25 Wage Spending Flexibility in Low-Income Countries1

Marc Robinson
Published Date:
October 2007
  • ShareShare
Show Summary Details
Matt Davies, Marijn Verhoeven and Victoria Gunnarsson 

A fundamental characteristic of performance budgeting is the shift from input control toward accountability for results. There are many types of performance budgeting with much variation in their complexity and ambition (see Chapter 1). All, however, carry with them a presumption that those in charge of policy implementation at the disaggregated micro level—primarily, managers of government departments and agencies—are held to account for what they are achieving rather than for remaining within prescribed input limits. The basic idea is that such managers understand better than decision-makers at the central level how to organize their inputs to achieve specific objectives. In contrast, centralized control of budgets at the line-item level is believed to stifle innovation at the micro level, move incentives toward spending rather than economy, and dilute the focus on policy objectives (Roberts, 2003).

Human resources are a key input for government output, and effective performance budgeting includes devolving employment and compensation decisions to managers at the micro level. They need to be able to trade-off labor inputs against other inputs, and to vary the type of labor and its compensation to be able to efficiently produce those outputs that contribute most to policy objectives.

Thus, effective performance budgeting requires flexibility in the use of human and other resources at the micro level. In particular, such micro-flexibility is important for the efficiency and effectiveness of spending because it: (1) facilitates a more efficient mix of labor and non-labor inputs, thereby contributing to technical efficiency; (2) permits the reallocation of human resources between different programs and purposes to yield a more optimal mix of outputs, thereby contributing to allocative efficiency; and (3) permits remuneration and employment practices which motivate government employees more effectively

In low-income countries (LICs), employment and compensation decisions are often made separately from other input decisions, and are typically highly centralized. Non-wage inputs are determined through the budget process (in a number of cases, the decision-making on capital spending is also different from that on current spending). But it is typical for decisions on hiring, firing, transfers, promotions, and compensation to be decided by a central ministry or high-level commission (see below). Stringent procedures for such human resource decisions often not only apply to central government agencies, but sub-national governments and autonomous entities as well. As a result, overall discretion for decision-making in the human resource area is often limited for managers of departments and agencies in government.

It is not obvious that wage spending in LICs is sufficiently flexible to ensure that performance budgeting is effective. Inflexibilities at the micro level can be a major obstacle for effective performance budgeting in LICs. In particular, three forms of such inflexibilities often exist in the public sector: (1) the lack of capacity to terminate employment (and sometimes also to hire new employees—for instance, in the case of a hiring freeze or centralized decision-making on hiring); (2) the inability to transfer government workers between agencies or between programs and activities; and (3) rigidities in the compensation structure and level.

Such spending inflexibilities at the micro level may also jeopardize government control of overall wage spending. Lack of control over the wage bill poses a serious challenge for maintaining overall aggregate budget flexibility, macroeconomic stability, and fiscal sustainability (see, for example, Maguire, 1993). In addition, inflexibilities in public sector employment and pay policy can have adverse macroeconomic implications through their impact on private sector labor supply and wage levels.

Achieving and retaining control of the wage bill is critical, and has been the dominant concern in many LICs. The concern about wage bills stems from the perception that in countries with weak fiscal discipline and governance structures, the wage bill is easily expanded during economic good times and in response to political pressures, but much harder to bring down during periods of economic adversity (Fedelino et al., 2006). The downward stickiness of nominal wages in the public sector and the lack of instruments for downsizing the public payroll contribute to the perception that increases in the government wage bill are unlikely to be reversed. Therefore, frequently a distinction is made between decision-making on non-wage recurrent spending and wage spending. Changing the composition of non-wage recurrent spending is typically seen to have a limited permanent impact on overall spending levels,2 while decisions to increase employment and compensation, even within a fixed budget constraint, are seen to enhance long-term spending pressures.

This chapter seeks to better understand the relationship between wage bill control, the flexbility of wage spending, and options for moving toward performance budgeting in LICs. We first review the relevant literature on the size and determinants of the wage bill. Although economic theory suggests that wage bills as a share of gross domestic product (GDP) tend to rise over time, empirical studies do not find evidence for this. We then look at data for 1990–2000 for 37 LICs with IMF-supported programs to increase our understanding of the degree of wage spending inflexibility in countries undergoing fiscal adjustment. Our empirical analysis confirms that countries have typically managed to exert control over the size of the wage bill over time. But we also find evidence that the wage bill is less flexible than other spending categories, particularly in the short run. The chapter then reviews how institutional processes contribute to spending inflexibilities. In particular, we investigate institutional constraints to various forms of wage spending flexibility and how these can facilitate or frustrate attempts to improve civil service performance and the implementation of performance budgeting.

The chapter draws tentative conclusions on how wage spending inflexibilities at the micro level can be removed to support a move to performance budgeting and greater control of the overall wage bill in LICs. It reviews evidence from LICs and OECD countries on the success of civil service reform and performance budgeting and how the two processes interact. Given this experience, it concludes that a number of mechanisms used to increase wage spending flexibility at the spending agency level in OECD countries could aid control of the wage bill as well as support efforts to improve the performance of the civil service in LICs. However, given the constrained institutional and volatile political environments found in many LICs, we advocate a gradual approach to decentralization of civil service management.

Determinants of the wage bill—theory and analysis in the literature

Early literature on wage bill determinants focused on long-term trends, and concluded that the government wage bill was likely to grow as a proportion of national income over time. The first articulation of this was by Wagner in the nineteenth century. He postulated that economic development would lead to a rising share of public expenditure due to the increase in the demand for, and relative cost of, public administration and regulation in a more developed economy. This basic notion was rephrased by Baumol (1967), who focused on public sector productivity. He suggested that while labor productivity in the goods-producing sector would tend to increase, productivity in the public sector (and the services sector in general) would tend to lag due to the labor-intensive nature of services and the limited options for labor productivity enhancement in such service activities as education (where Baumol refers to limits to reasonable class sizes and student-teacher ratios as key factors constraining labor productivity improvements). This would imply that in order for the government to maintain a given level of services per person and as wages rise reflecting productivity increases in the goods-producing sector, its wage bill must grow as a share of national income.3

Average wage bill spending as a share of GDP is lower in LICs than more developed countries. Table 25.1 shows that LICs, on average, have a wage bill of 4.8 percent of GDP, which is about 1 percentage point of GDP below average levels prevailing in middle-income countries (MICs) and high-income countries (HICs). However, measured as a share of overall government expenditure, the government wage bill is higher in LICs than in MICs and HICs. Also, in countries where the level of development is higher, the relative government wage tends to be lower while public employment is higher. But the large variation in regional averages suggests that these stylized facts obscure large differences in country experiences. Also, care needs to be taken in interpreting this empirical information because of data problems (Box 25.1).

Table 25.1Government wages and salaries, 2000a
Wages salaries as % of GDPWages and salaries as % of government expenditureAverage wage to per capita GDP ratioPublic employment as % of total population
Asia and the Pacific4.825.32.44.6
Western Hemisphere5.428.51.94.4
Middle East and Central Asia10.237.62.0
Sub-Saharan Africa7.
Low-income countries4.827.13.92.3
Middle-income countries6.
High-income countries5.720.11.38.5

For 2000 or most recent year available. Most data cover the general government. Public employment numbers cover public enterprises as well as the general government.

Sources: IMF World Economic Outlook database; IMF Government Finance Statistics database; World Bank Administrative and Civil Service Reform database; World Bank World Development Indicators database; ILO LABORSTA database.

For 2000 or most recent year available. Most data cover the general government. Public employment numbers cover public enterprises as well as the general government.

