1. Several years ago the author pursued the thesis that the propensity for personal saving to fall in recessions and rise in recoveries before the mid1950s reflected the dominant role of business saving by the unincorporated firm. This question was examined before data on the aggregate were generally inconsistent with traditional theory. Few data were available. However data for 1953 and 1954 - the only years for which unpublished data were available from the Federal Reserve Board - suggested that, if consumer installment debt held by spending units operating unincorporated firms was separated from debt held by other units, saving of spending units per se rose in 1954 as income fell. This suggests the reason NIA personal saving fell in 1954 was the dominant role of saving of the unincorporated firm. One can only speculate as to whether—and if so, for how long—this pattern prevailed when the data appeared consistent with traditional theory.

2. The tendency to use annual data in this paper stems from the observation that many are aware that quarterly data are often inconsistent with traditional theory, but that few seem aware that this is also true of annual data.

3. In 1958 the author suggested the consumption function be disaggregated and durables treated separately. The aggregate framework of this paper may be explained as follows: Most would agree that instability in the consumer sector can be viewed as instability in the allocation of total saving between real assets (primarily consumer durable) and financial assets (primarily personal saving). The traditional disaggregate approach, focusing on consumer durables, can be viewed as an attempt to explain that part of saving going to real assets. It treats personal saving as a residual. The alternative approach offered here represents a slight, but significant variant of that theme. The new approach attempts to explain and forecast that part of saving going into personal saving and treats consumer durables as a residual. Hence the present approach has much in common with both the aggregate and disaggregate approach.

It is not suggested, however, that this approach is necessarily superior to a disaggregate approach. Yet this approach has generated superior forecasts to some well-known disaggregate models (such as that of the Research Seminar in Quantitative Economics at the University of Michigan and others reported in Appendix A), presumably because these models remain contaminated by the econometric problem discussed shortly.

4. If estimates of the consumption function are limited to periods 17 years or longer and the initial year of computation is 1948 rather than 1947, the probability that the period selected will yield results inconsistent with the habit hypothesis rises from 0.56 to o.64. If 1952 is used as the initial year to avoid both post-World War II and initial Korean War adjustments, only one period provides estimates of the relative values of the SR-MPC and LR-MPC consistent with traditional theory. See Figure 1.

5. This suggests a theory of the short-run is relevant, not for any short period, but whenever the rate of change of income is significantly depressed or inflated. This notion is similar to that implied by the econometric formulation of the original Duesenberry-Modigliani model which suggested the habit hypothesis operative only when income is below a past peak.

6. The author was on the staff at Business Week when these reports were received throughout the third and fourth quarter, 1964.

7. This forecast vas in response to a request from the U.S. Treasury for the position of the American Bankers Association on the proposed surtax. The author was, at that time, economic advisor and deputy manager of the ABA. The forecast was part of the author’s transmittal of the views of the Economic Policy Committee, of which the author was secretary, to Charles E. Walker.

8. This was reported, in part, in Glenn Stoup, "WSU Prof Believes Credit Debt May Bring Recession," Journal Herald, Dayton, Ohio, April 21, 1969