II Methodology for Projections

Edgardo Ruggiero, Peter Heller, Menachem Katz, Robert Feldman, Richard Hemming, Peter Kohnert, Ziba Farhadian, Donogh McDonald, Ahsan Mansur, and Bernard Nivollet
Published Date:
September 1986
  • ShareShare
Show Summary Details

Pattern of Government Social Expenditure, 1980

An important problem faced in undertaking this study, particularly given its cross-country focus, was how to define the social expenditure of the government. Clearly, one should include expenditure on education, medical care, pensions, unemployment compensation, and income maintenance for the poor. The dividing line then becomes blurred. Should one include expenditure on family allowances? veterans benefits? workmen’s compensation? Should pensions of central government or public enterprise employees be included or treated analogously to the pension expenditure associated with private sector employees? Unfortunately, any cross-country comparison of government social expenditure by program is fraught with methodological difficulties, particularly for the governments of the seven major industrial countries. The structure of programs, the degree of public sector involvement, the instruments of intervention, and the level of benefits vary widely—thus rendering any comparison subject to numerous caveats.

In the individual country studies, the typical definition of social expenditure by government was that used by the country itself, though with the constraint that certain categories of government expenditure were to be included (e.g., medical care, education, welfare payments, pension expenditure, and unemployment insurance). Expenditures at both the general and local government levels were generally included. Pension expenditure was included if the government intermediated its financing—admittedly yielding some incomparabilities across countries in that the pension schemes of some public enterprises might be included in one country and not in another. Similarly, as countries differ in the extent to which the private sector participates in the financing of education and medical expenditure, differences will arise in the magnitude of government social expenditure across countries. Direct housing expenditures were omitted, given the difficulties of fully capturing the off-budget credit and tax subsidies accorded to the housing sector in some countries. Tax expenditures were excluded. These differences do not vitiate the analysis, but rather emphasize the relative importance of examining the movements across time in the level of a country’s social expenditure in response to demographic or other factors, rather than the differences across countries in expenditure ratios at a given point in time.

With these cautions in mind, the variation in the magnitude of the government social expenditure ratio is striking. It ranges from about 15–18 percent in Japan and the United States to as high as 31 percent in France and the Federal Republic of Germany (Chart 3 and Table 15). Much of this variability stems from the difference in the pension expenditure ratio, ranging from less than about 4 percent of GDP in Canada and Japan to over 13 percent in the Federal Republic of Germany (though some of this variability is spurious, reflecting classification difficulties between pension and other expenditures).

Chart 3.Social Expenditure as a Percentage of GDP by Sector, 1980

Source: Tables 15 and 16.

The differences in ratios may also be misleading to the extent that they do not take account of tax expenditure and private sector social expenditure. It is useful to indicate the relative importance of these factors. Only limited data are available on tax expenditure 15 and for only five of the countries concerned (Canada, France, the Federal Republic of Germany, the United Kingdom, and the United States). In three of the five countries (France, the United Kingdom, and the United States), inclusion of tax expenditure on social purposes adds significantly to social expenditure (by approximately 1–2 percent of GDP). Inclusion of private sector expenditure on medical care and education and private pension plan disbursements greatly reduces the variability between most of the seven major industrial countries, but significant differences still remain (Chart 3 and Table 16). For example, in 1980 Japan’s social expenditure ratio was still approximately half that of France and the Federal Republic of Germany. Inclusion of private sector expenditure suggests that only in the case of the United States and perhaps Canada, does a focus on government expenditure give a misleading picture of the magnitude of the social expenditure ratio for the society as a whole. In the case of the United States, almost a third of total social expenditure occurs at the private sector level; in Canada, a quarter.

Much of the growth in social expenditure ratios has occurred since 1960. In a recent OECD study, it is reported that between 1960 and 1981 social expenditure by the governments of the seven major industrial countries increased from an average of almost 14 percent of GDP to 25 percent.16 As Chart 4 illustrates, the expenditure ratio increased in a similar fashion in each of these countries.17 The pattern of growth, both over time and between programs, however, has varied among countries. On average, the income elasticity of social expenditure—the growth rate of nominal social expenditure relative to the growth rate of nominal GDP—fell from 1.9 for the period 1960–75 to 1.4 for the period 1975–81. But the United Kingdom was the only country where the change in the income elasticity was of this order (also see Table 18). The income elasticity of social expenditure increased in France and Japan, indicating that the expenditure ratio rose faster in the second period of slower economic growth, fell slightly in Italy, and fell sharply in Canada, the Federal Republic of Germany, and the United States. In the United States, the expenditure ratio was the same in 1981 as in 1975, while in Canada and the Federal Republic of Germany, the elasticity of less than unity in 1975–81 indicates that the expenditure ratio was actually lower in 1981 than in 1975.