Sources: IMF World Economic Outlook database; IMF Government Finance Statistics database; World Bank Administrative and Civil Service Reform database; World Bank World Development Indicators database; ILO LABORSTA database.

Box 25.1Data issues

Empirical analysis of government employment and wage levels is hampered by the lack of availability of consistent data between countries. Reliability of government employment statistics is an issue in many LICs. Where data are reliable, important definitional questions can arise. Countries differ in what they consider to be government employment, and therefore the same terminology is often used to refer to different categories. For example, some countries include teachers and health workers in the civil service, while others do not. Some countries consider local government employees paid from the central budget as local government staff, whereas others designate these as central government staff. International comparisons of government data on government wage levels are still more complicated because of the role of allowances and the presence of in-kind benefits, which are not consistently included in wage statistics.

However, the literature provides little empirical evidence that in LICs wage bills tend to increase as a share of GDP over time.4 In fact, partial data suggest that ratios of wage bills to GDP in LICs have remained broadly stable, or even declined, on average in recent decades. For example, Lienert and Modi (1997) show for 32 LICs in sub-Saharan Africa that, despite some large movements in individual countries, average wage bills declined by 1 percentage point of GDP to 6 percent of GDP during 1986–96, mainly driven by a decline in average real government wages. They also find that the government employment trends in sub-Saharan Africa were mixed. While a number of countries contracted government employment sharply, often in conjunction with real wage increases, an equal number moved in the opposite direction. In another example, Kraay and van Rijckeghem (1995) do not find a statistically significant time trend for the ratio of wage spending to GDP in a sample of 34 LICs during 1972–92. Finally, our data for a sample of 37 LICs that undertook macroeconomic reform show a decline in average wage spending from 7.2 percent of GDP in 1990 to 6.0 percent in 2000 (see below).

Kraay and van Rijckeghem (1995) find that wage bills in LICs are affected by a variety of factors. For a sample of 34 LICs, they consider a range of explanatory variables, covering macroeconomic conditions, fiscal pressures, labor supply and quality, demand for public services, and external pressures, and find the following statistically significant relationships:

  • Employment. The variable with the strongest (negative) effect on employment appears to be the presence of an IMF-supported adjustment program. Other variables with a positive, albeit small, effect are revenue effort and the real effective exchange rate. Urbanization and foreign financing also have a positive effect. Secondary school enrollment has a small negative effect as does population size, suggesting increasing returns to scale of a larger population.5

  • Real wages. Real government wages decline with real per capita GDP, presumably because the shortage of skills employed in government falls with economic development. In addition, real wages increase with revenue and decline with debt. IMF-supported structural adjustment programs have no significant impact on wages, but IMF-supported stabilization programs do have a significant negative impact.6 Real wages decrease with secondary school enrollment (because of its impact on the relative scarcity of skills employed by the government). The real effective exchange rate has a negative effect on real wages expressed in purchasing power parity (PPP) dollars, but not on real wages relative to GDP per capita.

  • Overall wage bill. The statistical relationship between real GDP per capita and wage spending as a share of GDP is negative, consistent with the impact on real government wages. In addition, secondary school enrollment has a negative effect, as the adverse impact on wages outweighs the associated increase in employment. Revenue effort and foreign financing both have a positive impact, mainly through allowing extra employment. The presence of an IMF-supported structural adjustment program has an overall negative effect, stemming from reductions in employment.

Despite the observed relative stability in average wage bills in LICs over the longer term, countries may find it difficult to contain wage spending at sustainable levels. Many LICs face an urgent need to enhance public service delivery, including meeting the Millennium Development Goals, and need to achieve this within a sustainable fiscal framework. Fedelino et al. (2006) discuss how undue increases in wage spending can jeopardize the achievement of macroeconomic objectives, and that in such cases IMF-supported programs have included conditionality on the size of wage bills. Such conditionality, often in the form of ceilings on the size of the wage bill, were included in as much as half of the 42 LICs with arrangements under the IMF’s Poverty Reduction and Growth Facility between 2003 and 2005.7 Fedelino et al. emphasize that wage bill ceilings provide only a short-term fix and ultimately need to be supplanted by civil service reform, and caution that countries may seek to meet wage bill conditionality (or targets for wage spending unrelated to IMF-supported programs) with short-term measures, which may lock in inappropriate civil service structures that are both unsustainable and costly in terms of economic efficiency. In the next section, we will explore wage bill flexibility in LICs in more detail, and particularly what can be inferred from the more short-run developments in wage bills about the cost of maintaining control over the wage bill through wage bill ceilings and other measures.

Wage spending flexibility—empirical analysis

This section examines how wage bills have been kept under control in LICs, in particular during periods of fiscal adjustment. The analysis up until now has revealed little about the relative movements of wage bills and overall expenditures. Inflexibilities in the wage bill may, in fact, be relative to other expenditure categories, especially in the short term. This would be particularly important in LICs undergoing periodic fiscal adjustment, and the analysis in this chapter focuses on this group of countries. If the wage bill is not flexible in the short term, other spending will have to bear the brunt of budget cuts. This could have a significant impact on public sector performance and economic growth. Our empirical analysis employs an explicit strategy. To test for the presence of inflexibilities during a fiscal adjustment process the analysis below looks at two areas to assess wage bill inflexibility: (1) the relative share of wages in primary expenditures and GDP in the longer term;8 and (2) the short-run responsiveness of wages to movements in annual levels of primary expenditure.9Box 25.2 describes the dataset used for the analysis.

Flexibility in the trend of wage spending

Over the period 1990–2000, average wage spending decreased both as a share of GDP and of primary expenditures (Table 25.2). Average wage spending declined from 7.2 percent of GDP in 1990 to 6.0 percent in 2000.10 In part, the decrease in wage spending reflects the decline of the average fiscal balance by around 3.2 percent of GDP, with average primary expenditures falling by 5.0 percentage points of GDP. But wage expenditure, on average, declined by more than primary spending, and the share of wages in primary spending was reduced from an average of 28.6 percent of primary spending in 1990 to 25.9 percent in 2000.

Table 25.2Summary statistics, 1990 and 2000a
Country groupVariableWage bill as % of primary expendituresWage bill as % of GDPPrimary expenditures as % of GDPBudget balance as % of GDP
Standard deviation13.
Lowest wage sharesMean10.
Standard deviation1.
Mid-sized wage sharesMean26.725.76.35.625.422.6-5.9-4.0
Standard deviation6.
Highest wage sharesMean51.
Standard deviation5.
Fiscal contractionMean23.726.77.15.834.222.5-8.6-4.2
Standard deviation11.
Fiscal expansionMean36.824.47.46.420.427.1-5.8-4.8
Standard deviation12.

1990 or earliest year available.

1990 or earliest year available.

The government wage bill was reduced most in countries that in 1990 had a relatively high level of wage spending (Figure 25.1). At the same time, countries with a relatively low initial level of wage spending increased their wage bills on average. This convergence of initial outliers to the middle also explains the reduced dispersion in wage spending across countries (Table 25.2 shows that the standard deviation of wages as a percentage of GDP decreased from 15.8 to 11.3 during 1990–2000, and from 13.7 to 7.7 as a proportion of primary expenditures). For example, wage spending in Benin dropped from 50.9 percent of primary spending in 1990, which was almost double the sample average, to 26.1 percent in 2000, which was just 0.1 percentage point above the sample average. Tanzania, on the other hand, increased its wage expenditures from 3.7 percent of GDP in 1990, or about half of the sample average, to 4.5 percent by 2000, while the sample average fell over the same period.

