Out of Work. Richard K Vedder
functions were estimated. The discrepancy between actual spending (say, for consumption) and predicted spending is an indication of the extent that spending in a given year deviated from the normal long-run trend consistent with that income or output level. It provides a means of measuring changes in autonomous consumption (or other components of spending). Second, we related the year-to-year changes in autonomous spending to observed productivity change using a simple bivariate regression. In every case (consumption spending, gross private domestic investment spending, government purchases of goods and services, and net exports), we found no statistically significant relationship, even at the 10 percent level using a one-tailed test. Indeed, a majority of the results had a negative sign, not the relationship predicted by Keynesian analysis. On the basis of this, we believe that the proposition that shifts in components of total spending are an important determinant of productivity change is not defensible. At the same time, however, it is true, over all, that the business cycle is related to productivity changes to a moderate extent. Cycles in innovation and capital formation may impact simultaneously on productivity (Schumpeterian and Smithian changes) and output growth. This in no way detracts from the usefulness of the adjusted real wage model in analyzing the proximate determinants of unemployment variations.
There is very strong statistical evidence of a relationship between the adjusted real wage and unemployment. Higher adjusted real wages, other things equal, mean higher unemployment. The three factors determining the adjusted real wage—money-wage levels, price levels, and labor productivity—are of roughly equal importance in their unemployment impact. There is some evidence that unemployment reacts to changes in the adjusted real wage almost immediately, with the bulk of the impact felt within two years. At the same time, there is some lingering effect felt even after that lag. The productivity variable has a cyclical component to it, but the empirical evidence suggests that shifts in the aggregate demand for goods and services play no significant role in explaining this component of the adjusted real wage.
In brief, the neoclassical and Austrian view of the determinants of unemployment seems to have a great deal of validity. We now turn to a closer look at the changes in unemployment in American history, with this theory in mind, hoping to ascertain in greater detail the underlying factors that have generated both changes in the adjusted real wage and variations in unemployment.
NOTES
1. Data before 1947 were originally compiled by Stanley Lebergott. See, for example, his Manpower in Economic Growth: The American Record Since 1800 (New York: McGraw-Hill, 1964.) The Lebergott estimates, slightly modified, serve as the basis of the official Bureau of Labor Statistics data. See U.S. Department of Commerce, Bureau of the Census, Historical Statistics of the United States, Colonial Times to 1970 (Washington, D.C.: Government Printing Office, 1975), p. 135. For recent unemployment statistics, see the Economic Report of the President 1991 (Washington, D.C.: Government Printing Office, 1991), p. 330.
2. See ibid., p. 346. In the postwar period, the growth of fringe benefits has led to a growing divergence between money-wage growth and the growth in total money compensation costs. The latter concept is used here, as fringe benefits are part of the true compensation package. Also, businesses presumably consider their total compensation costs in making employment decisions.
3. The critical wage series used in calculating our wage measure is found in Historical Statistics of the United States, Series D-724. They were calculated by the U.S. Office of Business Economics as a by-product of calculating the national income and product accounts. Because of lags in the model, data for the 1890s were needed. Data for that decade, obtained from Series D-735, ibid., were spliced to the data series for 1900 to 1946. The 1890s data were originally estimated by Lebergott, Manower in Economic Growth. The 1890–1946 data, in turn, were spliced to the data for the post-1947 era and turned to index number form, with 1982 = 100. Annual hours worked were obtained by using weekly hours data and multiplying by 50. The hours data for the 1890–1918 period came from Historical Statistics of the United States, Series D-767; for 1919–1946, the data are from Series D-803. The 1890–1918 data were originally compiled by Paul H. Douglas in his Real Wages in the United States (New York: Houghton Mifflin, 1930).
4. Series D-683.
5. John W. Kendrick, Productivity Trends in the United States (Princeton, N. J.: Princeton University Press for the National Bureau of Economic Research, 1961).
6. As reported annually in the Economic Report of the President (on p. 346 of the 1990 edition), and monthly in Economic Indicators (on p. 16.)
7. As reported in Historical Statistics of the United States, pp. 211-12, and in the Economic Report of the President 1990, p. 359. The pre-1946 data were spliced to the post-1946 data using the 1982–84 base years = 100.
8. The regression reported included two ARIMA adjustment terms, not reported, so as to deal with the existence of serial correlation. The reported results have no significant evidence of serial correlation. The unreported F-statistic is healthily large (over 114).
9. The 0.6 figure is derived by multiplying the regression coefficient, .315, by the median change (ignoring signs) in the adjusted real wage, 1.85, obtaining 0.583.
10. A detailed theoretical derivation of this form of the model is presented in the appendix.
11. Three ARIMA terms created to eliminate autocorrelation are not indicated. The results are not significantly impacted by the nature of the adjustments made to deal with the serial correlation issue.
12. “Changes in the Cyclical Sensitivity of Wages in the United States, 1891—1987,” American Economic Review 82 (1992): 122–40.
13. See Ludwig von Mises, Human Action, 3d rev. ed. (Chicago: Henry Regnery Company, 1966), pp. 350-57 for a flavor of the philosophical objections that the Austrians have to the empirical approach. Friedrich von Hayek said similar things. Speaking of the Austrian theory, he said that it “cannot by its very nature be tested by statistics.” See Chiaki Nishiyama and Kurt R. Leube, The Essence of Hayek (Stanford: Hoover Institution Press, 1984), p. 7.
14. Our favorite is Lawrence Leamer, “Let’s Take the Con Out of Econometrics,” American Economic Review 73 (1983): 31–43.
15. See his “The Rhetoric of Economics,” Journal of Economic Literature 21 (1983): 481–517. For a more extended treatment see McCloskey’s The Rhetoric of Economics (Madison: University of Wisconsin Press, 1985).
16. Christina Romer, “Spurious Volatility in Historical Unemployment Data,” Journal of Political Economy 94 (1986): 1–37; Michael Darby, “Three-and-a-Half Million U.S. Employees Have Been Mislaid: Or, an Explanation of Unemployment, 1934–41,” Journal of Political Economy 84 (1976): 1–15.
17. Reviewing several attempts to estimate unemployment rates for the critical decades of the 1920s and 1930s, Gene Smiley seems to conclude that Lebergott’s approach was quite valid. The Smiley evaluation was written before the Romer estimates were compiled. See Gene Smiley, “Recent Unemployment Rate Estimates for the 1920s and 1930s,” Journal of Economic History 43 (1983): 487–493.
18. Paul A. David and Peter Solar, “A Bicentenary Contribution to the History of the Cost of Living in America,” in Research in Economic History, Paul Uselding, ed., vol. 2 (Greenwich, Conn.: JAI Press, 1977), pp. 1-80.
19. The data, originally compiled by the U.S. Department of Labor, come