Chart 4.Government Social Expenditure, 1960–81

Similarly, in terms of the expenditure ratios for the different programs, there has been, on average, a long-run shift from expenditure on education to expenditure on medical care and pensions. However, this average experience is exemplified by that of Japan, the only country where such a shift has occurred. Elsewhere, the proportion of social expenditure devoted to education fell only in the United Kingdom and the United States, the proportion devoted to medical care increased everywhere except in the United Kingdom, and the proportion devoted to pensions increased only in Italy, Japan, and the United Kingdom. For each program, the ratio of social expenditure to GDP still increased (see Table 18).

What have been the sources of growth in social expenditure? The OECD’s study addresses this issue by breaking down the growth rate of nominal social expenditure into five components: (1) general inflation: (2) cost increases in excess of general inflation (the relative price effect); (3) demographic shifts (viz., changes in the size of population groups deemed relevant to particular sectors)18; (4) changes in program coverage, reflecting changes in eligibility and utilization: and (5) changes in average real benefits.19

Table 1 decomposes the past growth of the government social expenditure of the seven major industrial countries. Without significant exception across programs and countries, the principal source of real growth has been increases in average real benefits: medical benefits per covered person, education outlays per student, and the purchasing power of cash benefits. The impact of changes in coverage has been relatively small, particularly over the 1975–81 period. The fall in coverage under unemployment compensation programs, given rapidly increasing unemployment, may appear anomalous, but it simply reflects the increasing proportion of unemployed persons (youth, women, the long-term unemployed) who were not covered by unemployment compensation. In addition to unemployment compensation, the demographic impact has been relatively strong in the case of pensions, although much smaller overall.

Table 1.Decomposition of the Average Growth Rate of Government Social Expenditure, 1960–75 and 1975–81
Of Which:
Initial Expenditure RatioNominal ExpenditureGDP DeflatorDeflated ExpenditureRelative PricesReal ExpenditureDemographyCoverageAverage real benefitFinal Expenditure Ratio
1960Annual growth rates (percent): 1960–751975
Medical care2.516.75.710.
Unemployment compensation10.518.95.712.4−0.312.
Total of above programs11.
Total social expenditure13.714.
1975Annual growth rates (percent): 1975–811981
Medical care5.
Unemployment compensation1.−
Total of above programs19.414.
Total social expenditure23.
Source: OECD, Social Expenditure: 1960–1990. Problems of Growth and Control (1985).

Excluding France.

Source: OECD, Social Expenditure: 1960–1990. Problems of Growth and Control (1985).

Excluding France.

In medical care and education, increases in relative prices have contributed significantly to expenditure growth. Of course, a major source of nominal expenditure growth—and by far the major source in the later period—has been general inflation. But inflation has exactly the same impact on nominal GDP and does not affect the ratio of social expenditure to GDP.20

Methodology for Projecting Social Expenditure

The growth-accounting framework used by the OECD in characterizing the forces underlying the past growth in social expenditure provides a basis for the projections in this study, although the requirements of the long-term projections being attempted call for some consolidation and simplification. By definition, the volume of expenditure associated with a particular social program is the product of the number of people benefiting from that program and the average real benefit (in volume terms) that each receives. If the average real benefit is held constant, real volume rises in line with the number of beneficiaries. Because the focus of this paper is on long-term demographic developments, an attempt is made to identify this pure demographic effect on expenditure for each social program and on total social expenditure. This is equivalent to the combined effect of growth in the size of the population group relevant to a particular program and changes in the proportion of the relevant population receiving program benefits.

