TABLE 3.-ANNUAL GROWTH RATES, 1950-77, OF GOVERNMENT SPENDING AND GROSS DOMESTIC PRODUCT IN

Source: January 1979, "Economic Report of the President," table B-77, p. 272.

The growth rates of Table 3 are presented here as a simple benchmark to which the reader may refer in interpreting more complex data on government spending. This introduction would be incomplete, however, without the summary measures of changes in size of the public sector calculates from constant-price data. Median ratios of government spending to GDP, in real terms, are displayed below:

Medians for each category are drawn separately from each array; it is only a coincidence that components add to the total for 1977. These ratios show that the relative importance of transfer outlays grew markedly over the period; that the group of thirteen countries scarcely increased the share of GDP denoted to collective consumption; and that the total public sector, in real terms, expanded from about a fifth of GDP in 1950 to one-third in 1977.

PART II. BEHAVIOR OF GOVERNMENT SPENDING FOR COLLECTIVE CONSUMPTION

A half-century ago, when government spending in the United States amounted to ten percent of gross national product (GNP), government consumption expenditure accounted for 85 percent of total government expenditure. By 1978, when the U.S. public sector had expanded to nearly one-third of GNP, government spending for collective consumption had declined to 63 percent of the total. In the intervening years government had become a major redistributor of income, in the United States and other countries examined here. An increasing proportion of public revenues was diverted from the provision of collective services to the financing of transfer programs. The nature and consequences of this shift in the composition of government spending will now be systematically explored.

Part II focuses on the behavior of consumption expenditure-in real and nominal terms. Methodological issues associated with the deflation of government expenditure are an integral part of the analysis. Transfer outlays are discussed in Part III. The indexes of Table 4 measure the absolute change in government consumption expenditure over the study period. Calculated from data expressed in national currencies, they facilitate comparison between countries; and, for each country, they simplify comparison of nominal and real expenditures. The original data are displayed in Appendix Table A-1, for intervening as well as terminal years. Volume indexes of Table 4 are rounded versions of those shown in the Appendix, and may differ slightly from the latter.

Data for 1929 from US Department of Commerce, The National Income and Product Accounts of the United States, 1929-1974, p. 339; for 1978, from Survey of Current Business, May 1979, tables 12 and 13.

TABLE 4.-GOVERNMENT SPENDING FOR COLLECTIVE CONSUMPTION, 1977: INDEXES OF AMOUNT, COST AND

1 Index values shown are rounded versions of calculated values.
Source: January 1979, "Economic Report of the President," table B-77, p. 272.
Nominal versus real gains

An extremely wide range of increases, 1950-1977, in nominal expenditure for collective consumption is evident from column 1. The smallest increase-7.6 times-occurred in Switzerland, and the largest-40.6 times-took place in Greece. An inquiry into the reasons for this widely divergent behavior cannot be undertaken here, although some details on the changing composition of government expenditure are introduced below.

Attention is directed instead to a major factor responsible for the increases in nominal expenditure-the cost increases displayed in column 2. These ranged from a low of 261 percent in the United States to a high of 976 percent in Greece. Switzerland had the second smallest increase-281 percent-among the thirteen countries. The median increase in this deflator for government consumption-681 percent was nearly twice as large as the rise in the GDP deflator-369 percent. Characteristics of the deflator are discussed below.

Column 3 displays the indexes which are central to this investigation-those which measure the increases in real expenditure for government consumption. The volume increases range from a low of 76 percent in the United Kingdom to 3 high of 300 percent in Canada. Switzerland is again second lowest-with a gain of 123 percent-while Greece (along with Denmark, Finland, and Germany) increased its volume by nearly as much as Canada.

The median increase in volume-176 percent-belongs to the United States, one of the six countries that experienced a rising real share of resources devoted to collective consumption. Of the remaining five, only Ireland failed to achieve an above-average increase in real expenditure for government consumption.

In Part V the increases in government consumption expenditure (and the growth in transfer outlays) will be related to changes in GDP. We will then be in a position to strike the ratios which measure changes in public-sector size. We turn next, however, to an examination of growth rates in government consumption, over the entire study period and successive sub-periods.

Annual growth rates of collective consumption

Over the full period, 1950-1977, government spending for collective consumption grew at an average annual rate of 3.8 percent, in real terms. The median growth rate of nominal expenditure was 12.0 percent. Within the array of growth rates nations occupy the same relative position as they did in the array of volume indexes (columu 3 of Table 4).

• The six countries are those with elasticity coefficients for real G. greater than unity. See the middle column of Table 1.

TABLE 5.-ANNUAL GROWTH RATES OF GOVERNMENT CONSUMPTION EXPENDITURE IN CONSTANT PRICES

Source: January 1979, "Economic Report of the President," table B-77, p. 272.