Figure 25.1(a)Wage bill as a percentage of primary expenditures by size of 1990–92 wage share of primary expendituresa

Figure 25.1(b)Wage bill as a percentage of GDP by size of 1990–92 wage share of primary expendituresa

a The sample was disaggregated into three groups (lowest, mid-sized, and highest) based on wage spending as a percentage of primary expenditures at the beginning of the period (1990–92). The lowest group includes the group of countries with wage spending as a percentage of primary expenditures less than one standard deviation below the sample mean in 1990–92 (Armenia, Guyana, Mozambique, and São Tomé and Príncipe). The highest group includes the countries with wage spending as a percentage of primary expenditures above one standard deviation above the mean in 1990–92 (Benin, Cameroon, Djibouti, Honduras, Kyrgyz Republic, Senegal, and Yemen). The remaining countries in our sample are in the mid-sized group (except for Georgia, Guinea-Bissau, and Rwanda, which are excluded from the figures because of data inconsistencies).

Figure 25.1(c)Wage bill as a percentage of primary expenditures by direction of trend in primary expendituresb

Figure 25.1(d)Wage bill as a percentage of primary expenditures by direction of trend in primary expendituresb

b Fifteen “fiscal expansion” countries had a higher level of primary spending as a share of GDP in 2000 than in 1990 (Benin, Bolivia, Burkina Faso, Cambodia, Ethiopia, Ghana, Honduras, Laos, Lesotho, Malawi, Mali, Nicaragua, Senegal, Tanzania, and Yemen). In the 23 remaining “fiscal contraction” countries, primary expenditure declined over the same time period.

Box 25.2The dataset

This chapter considers wage bill data for 1990–2000 in a sample of 37 LICs that undertook macroeconomic reform in the context of an IMF-supported program.a We use a database constructed by Gupta et al. (2002) that draws from the World Economic Outlook (WEO) and other IMF databases to obtain macroeconomic and fiscal data for 37 LICs which received a concessional loan from the IMF in the 1990s. The fiscal data are measured on a cash basis and at the general government level (except for Lesotho and Yemen, where the data cover only the central government).

By using a dataset for LICs that undertook macroeconomic reform, our analysis is focusing on cases where there was a likely need for spending flexibility. In other words, the analysis is considering cases where the flexibility of government spending in general, and of wage spending in particular, may be an issue. Moreover, by focusing on the whole period for 1990–2000 and not just the years where an adjustment program was in place, the database contains episodes of both fiscal consolidation and expansion.

The strength of the database is in providing comparable macro-fiscal variables across a wide range of LICs. Nevertheless, the analysis remains vulnerable to unobserved changes in the coverage of the wage bill and other fiscal variables. For that reason, large movements in wage-related variables within the dataset are investigated and described following the main empirical analysis.

a Albania, Armenia, Benin, Bolivia, Burkina Faso, Cambodia, Cameroon, Central African Republic, Chad, Djibouti, Ethiopa, The Gambia, Georgia, Ghana, Guinea, Guinea-Bissau, Guyana, Honduras, Kenya, Kyrgyz Republic, Laos, Lesotho, Macedonia (FYR), Madagascar, Malawi, Mali, Mauritania, Mozambique, Nicaragua, Niger, Rwanda, Senegal, Sao Tome and Principe, Tajikistan, Tanzania, Yemen, and Zambia.

For the countries that started with high wage expenditures, the motivation for reducing wage spending as a share of GDP and primary expenditure is likely to have been to protect macroeconomic stability and fiscal sustainability (see Fedelino et al., 2006).11 In those countries, the wage bill absorbed such a large proportion of expenditure that vital growth and productivity enhancing expenditures were being squeezed out. For the countries with a low initial level of wage spending, the upward adjustment of wages is likely to have been aimed at expanding public service delivery and enhancing motivation, recruitment, and retention.

However, in countries where primary spending contracted during the 1990s, wage spending was not reduced in line with overall spending trends, although it declined as a share of GDP. In these countries (fiscal contraction countries), the average wage share in primary expenditure was volatile in the early 1990s, to stabilize at around 27 percent in the second half of the decade. This was still well above the initial level of 24 percent of primary spending, implying that non-wage spending was cut by more than overall primary spending. However, these averages obscure important country differences; wages actually fell as a share of primary spending in 8 out of 23 contractionary countries. At the same time, countries which boosted primary expenditure over the decade (fiscal expansion countries) saw the share of primary spending devoted to wage spending fall, from about 36 percent in the early 1990s to 24 percent in 2000, while wage spending as a share of GDP did not show a clear trend.

This analysis of spending trends over the 1990s suggests that LICs in our sample managed to adjust wage spending, especially when the initial wage bill was relatively high and when seeking to reduce overall spending levels. On average, countries that contracted their primary spending between 1990 and 2000 also brought down wage spending as a share of GDP, but still saw wage spending as a share of primary spending rise. This implies that non-wage spending has had to adjust more, and may be an indication of the relative inflexibility of wage spending. It may also reflect optimal policy choices—for example, the long-term cost of losing government workers with scarce skills may outweigh the short-term benefits from cutting them from the government payroll. The next two sections look more closely at disaggregated data which underpins the trends in sample and group averages that we have considered until now.

Short-term responsiveness of the wage bill to primary expenditure

The short-term responsiveness of wage expenditure to primary spending can be assessed from correlations and elasticities. This section considers several indicators of responsiveness. First, the share of countries with a large correlation between annual changes in the wage bill (and other spending components) and annual changes in primary spending indicates how close the short-term co-movement is between the wage bill and primary expenditures. Second, to understand the proportional relationship between wages and spending, the elasticity of wage spending to primary spending reveals the size of the annual change in the wage bill in relation to the annual change in primary spending.

Annual movements in the wage bill as a share of GDP are positively correlated with movements in primary expenditures (Table 25.3). In half of the countries, the wage bill in percentage of GDP moves positively with changes in primary expenditures as a percentage of GDP. But, in 56 percent of the countries, there is a negative relationship between wage spending as a percentage of primary expenditure and primary spending as a share of GDP. This indicates that, as primary expenditure increases or decreases, wages tend to move in the same direction but not in a proportional amount. Table 25.3 also shows that movements in other expenditure components, such as capital expenditures and goods and services, are more strongly correlated to movements in expenditures than are movements in the wage bill.

Table 25.3Percentage of countries with strong pair-wise correlations, 1990–2000a
Primary expendituresGDP per capita growthTotal revenueBudget balance
AllWage bill50.08.35.619.430.611.12.850.0
Wage bill (% of prim. exp.)2.855.68.322.213.947.222.211.1
Capital expenditures80.
Transfers and subsidies38.92.813.919.438.98.38.316.7
Goods and services56.
Fiscal contractionWage bill56.50.08.721.743.58.74.339.1
Wage bill (% of prim. exp.)4.356.513.013.021.752.230.44.3
Capital expenditures73.
Transfers and subsidies39.14.38.721.726.
Goods and services59.
Fiscal expansionWage bill38.523.
Wage bill (% of prim. exp.)
Capital expenditures90.
Transfers and subsidies0.038.530.87.761.57.723.17.7
Goods and services50.

Correlation coefficients were calculated on a country-by-country basis. Countries with strong pair-wise correlations are defined as countries where the correlation in the variable pair during 1990–2000 is 0.5 or above, and which are also statistically significant at the 5 percent confidence level. If the correlation coefficient is larger than 0.5, the country is added to the “positive” column, while it is added to the “negative” column if the correlation coefficient is smaller than -0.5. Variables are measured in percentage of GDP, unless indicated otherwise.

The country groups are defined as in Figures 25.1(c) and (d).