The detailed demographic projections described in Chapter III provide the basis for the projection of the number of program beneficiaries. For the most part, coverage in the future is assumed to remain at the 1980 rate—the proportion of the relevant population eligible for benefits and the proportion of those eligible who utilize a program remaining roughly constant. It varies only to the extent that certain strong trends in coverage, or already legislated changes in coverage for particular countries, are taken into account.

Before considering the assumptions that might be made about the growth of average real benefits, it should be realized that even if the volume of expenditure grows at the same rate as real GDP, nominal social expenditure can form an increasing proportion of nominal GDP. This is true because unbalanced productivity growth may lead to a rise in the cost of public sector output relative to a rise in the cost of private sector output (a tendency referred to as a relative price effect (Baumol (1967)). In the case of medical care and education, the size of the relative price effect tends to be measured by reference to changes in the appropriate expenditure deflator as compared with changes in the GDP deflator. However, such a comparison can be misleading since the expenditure deflator tends to be an input price deflator, which may be a poor guide to the change in the corresponding output prices.21

Projections of expenditure clearly require assumptions to be made about the future growth of average real benefits and relative prices.22 Since such assumptions are necessarily tentative, no precision is lost by making a single assumption about the growth rate of average benefits for each program-this assumption reflecting changes in both volumes per recipient and relative prices. In medical care and education, these assumptions are particularly speculative, and the reasonableness of the assumptions used can only be judged by reference to the growth in average benefits in the recent past. For example, account should be taken of a wide range of factors affecting the supply of, and demand for, medical care, particularly relating to improvements in medical technology.

For pension benefits, current program provisions (taking into account recently legislated changes) provide a basis for projecting the growth rate of average benefits. In general, such programs provide for increases in benefits to reflect either changes in the cost of living or average earnings for some combination of both). Furthermore, retirement pension programs have yet to reach maturity in some countries and average benefits will increase as programs develop. The growth in average benefits, over and above general inflation, can either be fairly accurately modeled by using actuarial studies produced by the individual countries or closely approximated.

The detailed assumptions in the study are discussed in the chapters dealing with the individual social programs (Chapters IV through VII). It is presumed that no major structural changes will occur in social programs, except those resulting from legislation already enacted.

Methodology for Projecting Economic Growth

In attempting to project the ratio of social expenditure to GDP through the first quarter of the next century, assumptions must be made not only in regard to the evolution of social expenditure, but also in regard to future economic growth. The latter will reflect changes in employment and changes in output per worker. As a result of the demographic developments described in Chapter III, the population of working age will increase through 2025 in most countries. The only country in which this is not the case is the Federal Republic of Germany, although the projected increase is small in Italy, Japan, and the United Kingdom. These developments will affect the labor force and employment, the links to the former being participation rates and to the latter unemployment rates.

Participation rates are assumed to remain close to those currently observed in the labor force. Some allowance has been made for continuing trends toward increasing female participation, increasing higher education enrollments, and a lower retirement age, although not in all countries. In the United States, it is expected that changes in pension policy already enacted will reverse the current trend toward early retirement after 2000, and this is taken into account. Combination of the demographic developments from the baseline scenarios and the participation rate projections yields the growth rates of the labor force shown in the first column of Table 2. These growth rates differ little from the corresponding growth rates of the working-age population. They range from –0.5 percent in the Federal Republic of Germany to 0.6 percent in the United States. With the exception of Canada, which is also at the top end of the range, the other growth rates are positive but small.

Table 2.Labor Market Developments as Projected in the Baseline Scenarios, 1980–2025(In percent)
Unemployment Rate
Labor Force Growth2025Employment Growth
CountryAverage annual growth rate19801984Baseline scenarioPessimistic economic scenarioAverage annual growth rate
Germany, Fed. Rep. of−− 0.5
United Kingdom0.16.913.
United States0.
Sources: OECD, Economic Outlook, various issues; and Fund staff estimates.

Official United Kingdom projections assume a long-term unemployment rate of 6 percent under all growth scenarios.

Sources: OECD, Economic Outlook, various issues; and Fund staff estimates.

Official United Kingdom projections assume a long-term unemployment rate of 6 percent under all growth scenarios.

