1. Factors influencing public sector employment and wages

2. An efficiency wage model of the public sector

In this section, we present a simple model of public sector employment and wage determination. The primary purpose of this model is to derive the optimal employment and wage policy of the government when the labor market is characterized by efficiency wages, and to show how these policies are influenced by unemployment and fiscal constraints. We assume a public sector which produces public goods using efficiency units of labor, and a private sector which produces a final consumption good using efficiency units of labor and the public good. The government levies lump-sum taxes on the private sector to finance the public sector wage bill and must balance its budget. The government’s objective is to maximize a weighted average of after-tax incomes and employment. In this stylized world, we show that (a) high public sector employment and low wages may be the outcome of optimizing behavior by government, (b) the existence of an employment objective by the government always lowers public sector wages and increases the public-private employment ratio, and (c) fiscal constraints may or may not affect the optimal choice of wages, depending on whether or not the government has employment objectives, while they do affect the choice of government employment.

There are two reasons for choosing an efficiency wage framework. First, it allows us to formalize the idea that wages affect effort. Examples of this linkage are the moonlighting and corruption commonly associated with declining real average civil service wages and wage compression. ^10/ Second, from a modelling perspective, efficiency wages in the private sector provide a simple way to introduce unemployment to the model. This permits us to examine how the government’s behavior changes when, for policy reasons, it tries to reduce unemployment via job creation programs.

On the other hand, we abstract from issues of labor supply and the choice individuals make to seek private versus public sector jobs, by simply assuming that labor, although not effort, is inelastically supplied. This is not really problematic, as it is likely that governments never face a shortage of workers, although they often face a shortage of motivated or skilled workers. We assume further that there is a sufficiently large labor force that both the public and private sectors are always on their labor demand curves.

Formally, suppose that the government sector produces public services, G, using efficiency units of labor, e_G(w_G)L_G, where w_G and L_G are wages and employment in the government sector, and e_G is an effort function with e_G>0, e_G'>0, e_G''<0. Production of government services is subject to diminishing returns, and the production function is assumed to be of the form:

where 0<α<1. The government levies lump-sum taxes, T, on the private sector, with collection costs C=øT and the balanced-budget requirement implies that w_GL_G=(1-ø)T.^11/

The private sector also employs efficiency units of labor to produce final output Y, where the effort function e_p(w_p) may differ from that of the government sector. The key assumption here is that firms take the level of government services as given in their employment and wage decision. Output in the private sector is thus given by

where K is a constant reflecting the scale of output.

We assume that the government maximizes a linear combination of output net of tax collection costs (Y-øT) ^12/ and employment, so that its problem

We consider two variants on this problem. In the first, presented below, only the government sets efficiency wages; the private sector sets the wage equal to the marginal product of labor. Because there is no unemployment in this case, the government employment objective has no impact. In the second, which we relegate to Appendix I, both government and private sector pay efficiency wages. In this case, there is unemployment, which the government tries to reduce. For each variant, we consider two cases. In the first, the government faces no constraints on the revenue it can raise from taxation, while in the second case, it faces an exogenous revenue constraint. The first case serves as a benchmark of optimal behavior, while the second allows us to consider the deviations from this employment and wage benchmark caused by fiscal pressures.

1. Levels and trends

For the developing countries in our sample, average central government employment per thousand population during 1972-1992 was 22. ^13/ Over the same period, central government employment constituted about 28 percent of formal sector employment (Table 1). Employment per thousand population was less in Sub-Saharan Africa (18 per thousand), as one would expect based on levels of income and development. Central government employment in Sub-Saharan Africa constituted a higher share of formal sector employment (35 percent), though care is necessary in interpreting this result because of the quality of the data used for formal sector employment. ^14/ General government employment per thousand population was 71 in our sample of OECD countries –much above the average in developing countries. Only some of the difference is attributable to the broader definition of government employment for OECD countries. On the conservative assumption that central government constitutes only half of general government employment in developing countries, general government employment in OECD countries would still be almost twice as high as in developing countries. The share of general government employment in total employment, at 17 percent, was lower in OECD countries than the share of central government employment in total formal sector employment in developing countries, reflecting the small degree of formality of employment in developing countries.

Table 1.

Government Employment and Wages: Sample Statistics

(Unweighted averages of country means)

Source: Appendices II and III.

^1/Share in total employment for OECD countries and for Bolivia, Egypt, Indonesia, Kenya, Mauritius Morocco Myanmar and Suriname

Table 1.

Government Employment and Wages: Sample Statistics

(Unweighted averages of country means)

				1972-92	1972-80	1981 -92
Government Employment Per Thousand Population
OECD Countries (general government)
	Mean			71.4	66.4	81.5
	Standard deviation			27.2	26.3	34.3
	Number of countries			21	21	17
Developing Countries (central government)
	Mean			22.4	20.9	22.6
	Standard deviation			26.1	23.8	26.2
	Number of countries			34	27	34
		Sub-Saharan Africa
			Mean	18.2	13.5	18.5
			Standard deviation	23.5	11.2	23.6
			Number of countries	22	16	22
Share of Government Employment in Formal Sector Employment1/
OECD Countries (general government)
	Mean			17.0	15.9	19.2
	Standard deviation			5.5	5.3	6.7
	Number of countries			21	21	17
Developing Countries {central government)
	Mean			28.3	22.9	28.2
	Standard deviation			18.1	12.2	18.2
	Number of countries			23	16	23
		Sub-Saharan Africa
			Mean	34.9	27.6	34.8
			Standard deviation	15.3	9.2	15.2
			Number of countries	13	9	13
Ratio of Government Wages to GDP Per Capita
OECD Countries (general government)
	Mean			1.40	1.54	1.27
	Standard deviation			0.46	0.50	0.39
	Number of countries			18	16	18
Developing Countries (central government)
	Mean			5.00	4.64	4.77
	Standard deviation			4.17	3.42	4.25
	Number of countries			27	20	27
		Sub-Saharan Africa
			Mean	5.87	5.87	5.60
			Standard deviation	4.63	3.79	4.76
			Number of countries	19	12	19

Source: Appendices II and III.

^1/Share in total employment for OECD countries and for Bolivia, Egypt, Indonesia, Kenya, Mauritius Morocco Myanmar and Suriname

View Table

Table 1.

