The literature on the determinants of government employment is thin.17 Among empirical works, only Martin (1982) and Lindauer (1980) have attempted any econometric explanation of the determinants of government employment. What is interesting is that such analyses fit within the framework of efforts to test the validity of Wagner’s law, which posited the growth of the government sector over time. Most tests of Wagner’s law have focused on the growth of the share of government expenditure, in real or nominal terms, as a share of GDP.18 Yet, clearly, growth in the size of government employment as a share of the total labor force or population over time might constitute an equally valid alternative test of this hypothesis. If public sector wages and salaries are strongly correlated with the size of the public sector (and from the discussion on pages 12-13 it appears that they are), then government employment and pay could be a good proxy measure of Wagner’s law. This would be a strong result in the sense that the growth of the public sector in terms of expenditure has also occurred in many developed countries by means of subsidies and transfers or through the contracting out of employment and services rather than through direct employment.
Wagner suggested that numerous “workers” (his quotes) forming part of the complicated bureaucracy will have a lower efficiency, and hence their employment and pay will be an increasing burden on the economy.19 Studies by Rose (1980) and Martin (1982) focused on whether the share of government employment in population has risen over time, but they focused on OECD countries. Martin also examined the relative importance of the level of development (as proxied by per capita income), demographic structure (as proxied by the dependency ratio), and the female dependency rate as determinants of the share of general government employment in total employment. Lindauer’s study of African countries sought to explain per capita public employment over time, primarily as a function of the size of a country (as proxied by its population size) and per capita income.
Lacking time series observations, the alternative test here of Wagner’s law is essentially a test of whether the number of employees per capita rises with per capita income. This model also tests (1) whether there are economies or diseconomies of scale in government, in the sense of an increasing or decreasing share of government in total population as total population rises, and (2) whether the type of economic system—capitalist, mixed, or socialist—affects the government employment share.20 Government employment was examined both in its aggregate measures—general government and public sector employment—and in its disaggregated components: central government, state and local government, and nonfinancial public enterprises.
In these estimations, four specifications on per capita income were tested: (1) a direct linear relationship, (2) a hyperbolic relationship (for example, the inverse of per capita income), (3) a logarithmic relationship, and (4) a semilogarithmic relationship. The choice criterion was primarily the goodness of overall fit. A test was made of the possibility that the nature of the relationships might differ according to whether the country was developed or developing. For each equation, a test was made of whether the coefficient of each independent variable was higher or lower for countries that were above or below a given per capita income level. The per capita income cutoff was chosen to optimize the statistical fit of the relationship.21 An index variable was used to proxy the type of economic system. The economic system index variable ranged from a value of one for a capitalist economy to four for a completely socialist economy.22 Since the index values are arbitrary, only the sign of the coefficient of this variable is important as a qualitative indicator.23
The results of the analysis are indicated in Table 7. The clearest result is that government employment tends to increase on a per capita basis as per capita income rises. While the specification may depend on the precise employment variables under consideration, the sign of the relationship is generally unaffected. Only at the central government level does the relationship between employment per capita and per capita income differ between developed and developing countries. For countries with per capita income that is less than US5800, there is no significant relationship; above that level, there is a direct relationship between per capita income and central government employment per capita. The relationship between state and local government employment is strong, leading to a clear relationship between both general government and public sector employment per capita and per capita income. No relationship emerged between nonfinancial public enterprise employment per capita and per capita income. These results support Wagner’s hypothesis that government employment growth (and especially local government growth), in terms of the number of employees per capita, rises with per capita income.
Determinants of Government Employment
(t-statistics in parentheses)
In thousands of U.S. dollars.
In thousands.
n = number of observations in the sample.
The dependent variable is taken in logarithmic terms.
