neither Agree or strongly agree of respondent Neither Agree or strongly agree Top performers Top performers Second best Second best category category Third best Third best category category 0 25 50 75 100 0 25 50 75 100 Percent of respondents Percent of respondents Source: World Bank survey. Assessing the Evidence from the Case Studies: Pay Differentiation In China the Internal Direct pay flexibility levers More anecdotal evidence from Brazil, Chile, The overall evidence from the case studies and Russia points to similar conclusions. In Revenue Service is that pay differentiation can be a success- Brazil the highest-paid careers, particularly can better compete ful way, within the available fiscal space, to subsidized careers that pay in a single pack- improve the quality of staff in some prioritized age with no allowances and therefore full par- in the market for public entities. Differentiation causes resent- ity between pay and pension, were reported the best talent ment among staff of nonprivileged entities, to have the highest application rates in merit and it is difficult to conclude whether these exams. In Chile differentiation to create “high coming out of local negative effects outweigh the positive effects pay” and “low pay” offices was a conscious universities thanks to in the high-pay agencies. Presumably, the decision of successive governments in order initial prioritization, to the extent that it was to recruit and retain the best-quality staff for its higher pay scale driven by technocratic rather than political priority functions such as revenue collection, reasons, indicates that the government valued finance and treasury, and audit. The Internal productivity improvements in these agencies Revenue Service, which has its own, higher pay as particularly important. scale, is a notable example. Interviewees noted In Indonesia the World Bank survey asked that this public sector agency can better com- some questions on staff perceptions of effort, pete in the market for the best talent coming engagement with the mission of the agency, out of local universities. The critical functions and the quality of new recruits. The general allowance and the high degree of variation view of respondents was that staff in BR agen- in pay based on individual contracts for the cies work harder than staff in non-BR agencies contrata staff have enabled the government to and had higher morale, and that BR agencies attract and retain highly skilled personnel in received higher-quality recruits (figure 18). Of key posts across all institutions. In Russia the the respondents from BR agencies, 63 percent evidence also suggests a positive, if much more agree or strongly agree that others contribute limited, effect on staff recruitment due to inef- more than expected, significantly higher than ficiencies in competitive recruitment proce- all other agencies. In addition, 68 percent of dures and human resource management that staff in BR agencies said that their agency was watered down the pay incentive effect. The a better or much better place to work than effect seems to be stronger on retention; staff private sector firms in similar areas of work, turnover in the Ministry of Finance, in which compared with 40  percent in non-BR agen- the salary level and compression ratio are cies. Only 11  percent of survey respondents among the highest in the federal government, in the BR agencies, compared with 27 percent is very low. of respondents in the non-BR agencies, either When weighing the positive effects of differ- disagreed or strongly disagreed with the state- entiation on recruitment and retention against ment that their agency is able to recruit high- possible negative effects due to pay inequities, quality staff. it is important to recognize that complexity 40 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  41 Perceptions of Effort, Morale, and Quality of Recruits between BR and Non-BR Agencies Figure in Indonesia 18 “Others of the same rank in the ministry/agency contribute more than is expected of them” “How is your ministry/agency compared with private sector firms in a similar area of work?” Percent of respondents Agree or Worse or much worse 80 strongly agree Neither Better or much better BR agencies 60 40 Non-BR agencies 20 0 25 50 75 100 0 BR agencies Non-BR agencies Percent of respondents “Graduates from elite universities consider “Your ministry/agency easily a career at your ministry/agency as best recruits high-quality staff” possible public sector option” Disagree or strongly disagree Disagree or strongly disagree Neither Agree or strongly agree Neither Agree or strongly agree BR agencies BR agencies Non-BR Non-BR agencies agencies 0 25 50 75 100 0 25 50 75 100 Percent of respondents Percent of respondents Source: World Bank survey. and lack of transparency in pay is the norm agencies. The survey revealed very strong per- in the countries reviewed in this study. The ceptions of unfairness within the agencies that plethora of allowances implies that salaries for do not receive the BR allowance, with less than similar jobs vary for a variety of individual-spe- 25 percent of respondents agreeing or strongly cific, agency-specific, and geographic reasons. agreeing that pay in their agencies is fair com- Therefore, a government’s decision to priori- pared with others doing the same job in other tize certain functions and pay them more to agencies (figure 19). Interestingly, a similar attract talent adds another layer to an already perception of unfairness exists in the non– highly inequitable system. This added inequity Ministry of Finance BR agencies even though may be only marginally more demotivating to they receive higher pay than the non-BR agen- staff in the less privileged agencies, who have cies, reflecting their lower pay than the Minis- after all been living with this perceived unfair- try of Finance. ness for much of their working lives. The survey respondents were also asked This point came out in the Indonesia sur- whether their pay was fair compared with vey. Greater individual and agency productiv- others doing the same job in their agency. ity in the BR agencies has indeed come at the Interestingly, pay is considered to be less fair expense of increased pay inequities between within the non-BR agencies than within the 42 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E The impact of BR agencies, despite the significant variation enable particularly high public sector wages for in BR allowances for staff in similar ranks certain privileged groups of civil servants may differentiation on within a BR agency (see figure 19). Surpris- attract a less desirable type of worker. A compari- productivity is highly ingly, higher-ranked staff in the BR agencies son of public and private sector salaries in Brazil were more likely to say that their pay was fair reveals that the public sector has become a wage contextual—too compared with others doing similar jobs in leader for certain careers, raising concern that much differentiation their agency, despite the fact that pay inequi- high public sector wages might attract recruits ties increase with seniority. These findings sug- to the public service who are more motivated by can create disruptive gest that the existing pay system was already money than by the desire for public service. Con- competition viewed as unfair prior to the introduction of siderable media coverage in Brazil has recently BR pay, and that therefore the additional ineq- focused on certain public employees being paid between agencies uity introduced by BR may not be more demo- at levels far beyond what is justifiable. tivating in the lower-paid agencies. Clearly, the impact of differentiation on Indirect pay flexibility levers productivity is highly contextual. Too much The hypothesis that pay differentiation induces differentiation can create disruptive competi- managers to improve the performance dia- tion between agencies, as recent experience logue with staff found some support in Indo- in