Government Employment and Wages: Sample Statistics

(Unweighted averages of country means)

				1972-92	1972-80	1981 -92
Government Employment Per Thousand Population
OECD Countries (general government)
	Mean			71.4	66.4	81.5
	Standard deviation			27.2	26.3	34.3
	Number of countries			21	21	17
Developing Countries (central government)
	Mean			22.4	20.9	22.6
	Standard deviation			26.1	23.8	26.2
	Number of countries			34	27	34
		Sub-Saharan Africa
			Mean	18.2	13.5	18.5
			Standard deviation	23.5	11.2	23.6
			Number of countries	22	16	22
Share of Government Employment in Formal Sector Employment1/
OECD Countries (general government)
	Mean			17.0	15.9	19.2
	Standard deviation			5.5	5.3	6.7
	Number of countries			21	21	17
Developing Countries {central government)
	Mean			28.3	22.9	28.2
	Standard deviation			18.1	12.2	18.2
	Number of countries			23	16	23
		Sub-Saharan Africa
			Mean	34.9	27.6	34.8
			Standard deviation	15.3	9.2	15.2
			Number of countries	13	9	13
Ratio of Government Wages to GDP Per Capita
OECD Countries (general government)
	Mean			1.40	1.54	1.27
	Standard deviation			0.46	0.50	0.39
	Number of countries			18	16	18
Developing Countries (central government)
	Mean			5.00	4.64	4.77
	Standard deviation			4.17	3.42	4.25
	Number of countries			27	20	27
		Sub-Saharan Africa
			Mean	5.87	5.87	5.60
			Standard deviation	4.63	3.79	4.76
			Number of countries	19	12	19

Source: Appendices II and III.

^1/Share in total employment for OECD countries and for Bolivia, Egypt, Indonesia, Kenya, Mauritius Morocco Myanmar and Suriname

While caution is required in interpreting the results because of variation in the countries included in the sample over time, the simple averages seem to indicate that there is a positive trend in both employment per thousand population and employment as a share of formal sector employment. Comparing 1972-1980 and 1981-1992, we find an increase in employment per thousand population from 21 per thousand to 23 per thousand in developing countries. Employment as a percent of formal sector employment increases from 23 to 28 percent. These trends are also present for Sub-Saharan Africa and the OECD. They are consistent with a number of views, including the view that bureaucracies expand over time and that governments hire counter-cyclically, though they are not consistent with the expected response to the debt crisis. ^15/ In the next sub-section, we confirm the presence of a positive time trend in government employment for developing countries, controlling for the country composition of the sample.

Data for employment at levels of government other than central government are very scarce for developing countries. Central government employment constitutes 44 percent, local government 29 percent, and public enterprise employment 25 percent of total public sector employment for the 9 countries for which data are available on all levels of government. There is, however, a wide variation in the relative importance of the different levels of government across these 9 countries. The variation in the relative importance of central and local governments is primarily due to the system of government (federal versus centralized). The centralized governments of Bolivia, Botswana, Egypt, and Kenya have relatively large central government employment (51 percent of the total public sector employment on average). The federal governments of Argentina, India, and Nigeria have relatively large local government employment (49 percent of the total on average). Countries with large public enterprise employment are Gabon, Myanmar, and Nigeria (38 percent of the total on average) (Appendix III, Table 1).

It is also possible to examine the correlations between annual employment growth rates at the various levels of government and with the formal private sector. These results are presented in Table 2. Surprisingly, the correlation between central government and formal private sector employment growth is virtually zero, contrary to what would be implied by counter-cyclical hiring on the part of the central government. The correlation between local and central government employment growth is positive at 0.22, suggesting that the phenomenon of devolution of responsibility to local governments is not very significant, but instead that similar forces contribute to the growth in employment at both the central and local government level.

Table 2.

Developing Countries: Public and Private Sector Employment: Relative Employment and Correlations in Employment Growth

Source: Appendices II and III.

^1/Standard deviation in parentheses; number of observations below parentheses.

^2/Correlations are for annual growth rates less than 30 percent in absolute value; number of observations in parentheses.

^3/Total private sector employment for Bolivia, Egypt, Indonesia, Kenya, Mauritius, Morocco, Myanmar and Suriname.

Table 2.

Developing Countries: Public and Private Sector Employment: Relative Employment and Correlations in Employment Growth

	Central Government	Local Government	Public Enterprises	Private Sector
Relative Employment in the Public Sector1/
(Denominator)
Central Government	1
Local Government	6.59	1
	(7.52)
	148
Public Enterprises	1.74	1.38	1
	(1.15)	(1.44)
	179	103
Formal Private Sector 3/	0.59	0.19	0.64	1
	(0.65)	(0.35)	(0.72)
	252	111	120
Correlation in Employment Growth Across Sectors 2/
Central Government	1
Local Government	0.22	1
	(129)
Public Enterprises	0.08	0.44	1
	(155)	(90)
Formal Private Sector 3/	-0.001	0.06	0.16	1
	(214)	(98)	(106)

Source: Appendices II and III.

^1/Standard deviation in parentheses; number of observations below parentheses.

^2/Correlations are for annual growth rates less than 30 percent in absolute value; number of observations in parentheses.

^3/Total private sector employment for Bolivia, Egypt, Indonesia, Kenya, Mauritius, Morocco, Myanmar and Suriname.

View Table

Table 2.

Developing Countries: Public and Private Sector Employment: Relative Employment and Correlations in Employment Growth

	Central Government	Local Government	Public Enterprises	Private Sector
Relative Employment in the Public Sector1/
(Denominator)
Central Government	1
Local Government	6.59	1
	(7.52)
	148
Public Enterprises	1.74	1.38	1
	(1.15)	(1.44)
	179	103
Formal Private Sector 3/	0.59	0.19	0.64	1
	(0.65)	(0.35)	(0.72)
	252	111	120
Correlation in Employment Growth Across Sectors 2/
Central Government	1
Local Government	0.22	1
	(129)
Public Enterprises	0.08	0.44	1
	(155)	(90)
Formal Private Sector 3/	-0.001	0.06	0.16	1
	(214)	(98)	(106)

Source: Appendices II and III.

^1/Standard deviation in parentheses; number of observations below parentheses.

^2/Correlations are for annual growth rates less than 30 percent in absolute value; number of observations in parentheses.

^3/Total private sector employment for Bolivia, Egypt, Indonesia, Kenya, Mauritius, Morocco, Myanmar and Suriname.

Wages at the central government level have on average been 5 times as large as GDP per capita, and they have been a particularly large multiple (5.9) in Sub-Saharan Africa (Table 1). The higher multiple in Sub-Saharan Africa could be explained by a high level of productivity of civil servants (who have scarce human capital) relative to that of the population (which is mostly agricultural); it could also reflect large rents. In OECD countries, wages as a multiple of GDP per capita were much lower, 1.4 in our sample. Although the developing country averages for the 1970s and 1980s seem to indicate that wages relative to GDP were rising, this result is due primarily to differences in the sample of countries included in the average for these periods. In the next sub-section, we give evidence that within countries, there has been a relatively consistent negative time trend in wages relative to GDP per capita, as would be consistent either with declining relative scarcity of human capital in the civil service during the course of development or erosion of civil service pay over time.

Wage comparisons between different levels of government and with the private sector are especially difficult, given the scarcity of data and variety in definitions. Moreover, comparison of average wages ignores differences between types of jobs at the same level of government, as well as important sectoral differences in both the public and private sector. We confine ourselves to the presentation of some averages and correlations of wage growth rates in Table 3. On average, central government employees command a wage premium of 22 percent over their counterparts in local governments, 10 percent over those in public enterprises and 9 percent over the private sector. Perhaps the most striking finding emerges from the comparison of real wage growth rates in the central government and in the private sector. ^16/ The simple correlation is positive, but quite low at 0.13, Local government and public enterprise wage growth rates are more strongly correlated with private sector wage growth (0.49 and 0.43 respectively), although the corresponding sample sizes are much smaller, and results should be interpreted cautiously.