Correlation coefficients were calculated on a country-by-country basis. Countries with strong pair-wise correlations are defined as countries where the correlation in the variable pair during 1990–2000 is 0.5 or above, and which are also statistically significant at the 5 percent confidence level. If the correlation coefficient is larger than 0.5, the country is added to the “positive” column, while it is added to the “negative” column if the correlation coefficient is smaller than -0.5. Variables are measured in percentage of GDP, unless indicated otherwise.

The country groups are defined as in Figures 25.1(c) and (d).

The pair-wise correlations also show that countries which fiscally contracted during the 1990s reduced their wage bills as a share of GDP, but not in equal proportion to the decline in primary spending. The correlation between the wage bill as a percentage of GDP and primary spending is strong and positive in more than half of the countries, while it is negative and strong between the wage bill as a share of primary spending, and primary expenditure is negative and strong also in more than half of the countries. Furthermore, in only 39 percent of the countries do expansionary fiscal conditions have a large positive impact on the wage bill as a share of GDP, while in 23 percent of the expansionary countries there was a strong negative impact on the wage bill as a percentage of GDP. Hence, in 54 percent of the expansionary countries, the share of primary expenditures composed by wages was strongly reduced as primary expenditures grew. These results are consistent with the earlier findings on average spending trends. The relationship between wage spending and other macro-fiscal variables, such as GDP growth, total revenue and the budget balance, is similar to that of the wage bill and primary expenditures, although slightly weaker.

The elasticity of the wage bill to changes in total primary expenditures and total revenue is smaller than for other expenditure aggregates (Table 25.4). The average elasticity of the wage bill to primary spending is 0.51 (that is, a 1 percent increase/decrease in primary spending is associated with an increase/decrease in the wage bill of 0.51 percent) while the elasticities of capital spending and spending on other goods and services to primary expenditure are 0.92 and 0.68, respectively. The elasticity of spending on transfers and subsidies is 1.07, which means it is more than proportionally responsive to movements in overall primary spending. The results are robust to dividing the sample based on trends in primary spending over the period, which implies that the size of the average response to an increase and decline in primary spending is symmetric. Moreover, although wage bills are somewhat responsive to changes in revenue, spending on transfers and subsidies as well as goods and services are more flexible to revenue changes.

Table 25.4Elasticities of government spending, 1990–2000a
GroupbPrimary expendituresGDP per capita growthTotal revenueBudget balance
AllWage bill0.51-0.070.570.07
Wage bill (% of prim. exp.)-0.49-0.11-0.00-0.06
Capital expenditures0.920.030.210.12
Transfers and subsidies1.
Goods and services0.680.100.690.17
Fiscal contractionWage bill0.49-0.060.54-0.01
Wage bill (% of prim. exp.)-0.51-0.10-0.03-0.10
Capital expenditures0.940.040.030.03
Transfers and subsidies1.
Goods and services0.590.130.710.15
Fiscal expansionWage bill0.59-0.080.620.25
Wage bill (% of prim. exp.)-0.41-0.090.00-0.00
Capital expenditures0.940.030.700.31
Transfers and subsidies0.92-
Goods and services1.14-0.000.820.85

Elasticities are not calculated on a country-by-country basis but on the pooled panel of countries. The elasticity is defined as the regression coefficient from a single regression of each variable pair in logs. Variables are measured in percentage of GDP, unless indicated otherwise.

The country groups are defined as in Figures 25.1(c) and (d).

Elasticities are not calculated on a country-by-country basis but on the pooled panel of countries. The elasticity is defined as the regression coefficient from a single regression of each variable pair in logs. Variables are measured in percentage of GDP, unless indicated otherwise.

The country groups are defined as in Figures 25.1(c) and (d).

Conclusions from the empirical analysis

The empirical analysis of 37 LICs that undertook macroeconomic reform during the 1990s under IMF-supported programs yields the following key findings:

  • Wage spending trends have, by and large, supported fiscal adjustment. In our sample of countries, wage expenditure as a share of GDP tends to go down during periods of fiscal contraction, and particularly when wage spending is initially high. The response of wage spending to a fiscal expansion is more ambiguous—wage spending as a share of GDP actually fell on average in the group of countries where primary expenditure increased during 1990–2000, and in 23 percent of the sample countries the wage bill-to-GDP ratio declined signicantly in years in which the share of primary spending in GDP rose.

  • At the same time, there is evidence of some wage bill inflexibility, at least in the short term. Wage bills tend to be less volatile than other components of primary expenditures and less correlated with and elastic to annual movements in primary expenditure. Wage bills across the sample were as flexible—as measured by the coefficient of variation—as primary expenditure over the entire period. Nevertheless, the elasticity of wage spending to changes in primary expenditures is smaller than that of other expenditure items, in particular capital spending and goods and services, suggesting that in the short term the wage bill may be less easy to adjust than other expenditure components.12

Such wage spending inflexibility may, in part, reflect national wage and employment policies. In situations where fiscal consolidation is needed, the cost of adjusting other spending components may be smaller, as reducing wage spending during downturns may result in the loss of highly skilled workers which were attracted and trained at high cost to the government. Similarly, quickly expanding wage spending by stepping up hiring or granting large wage increases during upswings may be difficult to undo later, and thereby jeopardize fiscal sustainability, while perhaps not yielding significant pay-offs in terms of better public service delivery.

However, the short-term inflexibility of wage spending also reflects institutional arrangements in many LICs. The next section examines how such arrangements can promote inflexibilities in wage spending at the micro level. A key question concerns the institutions and institutional processes that contribute to inflexibilities in wage expenditure, and the extent to which they are amenable to reform. In particular, what would be needed to establish an appropriate institutional environment for performance budgeting?

Institutions, wage spending flexibility, and performance budgeting

Government decision-making on wages and employment is the product of political and bureaucratic institutions and processes. These affect the wage bill through their impact on aggregate employment and wage policies, while also affecting the skill composition and motivation of the government workforce, and thereby its performance. An examination of these institutions and how they can affect wage bill flexibility and civil service performance is key for understanding how and where performance budgeting can drive changes in input composition and performance.

Although institutions differ in each country, and are often markedly different between regions, a number of typical institutions that particularly influence the flexibility of wage spending can be identified. Following Kiragu et al. (2004), we analyze relevant institutions and processes in four groups: administrative, political, trade union, and donor institutions.

Administrative institutions and processes

Centralized personnel management agencies and processes

Most governments have a department or agency dedicated to personnel management issues. These often were established in an effort to limit political influence on the civil service, and can be important in curbing the effects of corruption in government departments and agencies. In some countries these centralized agencies for personnel management may be part of the mainstream architecture of government (for example, in Tanzania); in others they may be semi-independent (for example, in South Asian countries such as Sri Lanka and Pakistan). Whichever model is chosen, these institutions are often responsible for administering a highly regulated set of positions, job classifications, and grading and promotion criteria that individual departments are unable to change without the approval of the central agency. These can introduce inflexibilities in a number of different ways:

  • Promotions based on seniority rather than merit place a natural upward pressure on the wage bill and provide a disincentive to performance. They are more likely to be relied on where the central agency regulates promotion decisions, perhaps because the central agency is more distant from performance. To the extent that promotions become automatic, these systems build an autonomous buoyancy into the wage bill.

  • Management of the number of established posts, which is often immune to change, especially in a downward direction. In some regions the overall number of established posts also has sub-ceilings for particular cadres that over time become entitlements for particular specializations. These cadre systems (for example, in South Asia and Ethiopia) may be based on previous models of government and, if they are not amenable to change, can inhibit the ability of institutions to adapt to changing functions or the nature of demand.