Brazil suggests. Spurred on by perceived nesia but could not be examined effectively in inequities, each career group bargains fiercely the other cases. In Indonesia the BR process is for increased allowance provisions to raise its meant to go beyond compensation reforms to compensation—a highly contentious process, emphasize managing for results. Experts who as indicated by the numerous strikes, both at were interviewed noted the linkages between the national and state level. These resentments BR pay increases and organizational changes also find their way into the labor courts, where in the Ministry of Finance and the exter- public employees sue, often successfully, to nal audit agency (BPK). These pay increases achieve remunerative equity with others in the gained the buy-in of staff on restructuring and public service based on the principle of “equal also put pressure on the concerned minister to pay for equal work.” show results. In the words of one expert, BR Taken to an extreme, the incentive effects of pay helped “oil the wheels.” differentiation can also create a sorting prob- These expert views are confirmed by lem. The possibility that differentiation can responses from the staff survey (figure 20). Perceptions of Pay Inequity across Ministries and Agencies and within BR Ministries in Figure Indonesia 19 “Your pay is fair compared with staff doing similar jobs in other ministries/agencies” “Your pay is fair compared with staff doing similar jobs in your ministry/agency” Percent of respondents Agree or Percent of respondents Agree or 80 strongly agree 80 strongly agree 60 60 40 40 20 20 0 0 Ministry of Other Non-BR Ministry of Other Non-BR Finance BR agencies agencies Finance BR agencies agencies Source: World Bank survey. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  43 Figure Staff Perceptions of Management and Willingness to Accept Restructuring in Indonesia 20 “Management is more focused on serving the country” “In your ministry/agency, staff are willing to accept changes, such as restructuring” Percent of respondents Agree or Percent of respondents Agree or 100 strongly agree strongly agree 100 75 75 50 50 25 25 0 0 Ministry of Other Non-BR Ministry of Other Non-BR Finance BR agencies agencies Finance BR agencies agencies Source: World Bank survey. Staff in the BR agencies were more likely to respondents from BR agencies were more agree with the statement that management likely to agree with the statement that employ- has become more focused on serving the ees in their agencies were willing to accept country’s interests. Senior staff who were changes such as restructuring. On both of most exposed to management were also most these questions, however, the differences were consistent in their responses about manage- mostly from the Ministry of Finance, which ment commitment. Successful reform also has been undertaking BR for the longest requires employees to embrace change, and period. Summary of Empirical Analysis In theory, PRP can Table 7 summarizes the evidence from the countries. Similarly, differentiation could in empirical analysis. Following the approach of theory disrupt public sector–wide wage nego- reduce wage bill the literature review, the overall evidence is tiations, thereby also limiting pressures for pressures because it recorded as “positive” if there are clear percep- general wage increases. tions of performance improvements among The case studies showed that in Thailand, provides the employer the staff interviewed; “negative” if the evidence PRP added 3 percent to the wage bill with little the tactical option of indicates no effect or negative effect of pay indication of any commensurate improvement flexibility; and “neutral” if there were opposite in performance, implying that it could have proposing that pay countervailing effects. The nature of the job been a politically easier way to justify a wage increases be provided has a significant bearing on the impact that increase. A similar motivation was in play in Bra- PRP has on performance through direct incen- zil at the federal level. By contrast, in Chile, PRP only as enhanced tive effects but not in how other aspects of pay was initially a response to pressure to increase performance bonuses flexibility work. public sector pay, which the government tried There is potential for the effective use of to manage by linking it to performance. The PRP in craft jobs through both individual- and envelope for public sector pay has not decreased group-based bonus schemes linked to out- since PRP started, but pressure for across-the- puts and outcomes. However, for coping jobs board pay increases other than those linked the evidence suggests that only group-based to inflation has receded since its introduction. bonuses are effective even if they require a Similarly, differentiation may have limited the forced distribution of unit-level performance power of unions in Chile, but it fueled competi- ratings with their accompanying rivalries. In tion for pay increases between the unions repre- either case, financial incentives must be signifi- senting the various careers in Brazil. cant (at least a month’s salary is a useful rule of thumb), and the measures should try to cap- The degree of delegated authority ture longer-term and sustained performance The case studies provided insufficient evidence improvements to counter gaming. to assess the role of organizational autonomy, or delegation, on pay flexibility. The only Fiscal impact examples of autonomy are from Brazil and The case studies provided conflicting evidence Chile. In Chile this delegation was in the on the affordability of pay flexibility. In theory, form of managerial authority to set the con- PRP can reduce wage bill pressures because it tractual terms of most civil servants who are provides the employer the tactical option of designated contrata staff and in the discretion proposing that pay increases be provided only to allocate the critical functions allowance. as enhanced performance bonuses—which The head of an agency can decide how much are less costly since they are normally not pen- to pay and in which category to hire contrata sionable (Marsden and French 1998). This was employees by arbitrarily defining their grade clearly a motivation for PRP in a number of on the pay scale and thereafter modifying this 44 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  45 Table Summary of Evidence about the Impact of Pay Flexibility on Civil Service Performance 7 PRP Hypothesis 1. Direct pay flexibility levers Hypothesis 1a. PRP can directly affect individual effort Craft job Coping job Literature review Reasonable evidence concerning teaching, health care, No relevant evidence in relation to coping jobs and other craft jobs (within and outside OECD settings) even though gaming is a persistent phenomenon Case studies: Chile Positive for revenue authority Negative since all staff got the bonus and therefore there was no incentive effect Brazil (federal government) Positive for revenue authority Negative since all staff got the bonus Brazil (Minas Gerais) Positive for revenue authority and police Weakly positive on effort: Weeded out the outliers Malaysia Weakly positive for revenue authority Negative since forced distribution hurt teamwork and made for contentious management-staff relations Thailand Not examined in the case study Negative since all staff got the bonus; the bonus was also very small Philippines Weakly positive for revenue authority Negative: Staff are broadly supportive of the scheme but there was no reported impact on effort Indonesia Positive for revenue authority Weakly positive on effort: Reduced absenteeism Korea, Rep. Not examined in the case study Positive: Broad support in the central government for the individual bonus scheme. Most agencies chose individual ranking and none chose a group-based equal bonus Hypothesis 1b. PRP can have a direct effect in improving the recruitment and retention of better-quality staff Literature review Sorting effect noted