Table 3.

Developing Countries: Public and Private Sector Wages: Relative Wages and Correlations in Wage Growth

Source: Appendices II and III.

^1/Standard deviation in parentheses; number of observations below parentheses.

^2/Correlations are for annual growth rates less than 20 percent in absolute value; number of observations in parentheses.

Table 3.

Developing Countries: Public and Private Sector Wages: Relative Wages and Correlations in Wage Growth

	Central Government	Local Government	Public Enterprises	Private Sector
Relative Waaes in the Public Sector1/
(Denominator)
Central Government	1
Local Government	1.22	1
	(0.38)
	70
Public Enterprises	1.1	0.8	1
	(0.34)	(0.1)
	69	33
Private Sector	1.09	1	1.21	1
	(0.54)	(0.23)	(0.33)
	169	71	59
Correlation in Real Waqe Growth Across Sectors 2/
Central Government	1
Local Government	0.38	1
	(58)
Public Enterprises	0.56	0.4	1
	(52)	(27)
Private Sector	0.13	0.49	0.43	1
	(95)	(41)	(27)

Source: Appendices II and III.

^1/Standard deviation in parentheses; number of observations below parentheses.

^2/Correlations are for annual growth rates less than 20 percent in absolute value; number of observations in parentheses.

View Table

Table 3.

Developing Countries: Public and Private Sector Wages: Relative Wages and Correlations in Wage Growth

	Central Government	Local Government	Public Enterprises	Private Sector
Relative Waaes in the Public Sector1/
(Denominator)
Central Government	1
Local Government	1.22	1
	(0.38)
	70
Public Enterprises	1.1	0.8	1
	(0.34)	(0.1)
	69	33
Private Sector	1.09	1	1.21	1
	(0.54)	(0.23)	(0.33)
	169	71	59
Correlation in Real Waqe Growth Across Sectors 2/
Central Government	1
Local Government	0.38	1
	(58)
Public Enterprises	0.56	0.4	1
	(52)	(27)
Private Sector	0.13	0.49	0.43	1
	(95)	(41)	(27)

Source: Appendices II and III.

^1/Standard deviation in parentheses; number of observations below parentheses.

^2/Correlations are for annual growth rates less than 20 percent in absolute value; number of observations in parentheses.

2. Wages, employment, income, and time

The simple averages presented in the previous sub-section are problematic, since the time-averages mask time-series variation while the cross-country averages mask the cross-country variation in employment and wages. To remedy this deficiency, we estimate simple regressions of the within-country variation of employment and wages on time, and the crosscountry variations of employment and wages on real income per capita. The results, presented in Table 4, should be interpreted only as simple correlations.

Table 4.

Developing Countries: Central Government Employment and Wages: Co-movements With Time and Real Income Per Capita

^1/Coefficient on Time In Fixed-Effects Regression of dependent variable on constant and time trend.

^2/Coefficient on Income In OLS Regression of mean dependent variable on constant and mean of Real GDP Per Capita.

^3/T- statistic below estimate, In parentheses.

Table 4.

Developing Countries: Central Government Employment and Wages: Co-movements With Time and Real Income Per Capita

	Within-Country Variation Over Time 1/3/						Cross-Country Variation With Real GDP Per Capita (1985 US. Dollars) 2/3/
	Employment Per Thousand			Wages/GDP per Capita			Employment Per Thousand		Wages GDP per Capita
	Coefficient on Time	Number of		Coefficient on Time	Number of		Coefficient on Income Per capita	Number of Countries	Coefficient on Income Per capita	Number of Countries
		Countries	Observations		Countries	Observations
1972-1992	0.10	34	402	-0.08	27	305	0.02	34	-0.53	27
	(2.44)			(-4.72)			(5.53)		(-5.05)
GDP Per Capita < $777	0.02	26	252	-0.10	21	206	0.04	26	-0.58	21
	(0.49)			(-4.12)			(3.68)		(-232)
GDP Per Capta > $777	0.10	12	150	0.00	9	99	0.02	12	-0.38	9
	(1.06)			(-0.14)			(1.87)		(-1.52)
1972-1981	0.33	29	153	-0.25	21	107	0.01	29	-0.52	21
	(294)			(-7.42)			(3.36)		(-5.31)
GDP Per Capita < $777	0.40	22	100	-0.29	16	79	0.04	22	-0.15	16
	(6.38)			(-7.77)			(3.57)		(-1.25)
GDP Per Capita > $777	0.07	10	53	0.01	7	28	0.01	10	-0.42	7
	(0.21)			(0.40)			(0.73)		(-1.83)
1982-1992	0.14	34	249	0.05	27	198	0.02	34	-0.46	27
	(200)			(1.44)			(1246)		(-3.33)
GDP Per Capita < $777	-0.11	23	152	0.10	19	127	0.02	23	-0.50	19
	(-1.60)			(1.60)			(230)		(-208)
GDP Per Capita > $777	0.45	12	97	-0.01	9	71	0.02	12	-0.34	9
	(3.50)			(-0.28)			(216)		(-1.26)

^1/Coefficient on Time In Fixed-Effects Regression of dependent variable on constant and time trend.

^2/Coefficient on Income In OLS Regression of mean dependent variable on constant and mean of Real GDP Per Capita.

^3/T- statistic below estimate, In parentheses.

View Table

Table 4.

Developing Countries: Central Government Employment and Wages: Co-movements With Time and Real Income Per Capita

	Within-Country Variation Over Time 1/3/						Cross-Country Variation With Real GDP Per Capita (1985 US. Dollars) 2/3/
	Employment Per Thousand			Wages/GDP per Capita			Employment Per Thousand		Wages GDP per Capita
	Coefficient on Time	Number of		Coefficient on Time	Number of		Coefficient on Income Per capita	Number of Countries	Coefficient on Income Per capita	Number of Countries
		Countries	Observations		Countries	Observations
1972-1992	0.10	34	402	-0.08	27	305	0.02	34	-0.53	27
	(2.44)			(-4.72)			(5.53)		(-5.05)
GDP Per Capita < $777	0.02	26	252	-0.10	21	206	0.04	26	-0.58	21
	(0.49)			(-4.12)			(3.68)		(-232)
GDP Per Capta > $777	0.10	12	150	0.00	9	99	0.02	12	-0.38	9
	(1.06)			(-0.14)			(1.87)		(-1.52)
1972-1981	0.33	29	153	-0.25	21	107	0.01	29	-0.52	21
	(294)			(-7.42)			(3.36)		(-5.31)
GDP Per Capita < $777	0.40	22	100	-0.29	16	79	0.04	22	-0.15	16
	(6.38)			(-7.77)			(3.57)		(-1.25)
GDP Per Capita > $777	0.07	10	53	0.01	7	28	0.01	10	-0.42	7
	(0.21)			(0.40)			(0.73)		(-1.83)
1982-1992	0.14	34	249	0.05	27	198	0.02	34	-0.46	27
	(200)			(1.44)			(1246)		(-3.33)
GDP Per Capita < $777	-0.11	23	152	0.10	19	127	0.02	23	-0.50	19
	(-1.60)			(1.60)			(230)		(-208)
GDP Per Capita > $777	0.45	12	97	-0.01	9	71	0.02	12	-0.34	9
	(3.50)			(-0.28)			(216)		(-1.26)

^1/Coefficient on Time In Fixed-Effects Regression of dependent variable on constant and time trend.