A closer look at Table 5 reveals the presence of a slowdown since the NineteenSixties, and a sharp decline in the final two years of the study period. The median growth rate rose from 3.7 percent in the Fifties to 4.5 percent in the Sixties, then dropped to 4.2 percent in the Seventies (through 1977). In 1976 the increase in real expenditure averaged 3.3 percent, and in one case (Denmark) the 1976 volume was slightly less than in the preceding year. The concluding year 1977 was marked by a further decline in the median growth rate-to 2.4 percent and included two cases of "no growth" (Switzerland and the United Kingdom) in the volume of consumption spending by government.

Deflating government expenditures: some methodological issues

The cost indexes used to obtain real expenditure for collective consumption were calculated from published data on expenditure in current and constant prices. Implicit price deflators for government consumption were not published by the U.N. until 1976. The reader is cautioned that the deflators used in this study may contain small amounts of error due to re-basing and splicing of deflators. This section addresses several of the issues associated with the problem of deflating government expenditures.

First, some comments are in order on the propriety of using separate deflators for the numerator and denominator of ratios representing public-sector size. In their seminal study of British public expenditures Peacock and Wiseman made only limited use of separate deflators for government expenditure. For their study period, 1890-1955, the divergence between nominal and real ratios of government spending to GNP was so slight that the issue of separate deflators did not arise.' They have since acknowledged that, for other periods and places, "deflation. can make a considerable difference." The need for separate deflation arises from the recognition of a productivity gap between the public and private sectors which has, in turn, given rise to the "relative price effect" of public expenditure analysis.' This writer agrees that the "issue" of separate deflators is a non-issue. In deflating a value series, the statistician is obligated to use the price (or cost) index that most clearly approximates the content of the value series. That principle has guided the conduct of this investigation. If the deflated data are then used to form a ratio measuring real size of the public sector, the validity of the ratio is unaffected by separate deflation of its terms.

A more serious question arises from the nature of the "government consumption" deflator. Like the GDP deflator, of which it is a major component, it is a Paasche index weighted by current. year quantities. The imputs are (a) services

? Peacock and Wiseman (1961), table A-6, p. 166. A later edition of their influential book (1967) does not contain any revision of the statistical data.

Peacock and Wiseman (1979), p. 12.

Peacock has also elaborated the need for separate deflation in an unpublished paper "The Expanding Public Sector" (his Madrid lecture of November 1978) wherein he states: "There is no reason why the deflator for GNP should be the same as for government expenditure . . . most studies of long-term expenditure growth have largely ignored this problem." (p. 6)

of government employees and (b) purchases from private firms supplying government agencies.10

Changes in quality of goods and services purchased by government cannot be captured by this deflator. Higher prices paid for these inputs drive up the index value of the deflator, but do not ncessarily raise the real cost of "government output." A more costly imput has been purchased, and presumably the public gets the benefit of a more valuable "output." In any case, because of this deficiency, the 'government consumption' deflator tends to overstate the rise in cost and to understate the increase in real expenditure.

There is no obvious solution to this problem. The US Department of Commerce has constructed a "fixed weighted price index" for total GDP, but detailed indexes for components of GNP are not yet available. As for the other countries included in this study, recent revisions of data submitted to the UN have significantly enhanced the quality of deflators needed to develop a real expenduture series for the public sector.

The Paasche index used here to deflat government consumption expenditure has been found wanting. That does not means, however, that a Laspeyres index is necessarily better. One recent study of government spending in 34 countries employed the consumer price index, which uses base-year weights, for the total and consumption component of government spending."1

The result is an exaggeration of the growth of real expenditure. There is room for disagreement over the best method of deflating a value series, but not over the use of inappropriate deflators.

Problems peculiar to the deflation of transfer outlays and total government spending are discussed in the next two chapters. The remaining issue to be considered here is whether the "consumption" deflator is affected by the 'productivity gap' between the private and public sectors.

The hypothesis of a productivity gap is based on the following premises: 1. That government activities are, for the most part, labor-intensive; 2. That technological advances have little, if any, impact on the quantity of labor needed to provide public services; and

3. That, in the absence of competitive market pressures, government agencies have little incentive to improve efficiency or productivity.

If the hypothesis is accepted, it follows that the real cost of government serv ices must rise relative to costs in the more productive private sector. With the rapid unionization of public employees, it is also likely that government wage rates will tend to keep up with those of private industry where the productivity factor is now a standard feature of collective bargaining.