in OECD settings Case studies: Brazil (Minas Gerais) Some sorting effect Hypothesis 2. Indirect pay flexibility levers (PRP can act indirectly by providing incentives for improved management) Case studies: Chile Weakly positive: Gradually improved results-based management; encouraged more delegation of human resource management authority Brazil (Federal) Weakly positive Brazil (Minas Gerais) Positive: Complemented performance budgeting reforms; encouraged delegation Malaysia Neutral: Complemented results-based management but prompted a backlash from the unions Thailand No evidence Philippines Positive: Improved goal setting and monitoring; complemented and strengthened results-based management Indonesia No evidence Korea, Rep. Positive: Has complemented results-based management and performance budgeting Differentiation Hypothesis 3. Direct pay flexibility levers (Differentiation can have a direct effect in improving the recruitment and retention of better-quality staff) Case studies: Indonesia Positive: Improved recruitment and effort in high-pay agencies; unclear if pay inequity reduced effort in low-pay agencies Chile Positive: Improved recruitment and effort in high-pay agencies; unclear if pay inequity reduced effort in low-pay agencies Brazil Neutral: While recruitment and retention improved in high-pay agencies, it also set up harmful competition between agencies Russian Federation Weakly positive: Improved retention and led to relatively low vacancy rates in high-pay ministries Hypothesis 4. Indirect pay flexibility levers (Differentiation can act indirectly by providing incentives for improved management) Case studies: Indonesia Positive: Raised stakes for management and opened up space for reforms in Ministry of Finance Chile No evidence Brazil No evidence Russian Federation No evidence grade as needed. The only limitations are that which can be up to 100 percent of gross pay, the grade of the contrata employee cannot be is given to contracted or permanent staff for higher than the highest permanent staff grade critical functions within an agency at the dis- defined by law for each organization and that cretion of the minister (again, within a budget personnel expenditures remain within Budget envelope). The allowance can be discontinued Office limits. The critical functions allowance, at any time by the agency head or minister. 46 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E PRP was also a vehicle for delegation in agreements that granted autonomy to those Chile. The introduction of PRP at the indi- entities agreeing to contractual account- vidual and unit level was always implemented ability in developing remuneration criteria through delegation to agency heads or minis- linked to performance targets. In fact, the ters. Its management has also gradually moved results-based management reforms and PRP from a centralized approach to partial del- triggered delegation of human resource man- egation to the ministry in charge of the policy agement authority, since agencies were able to area, but has not yet been delegated to the convince the central finance and personnel agency level. authorities that autonomy was necessary for There was fair degree of administrative them to be able to deliver on their contracted delegation in Minas Gerais through results outputs. Conclusions and Policy Recommendations How did the hypotheses hold up? from the literature review and from Brazil, The review of the empirical evidence exhibits Chile, Indonesia, and Malaysia is reasonably the diversity of experiences with pay flexibility, conclusive, though the effect depends on the which means that the answer to the central ques- nature of the PRP scheme. The main enablers tion of this study, “Can pay flexibility improve are a combination of an individual- and the performance of public bureaucracies?” is group-based option and a significant finan- yes, under some conditions and in certain con- cial incentive—at least a month’s salary for the texts. The earlier figure is repeated below to concerned staff as a rule of thumb. The find- show how PRP and differentiation are hypoth- ings suggest that the risks of too large a finan- esized to impact directly and indirectly on indi- cial incentive and problems of free-riding in vidual productivity/organizational citizenship group-based bonus schemes may be overstated. and better staff through sorting (figure 21). Some perverse incentives and unforeseen With Hypothesis 1a (PRP can directly consequences are inevitable in PRP schemes. affect individual effort), there are reasonable Since people respond to incentives, at a mini- grounds for concluding that in low-income mum some effort distribution will take place countries PRP can have a direct incentive effect toward tasks that are measured and away from on improving effort for teaching, health care, tasks that are not measured. Overt manipula- revenue collection, and other craft jobs that tion of targets, and other gaming behavior, is have more measurable outputs. The evidence also likely. The extent to which these problems Figure The Link between Pay Flexibility and Performance 21 Direct pay Indirect pay Functional exibility policy exibility policy improvements levers levers Greater effort by managers Greater effort by H2 staff (higher H1a engagement and PRP organizational citizenship) H1b H4 Better-quality staff Differentiation (sorting) H3 47 48 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E The extent to can be mitigated depends on the validation With Hypothesis 1b (that PRP can have a and review processes in place by managers and direct effect in improving the recruitment and which manipulation central oversight agencies, and on learning retention of better-quality staff), the literature and gaming can be and improvements in performance indicators review suggested some modest effect in OECD so as to encourage sustained, long-term perfor- settings and some support from the Brazil mitigated depends mance gains. In Minas Gerais the evolution of (Minas Gerais) case study. on the validation and the scheme in the revenue authority and the Hypothesis 2 (that PRP can act indirectly by innovative design of the general PRP scheme, providing incentives for greater effort by man- review processes in in which the size of the incentive increases with agers) was supported by evidence from Korea, place by managers the number of good performance evaluations, Minas Gerais, and the Philippines. The cases are instructive. In Korea senior civil servants show that PRP encourages frontline manag- and central are required to identify several key perfor- ers to pay attention to the mission and goals oversight agencies mance targets rather than one or two perfor- of their organization and to bring staff to work mance goals, and the general performance together to achieve these goals, and it can com- appraisal system for middle- and lower-level plement other reforms like performance-based officials reviews performance goals in three budgeting and results-based management. aspects—task completeness, achievement time- Managers can of course use other vehicles, liness, and job difficulties—in order to mini- including just becoming more effective man- mize gaming or evaluation errors. In Chile the agers, but the evidence suggests that monetary performance targets have evolved over time to incentives encourage managers and staff to be more robust and to be more result related have more communication around the perfor- as opposed to process related. mance agenda and that these conversations For core policy and administration jobs, the help renegotiate the “effort bargain” (see box direct effects of PRP are weak and limited to 2). identifying egregiously underperforming staff. These management changes are highly con- The lack of objective and defendable output ditional on context, as the contrast between measures discourages frontline managers from the experiences of Malaysia and Minas Gerais distinguishing between staff using subjective shows, and they are