^2/Coefficient on Income In OLS Regression of mean dependent variable on constant and mean of Real GDP Per Capita.

^3/T- statistic below estimate, In parentheses.

The first two equations show the correlations between central government employment per thousand population and time, and central government wages relative to income per capita and time. The positive time trend in employment can be seen to be much stronger in the 1970s than in the 1980s for the full sample of countries. The correlation with time was much stronger in the 1970s for those countries with real GDP per capita below the sample mean of US$ 777 per capita, than for those above the mean. In the 1980s, this pattern was reversed, with poorer countries actually showing a negative correlation with time, while the richer country correlation remained positive. The data show a negative time trend in wages relative to GDP per capita in the full sample, which is due to a strong negative correlation for the poorer countries in the first half of the sample.

The second two equations show the co-variation of country averages of employment and wages with average real GDP per capita. There is a significant positive correlation between employment and income reflecting the “luxury” good nature of public services and employment. This correlation is particularly significant for the poorer countries in the sample in the 1970s. Finally, the wage equation captures the correlation between income per capita and wages relative to GDP per capita. Across countries, there is a strong negative association between wages relative to income per capita and real income, which is especially strong for the poorer countries in the sample in the 1980s. This points to a possible role for declining relative scarcity of human capital as per capita income increases.

3. A decomposition of changes in the real wage bill

In this section, we informally examine the impact of adjustment programs supported by the IMF on the real wage bill and its components. We are interested in whether real reductions in the wage bill typically occur through reductions in government employment or in real wage rates, and what effect economic adjustment programs supported by the IMF or the World Bankhave on this pattern. ^17/ Figures 1 through 3 plot the percentage change in real central government wages (measured in local currency and deflated by the CPI) against the percentage change in central government employment, by country. All points below the diagonal line correspond to years in which the real wage bill fell, while those above correspond to periods of real growth in the wage bill. ^18/

Central Government Employment and Real Wages Countries During Non-Program Years 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Including non-program countries. Countries include: Argentina (1), Botswana (4), Burkina Faso (5), Cameroon (6), Chad (7), Congo (8), Gabon (10), Gambia (11), Guyana (13), India (15), Kenya (17), Madagascar (18), Malta (21), Mauritius (23), Morocco (24), Myanmar (25), Rwanda (28), Seychelles (30), Somalia (31), Sri Lanka (32), Suriname (34), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Central Government Employment and Real Wages Countries During Non-Program Years 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Including non-program countries. Countries include: Argentina (1), Botswana (4), Burkina Faso (5), Cameroon (6), Chad (7), Congo (8), Gabon (10), Gambia (11), Guyana (13), India (15), Kenya (17), Madagascar (18), Malta (21), Mauritius (23), Morocco (24), Myanmar (25), Rwanda (28), Seychelles (30), Somalia (31), Sri Lanka (32), Suriname (34), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

Central Government Employment and Real Wages Countries During Non-Program Years 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Including non-program countries. Countries include: Argentina (1), Botswana (4), Burkina Faso (5), Cameroon (6), Chad (7), Congo (8), Gabon (10), Gambia (11), Guyana (13), India (15), Kenya (17), Madagascar (18), Malta (21), Mauritius (23), Morocco (24), Myanmar (25), Rwanda (28), Seychelles (30), Somalia (31), Sri Lanka (32), Suriname (34), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

Figure 1 represents countries during periods when they did not have an adjustment program supported by the IMF. Not surprisingly, the majority of the country averages lie to the right of the origin, reflecting the predominant trend of growth in employment. The number of points above and below the horizontal axis reflect the relative frequency of real wage increases and decreases. Remarkably, wage bill declines were always at the expense of real wages, and employment actually increased in most cases (points in the lower-right quadrant and below the diagonal line). This is suggestive of the difficulties which governments have in reducing employment when, in response to fiscal imbalances, the real wage bill must be reduced.

Figures 2 and 3 repeat Figure 1, but only for periods during which short-term stabilization programs (Figure 2) and medium-term structural adjustment programs (Figure 3) were in place. Among countries with short-term stabilization programs, few countries experienced employment declines, notwithstanding the high incidence of declining real wage expenditures. However, several countries did reduce employment while having a medium-term structural adjustment program. ^19/ Below we explore the impact of adjustment programs further through regression analysis.

Central Government Employment and Real Wage During Stabilization Programs 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Countries include: Argentina (1), Cameroon (6), Congo (8), Gabon (10), Guyana (13 India (15), Kenya (17), Madagascar (18), Mauritania (22), Mauritius (23), Morocco (24), Myanmar (25), Senegal (29), Somalia (31), Sri Lanka (32), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Central Government Employment and Real Wage During Stabilization Programs 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Countries include: Argentina (1), Cameroon (6), Congo (8), Gabon (10), Guyana (13 India (15), Kenya (17), Madagascar (18), Mauritania (22), Mauritius (23), Morocco (24), Myanmar (25), Senegal (29), Somalia (31), Sri Lanka (32), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

Central Government Employment and Real Wage During Stabilization Programs 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Countries include: Argentina (1), Cameroon (6), Congo (8), Gabon (10), Guyana (13 India (15), Kenya (17), Madagascar (18), Mauritania (22), Mauritius (23), Morocco (24), Myanmar (25), Senegal (29), Somalia (31), Sri Lanka (32), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

Central Government Employment and Real Wages During Structural Adjustment Programs 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Countries include: Chad (7), Gambia (11), Guyana (13), Kenya (17), Madagascar (18), Mauritania (22), Senegal (29), Somalia (31), Sri Lanka (32), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Central Government Employment and Real Wages During Structural Adjustment Programs 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Countries include: Chad (7), Gambia (11), Guyana (13), Kenya (17), Madagascar (18), Mauritania (22), Senegal (29), Somalia (31), Sri Lanka (32), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

Central Government Employment and Real Wages During Structural Adjustment Programs 1/

Citation: IMF Working Papers 1995, 070; 10.5089/9781451849110.001.A001

1/ Countries include: Chad (7), Gambia (11), Guyana (13), Kenya (17), Madagascar (18), Mauritania (22), Senegal (29), Somalia (31), Sri Lanka (32), Togo (35), Zaire (36).

