Sources of Data

The collection of data on the subject of government employment and wages proved extremely difficult. Neither the International Labor Organization (ILO) nor any of the other United Nations organizations collects statistics on either subject in a standardized way. The Organization for Economic Cooperation and Development (OECD) has occasionally done work in this area but only on a limited basis, including a recent study on the general magnitude of government employment in the OECD countries during the 1970s.¹ Over the last 20 years, a handful of academic studies have been made on the subject.²

Given the dearth of studies in the area, one is forced to rely almost entirely on data from national sources. Generally, these are of four kinds: (1) statistical yearbooks; (2) data from budget documents; (3) data provided by personnel ministries, often in such publications as the establishment register; and (4) occasional studies and reports relating to the reform of civil service employment or wage policies within a particular country. However, in the absence of any standardized international effort to collect statistics on government employment, there is no generally accepted set of definitions as to what data should be collected routinely, or how data should be defined. As a result, the variance in the breadth and depth of statistics across individual countries is extremely large. For some countries, information could not be collected on the size of government employment. For other countries where there is a wealth of information, there is often a problem in interpretation of the available statistics.

To obtain as comprehensive a picture as possible on this subject, a letter was sent to almost all member countries of the International Monetary Fund (IMF) asking for any documents or information on the magnitude of government employment, aggregate wage and salary payments, and on the structure of wages and salaries in the government. Table 19, in Appendix I, indicates the countries to which the letters were sent and the number of countries that responded to the initial inquiry and/or to the subsequent request.

All things considered, it is extraordinary how impoverished the data base is. One would think that on a subject of this kind, governments would be able to provide at least some statistics on the size and distribution of government employees and salaries. Yet, it was quite apparent, even from the countries that responded to the request and made an effort to provide this information, that only a handful of countries were able to provide easily statistics on these employment and wage variables. Considering the importance of government employment and wages and salaries in the economies of almost all member governments, it is evident that this issue of paucity of data should be dealt with systematically and remedied in the future. Considerably greater resources need to be invested, perhaps by the ILO, the IMF, or the World Bank, to stimulate an improvement in the statistical data base on government employment and wages.

The Measurement of Government Employment

Several conceptual issues arise in the definition of public sector employment. These may be characterized in terms of the definition of a unit of government, the definition of what constitutes a government employee, and the classification of employees by function.

In principle, the problem of defining the unit of government is the same for the employment issue as it is for the definition of public sector expenditure or revenue. The same institutions or units of government that are used to define the central government, the state and local governments, or the nonfinancial public enterprise sector for financial analyses should also be used for defining these levels of government in terms of employment. The same problems of ensuring comparability in these definitions across countries arise in either case. In this study, the classification used for preparing the Fund’s Government Finance Statistics (GFS) Yearbook has been used wherever possible.

The GFS approach requires careful disaggregation of data by institutions, and sometimes by programs within institutions, to ensure a consistent definition of governmental units and functional program categories. The frequent lack of employment data at a sufficiently disaggregated level may force a study such as this one to use a different definition of employment for a particular level of government. This is particularly a problem in many of the Latin American countries where the central government includes large numbers of decentralized agencies.

Cross-country comparisons are also complicated by the different ways governments implement comparable policies. For example, some countries directly operate and manage the health and medical system through the government. In other countries, government is heavily involved in the financing of the medical system (e.g., the Netherlands) but allows the operation and ownership to be within the private sector. Yet when the system is a private one, financed indirectly through government subsidies and government transfer payments, the employees are outside the government sector. These institutional alternatives would imply considerable differences in the size of measured government employment in comparing countries where, in a meaningful sense, the employment in both countries may be equally reliant on government financing.

In many respects, the appropriateness of the definition depends on the question posed. For comparing size of government employment, a GFS-type definition may not always be satisfactory, as seen in the above example. For other policy questions, such as the impact of government wage rates on the economy, the GFS-type definition may be quite appropriate. For example, in a country with a private medical care system, wage rates in the private sector indeed may be independently determined from the wage rates that would prevail in a government-run medical institution. While this analysis used the GFS definitions of units of government, the problems that this can pose in some intercountry comparisons should be acknowledged. Where there was a serious problem or issue at this level, it has been noted in Appendix II, which describes the sources of data for the study.

Differences in the structure of government also create problems in cross-country comparisons of the size of government employment in total and in certain sectors. For example, in most federal countries, important education, health, police, fire fighting, and administrative responsibilities are delegated to the state and local governmental levels. It is therefore meaningless simply to compare the size of the central government across countries without taking into account that the central government in one country may perform many of the functions that in another country are performed at the state and local governmental level—this is particularly a problem when one is comparing the number of government employees in a particular functional sector. In comparisons of this kind, the number of state and local government employees in the health, education, and police sectors at the state and local governmental level have been added, where possible, to those at the central governmental level to produce more accurate measures of the extent of government involvement in these sectors.

The same problem of functional allocation also arises for many of the functions carried out under the auspices of nonfinancial public enterprises. The post office and railroad are examples of activities that, in some cases, are operated by the central government and in other cases by nonfinancial public enterprises. Reliance on the GFS definition of what constitutes central government and what constitutes the nonfinancial public enterprise sector in a country often leads to clear differences in definition across countries.