also gradual. In Malaysia assessments. As a result, PRP becomes a salary the imposition of an individual forced rank- supplement with no incentive effect. The ques- ing for PRP coupled with powerful unions and tion then becomes a different one of whether contentious staff-management relations lim- this form of salary supplement is better than ited the potential of these managerial improve- others. ments. In Minas Gerais management and Implementing PRP in coping jobs through staff developed a much more cooperative and a forced differentiation of individual per- accommodative relationship that was helped formance ratings is risky and can harm staff by a more nuanced PRP scheme that encour- morale. Korea has done reasonably well under aged longer-term performance improvements. the system, probably because of relatively high In the Philippines staff clearly perceive that trust levels and the general high functionality PRP has improved management, but there is as of the Korean bureaucracy. The experience of yet no discernible effect of this improved man- Malaysia highlights the risks of this approach agement on individual productivity and orga- and the considerable turmoil in a bureau- nizational citizenship. cracy that PRP in coping jobs can engender, Hypothesis 3 (that differentiation can and it may be more representative of develop- directly improve staff quality) found support. ing country contexts. Similarly, evidence from Higher average pay and greater pay decom- the Philippines points to the overall negative pression, so that pay increases significantly impact on staff motivation of forced distribu- over the course of a career, have a clear effect tion because of the lack of credibility of super- on staff quality in the privileged agencies visor performance assessments and concerns through improved recruitment and retention. of favoritism. The tradeoff is increased inequity in pay across PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  49 the public sector. Inequity and lack of trans- pensionable, depending on the particular PRP may be a parency is, however, the norm in many devel- career. The performance element is therefore oping and emerging market countries and so of secondary importance, and it may be a way politically feasible, the performance improvements for high-pay of justifying salary increases under the guise of and fiscally less costly, groups of staff may not necessarily be offset by performance improvements. A similar motiva- way of increasing commensurate performance declines in low- tion existed in Chile, where steady economic pay groups. growth created pressures to raise real public public sector salaries Hypothesis 4 (that differentiation can act sector wages, which are more easily sold to the indirectly by providing incentives for greater public as PRP. effort by managers) found some support from Even when the government motivation is the case studies. The evidence from Indonesia performance rather than cost savings, politi- in particular suggested that management in cal resistance from public sector unions has the Ministry of Finance was under the national often proved fatal to PRP. This was most obvi- spotlight to make performance improvements ously the case in Malaysia, but has also been because of its higher pay, which was widely jus- an important factor in Brazil and Chile. The tified and publicized as being linked to per- strength of public sector unions is therefore an formance improvements even though there important variable for PRP implementations, was no such explicit linkage in design. This with countries with weaker unions, such as pay change then opened up space for human Korea (unions were banned under the authori- resource improvements, and this greater man- tarian regime until the late 1990s) having more agement focus was complemented by greater success. Interestingly, even in the presence acceptance by staff of organizational changes of strong unions governments can have some such as restructuring. bargaining space. Staff interviewed in Minas Gerais reported a relatively harmonious rela- Implications for pay policy in tionship with the relevant unions, indicating developing countries that part of the success of public sector reforms What are the broader lessons for pay policy, was also linked to a different attitude toward particularly in the more challenging low- the role of unions in government. Staff attrib- income country contexts in Africa, South Asia, uted cordial relations in part to the proactive and elsewhere? role taken, in particular, by successive secretar- While this study has focused on the impact ies of finance, who customarily met personally of pay flexibility, set out as a series of hypoth- at least twice a year with union representatives. eses for testing, an equally significant question Taking into account both political factors concerns political feasibility and motivation . The and technical considerations, six main mes- case studies have shown that improving per- sages are evident. First, pay flexibility can improve formance may not be the primary motivation performance . Given the dearth of success stories for introducing PRP. Instead, PRP may be a in public administration reform in developing politically feasible, and fiscally less costly, way countries, and the generally negative view of of increasing public sector salaries. In Brazil PRP in particular that has prevailed in public at the federal level, for example, the phrase administration academic and policy circles, “variable pay” is often used interchangeably this is a powerful finding of the study. with the term “performance pay,” pointing There is a long tradition of skepticism about to the fiscal motivations behind its introduc- the introduction of seemingly OECD-like tion. With the aging profile of civil servants reforms in developing countries, and pay flex- (the average age has increased from 43 to 46 ibility could be placed in that category. Some over the past decade) and increasing pension technical criticisms of performance pay go liabilities, the government has taken a harder back to their use for teachers in British schools line with the employee unions on basic sal- in the nineteenth century (Gratz 2009). Argu- ary increases and has instead increased pay ments are made that performance measures by using allowances that may or may not be can induce tunnel vision, myopia, and measure 50 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Pay flexibility can fixation (Propper and Wilson 2003). Pink examples from Korea and Minas Gerais show (2009) has developed the critique of monetary that it is the combination of reforms—PRP, complement and and other extrinsic incentives into a broader results-based management, and performance- provide the behavioral theory, hypothesizing that they are both coun- based ­budgeting—that is the key to improved terproductive, since they frequently under- performance. impetus for budgeting mine intrinsic incentives, and unnecessary, This essential ingredient of management and other performance since intrinsic incentives can be harnessed has several implications for the technical and used to maximize individual productiv- design of the PRP scheme and for the sequence management reforms ity. These arguments are not without merit, of of reforms. PRP schemes for coping jobs should course, and there is little doubt that intrinsic encourage such management changes. A large motivations are of particular significance in group-based bonus PRP scheme is preferable, public service (Banuri and Keefer 2013). How- despite the difficulties in establishing unit- ever, the findings of this study suggest that level performance targets, because it bypasses simplistic universal arguments against pay the problem of distinguishing between indi- flexibility are not justified empirically and par- viduals’ performance and puts the spotlight ticularly not when considering self-selection on management improvements as the key link- and sorting, since an explicit system of perfor- ing