• Download Figure
• Download figure as PowerPoint slide

1. Choice of explanatory variables

While the above theoretical discussion does not lead to sharply-defined testable implications, it is suggestive of a number of explanatory variables for the determination of public sector employment and wages. In our basic regressions, we consider two groups of explanatory variables. The first group consists of “structural” variables intended to control for basic indicators of the level of development and the labor market situation of each country. The second group of “fiscal” variables are intended to capture some of the effects of budgetary pressures on employment and wage policy (corresponding to the effects of T on wages and employment in the constrained government model above). These variables include revenues and foreign financing. We only include the second group of variables for developing countries, the presumption being that OECD countries can easily determine revenues and foreign financing. Because of the difficulties with data coverage for the developing countries in our sample, the regressors in our basic equations were chosen to ensure as large a sample as possible, and thus often must proxy for a number of different effects, as discussed below. In particular, the scarcity and poor quality of the data on private sector labor-market indicators such as formal private sector employment, unemployment and wages for developing countries prevents us from including these indicators in our regressions.

We do not attempt to estimate underlying structural parameters, but confine ourselves to the estimation of reduced-form relations between employment and wages and a set of explanatory variables. This is because we regard our efficiency wage model as too stylized for estimation of its structural parameters to be meaningful, and because the identification of its parameters would require implausible linearizations of the above relations.

The structural variables we consider are real GDP per capita in 1985 U.S. dollars, ^20/ secondary-school enrollment, the proportion of population living in urban areas, population, a dummy equal to one for federal states, a proxy for private sector employment, and a time trend.

Real GDP per capita captures the resource constraint of the economy. Richer countries are expected to have a higher level of government services. In this context, we are also interested to know whether government services are a luxury good, with an increasing share in total resources as national income increases. The impact of income per capita on employment will depend on the income elasticity of demand as well as the productivity of labor in government employment. ^21/ Income per capita, while capturing the resource constraint of the economy, will be a poor measure of the resource constraint of the government if the tax administration capacity is not sufficiently developed. We take up this issue below.

Secondary-school enrollment is a proxy for abundance of human capital. Secondary-school enrollment can have an impact on government employment, by increasing the need for government employment as a last resort. Also, since relative wages of skilled workers are lower in countries with abundant human capital, and assuming civil service jobs are high-skill jobs, civil service wages will be negatively correlated with abundance of human capital. In countries where civil service jobs require little skill relative to other jobs, the correlation will be positive.

Urbanization is likely to affect both employment and wages. Higher rates of urbanization are likely to create a demand for more publicly-provided urban infrastructure, such as roads and sewers, as well as services such as police protection. While the provision of these services would be more expensive in the country-side because of the lack of returns to scale, ^22/ leading to even greater public employment, urban areas tend to be favored over rural areas in practice. ^23/ Urbanization may also contribute positively to government employment through political pressures caused by the existence of urban, and hence more visible and politically influential, unemployment. However, the direction of the causation could be the reverse, from government employment to urbanization, as workers migrate from rural areas to take advantage of employment opportunities in, for example, the capital, or as improved health services and health enhance the ability of workers to migrate.^24/ On the wage side, urbanization may have a positive effect on the absolute level of government wages, since the alternative wage for workers is better than in a predominantly agricultural setting.

Population has a potential impact on employment, but not on wages. First, countries with large populations tend to have decentralized fiscal systems, and may, thereby, have smaller employment at the central government level. Second, population proxies for scale in the production of government services, and may therefore have a negative impact on employment per thousand population if there are increasing returns to scale.

The federalism dummy captures the fact that in federal states, central government employment can be lower, since responsibility for the provision of public goods may be devolved to local and state levels of government to a greater extent than in centralized states. It is not necessary to include the federalism dummy in the OECD regressions, since data for general government are used. The inclusion of a time trend allows us to test the hypothesis that public sector employment has an intrinsic tendency to grow over time, with the caveat that the time trend also captures the impact of any omitted variables which move over time.

We include either the real effective exchange rate (developing countries) or the ratio of private sector employment to the population (OECD countries) as a proxy for counter-cyclical hiring pressure on the government. The real effective exchange rate ^25/ is expected to be highly correlated with private sector employment, while being much more widely available for developing countries. ^26/ We expect that higher private sector employment (or a lower real effective exchange rate) will lead to both lower government employment and higher wages, because high private employment frees the government from its role as employer of last resort, thereby providing resources for wages.

As fiscal variables, we include a number of factors capturing various aspects of immediate budgetary pressures. The budget deficit/GDP ratio is the most obvious indicator, the presumption being that high deficits trigger fiscal restraint and lower employment and wages. However, the direction of causation is ambiguous, as high employment and wages could lead to higher deficits. ^27/ We therefore include the total revenue - to-GDP ratio, as a measure of fiscal pressure, in lieu of the deficit, in developing country regressions. This variable can be considered exogenous in developing countries, where the structure of the economy and the weakness of tax administration implies that the government is not free to choose an optimal tax burden. We omit this variable in the OECD regressions, because decisions on the size of the tax burden and of civil service employment and wages are likely to be made simultaneously in those countries. ^28/

We include the ratio of net disbursements of foreign public and publicly guaranteed loans (net of amortization) to GDP as a proxy for the rigidity of the government’s budget constraint. A government with access to foreign financing might be more likely to expand employment and wages than a government which must rely exclusively on domestic revenue sources. For OECD countries the presumption is that foreign financing and wage and employment decisions are made simultaneously, and the variable is not included.

We include the debt-to-GDP ratio, as measured by total public and publicly-guaranteed external debt in developing countries and total central government debt in OECD countries (foreign and domestic). This variable is an indicator of fiscal pressure as it is a proxy for the desire to correct imbalances or for interest payments. We experimented with the share of interest payments in GDP, but it entered with a counter - intuitive positive sign in both wage and employment equations in developing countries, and was therefore not included in the regressions. We decided against using the United States prime rate as a proxy for interest, after finding a virtually zero correlation between the prime rate and the share of interest payments in GDP in our developing country sample, presumably because of long grace periods or default. ^29/

Finally, we include dummies for SAF/ESAF programs and for stand-by programs to capture the effects of two different types of adjustment programs. It might be expected that stand-by programs, because of their short-term nature, would only affect wages, while the longer-term SAF/ESAF programs would be able to affect employment. The impact of the longer-term SAF/ESAF programs on wages would depend on whether initial wages were judged to be too low or too high. It is unclear whether these programs should have contemporaneous or lagged effects. We experimented with both, and found that the results with lagged dummy variables for IMF-supported programs were qualitatively the same as those with contemporaneous dummy variables. ^30/

The coefficients on program dummies in the cross-country averages (the so-called “between”) and fixed-effects (the so-called “within”) specifications will have different interpretations. It is natural to think of the coefficient in the “between” equation (if positive) as the likelihood that countries which often have IMF-supported programs also have high wage expenditures. The association could reflect causality from the level of wage expenditures to the probability of having an adjustment program supported by the IMF, as when countries with excessive government wage expenditures need to engage in stabilization or structural adjustment programs. The coefficient in the “within” equation is more clearly causal from a program to wage expenditures. It is unlikely that high wage expenditures in a given year lead to the presence of a program in that year, as problems typically build up over time, before recourse to a program is taken. It is nevertheless possible that high wage expenditures, or sometimes very low wage levels, precipitate a crisis and lead to the implementation of an IMF-supported program, so that “within” estimates need to be interpreted with caution.

