IMPLICATIONS FOR FURTHER RESEARCH

Implications for multiplier analysis abound and this will be one area of further investigation. One preliminary hypothesis is that for periods of perhaps three years, it may be appropriate to use the long-run consumption function to compute the multiplier, but that for tracking the economy in the short-run, explaining and forecasting changes in measured SR-MPCs may prove critical. It would appear that the process of adjustment may at times, be much faster than usually assumed and at other times, much slower. Another preliminary working hypothesis which appears fruitful --- although it is an oversimplification—is that it is the amount being multiplied, and not the multiplier value, that fluctuates.

Given the growing acceptance of the permanent income hypothesis (PIH) over the relative income hypothesis (RIE) rejected in this paper, some have suggested problems identified here should disappear if one shifts to the framework provided by the PIH. The assumption is that the SR-MPC remains less than the LR-MPC in the PIH, removing any need to reformulate theory. In that case, one need only reject the RIH. This line of reasoning may carry several problems. In the first place, the PIH and RIH, because of the difference in dependent variables, are not alternatives—as is often assumed. Indeed, it is suggested that those who, like Eisner and Brunner, cite the PIH to explain spending behavior inconsistent ‘with the RIH—such as 1968 -- are on weak foundations.

It is further suggested that there may be an even stronger case against the PIH. There is evidence that the PIH, like the RIH, can be rejected due to inconsistency with data. This possibility will be another area of proposed research. Here the working hypothesis is summarized.

If as is customary in literature, one assumes that depreciation is a constant fraction of the stock of durables, it is not difficult to show why the data are generally inconsistent with the PIH view of short-run adjustments, even if data on short-run changes in the flow of services from the durable stock are not readily available. Consider for example, a period such as 1967-68 (or 1954-55 or 1958-59). If it is appropriate here to disregard the reduction in the flow of services between two years which results from depreciation of the stock existing at the end of 1967, we may reason as follows: The change in the flow of services from the durable stock in 1968 is in this case, derived from spending on the stock in 1968. Specifically, the increase in utility flow will be some constant proportion of spending in 1968. If spending on durables during such years increases faster than disposable income, then it follows that the flow of services emanating from the 1968 additions to stock increases faster than disposable income. A central feature of the PIH is that permanent income is more stable than annual income. Another feature of the PIH is that, in the short-run, the flow of services changes even slower than disposable income. But if the flow of services rises faster than disposable income and if disposable income rises faster than permanent income, the ratio of the measured change in flow of services to the change in permanent income exceeds the permanent LR-MPC.

All of this follows, of course, from the assumption that depreciation is a constant fraction of the stock. This may not be the case [ Burress, 1961]. This has been discussed with Milton Friedman and he reports that one of his students, Michael Darby, recently completed a dissertation on this topic. Darby’s results, according to Friedman, are in many ways similar to Burress’

In the search for hypotheses to explain why the cyclical income elasticity of personal saving was positive in the 1930s and 1940s, but negative since the mid-fifties, it has been concluded that the cyclical marginal propensity to save from transfer payments, such as unemployment compensation, may be commonly underestimated. The reasoning is as follows: Consider two persons, both having just experienced cyclical unemployment. Assume neither receives unemployment compensation or other transfers. Both presumably dissave to finance that standard of living considered nonpostponable. Assume further that one then receives unemployment compensation and the other does not. It does not seem reasonable to assume that absolute dissaving of the worker receiving unemployment compensation be unchanged after receiving the transfer. To the extent that unemployment compensation leads to a reduction in dissaving, the MPS from such transfers is probably greater than usually assumed. Lester Taylor reported that the MPS from transfers is much higher than usually assumed and that no one has offered a hypothesis to explain his findings. It is possible that the above line of reasoning may contribute to an understanding of Taylor’s data. However, this must be studied with greater care. In particular, one must make certain that estimates provided by his model do not retain the same econometric problem mentioned earlier in his and Houthakker’s estimate of te SR-MPC.

It has long been noted that sample survey data are consistent with the hypothesis, assumed to be suggested by aggregate data, suggesting the SR-MPC is less than the LR-MPC. Indeed, Friedman has produced an impressive analysis of sample survey and budget data consistent with this central proposition of the PIH. The sample survey data inconsistent with traditional theory must then be explained away or dismissed. One working hypothesis for further research is that a very different potential contribution of theory to explaining both sample survey and aggregate time series data emerges if one assumes the SR-MPC can exceed the LR-MPC.

If this is one direction in which it is expected this work will move, it should be noted that this is also the origin of questioning accepted hypotheses. In making out income tax returns for two families for 1949-55 period, Burress kept balance sheets, primarily as an exercise in bookkeeping, but also to keep track of interest expense. (Interest and other non-wage income was never a factor for these families.) Except for discontinuities, these families dissaved continuously if durables were neglected; they reacted to increased income by increased dissaving. Yet they had excellent credit ratings. While at the American Bankers Association Burress discussed this practice with consumer loan officials and they uniformly suggested such behavior was common. They indicated they would readily assist in collection of data.

Surely there is an inconsistency between these observations and both Duesenberry’s comment in 1949 and Evans’ comment in 1969 suggesting families cannot dissave continuously without experiencing bankruptcy. Such statements follow from a priori assumptions. Burress shows (1963) why one may not assume, a priori, continuous dissaving leads to bankruptcy and why, in fact, it may lead to an excellent credit rating. But if one assumes such behavior impossible, it is likely he will throw out sample survey or other data consistent with this hypothesis.

Following the same line of reasoning, Burress also suggested that many middle—low income consumer units may dissave continuously, not because income is temporarily depressed—the usual explanation of their high APCs—but because income is continuously rising rapidly. Such consumers may react to increased income by borrowing more. Consumer units may follow such a pattern through certain age and income brackets with the result that there be continuous long-run or secular dissaving in those income brackets, even if such dissaving at that phase of the life cycle and income bracket is later offset by saving in other age and income brackets. Is it possible, for example, that income tax cuts in these brackets might lead to increased dissaving, not increased saving? Note this possibility is inconsistent with the usual analysis of the effect of redistribution on the aggregate saving-income ratio(Goode, p. 651. It is also inconsistent with the usual interpretation of the PIH, i.e., the notion that the permanent APC is the same in all income brackets, or the proportionality hypothesis. As explained by Burress [December, 1972], the widely held assumption that the PIH implies proportionality between spending and income in budget data represents misinterpretation of Friedman. As a result, several tests and policy implications of the PIH must be questioned. This interpretation of the PIE has been confirmed in a discussion with Friedman.

Errors in measuring, explaining and forecasting _²PDSI(t) with an assumed value of k, the effective maturity of installment debt outstanding, can obviously be viewed as fluctuations in the value of k. Hence, one important task to be undertaken is an investigation of the determinant of k. Friedman has pointed out that if effective maturity were in fact 3.0 years, then the expression for repayments should be, not _ of the sum of extensions in t, t-1 and t-2, but 1/6 of the extensions in t and t-4 plus _ of the extensions in t-2 and t-3. As Friedman also points out, the good fit with a k value of 3.0 years may also reflect merely a weighted average of some maturities substantially less than 3.0 years and others that are more. An intensive investigation of k is under way by Frieda Mendelssohn, a student in the Dissertation Workshop at Washington University, although her thesis topic is much broader.