PRP to better performance. The perfor- mance pay can attract extrinsically motivated mance measures in such a scheme are likely applicants to the civil service. to be process oriented, and to the extent that Pay flexibility reforms are not a silver bullet, these policy jobs also entail providing services as no public sector reforms can be, and involve to government service delivery units (releases tradeoffs and risks. Poorly designed PRP of funds, regulatory oversight, personnel over- schemes can cause more harm than good, and sight, and so forth), these process indicators there is the potential for harmful task realloca- can be supplemented with customer satisfac- tion and gaming, given that even in craft jobs tion measures. The risks of gaming, however, no performance measures can capture all the will be higher, and the scheme will therefore dimensions of each job. The evidence reviewed, need to be more vigilantly managed. The however, suggests that these tradeoffs can be incentive has to be large (at least equivalent managed with the necessary sophistication if to a month’s salary) to be sufficient to induce risks of perverse behavior are correctly identi- managers to change their practices. fied. The extent of the appropriate pay flex- Introducing pay flexibility across the board ibility measures very much depends on local in a public sector where management is highly context. At a minimum, the study suggests that dysfunctional is inadvisable. An approach these schemes can be implemented in craft could be to introduce it asymmetrically where jobs with the necessary monitoring regime in there is some basic level of managerial com- place to detect and respond to gaming behav- petence—for example, using a PRP scheme ior. Despite much academic and professional having the design features identified above (a skepticism, there is therefore every reason to group-based bonus only, a financial incentive keep pay flexibility in the reform toolkit. equivalent to at least a month’s salary, and an Second, pay flexibility works most strikingly in incentive designed to encourage longer-term, changing managerial behaviors. It can focus gov- sustained performance). The focus could be ernment attention on management improve- agencies where there has been at least a seri- ments under the broad rubric of “the effort ous attempt to set goals and to improve human bargain”—vision, strategy, agency strategic resource management practices that flexible objectives and key performance indicators, pay policies can build on and strengthen. and better dialogue with staff to achieve In terms of sequencing, it is advisable to these through teamwork and task alloca- launch results-based management (or other tion. It can also complement and provide the variants of performance management) first behavioral impetus for budgeting and other before introducing PRP. In all the case coun- performance management reforms. The tries with positive experiences, PRP was PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  51 introduced after, and energized, an earlier sustained good performance over the employ- The prospect of set of reforms to improve goal setting and ee’s career. Similarly, in Indonesia pay differen- monitoring and managing for results. This tiation for BR agencies introduced both higher continuing salary sequence of reforms enabled the PRP scheme average pay for all staff but particularly higher enhancements over to build on a set of managerial resources—per- pay for senior staff. time with a strong formance indicators, progress review mecha- Fifth, the strategy and implementation of pay nisms, and so forth—that had been established flexibility reforms have to take into account the component of peer when there was no “money on the table” that extent of fragmentation and complexity of the exist- recognition remains could have corrupted the process. Often these ing public sector pay structure in the country. While, management reforms had limited impact pre- as noted above, the evidence suggests that pay important to staff cisely because there were no stakes attached to simplification is not necessary, or advisable, as them, and PRP was therefore needed to induce a prior step before introducing pay flexibility, managers to start taking their management the extent of “messiness” of the pay regime responsibilities more seriously. But introducing has implications for the pay flexibility strategy. PRP in an environment with no such history of The strategy should be different in relatively reforms risks having the money incentive over- neat systems (such as the Philippines) where whelm everything else. the pay structure is fairly simple and uniform Third, the path to improved public sector per- across the core administration, compared with formance does not necessarily need to go through a messy systems where pay varies for a whole host stage of “whole of government” pay rationalization of idiosyncratic reasons and where central or pay simplification. Many of the cases analyzed fiscal control and management coherence is in this study—for example, Brazil, Chile, and compromised. Indonesia—were symptomatic of the general In simpler systems there is less risk in more developing country phenomena of high vari- ambitious, across-the-board pay f lexibility ance in compensation for similar jobs based reforms, if there is an explicit recognition of on a variety of employer-related and personal possible perverse behavior and unintended factors. Pay inequity and lack of transpar- consequences, and experimentation and ency in compensation were ubiquitous. This learning-as-you-go are built in to the reforms. study suggests that there is no reason to seek Ideally, even in these systems pay flexibility to move from these haphazard or asymmetric would be introduced first for craft jobs, and pay structures to homogenous single pay spine within coping jobs first in organizations that arrangements before contemplating differen- have already made investments in improving tiation or PRP to improve performance. Per- management. This restriction is necessary to formance improvements are possible through limit the administrative burden of the neces- the “purposeful complexity” of pay flexibility sary validation and monitoring systems, and even when layered on top of a complex pay the system can then be gradually expanded as regime. This is an encouraging finding given the sophistication of this monitoring regime the technical and political challenges of com- increases. prehensive pay rationalization and the poor In complex systems there is the risk that pay track record of such reforms. flexibility degenerates into yet another element Fourth, in both “messy” and simple pay of the messy pay regime with few productivity policy contexts, flexible pay policy can work with gains and a further weakening of central fis- rather than instead of long-term career incentives. cal control and management coherence. This The prospect of continuing salary enhance- risk can be mitigated by limiting pay flex- ments over time with a strong component of ibility to a select few high-priority organiza- peer recognition remains important to staff. tional “islands,” chosen either because they The individual PRP scheme in Minas Gerais are the highest priority or because they are is a good example of this complementar- managed relatively well. These are the staff ity between short- and long-term incentives, whose productivity improvements are consid- since the size of the incentive increases with ered to be the most important for government 52 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E A government’s performance. Pay reforms are notoriously level of trust has been noted as an important sticky and hard to reverse, and limiting the ter- factor on numerous occasions in this report) capacity to collect rain is necessary to prevent the type of uncon- are all key issues that require more investiga- and validate data on trolled lobbying by all groups that has beset tion. The study could not