2. Empirical results

We estimate our basic wage and employment equations for the period 1972-1992. We use general government: data for employment and wages in OECD countries. In developing countries, we are restricted to using central government data because of data limitations. We omit the revenue - to-GDP ratio, the foreign financing-to-GDP ratio, the real effective exchange rate, and program dummies in the OECD regressions for the reasons given above. ^31/ We present results of both regressions based on country-means and on individual observations with country-dummies. The results for developing countries are presented in Table 5 and for OECD countries in Table 6. The left hand side panel gives the results for estimation using the country means (“between estimates”), while the right-hand side panel gives the results of pooled time-series cross-country regressions with country dummies (“within estimates”). The “between” regressions can be interpreted as long-run relationships, while the “within” regressions provide information on the short-run determination of wages and employment.

Table 5.

Developing Countries: Central Government Employment Wage, and Wage Bill Regressions

^1/T-statistic below estimate, in parentheses. Dependent variable in logx100. Standard errors are White-corrected for heteroscedastfclty.

Table 5.

Developing Countries: Central Government Employment Wage, and Wage Bill Regressions

	Reoressbns on Country Means (“Between”) 1/				Pegresalons wtth Country Dummies (“Within”) 1/
	Government Employment Per Thousand Popubtion	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$	Government Wage Bill as percent of GDP	Government Employment Per Thousand Popubtion	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$	Government Wage Bill as percent of GDP
Structural Variables
Constant	-130.094	583.685	629.153	441.191
	(-0.552)	(3.055)	(3.278)	(2.707)
Time					0.076	-1.305	-0.432	0.040
Time					(0.149)	(-2.034)	(-0.877)	(0.060)
Real OOP per	0.141	-0.442	0.518	-0527	0.014	0.125	0.677	0.025
Capita in 1965 US$ (log, x100)	(0.442)	(-3.637)	(4.692)	(-2.367)	(0.185)	(0.675)	(4.933)	(0.229)
Secondary School	1.545	-1.538	-1.513	-0.351	0.403	-0.809	-0.494	-0.368
Enrollment (percent)	(1.756)	(-2.583)	(-2.406)	(-1.520)	(2.299)	(-2.090)	(-1.940)	(-2.248)
Urban	0.952			0.861	2.467			-0.213
Population (percent)	(0.952)			(2.871)	(5.519)			(-0.352)
Popubtion (log, x100)	-0.109			-0.016	-0.309			0.099
	(-1.369)			(-0.327)	(-1.767)			(0.454)
Real Effective	-0.108	-30.131	-38.367	-0.453	0.137	-0.046	-0.276	-0.096
Exchange Rate (log, x100)	(-0.441)	(-0.913)	(-1.178)	(-1.260)	(3.376)	(-0.408)	(-3.557)	(-1.425)
Fiscal Federalism	-61.076			-79.006
Dummy	(-1.972)			(-4.923)
Fiscal Variables
Revenue/GDP	3.407	-0.282	0.351	3.710	0.344	0.912	0.471	1.484
(percent)	(1.692)	(-0.295)	(0.374)	(4.526)	(1.812)	(1.615)	(1.415)	(5504)
SAF/ESAF	49.150	-70.736	-81.945	10.848	-4.550	0.870	2.473	-14.006
Dummy	(1.139)	(-1.559)	(-1.529)	(0.406)	(-1.717)	(0.159)	(0.543)	(-2.558)
Standby/EFF	11.367	-18.520	-14.729	-14.848	3.528	-10.384	-7.698	-3.596
Dummy	(0.462)	(-0.482)	(-0.331)	(-0.647)	(2.169)	(-2.873)	(-2.944)	(-1.326)
Foreign	-0.004	1.339	2.189	-0.231	0.598	0.665	0.251	0.817
FlnancIng/GDP (percent)	(-0.0004)	(0.681)	(1.301)	(-0.061)	(2.875)	(1.272)	(0.789)	(2.146)
External	-0.199	0.389	0.293	0.232	-0.399	-0.195	-0.321	-0.080
Debt/GDP (percent)	(-0.529)	(1.970)	(1.269)	(0.923)	(-1.279)	(-2.109)	(-5.665)	(-1.075)
Number of Observations	29	26	26	31	331	276	274	433
R2	0.82	0.7	0.58	0.84	0.25	0.28	0.43	0.18

^1/T-statistic below estimate, in parentheses. Dependent variable in logx100. Standard errors are White-corrected for heteroscedastfclty.

View Table

Table 5.

Developing Countries: Central Government Employment Wage, and Wage Bill Regressions

	Reoressbns on Country Means (“Between”) 1/				Pegresalons wtth Country Dummies (“Within”) 1/
	Government Employment Per Thousand Popubtion	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$	Government Wage Bill as percent of GDP	Government Employment Per Thousand Popubtion	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$	Government Wage Bill as percent of GDP
Structural Variables
Constant	-130.094	583.685	629.153	441.191
	(-0.552)	(3.055)	(3.278)	(2.707)
Time					0.076	-1.305	-0.432	0.040
Time					(0.149)	(-2.034)	(-0.877)	(0.060)
Real OOP per	0.141	-0.442	0.518	-0527	0.014	0.125	0.677	0.025
Capita in 1965 US$ (log, x100)	(0.442)	(-3.637)	(4.692)	(-2.367)	(0.185)	(0.675)	(4.933)	(0.229)
Secondary School	1.545	-1.538	-1.513	-0.351	0.403	-0.809	-0.494	-0.368
Enrollment (percent)	(1.756)	(-2.583)	(-2.406)	(-1.520)	(2.299)	(-2.090)	(-1.940)	(-2.248)
Urban	0.952			0.861	2.467			-0.213
Population (percent)	(0.952)			(2.871)	(5.519)			(-0.352)
Popubtion (log, x100)	-0.109			-0.016	-0.309			0.099
	(-1.369)			(-0.327)	(-1.767)			(0.454)
Real Effective	-0.108	-30.131	-38.367	-0.453	0.137	-0.046	-0.276	-0.096
Exchange Rate (log, x100)	(-0.441)	(-0.913)	(-1.178)	(-1.260)	(3.376)	(-0.408)	(-3.557)	(-1.425)
Fiscal Federalism	-61.076			-79.006
Dummy	(-1.972)			(-4.923)
Fiscal Variables
Revenue/GDP	3.407	-0.282	0.351	3.710	0.344	0.912	0.471	1.484
(percent)	(1.692)	(-0.295)	(0.374)	(4.526)	(1.812)	(1.615)	(1.415)	(5504)
SAF/ESAF	49.150	-70.736	-81.945	10.848	-4.550	0.870	2.473	-14.006
Dummy	(1.139)	(-1.559)	(-1.529)	(0.406)	(-1.717)	(0.159)	(0.543)	(-2.558)
Standby/EFF	11.367	-18.520	-14.729	-14.848	3.528	-10.384	-7.698	-3.596
Dummy	(0.462)	(-0.482)	(-0.331)	(-0.647)	(2.169)	(-2.873)	(-2.944)	(-1.326)
Foreign	-0.004	1.339	2.189	-0.231	0.598	0.665	0.251	0.817
FlnancIng/GDP (percent)	(-0.0004)	(0.681)	(1.301)	(-0.061)	(2.875)	(1.272)	(0.789)	(2.146)
External	-0.199	0.389	0.293	0.232	-0.399	-0.195	-0.321	-0.080
Debt/GDP (percent)	(-0.529)	(1.970)	(1.269)	(0.923)	(-1.279)	(-2.109)	(-5.665)	(-1.075)
Number of Observations	29	26	26	31	331	276	274	433
R2	0.82	0.7	0.58	0.84	0.25	0.28	0.43	0.18

^1/T-statistic below estimate, in parentheses. Dependent variable in logx100. Standard errors are White-corrected for heteroscedastfclty.