A widely quoted exposition of this hypothesis is that of Baumol (1967) in which the lagging productivity of the public sector is credited with major responsibility for "the urban crisis" of the mid-Sixties. An elegant mathematical formulation of the productivity gap, Baumol's article gives scant recognition to erosion of the tax base, increasing concentration of low-income families in older cities, higher costs of public-safety programs, and other causes of the urban fiscal predicament. "Baumol's disease", as his theory is sometimes called, is nevertheless a useful addition to the literature on public expenditure growth, and helps to explain why the nominal size of the public sector tends to increase over time.12

Measures of productivity for the private sector are useful tools for analyzing the effects of mechanization and technological progress. A rise in output per worker usually means that the capital-labor mix has shifted toward greater use of equipment and less reliance on labor input. The alert trade-union negotiator, however, is quick to claim a share of the productivity gain on the ground that output-per-manhour has risen-he does not have to prove that his members are working harder.

10 The relative importance of these components has undoubtedly changed over the study period, but the source material does not provide a basis for decomposing the aggregate. For the United States, however, estimates can be made from the National Income and Product Accounts: between 1955 and 1976 the proportion of government consumption represented by employee Compensation in constant dollars, rose from 46 percent to 53 percent. (Beck 1979a).

Wagner and Weber, 1977. The authors explain that this was the only index available in the source material for the entire observation period.

12 Baumol and Oates (1975) contains a refinement of the "cost-disease problem" (in Part III, Provision of Public Services, pp. 235-66), but no empirical evidence on the productivity gap. Several writers in this field, inspired mainly by Peacock and Wiseman's pioneering study, have attempted to quantify the pro ductivity gap. Among the more interesting studies are those of Andic and Veverka (1964), Gupta (1968). Holmans (1968), Tussing and Henning (1974), and Veverka (1963).

In the public sector the concept of productivity just described has little or no relevance. The vast majority of public employees are providing a service which cannot be reduced to homogeneous units of output. What is relevant here is that the public sector provides fewer opportunities than the private sector for mechanization of operations.

In public education, for example, the advent of teaching machines has meant that fast learners are no longer held back by their classmates; but the teacher's productivity is unaffected by these mechanical aids to instruction."

Professors at public universities can greatly increase their research output by tapping the data base stored in a computer. If the more 'productive' scholar receives an increase in salary, does the increase represent a gain in the value of government product? 14

If there is a productivity gap between the public sector and private production, the effects are imbedded in the deflator for each sector of the economy. Neither the "government consumption" deflator nor the GDP deflator with which it is compared can be adjusted for changes in the quaiity of inputs. Hence, in both cases, the increase in volume is misstated by an indeterminate magnitude.

III. GROWTH OF GOVERNMENT TRANSFER OUTLAYS

Attempts to explain public-sector growth, in real terms, seldom deal explicitly with the problems posed by government transfer outlays. An integral part of total expenditure, these outlays are always included in the numerator of ratios representing nominal size of the public sector. They are usually absent, however, from measures depicting real size of the public sector.

An example is a recent report of the OECD (Organization for Economic Cooperation and Development) on government spending since the mid-1950s in member countries.15 This long-awaited report, summarizing research that began more than a decade ago, contains much helpful information on the behavior of nominal expenditure. The reader will search in vain, however, for an answer to the question: "Did the real size of the total public sector increase, decrease, or remain constant over the study period?"

Evidence on this point is particularly urgent in view of the OECD concern with "the appropriate allocation of resources to different public and private needs."

9316

Government transfer expenditure does not involve the movements of resources to the public sector. Within the private sector a re-allocation of resources occurs as revenue collected from taxpayers is re-distributed to beneficiaries of the transfer programs. The OECD report is silent on this issue.

The enterprising scholar, who does not ordinarily have access to the kind of data assembled by the OECD Secretariat, will appreciate the statistics published in this report, on taxes as well as government spending. But he will find therein only one table (page 18) dealing with real expenditure, for "public final consumption expenditure" as a percentage of GDP, in 20 OECD countries. Summary

The data are unweighted averages and, despite differences in timing and coverage, generally correspond to the findings of this study: (1) that the nominal ratio of G. to Y rose modestly over the study period, and (2) that the real ratio was approximately stable.

1 In New Jersey public schools, teachers and administrators are now being judged on the performance of their pupils in standardized reading and "math" tests. Test scores have a place in evaluating a teacher or a school system, but they are not measures of productivity.

14 As used in the national income accounts of the United States, "government product" is simply another name for compensation of general government employees. A deflator for this component of government spending is not published by the Department of Commerce, but estimates for 1978 are available in Beck (1979b), table 1. Since 1953 this deflator has risen less rapidly-2064 percent-than the price index of goods purchased from private firms-253 percent.

OECD, Public Expenditure Trends, July 1978 (Studies in Resource Allocation, no. 5.)

10 Ibid., p. 5. The Working Party responsible for the Report warns that the statistical data must be interpreted with caution; but it provides little or no guidance on the allocative effects of transfer outlays.