effectively explore Brazil. These agencies should be required to how organizational autonomy interacts with performance and achieve explicit standards in management pay flexibility. A government’s capacity to col- to coordinate pay practices (on recruitment, goal setting and lect and validate data on performance and monitoring, and staff performance evalua- to coordinate pay flexibility across the public flexibility across tions) to qualify for pay flexibility. sector is also presumably an important factor the public sector Sixth, many questions remain, and much more determining the success of these schemes. The research is needed . As noted earlier, the breadth need for a robust monitoring regime raises is an important of contextual coverage of this study and the the question of cost effectiveness and whether factor determining range of evidence used have come at the the added bureaucratic burden associated the success of expense of empirical depth, in part because with performance incentives can pay for itself analyses of the core public administration do though higher productivity. these schemes not easily lend themselves to rigorous impact All these questions that must be analyzed in evaluations. The study cannot make causal specific country contexts. In the past, pay flex- claims, and the use of staff perceptions creates ibility for the core public administration has potential biases because respondents may have often been ruled out a priori in World Bank an incentive to preserve pay flexibility and so and other donor advice. This study instead may be more positive in their responses. Every calls for a different approach, arguing that effort should therefore be made at the country these reforms be assessed based on the design level to track improvements in deliverables to features of the scheme, the jobs to which it is ensure that the application of pay flexibility is being applied, potential tradeoffs in terms of tested against services or outputs that matter. increased pay inequity, the unintended con- There is considerable heterogeneity in the sequences and perverse gaming behavior that impact of pay flexibility reforms based on con- these are likely to generate and whether they textual factors that go beyond the two—type of can be managed, and other contextual factors. public sector job and design features of the pay We hope that this more nuanced approach flexibility scheme—that this study has looked to pay reform will assist policy makers and at. How pay flexibility interacts with existing development partners in the critical agenda of formal and informal rules and culture (the improving public sector performance. Notes 1. Rafferty and others (2005) review the liter- 5. The degree of delegation varies widely ature on organizational commitment and across the OECD countries, with delega- citizenship, finding empirical associations tion largely, but not entirely, correlated between commitment and increased job with the broad introduction of “new public satisfaction (Vandenberg and Lance 1992); management” ideas (OECD 2011, 127). increased job performance (Mathieu and 6. Inevitably, the classification of studies is Zajac 1990); improved sales (Barber, Hay- somewhat subjective. Studies were rated day, and Bevan 1999); lower employee as positive if they showed general evidence turnover (Cohen 1991); less intention to on the basic functionality of incentive leave (Cohen 1993; Balfour and Wechsler schemes, even if additional results qual- 1996); lower absenteeism (Cohen 1993; ify the effect—for example, studies on Barber, Hayday, and Bevan 1999). Oster- crowding-out of intrinsic motivation gen- loh and Frost 2002 note that citizenship erally still find positive effects of explicit entails employee behaviors that maintain incentives. the organization’s social system and are 7. Interpreted as purely theoretical papers important to its smooth running (Hous- or studies with a weak research design (for ton 2009). example, selection on the dependent vari- 2. The literature uses a variety of terms for able only, no meaningful variation, and no such financial incentives: performance explicit consideration of counterfactuals). pay, performance-based pay, performance- 8. Studies that mostly describe reforms based incentives, and pay-for-performance. implemented in a small number of cases 3. The expert interviews and perception sur- without comparing them with cases with- veys ask questions on recruitment, reten- out performance pay. tion, career development, performance 9. Studies based on a small number of cases management, pay levels and dispersion, but having at least an implicit consider- and performance incentives. The survey ation of a counterfactual and with some in Indonesia was conducted in 14 central minimal data analysis. government ministries and agencies and 10. Studies with an explicit counterfactual covered 4,000 staff. The survey in the analysis, using a representative sample Philippines covered 7 central government of cases with and without treatment and departments and agencies and approxi- often using statistical techniques to limit mately 2,500 staff. threats to causal inference. 4. Wilson had originally used this framework 11. We think a minimal level of internal and to classify organizations and not jobs, the external validity is necessary to draw reli- implicit assumption being that organiza- able conclusions from the evidence pre- tions were homogenous in the tasks that sented in a study, especially when policy they performed. recommendations are concerned. For that 53 54 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E reason we opted to classify studies as “high 13. Comparing only agencies with fingerprint quality” only if their analysis was based on machines or only agencies without them, quasi-experimental methods and the ana- respondents in agencies with BR status lyzed sample was somewhat representative are much less likely to report that at least of the theoretical population under study. 20  percent of their coworkers leave work 12. Deconcentrated units of federal executive early. authorities in the regions are referred to as territorial authorities. References Ahmad, S., and R. G. Schroeder. 2003. “The Discussion Paper No. 2062. Bonn, Germany: Impact of Human Resource Management Institute for the Study of Labor. Practices on Operational Performance: Rec- Banuri, S., and P. Keefer. 2013. Intrinsic Motiva- ognizing Country and Industry Differences.” tion, Effort and the Call to Public Service. Wash- Journal of Operations Management 21: 19–43. ington, DC: World Bank. Ahmed, Z., F. Shields, R. White, and J. Wil- Barber, L., S. Hayday, and S. Bevan. 1999. From bert. 2010. “Managerial Communication: People to Profits (IES Report 355). London: The Link between Frontline Leadership Institute for Employment Studies. and Organizational Performance.” Journal Basinga, P., P. J. Gertler, A. Binagwaho, A. L. B. of Organizational Culture, Communications & Soucat, J. R. Sturdy, and C. M. J. Vermeersch. Conflict 14 (1): 107–20. 2010. Paying Primary Health Care Centers for Andrade, N. H. S. 2011. A introduçao dos Performance in Rwanda. Washington, DC: especialistas em politicas públicas e gestão World Bank. governamental na administraçao pública Behrend, H. 1957. “The Effort Bargain.” Brasileira: Análise comparativa da carreira Industrial and Labor Relations Review 10 (4): nas esferas estadual e federal. IV Congresso 503–15. CONSAD de Gestão Pública, Brasília, DF. Bender, K. A., and R. F. Elliott. 2003. Decentral- Armstrong, M. 2010. Armstrong’s Essential ised Pay Setting: A Study of the Outcomes of Col- Human Resource Management Practice. Lon- lective Bargaining Reform in the Civil Service in don, Kogan Page. Australia, Sweden and the U.K. Ashgate, UK: Asch, B. J. 1990. Navy Recruiter Productivity and Farnham. the Freeman Plan. Santa Monica, CA: RAND Bertelli, A. M. 2006. “Motivation Crowding and Corporation. the Federal Civil Servant: Evidence from the Atkinson, A., S. Burgess, B. Croxson, P. Gregg, U.S. Internal Revenue Service.” International C. Propper, H. Slater, and D. Wilson. 