Table 6.

OECD Countries: General Government Employment Wage, and Wage Bill Regressions

^1/T-statistic below estimate, in parentheses. Standard errors are White-corrected for heteroscedasticity.

Table 6.

OECD Countries: General Government Employment Wage, and Wage Bill Regressions

		Repressions on Country Means (“Between”) 1/				Repressions with Country Dummies (“Within”)1/
		Government Employment Per Thousand Population	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$ log, x100)	Government Wage Bill as percent of GDP	Government Employment Per Thousand Population	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$ log, x100)	Government Wage Bill as percent of GDP
Structural Variables
	Constant	-140.671	654.700	481.800	-8.800
		(-1.630)	(2.721)	(3.354)	(-0.716)
	Time					1.096	-1.100	-0.377	-0.060
						(2.906)	(-1.321)	(-0.801)	(-1508)
	Real GDP per	0568	-0.557	0.470	0.015	0.063	-0.317	0.667	-0.008
	Capita 1985 US$ (log, x100)	(3.242)	(-1.439)	(1.999)	(0736)	(0.518)	(-0.818)	(3.092)	(-0.469)
	Secondary School	0.321	-0.197	-0.042	0.044	0.555	-0.836	-0.677	0.037
	Enrollment (percent)	(2.156)	(-0.448)	(-0.118)	(2.087)	(3.642)	(-3.533)	(-5.419)	(2.398)
	Urban Population	0.129			0.069	0.861			0573
	(percent)	(0.411)			(2.054)	(2519			(3.926)
	Population	-0.049			-0.009	-0.774			0.090
	(log, percent)	(-1.476)			(-2.465)	(-2.305)			(1.822)
	Private Employment	-1.385	0.895	1.430	-0.021	-1.590	-3.690	-1.430	-0518
	(percent of population)	(-1.059)	(0.216)	(0.576)	(-0.117)	(-3.891)	(-3.039)	(-1.966)	(-4.591)
Fiscal Variables
	Debt/GDP	-0.093	-0.564	-0.416	-0.018	-0.042	-0.136	-0.116	-0.028
	(percent)	(-0.441)	(-1.106)	(-1.154)	(-0.662)	(-1.535)	(-1.682)	(-2.407)	(-6579)
Number of Observations		21	18	18	18	243	153	153	153
	R2	0.40	052	0.47	0.44	0.61	0.55	059	0.47

^1/T-statistic below estimate, in parentheses. Standard errors are White-corrected for heteroscedasticity.

View Table

Table 6.

OECD Countries: General Government Employment Wage, and Wage Bill Regressions

		Repressions on Country Means (“Between”) 1/				Repressions with Country Dummies (“Within”)1/
		Government Employment Per Thousand Population	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$ log, x100)	Government Wage Bill as percent of GDP	Government Employment Per Thousand Population	Government Wages as Percent of GDP Per Capita	Government Wages in 1985 US$ log, x100)	Government Wage Bill as percent of GDP
Structural Variables
	Constant	-140.671	654.700	481.800	-8.800
		(-1.630)	(2.721)	(3.354)	(-0.716)
	Time					1.096	-1.100	-0.377	-0.060
						(2.906)	(-1.321)	(-0.801)	(-1508)
	Real GDP per	0568	-0.557	0.470	0.015	0.063	-0.317	0.667	-0.008
	Capita 1985 US$ (log, x100)	(3.242)	(-1.439)	(1.999)	(0736)	(0.518)	(-0.818)	(3.092)	(-0.469)
	Secondary School	0.321	-0.197	-0.042	0.044	0.555	-0.836	-0.677	0.037
	Enrollment (percent)	(2.156)	(-0.448)	(-0.118)	(2.087)	(3.642)	(-3.533)	(-5.419)	(2.398)
	Urban Population	0.129			0.069	0.861			0573
	(percent)	(0.411)			(2.054)	(2519			(3.926)
	Population	-0.049			-0.009	-0.774			0.090
	(log, percent)	(-1.476)			(-2.465)	(-2.305)			(1.822)
	Private Employment	-1.385	0.895	1.430	-0.021	-1.590	-3.690	-1.430	-0518
	(percent of population)	(-1.059)	(0.216)	(0.576)	(-0.117)	(-3.891)	(-3.039)	(-1.966)	(-4.591)
Fiscal Variables
	Debt/GDP	-0.093	-0.564	-0.416	-0.018	-0.042	-0.136	-0.116	-0.028
	(percent)	(-0.441)	(-1.106)	(-1.154)	(-0.662)	(-1.535)	(-1.682)	(-2.407)	(-6579)
Number of Observations		21	18	18	18	243	153	153	153
	R2	0.40	052	0.47	0.44	0.61	0.55	059	0.47

^1/T-statistic below estimate, in parentheses. Standard errors are White-corrected for heteroscedasticity.

The first column gives the results for employment per thousand population; the second and third columns give the results for two different measures of wages, one relative to GDP per capita, and the other in 1985 U.S. dollars; ^32/ and the fourth column gives the results for the wage bill (GFS data). ^33/ The wage bill equation is presented mainly to highlight the additional information available when its components–employment and wages– are considered separately. Real GDP per capita, government wages in 1985 U.S. dollars, the real effective exchange rate, and population are in logarithms. The other variables are ratios to population or to GDP. In addition, all dependent variables are expressed in logarithms for developing countries. While this is not the most natural representation (interpretation is difficult), the advantage is that the employment and wage equation (wages relative to GDP per capita) add up to the specification for the wage bill. ^34/

Employment: developing countries. Regarding government employment in developing countries, the evidence is mixed regarding the importance of the government budget constraint. The coefficient on the revenue-to-GDP ratio is positive and significant at the 10 percent level across and within countries in the equation for employment per thousand population. GDP per capita is not significant across countries or within countries. The finding that revenues affect employment “within” countries (that is, in the short run) is surprising in view of expected rigidity in government employment. The effect does not appear large quantitatively, however, as a 1 point increase in the revenue-to-GDP ratio ^35/ leads to an increase in employment per thousand population (evaluated at the sample mean of 22) of only 0.05. Foreign financing has a significant effect on employment in the “within” equation, and debt has an insignificant negative impact on employment across countries and within countries. The dummy variable for a structural adjustment program is insignificant in the “between” and significant and negative in the “within” equation, indicating probable causation from structural adjustment programs to employment (see Section V.1). The effect is important: when a structural adjustment program is in place, government employment is 0.7 less per thousand population. Surprisingly, there is a significant positive coefficient for stand-by programs in the “within” equation for employment. ^36/

