¹ Gordon, Robert T. and Wilcox, James A., “Monetarist Interpretations of the Great Depression: An Evaluation and Critique,” in Brunner, Karl, ed. The Great Depression Revisited (Boston, 1980), p. 96.Google Scholar

² We might also suppose that I = Ī + a' W, where Ī is the truly autonomous component of investment and a' expresses the effect of wealth through Tobin's q, the required marginal efficiency of new capital or some other channel. The addition of such an equation to this system would not alter the formal structure of Equation (4). However, it would reduce the magnitude of the coefficient A and increase a, thus shifting the IS curves to the left and reducing their slopes as drawn in Figure 1.

³ U.S. Bureau of the Census, Historical Statistics of the United States, Colonial Times to 1970, Part 1 (Washington, D.C., 1975), p. 135.Google Scholar

⁴ Ibid.

⁵ The division is not necessarily exhaustive. The equal weights assigned to population growth and labor growth seem roughly consistent with the fact that 36.6 percent of all real, non-farm, domestic capital formation from 1899 to 1929 was investment in non-farm residences; see Kendrick, John W., Productivity Trends in the United States (Princeton, 1961), pp. 320–21. A fuller measure of population-sensitive investment would add to residential construction the creation of schools, commercial buildings, utilities, transportation systems, and so on.Google Scholar

⁶ In addition to being periods of relatively high employment, each of the periods compared contains the peak of a long swing in demographic and investment data.Google Scholar

⁷ The higher levels of the calculated investment ratio might be due to an upward bias in the crude estimate of the marginal capital/output ratio is Col. (1) or a tendency for the actual investment ratio to fall below the warranted whenever output falls below “full employment.”Google Scholar

⁸ [Board of Governors of the Federal Reserve System], “The Quarterly Econometric Model, 1978” (Washington, D.C., [1978]). The coefficients shown (with t values underneath) are sums of coefficients on lagged quarterly values going back two or three years. The entire equation has been dramatically simplified and many channels of causation not relevant for my paper have been omitted. The coefficients of S were restricted a priori to sum to one half of the total of the coefficients of (W – S).Google Scholar

⁹ Computed from Historical Statistics, p. 15.Google Scholar

¹⁰ This value of S/W is the mean of the first four values in Table 2.Google Scholar

¹¹ In order to use Equation (9) as a substitute for Equation (1) in the algebra of constructing an IS curve (Equation 4), it is necessary to make two adjustments. First, the coefficient b must be multiplied by the value of POP₁ and by the ratio YD/Y. The average ratios of DPI less transfer payments to GNP were, 1900–09: 0.82; 1921–29: 0.81; 1958–77: 0.62. Second, transfer payments excluded from YD must be added to the constituents of A. This is appropriate, since the full version of Equation (9) estimates a propensity to consume out of transfer payments equal to 1.

I take (I + G + F)/Y* = 0.24 in 1900–09 and 0.22 in 1921–29. An average government deficit of 1.5 percent raises this ratio to about 0.235 in 1958–77. Transfer payments add an additional 1 percent of GNP to this number in the 1920s and a hefty 20 percent in 1958–77. The redefined values of A/Y* are 0.24, 0.23, and 0.435 respectively. The values of the multiplier, 1/(1 – b) in equation (4), are 2.3, 2.2, and 1.7. The zero wealth intercepts of the IS curves, A/Y*/l – b, are 0.544, 0.508, and 0.739. The slopes of the IS curves, (1 – b)/a, are 9, 9.2, and 11.2.

The 4.5 percent shortfall cited in the text is a calculation of the distance between the point where curves with the first and second set of intercepts and slopes would intersect an LM curve assumed to be horizontal over the relevant range at the level W/Y* = 4.09. This distance in turn is the value necessary to assure an intersection in 1900–09 at Y = Y*. (The actual average of W/Y for 1900 and 1912 from Table 2 is 4.06.) The potential shortfall has been exaggerated to the extent that the income loss between E₁ and F in Figure 1 exceeds the loss between E₁ and E₂. Subsequent calculations of the vertical shift in the LM curve required to restore high employment depend only on the distance from E₁ to E₃ and are not affected by this simplification.

¹² The required increase in W/Y* is (9.2) x (.045), the sope of the IS curve times the shortfafl from high-employment. See previous footnote.Google Scholar

Applying the same techniques to the transition from the 1920s to 1958–77 implies that W/Y* had to fall by 1.59 to maintain high-employment equilibrium and by 1.80 to produce the actual outcome in which unemployment averaged 5.5 percent versus 3.7 percent in the 1920s. The actual decline in W/Y from 1929 alone to 1958–77 is 1.97 (see Table 2).

The string of close fits between actual and calculated relative wealth and income levels, despite sometimes crude methods of calculation with sometimes crude data, suggests that the coefficients of Equation (9) are useful descriptions of behavior in the two earlier periods.

¹³ This figure is the ratio of the average error term to average observed starts for 1926–28 in a housing starts regression estimated by Hickman, Bert G., “What Became of the Building Cycle?” in David, Paul A. and Reder, Melvin W., eds., Nations and Households in Economic Growth (New York, 1974), p. 302.Google Scholar

¹⁴ Gordon and Wilcox, “Monetarist Interpretations,” pp. 77–79;Google ScholarGrebler, Leo, Blank, David M. and Winnick, Louis, Capital Formation in Residential Real Estate (Princeton, 1956), pp. 99, 112, 120, 333, 426–27.Google Scholar

¹⁵ Historical Statistics, pp. 1002–04.Google Scholar

¹⁶ Mintz, Ilse, Deterioration in the Quality of Foreign Bonds Issued in the United States: 1920–1930 (New York, 1951), pp. 46, 48, 90–91, 95.Google Scholar

¹⁷ Mintz, Foreign Bonds, pp. 63–86 states this interpretation ably.Google Scholar