2004. Public Management Journal 9 (23): 3–23. Evaluating the Impact of Performance-Related Bruns, B., D. Filmer, and H. A. Patrinos. 2011. Pay for Teachers in England. Working Paper Making Schools Work: New Evidence on Account- No.04/113. Bristol, UK: Centre for Market ability Reforms. Washington, DC: World Bank. and Public Organisation. Burgess, S., and P. Metcalfe. 1999. Incentives in Balfour, D., and B. Wechsler. 1996. “Organ- Organisations: A Selective Overview of the Litera- isational Commitment: Antecedents and ture with Application to the Public Sector. Bristol, Outcomes in Public Organisations.” Pub- UK: University of Bristol, CMPO and CEPR. lic Productivity and Management Review 29: Burgess, S., C. Propper, M. Ratto, S. Scholder, 256–77. and E. Tominey. 2010. “Smarter Task Assign- Bandiera, O., I. Barankay, and I. Rasul. 2006. ment or Greater Effort: What Makes the Dif- Incentives for Managers and Inequality among ference on Team Performance?” Economic Workers: Evidence from a Firm-Level Experiment. Journal 120 (547): 968–89. 55 56 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Cadsby, C. B., F. Song, and F. Tapon. 2007. Dickens, W. T., and L. F. Katz. 1987. Interindus- “Sorting and Incentive Effects of Pay-for- try Wage Differences and Industry Characteristics. Performance: An Experimental Investiga- NBER Working Paper 2014. Washington, tion.” Academy of Management Journal 50 (2): DC: National Bureau of Economic Research. 387–405. Dowling, B., and R. Richardson. 1997. “Evaluat- Christensen, D. B., N. Neil, W. E. Fasset, D. ing Performance-Related Pay for Managers H. Smith, and G. A. S. Holmes. 2000. “Fre- in the National Health Service.” International quency and Characteristics of Cognitive Ser- Journal of Human Resource Management 8 (3): vices Provided in Response to a Financial 348–66. Incentive.” Journal of the American Pharmaceu- Duflo, E., R. Hanna, and S. P. Ryany. 2010. tical Association 40 (5): 609–17. Incentives Work: Getting Teachers to Come to Clotfelter, C., R. A. Diaz, H. Ladd, and J. Vigdor. School. Cambridge, MA: Department of Eco- 2004. “Do School Accountability Systems nomics and J-PAL, MIT, and Kennedy School Make It More Difficult for Low-Performing of Government. Schools to Attract and Retain High-Quality Eldridge, C., and N. Palmer. 2009. “Perfor- Teachers?” Journal of Policy Analysis and Man- mance-Based Payment: Some Reflections on agement 23 (2): 251–71. the Discourse, Evidence and Unanswered Clotfelter, C., E. Glennie, H. Ladd, and J. Vig- Questions.” Health Policy and Planning 1 (7): dor. 2008. “Would Higher Salaries Keep 160–66. Teachers in High-Poverty Schools? Evidence Fairbrother, G., K. L. Hanson, S. Friedman, and from a Policy Intervention in North Caro- G. C. Butts. 1999. “The Impact of Physician lina.” Journal of Public Economics 92: 1352–70. Bonuses, Enhanced Fees, and Feedback on Coase, R. H. 1937. “The Nature of the Firm.” Childhood Immunization Coverage Rates.” Economica 4 (16): 386–405. American Journal of Public Health 89 (2): Cohen, A. 1991. “Career Stage as a Moderator 171–75. of the Relationship between Organisational Fairbrother, G., M. J. Siegel, S. Friedman, P. Commitment and Its Outcomes: A Meta- D. Kory, and G. C. Butts. 2001. “Impact of Analysis.” Journal of Occupational Psychology Financial Incentives on Immunization Rates 64: 253–68. in the Inner City: Results of a Randomized ———. 1993. “Age and Tenure in Relation to Controlled Trial.” Ambulatory Pediatrics 1 (4): Organisational Commitment: A Meta-Anal- 206–12. ysis.” Basic and Applied Social Psychology 14: Figlio, D. N., and L. W. Kenny. 2007. “Individ- 143–59. ual Teacher Incentives and Student Perfor- Courty, P., and G. Marschke. 2004. “An Empiri- mance,” Journal of Public Economics 91: 901–14. cal Investigation of Gaming Responses to Figlio, D. N., and J. Winicki. 2005. “Food for Explicit Performance Incentives.” Journal of Thought: The Effects of School Account- Labor Economics 22 (1): 23–56. ability Plans on School Nutrition.” Journal of de Assis, L. O. M. 2012. Efeitos organizacionais Public Economics 89: 381–94. da implantaçao de metas e remuneração variável Frey, B. S., and M. Osterloh. 1999. Pay for Per- por desempenho: O caso da segurança pública em formance: Immer Empfehlenswert? Münster, Minas Gerais. São Paulo: Fundação Getúlio Germany: Zeitschrift für Führung und Vargas. Organisation. Dewatripont, M., I. Jewitt, and J. Tirole. 1999a. Fryer, R. G. 2011. Teacher Incentives and Student “The Economics of Career Concerns, Part 1: Achievement: Evidence from New York City Public Comparing Information Structures.” Review Schools. NBER Working Paper 16850. Wash- of Economic Studies 66: 183–98. ington, DC: National Bureau for Economic ———. 1999b. “The Economics of Career Con- Research. cerns, Part 2: Application to Missions and Glewwe, P., N. Ilias, and M. Kremer. 2010. Accountability of Government Agencies.” “Teacher Incentives.” American Economic Jour- Review of Economic Studies 66: 199–217. nal: Applied Economics 2 (3): 205–27. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  57 Goodman, S., and L. Turner. 2010. “Teacher Incentive Contracts, Asset Ownership, and Incentive Pay and Educational Outcomes: Job Design.” Journal of Law, Economics & Evidence from the NYC Bonus Program.” Organization 7: 24–52. Working paper presented at the PEPG Houston, D. J. 2009. “Motivating Knights or Conference, “Merit Pay: Will It Work? Is It Knaves? Moving Beyond Performance- Politically Viable?” Cambridge, MA, Harvard Related Pay for the Public Sector.” Public Kennedy School, June 3–4. Administration Review 69 (1): 43–56. Grady, K., J. Lemkau, L. N., and C. Caddell. Hutchinson, S., and J. Purcell. 2003. Bring- 1997. “Enhancing Mammography Referral ing Policies to Life: The Vital Role of Front in Primary Care.” Preventive Medicine (26): Line Managers in People Management. Lon- 791–800. don: Chartered Institute of Personnel and Gratz, D. B. 2009. The Peril and Promise of Per- Development. formance Pay: Making Education Compensation ———. 2007. Learning and the Line: The Role of Work. Lanham, MD: Rowman & Littlefield. Line Managers in Training, Learning and Devel- Groshen, E. L. 1991. “Sources of Intra-Industry opment. London: Chartered Institute of Per- Wage Dispersion: How Much Do Employers sonnel and Development. Matter?” Quarterly Journal of Economics 106 Independent Evaluation Group. 2008. Public (3): 869–84. Sector Reform: What Works and Why? Washing- Han, S. 2010. “Government Officials’ Responses ton, DC: World Bank. to the Performance-Related Pay System: The Jacob, B. A. 2005. “Accountability, Incentives Application of the Grounded Theory.” Korean and Behavior: The Impact of High-Stakes Public Administration Review 44 (4): 29–58. Testing in the Chicago Public Schools.” Jour- Hasnain, Z., N. Manning, and J. H. Pierskalla. nal of Public Economics 89: 761–96. 2012. Performance-Related Pay in the Public Sec- Jacob, B. A., and S. D. Levitt. 2003. “Rotten tor: A Review of Theory and Evidence. Policy Apples: An Investigation of the Prevalence Research Working Paper 6043. Washington, and Predictors of Teacher Cheating.” Quar- DC: World Bank. terly Journal of Economics 118 (3): 843–77. Heckman, J., C. Heinrich, and J. Smith. 1997. Kahn, C. M., E. C. De Silva, and J. P. Ziliak. “Assessing the Performance of Performance 2001. “Performance-Based Wages in Tax Standards in Public Bureaucracies.” Ameri- Collection: The Brazilian Tax Collection can Economic Review 87 (2): 389–95. Reform and Its Effects.” Economic Journal Hillman, A., K. Ripley, N. Goldfarb, I. Nuamah, 111: 188–205. J. Weiner, and E. Lusk. 1998. “Physician Kellough, E. J., and H. Lu. 1993. “The Paradox Financial Incentives and Feedback: Failure of Merit Pay in the Public Sector: Persistence to Increase Cancer Screening in Medicaid of a Problematic Procedure.” Review of Public Managed Care.” American Journal of Public Personnel Administration 13: 45–64. Health 88 (11): 1699–701. Ketelaar, A., N. Manning, and E. Turkisch. ———. 