Turning to structural variables, the coefficient on the share of the urban population is significant and positive in the “within” equation, thereby lending some support to the idea that urbanization leads to a greater pressure for the provision of public services. A 1 percentage point increase in urbanization leads to a 0.40 point increase in employment per thousand in our developing country sample. Secondary-school enrollment, which could capture either demand or supply-side pressures, has a near-significant coefficient in the “between” and a significant, though small, coefficient in the “within” equation.^37/ Further, the real effective exchange rate has a significant positive impact on government employment in the “within” equation (i.e., an appreciation is associated with an increase in government employment), in line with the hypothesis of countercyclical hiring.^38/ The effect is small: a 1 percent increase in the real effective exchange rate (a reduction in competitiveness) leads to an increase of 0.02 in employment per thousand population. Fiscal federalism–a proxy for formal fiscal decentralization—has the expected significant negative impact on central government employment in the “between” equation. The coefficient on population is significant at the 10 percent level in the “within” equation, possibly indicating increasing returns to scale in the provision of government services. Urban population and total population do not, however, carry significant coefficients in the “between” equations. The time trend is not significant in the employment equation.

Employment: OECD countries. For OECD countries, mostly structural variables are considered, as government revenue and financing are no longer exogenous. Government employment in OECD countries across countries (the “between” equation) is a positive function of real GDP per capita and secondary-school enrollment. Between countries, a 1 percent increase in GDP per capita in constant U.S. dollars is associated with a 0.27 point increase in employment per thousand in the case of OECD countries (evaluated at the sample average of 71). This lends credence to a Wagnerian viewpoint that an expanding public sector is an inevitable consequence of economic development. Within countries real GDP per capita is no longer significant. Its significance appears to be absorbed by a significantly positive time trend, which may indicate that there is inherent expansion of the public sector over time.

Within countries, there is also a significant positive impact of secondary school enrollment and urbanization, a significant negative impact of population, a significant negative impact of private sector employment, consistent with countercyclical hiring, and a marginally significant negative impact of debt. Some of these effects are also economically significant. A 1 percentage point increase in secondary school enrollment is associated with a 0.56 point increase in employment per thousand population, while a 1 percentage point increase in urbanization leads to a 0.86 point increase in employment per thousand population. A 1 percentage point increase in the ratio of private employment to the population leads to a reduction of employment per thousand of 1.59 in OECD countries. ^39/

Wages. For developing countries, there is some evidence of an impact of the government budget constraint on wages in the “within” equations, where there is a significant (at the 10 percent level) positive impact of the revenue-to-GDP ratio on wages as a percent of GDP per capita. Within countries, a one percentage point increase in government revenue as a percent of GDP is associated with a 3 percentage point increase in government wages as a percent of GDP per capita (the sample average is 330). ^40/ For OECD and developing countries, debt has a significant negative impact (at either the 5 or 10 percent level) on wages within countries. ^41/ In developing countries, a one percentage point increase in external debt as a percent of GDP involves a 1 percentage point reduction in wages expressed as a percent of GDP per capita and a 0.32 percent reduction in wages in 1985 U.S. dollars. In OECD countries, within countries, a one percentage point increase in total government debt as a percent of GDP leads to a 0.14 percentage point reduction in wages as a percent of GDP per capita and a 0.12 percent reduction in wages in 1985 U.S. dollars.

SAF/ESAF programs have no significant association with wages across or within countries. Stand-by programs, on the other hand, have a significant negative impact on both expressions for wages within countries, but not across, indicating that a causal relationship may be present. When a standby arrangement is in place, government wages as a percent of GDP are 33 percentage points lower, while wages expressed in 1985 U.S. dollars are 7.4 percent lower. These results might be explained by noting that the civil service reforms which often accompany the longer-term, more structural SAF/ESAF programs do not necessarily include wage reductions, as these might compromise the efficiency of the civil service.

Turning to structural variables, wages are significantly and negatively influenced by secondary-school enrollment, in both OECD and developing countries, both across (except OECD countries) and within countries, and both when expressed as a percent of GDP per capita and when expressed in 1985 U.S. dollars. These findings are in line with the hypothesis that competition lowers wages in the public sector (presumed to have a higher skill-content than those in the remainder of the economy). They are also in line with the alternative hypothesis of government employment of university graduates and erosion of pay to accommodate this employment within the budget constraint (assuming there is a positive correlation between secondary and tertiary education). Within developing countries, a one percentage point increase in the secondary-school enrollment ratio is associated with a 3 percentage point decline in government wages as a percent of GDP per capita and with a 0.49 percent decline in wages in 1985 U.S. dollars. Within OECD countries, the figures are 0.84 and 0.68, respectively. Real GDP per capita has a significant negative impact on wages as a percent of GDP per capita in the “between” equation for developing countries and—as expected—a significant positive effect on wages in 1985 U.S. dollars in both the “between” and the “within” equations, in developing countries and OECD countries. The elasticity of wages in 1985 U.S. dollars with respect to GDP per capita in 1985 U.S. dollars is less than 1 in all 4 cases (though not always significantly so), as would follow if the shortage of skills employed in government declines with development (per capita income).

Private employment has a significant negative coefficient in the “within” wage equations for OECD countries, contrary to our prediction that lower private-sector employment generates pressures to reduce wages (to provide funds for higher government employment). For developing countries, on the other hand, the real effective exchange rate has the expected significant negative impact on wages in U.S. dollars in the “within” equation, in line with the hypothesis that the need to finance countercyclical hiring leads to (real) wage reductions when the real effective exchange rate appreciates. ^42/ In developing countries, within countries, a one percent increase in the real effective exchange rate is associated with a 0.28 percent decline in wages in 1985 U.S. dollars. This impact is not present in the relative wage equation, presumably on account of reductions in economy-wide wages in the face of reduction in competitiveness. The time trend is significant and negative for wages as a percent of GDP per capita in developing countries.

Wage bill. The results of wage bill regressions are generally in line with expectations. Across developing countries, real GDP per capita has a negative impact on the wage bill (presumably because of the impact on wage levels), while urban population and revenues have a positive impact. The dummy for fiscal federalism is significant. Within countries, secondary-school enrollment (presumably because of the impact on wage levels), and the SAF/ESAF dummy have a negative impact (the impact of the stand-by dummy is insignificant), while revenues and foreign financing have a positive impact. The presence of a SAF or ESAF is associated with a reduction of 0.8 percent in the wage bill. This reduction is of the same order of magnitude as found in Hewitt and Van Rijckeghem (1995). In contrast with a finding of Schuknecht (1994), the real effective exchange rate does not have a significant impact on wage expenditures as a share of GDP.^43/

Across OECD countries, the population has a negative impact on the government wage bill, while secondary - school enrollment and urban population have a positive impact. Within countries, private employment (countercyclical hiring) and debt have a negative impact and secondary-school enrollment and urban population have a positive impact.