Determinants of Government Employment
(t-statistics in parentheses)
| Independent Variables Dependent Variables | Logarithm of Per Capita Income (PCI) | Per Capita Income1 | Inverse of Per Capita Income | Logarithm of Population | Population2 | Economic System | Constant | R2(n)3 | |
|---|---|---|---|---|---|---|---|---|---|
| (Dependent variables as percentage of employment in nonagricultural sector) | |||||||||
| Central government employment4 | −0.35 | −0.19 | 0.01 | 4.3 | 0.57 | ||||
| (−6.0) | (−4.43) | (0.19) | (11.0) | (47) | |||||
| State and local government employment | |||||||||
| For countries with PCI ≤ US$1,200 | 0.39 | 0.05 | 1.26 | −2.6 | 0.60 | ||||
| (1.31) | (6.29) | (1.70) | (1.1) | (44) | |||||
| For countries with PCI > US$1,200 | 0.11 | 0.01 | 1.26 | −2.6 | 0.60 | ||||
| (−1.01) | (1.67) | (1.70) | (1.1) | (44) | |||||
| Nonfinancial public enterprise employment | |||||||||
| For countries with PCI ≤ US$600 | 5.78 | −0,03 | 3.17 | −3.84 | 0.72 | ||||
| (5.53) | (−2.3) | (2.74) | (−1.47) | (32) | |||||
| For countries with PCI > US$600 | 10.90 | 0.01 | 3.17 | −3.84 | 0.72 | ||||
| (2.23) | (1.26) | (2.74) | (−1.47) | (32) | |||||
| General government employment | |||||||||
| For countries with PCI ≤ US$l,400 | 3.95 | 0.01 | 3.55 | 14.7 | 0.49 | ||||
| (3.12) | (0.53) | (2.61) | (5.2) | (51) | |||||
| For countries with PCI > US$1,400 | −13.9 | 0.01 | 3.55 | 14.7 | 0.49 | ||||
| (−1.9) | (0.53) | (2.61) | (5.2) | (51) | |||||
| Public sector employment | |||||||||
| For countries with PCI ≤ US$600 | 2.24 | 0.04 | 10.6 | 5.63 | 0.40 | ||||
| (1.46) | (1.24) | (3.35) | (0.75) | (37) | |||||
| For countries with PCI > US$600 | 14.7 | −0.02 | 10.6 | 5.63 | 0.40 | ||||
| (1.93) | (−0.76) | (3.35) | (0.75) | (37) | |||||
| (Dependent variables in terms of number of employees per 100 inhabitants) | |||||||||
| Central government employment | |||||||||
| For countries with PC ≤ US$800 | 0.10 | −0.39 | −0.18 | 5.44 | 0.57 | ||||
| (0.26) | (−3.19) | (−1.22) | (5.63) | (50) | |||||
| For countries with PCI > US$800 | 0.82 | −0.60 | −0.18 | 5.44 | 0.57 | ||||
| (1.73) | (−1.92) | (−1.22) | (5.63) | (50) | |||||
| State and local government employment | 0.04 | — | 0.40 | −0.92 | 0.56 | ||||
| (7.23) | (0.9) | (1.54) | (−1.42) | (46) | |||||
| General government employment4 | 0.41 | −0.02 | 0.02 | 0.21 | 0.64 | ||||
| (9.44) | (−0.70) | (0.34) | (0.55) | (56) | |||||
| Public sector employment4 | 0.35 | 0.01 | — | 0.48 | 0.62 | ||||
| (6.84) | (0.11) | (-) | (0.68) | (34) | |||||
In thousands of U.S. dollars.
In thousands.
n = number of observations in the sample.
The dependent variable is taken in logarithmic terms.
