Public Spending And Social Progress

Studies | Federal Spending | Social

No. 232
Thursday, June 01, 2000
by Gerald W. Scully


  1. A fuller and more technical version of this paper will appear as Gerald W. Scully, "Government Expenditure and Quality of Life," Public Choice, forthcoming.
  2. See Gerald W. Scully, "Measuring the Burden of High Taxes," National Center for Policy Analysis, Policy Report No. 215, July 1998.
  3. Vito Tanzi and Ludger Schuknecht, "The Growth of Government and the Reform of the State in Industrial Countries," IMF Working Paper, 1995, International Monetary Fund, Washington, D.C.
  4. Tanzi and Schuknecht use the following social indicators: the unemployment rate, infant mortality, the death rate, years of primary schooling, U.N. rank in human development, income share of the lowest 40 percent, illiteracy rate, secondary school enrollment, life expectancy, prisoners per 100,000 population, divorce rate, emigration rate and some economic indicators. Ibid., p. 26.
  5. Ibid.
  6. See J. Burkett, "Systematic Influences on the Physical Quality of Life: A Bayesian Analysis of Cross-Sectional Data," Journal of Comparative Economics, 1995, 9, 145-63, and Rati Ram, "International Inequality in the Basic Needs Indicators," Journal of Development Economics, 1982, 10, 113-17.
  7. The data are contained in George T. Kurian, Illustrated Book of World Rankings 1997 (Armonk, N.Y.: M.E.Sharpe, 1997).
  8. In order to aggregate indicators that are measured by a variety of units, the indices that are not naturally scaled from 0 to 100 were so scaled. Thus, for attributes where the maximum value represents the highest achievement (e.g., life expectancy), the index was constructed as a ratio of the measured value of the attribute to the maximum value for the attribute in the data set. For attributes where the minimum value represents the highest attainment (e.g., infant mortality), the index was constructed as an inverse ratio of the value of the attribute to the maximum value. Such a scaling makes it possible to aggregate indicators that are measured by a variety of units.
  9. A literacy rate of 100 percent has to be treated with some caution. Countries define literacy differently (e.g., an ability to read a simple sentence, having attended school). The literacy rate for the United States is given as 95.5 percent. This high percentage is rather hard to believe, given the quality of public schooling. On the basis of a definition of "functional literacy" (e.g., ability to read a bus schedule, fill out an employment application), the literacy rate would be much lower. Moreover, IQ has a mean of 100 and a standard deviation of 15. This means that any population will have two-thirds in the interval 85 to 115 and 95 percent in the interval 70 to 130. It is hard to believe that anyone with an IQ below about 80 points is "functionally literate."
  10. See Amartya Sen, Commodities and Capabilities (Amsterdam: North-Holland, 1985) and his The Standard of Living (Cambridge: Cambridge University Press, 1987).
  11. Daniel J. Slottje, Gerald W. Scully, Joseph G. Hirschberg and Kathy J. Hayes, Measuring the Quality of Life Across Countries: A Multidimensional Analysis (Boulder, Colo.: Westview Press, 1991), pp. 24-58.
  12. Gary S. Becker, A Treatise on the Family (Cambridge: Harvard University Press, 1981). Earlier, Becker solved two important riddles. Beginning about 1870 for the advanced countries, the birth rate began a systematic decline that continues to this day. Less developed countries are now in various stages of a decline in the birth rate. This decline in birth rates is of such magnitude that rather than talk about the doom of a world population explosion and a Malthusian notion of widespread famine as population growth outstrips the food supply, there is talk now of a population implosion (i.e., a shrinkage in population as replacement levels fall below deaths). Demographers had no theory of the change in population growth before Becker provided an economic theory of the decision to have children and how many to have. The second riddle was why in the face of rising real incomes people were choosing to have fewer children. The explanation offered had two parts: (i) as income rises the value of time increases, children are the most time-intensive form of consumption, and people substitute out of time-intensive forms of consumption into less time-intensive forms, and (ii) parents substitute quality for quantity of children (i.e., they invest more resources in the children, now, than in the past). This notion of the substitution of quality for quantity is an early example of hedonic change in goods that make interperiod comparisons of living standards tenuous (e.g., the CD of Ludwig Van Beethoven's Eroica played on a good stereo in 1998 is of rather different sound reproduction quality than on a record played on an RCA Victrola of 1930 vintage, yet both are recording players).
  13. The various attributes that are used to construct the index of political rights and the index of civil rights are discussed in detail in Gerald W. Scully, Constitutional Environments and Economic Growth (Princeton, N.J.: Princeton University Press, 1992), pp. 106-16.
  14. These are scaled by Kurian from 0 (best) to 10 (worst), but I have rescaled them from 0 (worst) to 10 (best).
  15. James D. Gwartney and Robert A. Lawson, Economic Freedom of the World: 1997 Annual Report, Vancouver: The Fraser Institute, 1997, pp. 18-23.
  16. Some of the attributes capture income distribution effects. To verify this statement, the 12 indicators, IA1 to IA12, were correlated with three measures that capture some dimension of the income distribution. These measures are the human suffering index (Kurian, Illustrated Book of World Rankings 1997, pp. 375-76); poverty measured by the percent of people living on less than $1 (valued by the Purchasing Power Parity method) a day (World Development Report, 1997, pp. 214-15); and an Atkinson inequality index (aversion parameter equal to .95) (Scully, Constitutional Environments and Economic Growth, pp. 197-99). All 12 indicators are significantly (95 percent level) correlated with the human suffering index (N = 103), with a range from -.34 (IA5) to -.91 (IA8). Eleven of the 12 attributes are significantly correlated with the poverty measure (N = 53), with a range from -.28 (IA4) to -.62 (IA1). Nine indicators are correlated with the inequality index (N = 62), with a range from -.38 (IA9) to -.56 (IA4). The poverty and suffering indexes are significantly correlated (r = .62, N = 52), as are the suffering and inequality indexes (r = .49, N = 62). But the poverty and inequality indexes are not significantly correlated with each other (r = .07, N = 31).