1999. “The Use of Physician Financial 2007. Performance-Based Arrangements for Incentives and Feedback to Improve Pedi- Senior Civil Servants: OECD and Other Country atric Preventive Care in Medicaid Managed Experiences. OECD Working Papers on Public Care.” Pediatrics (104): 931–35. Governance 2007/5. Paris: OECD. Hochberg, Y. V., and L. Lindsey. 2010. “Incen- Knies, E., and P. Leisink. 2013. “Leadership tives, Targeting and Firm Performance: An Behavior in Public Organizations: A Study of Analysis of Non-Executive Stock Options.” Supervisory Support by Police and Medical Review of Financial Studies 23 (11): 4148–86. Center Middle Managers.” Review of Public Holmstrom, B. 1982. “Managerial Incentive Personnel Administration (online), November. Problems: A Dynamic Perspective.” Review of Koch, R. 1996. “Towards and Entrepreneurial Economic Studies 1: 169–82. Paradigm in Public Management: Interna- Holmstrom, B., and P. Milgrom. 1991. tional Trends.” In Entrepreneurial Manage- “Multitask Principal-Agent Analyses: ment in the Public Sector, edited by J. Wanna, J. 58 PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E Forster, and P. Graham, 33–40. Melbourne: London: London School of Economics, Cen- Macmillan Education. tre for Economic Performance. Kouides, R., N. Bennett, B. Lewis, J. Cap- Marsden, D., and S. French. 1998. What a Per- puccio, W. Barker, and M. LaForce. 1998. formance: Performance-Related Pay in the Public “Performance-Based Physician Reimburse- Services. London: London School of Eco- ment and Influenza Rates in the Elderly.” nomics, Centre for Economic Performance. American Journal of Preventive Medicine 14 Mathieu, J., and D. Zajac. 1990. “A Review and (2): 89–95. Meta-Analysis of the Antecedents, Corre- Kremer, M., and D. Chen. 2001. “An Interim lates, and Consequences of Organisational Report on a Teacher Attendance Incentive Commitment.” Psychological Bulletin of the Program in Kenya.” Cambridge, MA: Har- American Psychological Association 108: 171–94. vard University. Muralidharan, K., and V. Sundararaman. 2009. Krueger, A. B., and L. H. Summers. 1988. “Effi- Teacher Performance Pay: Experimental Evidence ciency Wages and the Inter-Industry Wage from India. NBER Working Paper 15323. Structure.” Econometrica 56 (2): 259–93. Washington, DC: National Bureau for Eco- Kular, S., M. Gatenby, C. Rees, E. Soane, and K. nomic Research. Truss. 2008. Employee Engagement: A Literature Norrgren, F., and J. Schaller. 1999. “Leadership Review. Kingston upon Thames, UK: Kings- Style: Its Impact on Cross-Functional Prod- ton Business School. http://eprints.kings- uct Development.” Journal of Product Innova- ton.ac.uk/4192/1/19wempen.pdf. tion Management 16 (4): 377–84. Lavy, V. 2008. Gender Differences in Market Com- Nunberg, B., and R. Taliercio. 2012. “Sabotag- petitiveness in a Real Workplace: Evidence from ing Civil Service Reform in Aid-Dependent Performance-Based Pay Tournaments among Countries: Are Donors to Blame?” World Teachers. NBER Working Paper No. 14338. Development 40 (10): 1970–81. Washington, DC: National Bureau for Eco- Odden, A., and C. Kelley. 2002. Paying Teachers nomic Research. for What They Know and Do. Thousand Oaks, ———. 2009. “Performance Pay and Teachers’ CA: Corwin Press. Effort, Productivity and Grading Ethics.” OECD (Organisation for Economic Co-opera- American Economic Review 99 (5): 1979–2011. tion and Development). 2005. Performance- Lazear, E. P. 1981. “Agency, Earnings Profiles, Related Pay Policies for Government Employees. Productivity and Hours Restrictions.” Ameri- Paris: OECD. can Economic Review 71 (5): 606–20. ———. 2011. Government at a Glance. Paris: ———. 1989. “Pay Equality and Industrial Poli- OECD. tics.” Journal of Political Economy 97: 561–80. Osterloh, M., and J. Frost. 2002. “Motivation Lee, H. 2010. “The Effect of Public Officials’ and Knowledge as Strategic Resources.” In Pay for Performance Satisfaction on Salary Successful Management by Motivation: Balanc- Satisfaction, Job Satisfaction and Organi- ing Intrinsic and Extrinsic Incentives, edited by zational Commitment.” Korea Governance B. S. Frey and M. Osterloh, 27–51. New York: Review 17 (2): 139–66. Springer-Verlag. Lindauer, D. L., and B. Nunberg, eds. 1994. Perry, J. L., T. A. Engbers, and S. Y. Jun. 2009. Rehabilitating Government. Washington, DC: “Back to the Future? Performance-Related World Bank. Pay, Empirical Research and the Perils of Marsden, D. 2004. “The Role of Performance- Persistence.” Public Administration Review 69 Related Pay in Renegotiating the “Effort (1): 39–51. Bargain”: The Case of the British Public Ser- Pfeffer, J. 1998a. The Human Equation: Building vice.” Industrial and labor Relations Review 57 Profits by Putting People First. Boston: Harvard (3): 350–70. Business School Press. ———. 2009. The Paradox of Performance-Related ———. 1998b. “Seven Practices of Successful Pay Systems: Why Do We Keep Adopting Them in Organizations.” California Management Review the Face of Evidence That They Fail to Motivate? 40 (2): 96–124. PAY F L E X I B I L I T Y A N D G OV E R N M E N T P E R F O R M A N C E  59 Pink, D. H. 2009. Drive: The Surprising Truth on Preventive Care Quality: Increasing Pro- about What Motivates Us. New York: Riverhead. vider Adherence to Evidence-Based Smok- Propper, C., and D. Wilson. 2003. The Use and ing Cessation Practice Guidelines.” Preventive Usefulness of Performance Measures in the Pub- Medicine 36 (3): 291–99. lic Sector. CMPO Working Paper Series No. Singh, P. 2010. Performance Pay and Information: 03/073. Bristol, UK: Centre for Market and Reducing Child Malnutrition in Urban Slums. Public Organisation. London: London School of Economics. Purcell, J., and S. Hutchinson. 2007. “Front- Springer, M. G., D. Ballou, L. Hamilton, V.-N. Line Managers as Agents in the HRM-Per- Le, J. R. Lockwood, D. F. McCaffrey, and M. formance Causal Chain: Theory, Analysis P. B. M. Stecher. 2010. Teacher Pay for Perfor- and Evidence.” Human Resource Management mance: Experimental Evidence from the Project on Journal 17 (1): 3–20. Incentives in Teaching. Nashville, TN: Vander- Purcell, J., K. Kinnie, S. Hutchinson, B. Rayton, bilt University, National Center on Perfor- and J. Swart. 2003a. People and Performance: mance Incentives. How People Management Impacts on Organiza- Tirole, J. 1994. “The Internal Organisation of tional Performance. London: Chartered Insti- Government.” Oxford Economic Papers 46 (1): tute of Personnel and Development. 1–29. ———. 2003b. Understanding the People and Per- Vandenberg, R., and C. Lance. 1992. “Satisfac- formance Link: Unlocking the Black Box. Lon- tion and Organisational Commitment.” Jour- don: Chartered Institute of Personnel and nal of Management 18: 153–67. Development. Weber, M. 1978. Economy and Society, Vol. 2. Rafferty, A. M., J. Maben, E. West, and D. Rob- Berkeley, CA: University of California Press. inson. 2005. What Makes a Good Employer? Wilson, J. Q. 1989. Bureaucracy: What Government Geneva: International Council of Nurses. Agencies Do and Why They Do It. New York: www.icn.ch/images/stories/documents/ Basic Books. publications/GNRI/Issue3_Employer.pdf. Winters, M. A., G. W. Ritter, J. P. Greene, and Rexed, K., C. Moll, N. Manning, and J. Allain. R. Marsh. 2009. “Student Outcomes and 2007. Governance of Decentralised Pay Setting Teacher Productivity and Perceptions in in Selected OECD Countries. OECD Work- Arkansas.” In Performance Incentives: Their ing Papers on Public Governance, 2007/3. Growing Impact on American K–12 Education, Paris: OECD. http://caliban.sourceoecd. ed. M. G. Springer. Washington, DC: Brook- org/vl=7179447/cl=20/nw=1/rpsv/cgi-bin/ ings Institution Press. wppdf?file=5l4qdflvl56d.pdf. World Bank. 2012. The World Bank’s Approach Roski, J., R. Jeddeloh, L. An, H. Lando, P. Han- to Public Sector Management 2011–2020: Better nan, and C. Hall. 2003. “The Impact of Results from Public Sector Institutions. Washing- Financial Incentives and a Patient Registry ton, DC: World Bank.