Determinants of Government Employment
(t-statistics in parentheses)
| Independent Variables Dependent Variables | Logarithm of Per Capita Income (PCI) | Per Capita Income1 | Inverse of Per Capita Income | Logarithm of Population | Population2 | Economic System | Constant | R2(n)3 | |
|---|---|---|---|---|---|---|---|---|---|
| (Dependent variables as percentage of employment in nonagricultural sector) | |||||||||
| Central government employment4 | −0.35 | −0.19 | 0.01 | 4.3 | 0.57 | ||||
| (−6.0) | (−4.43) | (0.19) | (11.0) | (47) | |||||
| State and local government employment | |||||||||
| For countries with PCI ≤ US$1,200 | 0.39 | 0.05 | 1.26 | −2.6 | 0.60 | ||||
| (1.31) | (6.29) | (1.70) | (1.1) | (44) | |||||
| For countries with PCI > US$1,200 | 0.11 | 0.01 | 1.26 | −2.6 | 0.60 | ||||
| (−1.01) | (1.67) | (1.70) | (1.1) | (44) | |||||
| Nonfinancial public enterprise employment | |||||||||
| For countries with PCI ≤ US$600 | 5.78 | −0,03 | 3.17 | −3.84 | 0.72 | ||||
| (5.53) | (−2.3) | (2.74) | (−1.47) | (32) | |||||
| For countries with PCI > US$600 | 10.90 | 0.01 | 3.17 | −3.84 | 0.72 | ||||
| (2.23) | (1.26) | (2.74) | (−1.47) | (32) | |||||
| General government employment | |||||||||
| For countries with PCI ≤ US$l,400 | 3.95 | 0.01 | 3.55 | 14.7 | 0.49 | ||||
| (3.12) | (0.53) | (2.61) | (5.2) | (51) | |||||
| For countries with PCI > US$1,400 | −13.9 | 0.01 | 3.55 | 14.7 | 0.49 | ||||
| (−1.9) | (0.53) | (2.61) | (5.2) | (51) | |||||
| Public sector employment | |||||||||
| For countries with PCI ≤ US$600 | 2.24 | 0.04 | 10.6 | 5.63 | 0.40 | ||||
| (1.46) | (1.24) | (3.35) | (0.75) | (37) | |||||
| For countries with PCI > US$600 | 14.7 | −0.02 | 10.6 | 5.63 | 0.40 | ||||
| (1.93) | (−0.76) | (3.35) | (0.75) | (37) | |||||
| (Dependent variables in terms of number of employees per 100 inhabitants) | |||||||||
| Central government employment | |||||||||
| For countries with PC ≤ US$800 | 0.10 | −0.39 | −0.18 | 5.44 | 0.57 | ||||
| (0.26) | (−3.19) | (−1.22) | (5.63) | (50) | |||||
| For countries with PCI > US$800 | 0.82 | −0.60 | −0.18 | 5.44 | 0.57 | ||||
| (1.73) | (−1.92) | (−1.22) | (5.63) | (50) | |||||
| State and local government employment | 0.04 | — | 0.40 | −0.92 | 0.56 | ||||
| (7.23) | (0.9) | (1.54) | (−1.42) | (46) | |||||
| General government employment4 | 0.41 | −0.02 | 0.02 | 0.21 | 0.64 | ||||
| (9.44) | (−0.70) | (0.34) | (0.55) | (56) | |||||
| Public sector employment4 | 0.35 | 0.01 | — | 0.48 | 0.62 | ||||
| (6.84) | (0.11) | (-) | (0.68) | (34) | |||||
In thousands of U.S. dollars.
In thousands.
n = number of observations in the sample.
The dependent variable is taken in logarithmic terms.
Focusing on the government employment variables as a share of the nonagriculturally employed, the sign of the relationship between the share of government and per capita income does differ between developed and developing countries. The central government employment share declines unambiguously as per capita income rises, with no difference in the magnitude of the relationship by group of countries. Conversely, the share of state and local government employment increases, although the increase is greater for a given change in per capita income, for the group of less developed countries (with per capita income of less than US$1,200). Given these offsetting effects, one finds that for countries with per capita income that is less than US$1,400, the general government employment share declines hyperbolically as per capita income rises; above that level, the employment share increases with increases in per capita income. For the smaller sample of countries for which data on nonfinancial public enterprise employment are available, the share of such enterprises among the nonagriculturally employed declines hyperbolically as per capita income rises. The effect of this latter relationship is to ensure that the share of public sector employment among the nonagriculturally employed declines with per capita income, with the rate of decline greater among countries at per capita income levels that are above US$600.
The scale of a country, as proxied by the size of population, proved to be negatively and significantly correlated with the share of central government employment in both nonagricultural sector employment and total population. The larger the population, the lower the central government employment share; the obvious corollary relationship, that the share of state and local government would increase, was true only vis-à-vis the share in nonagricultural sector employment. State and local government employment per capita is not significantly influenced by population size; perhaps as a result, neither is general government nor public sector employment.
The type of economic system also proved to be an important factor in explaining the share of government employment in nonagricultural sector employment. The more centrally planned the economy, the higher the share among the nonagriculturally employed of employees in the state and local government, nonfinancial public enterprise sector, general government, and public sector. However, on a per capita basis, the type of economic system does not appear to have a significant impact on the size of government or public sector employment.
The strength of the overall relationships is remarkable given the cross-sectional nature of the data base. With the exception of the equation explaining nonfinancial public enterprise employment per capita (where the R2 was insignificant), the R2 of the equations exceed 0.40 and range as high as 0.72.