  • Central hiring and firing regulations, in particular where civil servants have de facto tenure, can prevent reductions (or increases) in the government labor force in response to changing demand conditions or budgetary pressures. Williamson (2003) notes that these regulations in Uganda have prevented institutions attempting to implement performance budgeting from varying the skill mix.

Nationwide pay scales and settlements

It is widespread practice for public service pay to be set according to a single pay scale, differentiated by grade that covers the entire civil service or a specific cadre within it. When wage increases are granted—often annually, but not always in the context of the annual budget—they affect the whole scale, often without differentiation by grade. Wage increases may be formally linked to external indicators, such as consumer price inflation or earnings in the private sector. It may also be the result of a central collective bargaining agreement (see below on trade unions) or be an entirely discretionary decision by the government (see below on political institutions). In many systems civil servants are entitled to service-related increments in addition to the global increase in pay scales. The main issues to consider in connection with wage spending flexibility include the following:

  • Comprehensive pay scales in principle enable governments to exercise control over the overall wage bill, in particular where pay increases are a discretionary policy decision. Although political factors may not allow for this opportunity to be grasped, there is evidence from OECD countries in the 1980s that centralized systems are more successful at controlling wage bills (Meurs, 1993). Lienert and Modi (1997) note that in countries of the CFA franc zone in the mid-1990s, macroeconomic affordability of the wage bill was achieved by granting limited across-the-board nominal wage increases in the face of high inflation.

  • However, to the extent that this control is not, or cannot, be exercised (for example, where scale revisions are tied to an external indicator not under the control of the authorities), the broad applicability of the scales can exacerbate the fiscal pressure from the wage bill. For instance, in Tanzania, central wage settlements have caused other inputs to be squeezed out, particularly in labor-intensive sectors.

  • Where service-related increments are part of the system, these can contribute to the buoyancy of the wage bill, which can be problematic if fiscal contraction is required in an economic downturn.

  • The key problem of systems with nationwide pay scales and settlements is that they restrict managerial and labor market flexibility. Managers are rendered unable to reward good performance with pay and to vary wage rates according to local or sectoral market conditions (Meurs, 1993, reports the same finding in the context of OECD countries).

Fragmented remuneration

In many countries the basic salary of a civil servant is a relatively minor proportion of their overall remuneration. Complex systems of allowances and entitlements in cash (hardship, task-based, and cost-of-living allowances), in kind (free transportation and housing), and of an intangible nature (status, job security, and access to power and rent-seeking opportunities) significantly increase compensation over the basic wage. For instance, Indonesia operates a system of allowances for each department, funded from the development budget, based on the number of projects the department implements, which raise salaries by almost 100 percent (Manning, 2000). These allowances make pay policy harder to manage, and raise a number of issues in relation to wage spending flexibility:

  • The total value of allowances may be unknown, both in cash terms to the budget and in terms of relative importance in the total remuneration of government employees. As such they are unlikely to be subject to the same degree of monitoring and control as wages.

  • The allowances may be outside of the scope of pay revisions and negotiations, and contain explicit or implicit indexation (for example, a cost-of-living allowance directly indexed to a price index was recently proposed in Sri Lanka). The provision of housing, which is a widespread benefit for senior civil servants in LICs, is an example of an in-kind allowance that is implicitly indexed to an external variable. While this may not have a large short-term cash impact on the budget, it does have an opportunity cost and can significantly impact public-private sector wage differentials with a consequent impact on recruitment, turnover and, ultimately, performance.

  • To the extent that they are valuable to employees and differentiated between departments and agencies—as in the case of Indonesia—they can be an obstacle for the mobility of labor within the public sector and compromise the ability of agencies with lower allowances to attract adequately skilled staff. This is particularly damaging where the distribution of allowances is at odds with the government priorities because it reflects, for example, the scope for the ministry or agency to generate own revenues to fund allowances for its workers.

  • If funding for salary top-ups is time limited it can have implications for the human capital of an agency as experienced and talented staff leave as allowance top-ups disappear. In Nicaragua, this has also resulted in an upward pressure on the aggregate wage bill because a rigid labor code enforces the continuation of these allowances after the funding source has disappeared.

  • The formal wage bill in the government accounts may not accurately reflect total personnel-related costs. Pay decisions may also affect other line-items of the budget—for instance, public sector pensions that may be linked, formally or informally, to the public sector pay scales. Wage bills may also be hidden in current transfers to public sector agencies. In Nicaragua it is estimated that almost half of the wage bill is reported in non-wage budget items.

Political institutions and processes

All the components of the wage bill have great political significance, particularly in LICs. Public sector employment is a highly visible policy instrument and often an important tool of political patronage. Similarly, wage policy can be a valuable political tool particularly where the public sector is large or acts as a social safety net. At the extreme, for instance in Honduras (World Bank, 2002), the legislature has complete control over public sector wage setting and negotiations. In most countries the wage bill is the result of a balance between political and technical considerations. The main mechanisms through which political institutions affect the flexibility of wage expenditure are as follows:

  • Pay increases linked to the electoral cycle. The nationwide pay scales and settlement process described above provide a very visible mechanism for ruling political parties to provide pay increases to a large and politically influential group (see below on trade unions). As such they can result in large movements in the wage bill that may work against sound macroeconomic policy.

  • Political preferences for wage compression. Katz and Krueger (1993) note for OECD countries that the political motivations on pay tend to lead to greater wage compression than technical considerations would recommend. The political debate in many countries is both against excessively high wages and for “fair” wages to the lower-paid. A widening of the differential may under such circumstances not be politically expedient, irrespective of technical considerations. This is borne out by the evidence across sub-Saharan Africa, where, despite wage compression being an acknowledged constraint to improved civil service performance, compression ratios generally remained stagnant or deteriorated through the 1990s (Kiragu et al., 2004; Stevens and Teggemann, 2004).

  • Political- and electoral-based hiring. In varying degrees, patronage affects employment policies in many LICs. The skill set of workers hired for patronage reasons tends to be inconsistent with the needs of the government, and successive rounds of patronage-driven hiring as governments change can drive up government employment and the wage bill over time, as it may be hard to reverse hiring by previous governments. In addition, in some countries (for example, Yemen; see World Bank, 2002), the government has acted as an employer of last resort by absorbing redundant workers from public enterprises, army retirees, or recent university graduates. Employing workers in order to meet short-term political aims reduces the funds available for more productive expenditure. It also complicates the management of public agencies.

Trade unions

Powerful and large public sector unions can also contribute to wage spending inflexibility. Trade unions exist to serve the interests of their current members and thus have a vested interest to resist any reform that would reduce the size, job security, or real wage level of the civil service. In many countries they are also a very powerful political force and thus have substantial negotiating power. Kiragu et al. (2004) show that in three countries with strong trade unions (Benin, Burkina Faso, and Zambia), such unions have been a force against pay restraint and modernization of compensation determination frameworks. Stevens and Teggemann (2004) note that Zambia’s political context, including the lack of willingness to take on the trade unions, was a particular source of failure in the civil service reform program.

Trade unions typically support wage compression. As the bulk of their members hold relatively low-paying positions, their negotiating strategy is typically to maximize the average wage through increases at the lower end rather than substantial increases for the much smaller number of senior managers (Katz and Krueger, 1993). This observation is borne out in the analysis of Kiragu et al. (2004), which shows that in countries with strong trade unions, wage compression either stayed high or increased. Countries with weak trade unions, on the other hand, achieved periods of salary decompression (although this result was not found over the entire period).