Turning to the second issue, that of defining a government employee, many questions arise. Should one measure the number of employees in terms of man-years worked (as in the Netherlands), in terms of the total number of employees, whether full-time or part-time, or in terms of full-time equivalent employees (as in the Federal Republic of Germany and the United States)? How is a full-time equivalent defined in a country? Can one be certain that the same methodological procedures are used to convert part-time employees to full-time equivalent employees? How should the use of consultants as a “backdoor” form of employment be treated, such as in the defense sector of the United States? Another issue arises in that some countries have a corps of regular or permanent employees involved in public sector capital projects, while in other countries, ministries employ so-called daily paid workers for the implementation of capital projects. In principle, such workers, hired on a daily basis as a function of the level of capital expenditure in a given period, are not permanent government employees and do not appear anywhere in the statistics on government employment. Yet in some countries, reluctance to lay off such workers renders them almost the equivalent of permanent employees. Should these latter employees be included or not included in a measure of government? In general, they have not been included in this analysis. The same problem arises with respect to contractual employees where, in many cases, payments for contracted services do not appear in wage and salary budgets, and the employment implied by such contracting is not defined as a form of government employment, per se. An examination of the scale and importance of such employment (e.g., in printing and publishing, health services, communications, transport, construction, road repairs) could, and should, form an interesting avenue for research.

Another problem that arises in defining the size of the government labor force is the treatment of defense employees. Military employees are not included in establishment registers, and for security reasons the size of the military is generally not public information. Yet it is clear that the military may constitute a very significant portion of the total work force in a government. Every effort has been made therefore to include the number of military employees in the employment statistics in this paper. Where there are no national statistics on the size of the military force, reliance was placed on the most recent publication on military expenditures of the U.S. Arms Control and Disarmament Agency.³ A further problem relating to defense employment is whether to distinguish between permanent military employees and draftees. While both are clearly government employees, draftees are paid considerably lower salaries, and inclusion of such employees may lead to an understatement of the average wage in the central government sector. In this study, draftees and permanent military employees have been included in the defense sector without any distinction.

In principle, in deciding what constitutes a “government employee,” it would be preferable to use statistics on the actual number of employees, on a full-time equivalent basis, employed as of a given date by a governmental unit. In the absence of any such data, statistics on the formal establishment in specific ministries have been used. High vacancy rates would obviously lead to an overstatement of the magnitude of employment in a given functional sector.

A third issue is the difficulty of classifying government employees by function. This problem is, in principle, no different from that which arises in classifying expenditure on a functional basis. It is well recognized that the traditional institutional division of responsibilities may not correspond to a rigorous functional division, and that given ministries may provide services that overlap functional expenditure categories. It is often difficult in expenditure analysis to separate the different functional components of a ministry’s operations, and it proves even more difficult to separate the employees of a ministry by their different functions. Since the division of functional responsibilities varies widely across countries, it is often difficult to develop a clear comparable delineation of employment by function, and the statistics on employees by function presented here must be regarded with considerably more caution than the numbers on the total size of public sector employment. This is particularly the case for such broad and not easily defined sectors such as administration.

The Measurement of Wages and Salaries

The first problem confronting anyone trying to measure the amount of wages and salaries at any governmental level is the definition of what should be included in “wages and salaries.” It is common, particularly in developing countries, for many civil servants to receive food, car, and housing allowances as part of their conditions of employment. Yet it is exceptionally difficult to place a value on these nonwage benefits—certainly, it is rare to see a financial valuation of such remuneration in any government budget document. Nonsalary benefits are particularly a problem in countries with large military employment, since traditionally considerable nonwage services are provided to soldiers. Also a problem in this respect is the treatment of bonus payments, the accrued liabilities of the government for future pension payments, and other types of allowances. Where statistics on bonuses, allowances, or fringe payments were readily available, an attempt was made to include such statistics explicitly. Pension payments and accrued liabilities for pensions for present government employees, are not included.

A second problem that arises is the potential inconsistency between (1) statistics on employment that may include so-called daily paid workers and (2) statistics on wages that may relate only to permanent civil service employees and that exclude from the wage and salary budget any payments to daily paid workers. Daily paid employees may simply be paid out of capital funds, and there may be no statistics on the component of such capital funds paid out in wages and salaries.

A third issue relates to the availability of actual expenditure data. Wherever possible, actual wage and salary expenditure were used; however, there were cases where the only source of data on wages was a recent budget document. To ensure that the wage estimates related to comparable employment statistics, budget estimates had to be used rather than actual wage and salary payments.

Finally, the same problems that arose in classifying ministries on a functional basis for employment purposes also bedevil the calculation of wage and salary payments on a functional basis.

Measurement of the Salary of Specific Jobs

The obvious problem that arose in comparing the salaries of employees in similar employment classifications within and across countries was to ensure that the same job definition was being used. It is, of course, difficult to know whether a clerical officer in one country is, in fact, defined in the same way as a clerical officer in another country. However, after reviewing many government job definitions, the duties and responsibilities of a government clerical officer seemed sufficiently comparable to use the starting salary of this grade as a numéraire. It should be kept in mind that the level of responsibility and required skills may be different from those required in another country for a position with the same nominal title. It was also necessary to define the desired starting salary for any position. If the starting salary was unavailable, the average salary for the position was calculated.

Timing

Only one year was taken for each country. While the numbers employed may not change significantly from one year to another, the pay relative to private sector employees (especially when all prices are changing rapidly) may change quite sharply.⁴