    The significant link between political, civil, and economic liberty and income inequality (more freedom, more equality of income) has been shown before (Ibid, pp. 184-99).

    The attributes also are correlated with per capita GNP (PPP). For the 12 attributes (N = 112), the correlations range from .31 (A5) to .94 (A8). For the other attributes (N = 85), there is an inverse correlation with the crime rate (A13) of -.30, and positive correlations with political (A14) and civil (A15) rights of .63 and .72, respectively. The index of economic rights, which captures a variety of institutional and policy matters, has a correlation coefficient of .65 with per capita GNP.
  17. For instance, consumption by United States households in the lowest 20 percent of the income distribution averaged $13,957 in 1993, while their income averaged only $6,395. Unreported income, noncash benefits and spending out of household wealth are some of the reasons the poor can consume more than their reported income. See "Consumer Expenditures in 1993," U.S. Department of Labor, Bureau of Labor Statistics Report 885, December 1994; cited in Bruce Bartlett, "How Poor Are the Poor?" National Center for Policy Analysis, Brief Analysis No. 185, October 25, 1995.
  18. Principal components analysis measures the amount of independence in a group of variables. The original data, X, is transformed into a new data set, Z, where the variables are pairwise uncorrelated. Hence Z = XA, and A'X'XA = D(li), where D(li) is the diagonal matrix of eigenvalues, which represent the variances of the Z matrix. Specifically, li/Sli = Pi is the proportional contribution of each of the principal components to the total variation in the X matrix, and SPi = 1.

    For the hedonic estimation, real GNP (PPP, purchasing power parity adjusted) per capita was used as the instrumental variable. Thus, the regression IY = b0 + Sbi IA + e was estimated, and the bi's were normalized to sum to unity.

    For a more detailed analysis of the methodology of constructing aggregate indexes with principal component and hedonic weights, see D.J. Slottje, G.W. Scully, J.G. Hirschberg and K.J. Hayes, Measuring the Quality of Life Across Countries (Boulder, Colo.: Westview Press, 1991), pp. 59- 67, and Scully, Constitutional Environments and Economic Growth, pp. 120-27, 130-31.
  19. The range of "maximal" estimates is from 72.5 to 82.5 percent of the observed maximum government expenditure per person of $5,307.5.
  20. It is outside the scope of this paper, but some developing countries have experienced negative economic growth since the 1960s. They are able to get by for the time being due to unreported black market activity, foreign aid from wealthy welfare states and by consuming their capital stock. As a consequence, those countries have made very little social progress, in a few, conditions have worsened.
  21. The percentages were obtained by dividing GNP per capita by the average of the six estimates for the polynomial model of MB=0, and all 18 estimates of MB=ME, in Appendix Table II. Government expenditures are usually expressed as a share of gross domestic product (GDP). Until the 1980s national production of goods and services, or gross national product (GNP), included some items that represented foreign wealth. This was altered to GDP to reflect more accurately wealth generated domestically. GNP and GDP figures differ slightly, but not enough to create a serious distortion.
  22. Tanzi and Schuknecht include Chile and Korea as nations that have achieved a high level of social progress with comparatively small fiscal governments. In 1995, government expenditures as shares of GDP were 19.2 and 20.4 percent, respectively. The value of the average of the six indexes of social progress is 66.3 for Chile and 70.5 for Korea. For the narrower PQL index (equal weights), the value is 70.3 for Chile and 68.1 for Korea. Singapore has a PQL value of 86.8.
  23. It is likely that Hong Kong would also have a high value of social progress, had the data been available to calculate the indices.
  24. For nations farther down the ranks of per capita GNP, social progress indicators are so low that those nations would have to spend more than their entire GNP annually in order to reach MB=0. Note that 22 countries are chosen for the sample in this paper, but 26 countries are used in this calculation.
  25. James Gwartney, Robert Lawson, and Randall Holcombe in "The Size and Functions of Government and Economic Growth," Joint Economic Committee, Washington, D.C., April, 1998, pp. 24-26 have estimated "core" government expenditures as a fraction of GDP for several advanced nations. These core functions are expenditures related to the protection of persons and property, national defense, education, monetary stability and physical infrastructure. The range in the estimated core functions is from 9.1 percent for Germany to 13.8 percent for the United States, with an average value of 11.9 percent.
  26. William J. Baumol, "Macroeconomics of Unbalanced Growth: The Anatomy of Urban Crisis," American Economic Review, 1967, 57, 415-26.
  27. William A. Niskanen, Jr., Bureaucracy and Representative Government (Chicago: Aldens and Atherton, 1971).

Read Article as PDF