Donor community

The donor community can have a significant impact on the wage bill in LICs. Its influence is felt in two main ways:

  • Advocacy. In many LICs the donor community is a forceful voice in favor of civil service reform. It supports it through a mixture of advocacy, financial assistance, and program conditionality. The donors can be a valuable ally for a government that wants to implement public sector reforms in the face of political or trade union opposition. For instance, in Burkina Faso, a 1998 law on public administration reform was passed despite strong union opposition, largely through the support and urging of the donor community. In this sense the donors are a significant, although far from dominant, force for flexibility in the wage bill (see World Bank, 2002, for a review of the effectiveness of conditionality on civil service reform; Fedelino et al., 2006, for a review of wage bill ceilings in IMF-supported programs).

  • Project-based employment and pay. Donors also affect the wage bill through the projects they fund. The primary channel is through the use of salary top-ups and other project-related cash and in-kind allowances (see above on administrative institutions and processes). Another channel is through the additional contractual employment that donor projects may create in government departments. These contractuals are not formally part of the established strength of the civil service and their remuneration is generally not part of the government wage bill (the costs are normally to be found in the investment budget). However, typically these contractual workers are converted into civil servants once the project expires, thereby contributing to the size of the government payroll (for example, in Bangladesh). And the contractuals hired for foreign-funded projects may not have the skill mix needed most by government, in which case it also contributes to the mismatch between worker skills and government needs.

Implications for performance budgeting and civil service reform in LICs

The previous section has described how the nature and quality of the administrative and political institutions that have to manage the process in LICs often work against the flexibility that is required for control of the overall level of wage spending, performance budgeting, and other reforms aimed at enhanced government performance (see also Box 25.3). It also showed that the institutional arrangements, and in particular centralization of employment and wage decisions, have allowed governments to exercise control over overall wage spending. This section reviews the approaches taken in both OECD countries and LICs to civil service reform and performance budgeting, while retaining control of the overall wage bill.

The experience of OECD countries shows that strengthening managerial flexibility can be consistent with macroeconomic control of the wage bill. In fact, it has typically been severe macroeconomic pressure and concerns about the size of the wage bill that have prompted governments to tackle the difficult political issue of civil service size, cost, and performance. As a result, periods of fiscal stress in OECD countries are generally well correlated with public sector reform programs,13 often alongside or directly linked to performance budgeting reform. In the area of employment and wages, such reforms have focused on remuneration. Although individual country institutions differed widely, Marsden (1993) notes a number of commonalities in the approaches to public sector pay. The reforms sought to reduce indexation, include greater sensitivity to local labor market conditions, adapt pay systems to new management methods (for example, performance budgeting), and move towards performance-related pay (OECD, 1993, 1995). The mechanisms that are potentially relevant to LICs for increasing the wage bill and managerial flexibility include the following:

  • Delegated pay bargaining. Although the precise model and the level of delegation varies, the movement away from civil service-wide pay scales and pay agreements has been one of the fundamental changes observed in the OECD countries. One example is the one-line running cost budgets used in the UK. These cover both labor and other goods and services used in the administration of public services. This mechanism allows savings on non-wage inputs to fund necessary enhancements of the wage bill or vice versa. It is, however, only meaningful with delegated pay bargaining. There is some evidence from the UK (Marsden, 1993) that the dispersion of public sector pay has gone up with increased autonomy.

  • Different rates of pay and pay adjustment for different groups and locations. This, in effect, is a more nuanced set of pay scales so that remuneration is more closely linked to either regional or occupational labor markets. While it was attempted in a number of OECD countries (notably by the US federal government), it was not as widespread or as successful as was expected (Marsden, 1993). One reason was the perceived problems in withdrawing location and other specific pay enhancements. It was feared that withdrawal would lead to reduced labor mobility while retention would undermine control over the wage bill.

  • Performance-related pay. This was adopted for four main reasons: as a complement to management change; as a symbol of organizational change; as a motivator; and to increase pay flexibility by breaking the link between length of service and pay level. But to be successful, performance-related pay needs to extend beyond annual pay increases. Policies and controls on advancement need to be linked to performance rather than seniority, and in order to resolve the tension with wage bill flexibility, performance-related pay should not be incorporated into base pay. However, these conditions have rarely been met. By the end of the 1990s most OECD countries had some form of performance-related pay, but few actually had an extensive system (OECD, 2005). Many systems focus only on senior civil servants, in order to minimize the costs of implementation and maximize the value for money of substantial performance rewards. In cases where more junior staff are included in the performance-related pay system, they tend to have more modest incentives.

Box 25.3Nicaragua (2000–05)—institutions and wage spending

Wage setting in Nicaragua is complex and fragmented. Before 2003, the determination of wages in public institutions was entirely discretionary, and mainly determined at the spending agency level. A 2003 civil service law created a centralized pay scale based on job complexity. However, this pay scale does not apply to specialist sectors (such as health and education employees) and covers only 20 percent of central government employment. The specialist groups have separate negotiated wage settlements.

Arrangements regarding decentralized entities reduce control and transparency, and are a source of wage inflation. Wage bills in these institutions (for example, schools) are recorded as current transfers and are not included in the government’s wage bill. Decentralized entities set their own compensation levels with no reference to the ministry of finance. Their wage levels are twice that of the rest of the civil service and have seen the largest wage increases in recent years.

The legislature can, and does, override executive wage policy after its own negotiations with public sector trade unions. Direct wage negotiations between the strong public sector groups excluded from the civil service law and the assembly have resulted in large, unplanned wage increases.

A rigid labor code constrains flexibility. The code treats all benefits that have been granted as an acquired right of beneficiaries and also contains provisions that constrain the ability of the public sector to reduce employment. These provisions have caused particular problems with respect to donor projects where salary supplementation and project-based employment is forced to continue beyond the expiry of projects.

The result has been upward pressure on the total wage bill. Between 2000 and 2005, the formal wage bill remained around 4 percent of GDP, but total compensation of public employees increased from 7.5 percent to 8.5 percent of GDP. Over the same period, public sector wages increased by 34 percent while average wages in the formal private sector dropped by 6 percent.

The constrained institutional capacity and the sometimes volatile political environment add complexity to performance-based civil service reform in LICs. Even in the OECD countries, with their more sophisticated and effective institutions, there has been a tendency to overrate the potential of performance-based approaches to change behavior and culture (OECD, 2005). Reforms aimed at increasing managerial flexibility, as described above, need to be balanced with adequate control. This requires strong central management and coordination, and often a continuation of direct centralized control over at least some wage and employment decisions. Against this background, there are a number of questions that need to be addressed in LICs before undertaking a reform program along the lines pursued by OECD countries:

  • To what extent is a national public service with smooth mobility between agencies required or desired? The greater the degree of differentiation in pay scales and job ratings, and the greater the level of delegated pay setting, the less meaningful becomes the notion of a national public service.

  • To what extent is pay competition between government agencies desirable? Unconstrained competition between agencies for the best staff is unlikely to result in an optimal distribution of labor from the viewpoint of the government. Therefore, careful regulation of a system of delegated pay bargaining is essential (see World Bank, 2002).

  • What mechanisms will be available for the center to influence and control the wage bill at both the aggregate and the agency level? The wage bill remains a key macroeconomic and structural variable over which the central agencies cannot abandon their control, even with delegated wage setting. Any such system needs to recognize that, and retain central powers of oversight and, when necessary, intervention in agency level policies and agreements.

  • Will selective labor market-related pay rises put upward pressure on the aggregate wage bill? Revising pay scales for particular high-profile groups—for instance, nurses or teachers—appears a sensible way to improve the competitiveness of key public sector wages in a situation where resources are limited. However, there is a possibility, particularly where public sector unions are strong and the government is politically vulnerable, that this will create a benchmark for the rest of the public service to follow in subsequent pay negotiations.