In all seven countries it is assumed that there will eventually be some retreat from the recent peak rates of unemployment, with a tendency toward stabilization at rates close to those experienced in the late 1970s or early 1980s. Only in Italy is the baseline unemployment rate projected for 2025 higher than that experienced in 1980, while only in France is it substantially below the 1980 rate. These two cases may seem out of line with what is assumed to happen elsewhere.23 However, it should be noted that the growth rate of employment is dominated by demographic developments and is not very sensitive to what is assumed about the future unemployment rate. Table 2 shows that, in all but two countries, the rates of employment and labor force growth are identical, while in those two—France and the United States—the rate of employment growth is only marginally higher.

With the exception of France, the Federal Republic of Germany, and the United Kingdom, productivity growth in the baseline scenario is assumed to rise above the rates experienced over the past decade.24 However, as Table 3 shows, in these three countries and Japan, the declines are small. In Canada, Italy, and the United States, a significant productivity improvement is projected. In the first two cases the improvement appears of a reasonable magnitude in light of the very recent deterioration in productivity. The figures for the United States are high by the standards of any period in recent history, but they are taken from official government projections.25

Table 3.Productivity and GDP Growth as Projected in the Baseline Scenarios, 1980–2025(Average annual growth rates in percent)
Productivity Growth
1980–2025GDP Growth
Country1974–84Baseline scenarioPessimistic economic scenario1974–841980–20002000–20102010–2025
Baseline scenario
Germany, Fed. Rep. of2.42.011.511.
United Kingdom1.61.420.921.
United States0.
Sources: OECD, Economic Outlook, various issues: and Fund staff estimates.


Derived from assumed GDP growth rates and projected employment growth rates.

Sources: OECD, Economic Outlook, various issues: and Fund staff estimates.


Derived from assumed GDP growth rates and projected employment growth rates.

The employment growth rates reported in Table 2 are averages for the entire projection period. Because the age structure of the population is not projected to change steadily, labor force and employment growth rates will vary over time. This fact is reflected in the subperiod GDP growth rates shown in Table 3. These are the product of the corresponding subperiod employment growth rates and recent and projected productivity growth rates. The projected average GDP growth rate for the seven major industrial countries is 2.4 percent for 1980–2000, 2.1 percent for 2000–2010, and 1.9 percent for 2010–2025. Thus, on average, projected demographic developments are expected to drive down growth rates from the current average of about 3 percent (for the first half of 1985) to below the 1974–84 average of about 2 percent.

The pattern of growth rates will remain broadly similar to that experienced over the 1974–84 period. The only notable differences are the much improved economic performance in France and Italy (the former a result of increasing employment and the latter the result of productivity improvement), and the much lower GDP growth in Canada (the result of a marked fall in the labor force growth rate) despite improved productivity. Note also that real growth falls in Japan and the United States as employment growth slows.

The social expenditure scenarios elaborated in later chapters and the economic growth scenarios described above combine with the initial expenditure ratios presented in Table 15 to yield projections of future government social expenditure ratios. These scenarios are interrelated. For example, the growth in average benefits is generally related to the growth of income in the economy as a whole—either per capita GDP or average productivity—and the expenditure on unemployment compensation programs is a function of the assumed rate of unemployment. However, no explicit effects from social expenditure relate to economic growth, except insofar as future participation rates have been revised to reflect changes in retirement age. Thus, output is not related to the level of medical care, education, or pension benefits.26

Nor is there any link between the scenarios across countries. Of course, links are only likely to exist in the case of the social expenditure scenarios to the extent that economic developments may impose a need for restrictions on the growth of social expenditure and to the extent that economic developments are themselves linked. The failure to take explicit account of international linkages between the economies of the seven major industrial countries is unlikely to be a serious omission given the large element of common experience characterizing their long-term scenarios and the slow rates of change implied.

The social expenditure and economic growth scenarios could also be linked more directly through the aging process itself. Future expenditure on medical care, education, pensions, and other benefits respond to the aging of the population. The channels through which these responses occur, and their extent, emerge clearly from the social expenditure projections. Future economic growth also responds to aging of the population, but this factor emerges in the projections only as a change in the size of the labor force. However, as shown in Chapters III, when the population ages over the next 40 years, so does the labor force. Yet no attempt has been made to explore the economic implications of an aging labor force, and in particular, the relationship between average age and average productivity. While it seems likely that these factors are related, productivity may be influenced more by technological developments and other changes in working practices—relationships which also remain unexplored.

    Other Resources Citing This Publication