¹ If these items were included, the percentages would be considerably higher and probably by a bigger proportion today than in 1900.

² Rent is treated as payment for services purchased from other industries. There is a question about taxes, but I pass over it.

³ As indicated, lack of information on military assets and roads and streets is especially annoying because of their importance. Difficulty is encountered, also, with the estimation of depreciation reserves needed to calculate net values of government capital assets.

⁴ These percentages are weighted averages of the percentages for employment and capital assets referred to earlier. The weights are based on the ratio of payrolls to the imputed (4 per cent) return on capital.

⁵ The government expenditures mentioned earlier are even more inclusive, since they cover transfers to the public and sometimes even transfers to other governments (grants-in-aid), as well as payrolls, outlays, and purchases of currently used goods and services.

⁶ Leontief, Wassily, “Recent Developments in the Study of Interindustrial Relationships,” and accompanying discussion, American Economic Review, XXXIX, Papers and Proceedings (May 1949)Google Scholar, and The Structure of the American Economy (Cambridge: Harvard University Press, 1941).Google Scholar For a brief explanation, see Ruggles, Richard, An Introduction to National Income and Income Analysis (New York: McGraw-Hill Book Co., 1949), chap, viGoogle Scholar; Copeland, M. A., “Tracing Money Flows Through the United States Economy,” American Economic Review, XXXVII, Papers and Proceedings (May 1947)Google Scholar; Hart, A. G., Uses of National Wealth Estimates and the Structure of Claims (Studies in Income and Wealth, Vol. XII; New York: National Bureau of Economic Research, 1950).Google Scholar

⁷ Other problems involved in this type of measurement, which can merely be listed, are the degree of detail in the classification, that is, the number of industries to be distinguished in the table; the treatment of investment, depreciation, and imputed items; the treatment of intrasector flows; and, by no means least, the difficulties of obtaining data once decisions have been made on these and other conceptual and procedural problems.

⁸ Walker, M. L., Municipal Expenditures (Baltimore: Johns Hopkins University Press, 1930)Google Scholar; Ridley, C. E. and Simon, H. A., Measuring Municipal Activities (2d ed.;Chicago: The International City Managers’ Association, 1943).Google Scholar

⁹ Productivity is defined here as the ratio of output to total input, the latter being measured by the weighted combination of employment and real capital assets. Because real assets per worker changed but little on net balance over the fifty years, change in productivity so measured is not much different from change in the conventional index of output per worker.

¹⁰ Upson, L. D., “The Growth of a City Government,” School of Public Affairs and Social Work of Wayne University, Report No. 36, April 1942Google Scholar; Anderson, H. D., Our California State Taxes, Facts and Problems (Stanford University, Calif.: Stanford University Press, 1937), chap. i.Google Scholar

¹¹ Report of the Special Joint Committee on Taxation and Retrenchment, “State Expenditures. Tax Burden, and Wealth,” No. 68 in New York Legislative Documents (Albany, 1926).Google Scholar

¹² If it were applied, one problem would be to decide how much the degree of detail in the industrial classification utilized affected the results of the analysis, for the results hinge on that detail.

¹³ This is to avoid being diverted by interstate differences in the way in which functions are divided among types of government within the state.

¹⁴ Differences in the prices of commodities and services (other than labor) seem relatively small, to judge from the WPA-BLS measures of intercity differences in cost of living and from Koffsky's estimates of rural-urban price differences. See M. L. Stecker, “Intercity Differences in Costs of Living in March 1935, 59 Cities,” Works Progress Administration, Research Monograph XII, 1937, and later publications of the Bureau of Labor Statistics; Koffsky, N., Farm and Urban Purchasing Power (Studies in Income and Wealth, Vol. XI; New York: National Bureau of Economic Research, 1949). These data seem to suggest also that differences among city sizes are not great. However, the same is not true of salary levels, and salaries make up over half the expenditures on operation. We know that government salary levels rise substantially with increasing size of city, even after allowing (as we must) for the relatively greater importance of part-time work in the smaller municipalities and counties, and the lesser training of their personnel, and the same is probably true of rentals and similar costs. As we shall see, there is also a positive correlation between level of government expenditures and degree of urbanization. It follows, then, that interstate price differences do play a part.Google Scholar

¹⁵ Here also arise questions on the influence of prices and productivity, but, again, they do not seem sufficient to account for the finding.

¹⁶ In measuring income, for example, one might want to take into account income originating in a state, as well as income received by it, and Federal grants as well; all are related to the fiscal capacity that in part underlies government expenditures. Urbanization could be measured by the simple proportion of population in all or large urban places or by a more complicated procedure—to take account of the distribution of population among different-sized cities. And similarly sophisticated measures could be applied to density rather than the simple measure of population per square mile or rural population per square mile.

¹⁷ There is a question also of the kind of equation or surface to fit, at least to start with. The above results are based on a linear equation. A second degree parabola (less the product terms) yields a somewhat higher multiple correlation coefficient but otherwise, substantially the same results.

¹⁸ The price factor mentioned earlier plays a part in accounting for some of the correlation between income and expenditures and the magnitude of the income elasticity. It may be sufficient to note here, however, that when government employment rather than expenditures is used as the dependent variable the results do not appear to be substantially altered, to judge from some preliminary calculations.

¹⁹ A measure of urbanization devised to take into account size of city, larger cities being weighted more than smaller, would have been even more highly correlated with income.

²⁰ That is, the surface must pass through the point defined by the mean of the system. We have no state data on incomes in 1898–1902, it is true, but we have a rough national estimate and, therefore, the basis for a first approximation to the mean state income per capita. Both income and expenditures in the earlier period need, of course, to be put into the prices of the later period.

²¹ The one point we know to lie on the earlier surface does not also lie on the recent surface, but the difference between the two surfaces in the vicinity of that point is of reasonable sign and not great.