  • Will the introduction of more discretion and differentiation in pay increase patronage and corruption in the public service? Care needs to be taken to align the incentives and sanctions for performance with the level of delegation of pay setting and advancement decisions. In the highly politicized environment that many LIC civil services operate in, it is quite possible that well-intentioned pay reforms can serve to decrease rather than increase the productivity of the public service by increasing the scope for politically motivated, rather than efficiency motivated, decisions on pay and promotion.

Civil service reform in LICs has tended to tackle the different types of flexibility in sequence. Attempts at civil service reform in LICs can be separated into first and second generation reforms (see Stevens and Teggemann, 2004; Lienert and Modi, 1997). First generation reforms, which were often components of the structural adjustment programs of the late 1980s and early 1990s, focused almost exclusively on the macro-fiscal objective of controlling the wage bill. The main techniques employed were retrenchment, civil service censuses, and removal of ghost workers from the payroll. The focus on macroeconomic stabilization meant that real wages, despite their fall over the previous decades, were often further decreased. By contrast, second generation reforms focus more on a differentiated approach to pay, in particular concentrating on wage decompression; reform of government functions and machinery; and improving performance and incorporating other areas of public sector reform, notably public finance management systems. Although success has been achieved in individual cases (Box 25.4), overall progress with these reforms has been modest.

In practice, the process of civil service reform in LICs has not extended to performance budgeting. Roberts (2003) assesses seven country case studies of the implementation of performance budgeting in LICs. He notes that in practice, decision-making flexibility has been very limited below ministerial levels, and thus the ability for managers to vary inputs, including labor inputs, did not materialize. For instance: non-wage resource envelopes at the agency level were squeezed by central pay agreements in Tanzania (Ronsholt et al., 2003); central hiring and firing regulations inhibited changing the skill mix in Uganda (Williamson, 2003); also in Uganda, low real wages led to only partial employment of the strongest performers (Williamson, 2003);14 and in all seven countries of the sample, pay and promotion mechanisms did not provide the ability to reward good performers (Roberts, 2003).

But this does not mean that implementing performance budgeting models in LICs is a futile task. However, realism is required as to the extent of changes in the level and quality of outputs and outcomes that can be achieved through performance budgeting where significant wage spending inflexibilities exist. But by providing a framework for improving accountability and incentives, improving the flexibility and quality of wage and employment policies would have its own benefits, as well as establishing critical requirements for the introduction of performance budgeting in the future. There are a number of civil service reform mechanisms that have the potential for allowing managers more flexibility within a performance budgeting environment without placing unacceptable risks to macroeconomic control. These include: salary top-ups; delegation of promotion decisions within a rigid grade structure (to avoid unintended upward drift in average salaries); and selective pay scale revisions that reflect Poverty Reduction Strategy Paper priorities and respond to external labor market pressures.

Box 25.4Tanzania (1999–2004)—signs of success

The civil service department (CSD) and the public service commission (PSC) administer a traditional national civil service system with national pay scales. In agencies where measures to monitor and improve performance within a hard budget constraint have been established, measures to move away from the centralized pay and employment environment have been encouraged.

The flexibility of national pay scales has been increased by an experimental salary enhancement scheme. Donors are funding a salary top-up scheme to reward strong performers in key positions, with salary enhancements of up to 25 percent. This level of top-up was designed to be fiscally affordable when donor funding expires. However, implementation of the system was slower than envisaged, and government contributions have been slow to materialize.

Executive agencies have been freed from central personnel policies. Agencies with substantial own revenues have been granted executive agency status. There are 19 of these agencies that, on the basis of a three-year plan with clear performance indicators and audit and accountability standards, are released from normal personnel and budgetary procedures. This includes the ability to set their own wage levels and employment policies.

Administrative reforms have provided cost savings to finance the process. A computerized integrated personnel and payroll system was established in 2000. This has delivered substantial cost savings by identifying ghost workers and duplicate salaries, and preventing fraud. The sound records management provided by this system has also underpinned the salary enhancement scheme.

The political environment has been favorable. The stable and relatively non-competitive political environment with weak trade unions provided ample space for the political level to provide the strong backing required for effective civil service reform. The recent advent of a somewhat more competitive political environment and stronger trade unions has created new issues, in particular an increase in allowances which placed upward pressure on the wage bill.

The wide-ranging public sector reform program provides indications of the type of mechanisms that may be useful in increasing wage-bill flexibility alongside performance budgeting. This has been achieved without creating undue macro-fiscal pressure. In fact, Tanzania has had some success in raising real wage levels and improving incentives while keeping the wage bill stable at around 4 percent of GDP (although further real wage increases planned for coming years are expected to put an upward pressure on the wage bill).

The prevailing political and institutional environment will clearly condition the extent to which the type of mechanisms utilized by OECD countries can be adopted in LICs. Even in the OECD, “modernization depends on context” (OECD, 2005) and a particular mechanism is not readily transferable between different countries. Reforms that increase the flexibility of pay determination heighten the risk of reduced control of the overall level of the wage bill, particularly if the central agencies are not strong enough or politically empowered to enforce crucial elements of the overall package. At the microeconomic level, increasing managerial flexibility in the absence of an effective accountability, incentives, and control package run the risk of intensifying existing problems of political patronage and interference in the civil service. Widespread delegated pay bargaining or running cost budgets are therefore unlikely to be a viable solution to resolving wage inflexibilities in most LICs.

Future attempts at civil service reform in LICs need to draw heavily on the lessons learned from second generation reforms. These lessons are relevant, both for control of the aggregate wage bill and improving performance and flexibility at an agency level:

  • A favorable political environment is the most important prerequisite for civil service reform (Stevens and Teggemann, 2004). The government needs political space to push through reforms that will inevitably be unpopular in some quarters. Attempting to embark on a substantial reform program in a volatile political climate or with a fragile administration is likely to end in disappointment.

  • Establishment of a strong information base should precede the design of reforms. An accurate and comprehensive measure of the wage bill and its components is a vital starting point for any program of reform. The measure of the wage bill should include all labor-related costs including base salary, top-ups, allowances, pensions, and costs of contractual labor. Other information should include the composition of the civil service by employment status, grade, occupation, and age. Without this basic information it is impossible to know how structural reforms may affect the ultimate wage bill. Information on private sector remuneration is also important in order to be able to judge the effect that reforms may have on the competitiveness of government wages.

  • A realistic assessment of the institutional context is vital. It is crucial that the reform program components are appropriate, technically sound, and well-sequenced (Stevens and Teggemann, 2004). Attempting to reform the wage bill without addressing the institutional processes that underlie it is likely to result in failure. In particular, it is unlikely that a move toward fixed-term contracts and performance-based pay will be feasible or productive in the institutional context of most LICs (World Bank, 2002). Institutional processes that, for example, promote wage increases for selected groups of civil servants that spill over into wage demands by other groups need to be considered up front, rather than being dealt with as a complication during the reform process.

  • A reform program needs to be set in a realistic and funded medium-term framework. Budget constraints for the wage bill, including proposed increases to salary scales or top-ups, need to be realistic in both the short and medium term and have identified financing.

  • Appropriate administrative structures for the management of the public service are crucial, and dismantling centralized institutions for the management of wage and personnel policy may not be feasible in the short term. Institutional structures are important and need to be embedded in the core public service (Stevens and Teggemann, 2004). In some African countries (for example, Uganda) this has meant a return to a centralized and semi-independent civil service commission after they were abolished. This has helped to embed the new system and provide a bulwark against political interference in public sector personnel decisions.

  • Implementing performance management systems may not be the most effective means of improving performance. A World Bank (2003) study of three transition countries (Slovak Republic, Romania, and Kyrgyz Republic) and their differing levels of development and public administration legacies found that meritocracy was the only consistent influence on overall public sector performance. This was particularly so in the two more advanced countries. It also found that a well-functioning system of administrative procedures was crucial in laying the framework for meritocracy. In contrast, the implementation of performance management systems was found to be irrelevant to civil service performance, particularly in the absence of a well-functioning and meritocratic administrative framework.


We gratefully acknowledge invaluable comments from Marc Robinson and participants at the Seminar on Performance Budgeting held at IMF Headquarters (Washington, DC), December 5–7, 2005. All remaining errors, of course, are ours.

Although this is not necessarily the case—a reduction in maintenance spending may increase the need for maintenance and investment spending in the future.

Kraay and van Rijckeghem (1995) note the role of self-interested public sector employees in long-term wage bill trends. They develop an insider-outsider model of wage negotiation which suggests that public sector employees have the incentive to restrict the growth in employment to ensure that their wage increases and rents are maximized.

However, based on a comprehensive cross-country empirical analysis including developed as well as developing countries, Schiavo-Campo et al. (1997) find evidence of a positive relationship between government employment and GDP levels. They also find an association between high wage levels and low levels of government employment, and weak support for positive correlations between deficits and employment.

Schiavo-Campo et al. (1997), using slightly different techniques, found a negative relationship between wages and employment and weak support for positive correlations between deficits and employment.

Kraay and van Rijckeghem (1995) considered the situation in the mid-1990s, when structural adjustment programs supported by the IMF’s Structural Adjustment and Enhanced Structural Adjustment Facilities (SAF and ESAF) focused on medium-term measures aimed at structural reforms, including civil service reforms. IMF-supported stabilization (Stand-By) arrangements generally are of a short duration, and their policy instruments are limited to those that are macroeconomic in nature.

Fedelino et al. (2006) also conclude that there is no evidence that such wage bill ceilings have limited government wage spending when donors have made additional resources available.

Primary expenditure excludes interest spending, the major item of non-discretionary expenditure in most countries. This allows the analysis to focus on items of expenditure where there is at least some discretion for the government in setting expenditure levels in response to changing macro-fiscal circumstances

A similar strategy for assessing the flexibiliy of spending is employed by Mattina and Gunnarsson (forthcoming).

But, over the whole sample, changes in the average wage bill over the period are not statistically different from zero at the 5 percent significance level.

The convergence of outliers to the middle we observe in our dataset resembles mean reversal, the stochastic process in which outliers tend to return to a common population mean. This suggests that the convergence in wage spending in our sample may be the result of a statistical phenomenon and not deliberate policy choices. However, the key assumption for the observed data trends to be driven by mean reversal is that wage spending across countries would have a common, or “natural,” mean and that movements around this mean could be represented as a stochastic process. This key assumption seems obviously not to be met in this case.

Shortcomings of the dataset prevent a substantive analysis of whether movements in wage bills are driven by employment or wage changes. The dataset used for this study does not include wage level and employment data. However, findings from the literature suggest that countries have made adjustments to real wage levels to keep the wage bill under control (for example, by letting nominal wage increases lag behind inflation—see Lienert and Modi, 1997). Our sample provides some support for the conjecture that inflation and wage bill adjustment are linked. The correlation between annual consumer price inflation rates and the change in the wage bill as a share of GDP is -0.17 and that between inflation and the wage bill as a share of primary spending is -0.13. Both correlations are statistically significant at the 10 percent confidence level.

A striking example is the wide-ranging public sector reforms in New Zealand in the late 1980s, which in many ways led the way for the implementation of performance budgeting in OECD and other countries. The impetus for these reforms stemmed from a significant fiscal crisis that required major reforms both in the level of fiscal balances and the effectiveness with which public resources were utilized.

This is despite Uganda achieving one of the most drastic wage decompressions in sub-Saharan Africa. See Kiragu et al. (2004).


    BaumolW.J.1967Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis,American Economic Review Vol. 57 (3) pp. 41526.

    FedelinoA.G.Schwartz and M.Verhoeven2006Aid Scaling Up: Do Wage Bill Ceilings Stand in the Way?IMF Working Paper 06/106 (Washington: International Monetary Fund).

    GuptaS.B.ClementsE.Baldacci and C.Mulas-Granados2002Expenditure Composition, Fiscal Adjustment, and Growth in Low-Income Countries,IMF Working Paper 02/77 (Washington: International Monetary Fund).

    KatzL. and A.Krueger1993Public Sector Pay Flexibility: Labor Market and Budgetary Considerations,” in Pay Flexibility in the Public Sector OECD Public Management Studies (Paris: OECD).

    KiraguK.M.Rwekaza and M.Denyse2004Reforming Pay Policy: Techniques, Sequencing and Policies,” in Building State Capacity in Africa: New Approaches Emerging Lessonsed. B.Levy and S.Kpundeh (Washington: World Bank).

    KraayA. and C.van Rijckeghem1995Employment and Wages in the Public Sector—A Cross Country Study,IMF Working Paper 95/70 (Washington: International Monetary Fund).

    LienertI. and J.Modi1997A Decade of Civil Service Reform in Sub-Saharan Africa,IMF Working Paper 97/179 (Washington: International Monetary Fund).

    MaguireM.1993Pay Flexibility in the Public Sector—An Overview,” in Pay Flexibility in the Public Sector OECD Public Management Studies (Paris: OECD).

    ManningC2000Labor Market Adjustment to Indonesia’s Economic Crisis: Context, Trends and Implications,Bulletin of Indonesian Economic Studies Vol. 36(1) pp. 10536.

    MarsdenD.1993Reforming Public Sector Pay,” in Pay Flexibility in the Public Sector OECD Public Management Studies (Paris: OECD).

    MattinaT. and V.GunnarssonforthcomingBudget Rigidity and Expenditure Efficiency in Slovenia,IMF Working Paper (Washington: International Monetary Fund).

    MeursD.1993The Rationale for, and Implications of. Centralized Pay Determination Systems in the Public Sector,” in Pay Flexibility in the Public Sector OECD Public Management Studies(Paris: OECD).

    OECD1993Private Pay for Public Work: Performance Related Pay for Public Sector Managers (Paris: OECD).

    OECD1995Performance-Related Pay Policies for Government Employees (Paris: OECD).

    OECD2005Modernising Government: The Way Forward (Paris: OECD).

    RobertsJ.2003Managing Public Expenditure for Development Results and Poverty Reduction,ODI Working Paper 203 (London: Overseas Development Institute).

    RonsholtF. with R.MushiB.Shallanda and P.Assey2003Results-Orientated Expenditure Management Country Study—Tanzania,ODI Working Paper 204 (London: Overseas Development Institute). Tommaso and A.Mukherkee1997Government Employment and Pay: A Global and Regional Perspective,World Bank Policy Research Paper 1771 (Washington: World Bank).

    StevensM. and S.Teggemann2004Comparative Experience with Public Sector Reform in Ghana, Uganda and Tanzania,” in Building State Capacity in Africa: New Approaches Emerging Lessonsed. B.Levy and S.Kpundeh (Washington: World Bank).

    WilliamsonT.2003Targets and Results in Public Sector Management: Uganda Case Study,ODI Working Paper 205 (London: Overseas Development Institute).

    World Bank2002Civil Service Reform: Strengthening World Bank and IMF Collaboration(Washington: World Bank).

    World Bank2003PREM Unit, European and Central Asia Region2003Understanding Public Sector Performance in Transition Countries—An Empirical Contribution (Washington: World Bank).

    Other Resources Citing This Publication