FORECASTING RECORD WITH THE INSTALLMENT DEBT MODEL

It was reported at Business Week in 1964 that the Council of Economic Advisors was disturbed in the first few months after the 1964 tax cut because there was so little immediate impact on spending.⁶ Perhaps increased PD_SI—amounting to 45% of the tax cut—offers some explanation. This delayed effect of the tax cut was forecast in a journal article written as the tax cut was signed into law and published in September, 1964 [Burress].

It is well known that fiscalists were surprised that the 1968 surtax had so little immediate effect. Also Okun and Brookings Panel found the surtax apparently had more than the expected effect in 1969, but could find no explanation. Perhaps the reduction in PD_SI in 1968 -- amounting to 70 percent of the surtax—and the increase in PD_SI in 1969 -- amounting to 32 percent of the surtax—offer some partial explanation. It should be noted that in August, 1967, in response to a U.S. Treasury request for the American Bankers Association’s position on the proposed surtax, the notion that the SR-MPC would be inflated, in part due to reduced PD_SI in 1968, was used by Burress to forecast that saving might fall in 1968 and that there would be more inflationary pressure in 1968 than suggested by the standard forecast.⁷ Saving did fall in 1968. There was more inflationary pressure than anticipated by the standard forecast.

In March, 1969 the new model was an important input for forecasting a 1969-70 recession. This was reported in the press.⁸ In particular, it was suggested that as income fell, saving would rise and that the SR-MPC would exceed unity; that the expected slowdown would turn into a full-blown recession [Burress, March, 1969]. As it turned out, personal saving rose more than GNP fell between the 1969:IV peak and 1970:IV trough.

Despite this record, the new approach received virtually no attention from academic economists. The effect of this was to intensify once more the research and one result was the finding, in March, 1972, that the Houthakker-Taylor model produced values of the SR-MPC that exceeded unity. The first reaction of specialists to this finding was that there was some problem with the data. Only when it was reported that this result was consistent with a generally unrecognized body of reasoning that had a successful forecasting record did the work begin receiving some attention.

The next forecast with the model was made in late 1972. It is found in a paper written in December, 1973 and submitted to the staff of the National Bureau of Economic Research January 7, 1973. This is important now (September, 1973) because many believe the economy has been sailing on uncharted seas in the first half of 1973. However, when viewed through the installment debt model, it is clear that there were sufficient similarities between the outlook in late 1972 and earlier periods to permit reasonably successful forecasts.

In the December, 1972 paper for the NBER there appears the following passage: "The Michigan model suggests that from the end of 1972:III through 1973:II, real aggregate disposable income will rise at a 10 percent annual rate. Three previous periods in which real aggregate disposable income rose this fast were 1955, 1959 and 1968, and the rate of increase those years was no more than a half point higher than projected for these three quarters of 1972-73. But the Michigan model suggests the saving rate will. rise more than 36 percent during this period. In contrast, in each of the previous periods, in which the rate of increase in income was so rapid, the saving rate not only fell, but the absolute level of saving fell. There were of course some differences. These were recovery years, not years following periods as strong as 1972. What may be more important, _²PDSI(t) is much higher than in any of these earlier years—when, in fact, it was typically negative. But I suggest the odds favor only a modest increase in the saving ratio in the first half of 1973. If one accepted the remainder of the Michigan forecast, it would appear to follow that the economy will be even stronger than the RSQE staff suggests. Given that the RSQE staff saw so much strength in the first half of 1973 they built in an assumed 10 percent surtax at mid-year, it would appear that this line of reasoning should be further explored.

"From mid-1973 through mid-1974 the Michigan model suggests a dramatic slowdown in the rate of increase of real disposable income, ranging from +1.9 percent annual rate in 1973:III to +6.17 percent for 1973:IV with approximately 4 percent rate of increase for the first half of 1974. With this reduction in the rate of increase, they suggest a modest reduction in the saving rate. The modest reduction in the average saving rate between the first halves of 1973 and 1974 is less interesting than the printout which suggests that the saving rate is positively related to the rate of increase in disposable income. This is presumably generated on the assumption that the SR-MPC is less than the LR-MPC. The thrust of my work is that this is sometimes true, sometimes not. If the rate of increase in the economy and disposable income falls as much as suggested by the Michigan model in the second half of 1973, there is, I suggest, reason for concern. This concern is based on the large positive value of _²PDSI(t) for 1973 Plus the assumption that the consumer is somewhat more responsive to these changes than usually assumed."

On April 12, 1973, in response to a request from Business Week, the model was used to forecast how first quarter data would appear when released a week later. As reported in Business Week, nearly every economist interviewed believed saving would rise and the request for a forecast came just before the deadline in an article that was devoted to explaining why data would show that saving did rise in the first quarter. Hence there was little space devoted to a forecast that suggested saving would fall. But that was the position taken on the basis of the installment debt model and this possibility was mentioned in Business Week April 21.⁹ When the data came out April 19, they showed saying in the first quarter had fallen more than $4-billion.

In the same April 12 interview Business Week also requested a forecast for the first and second half as compared to 1972 calendar year levels and rates. The forecast offered was confirmed by a wire, a press release and reported, in part, in Business Week May 11. It was suggested that for the first half, the tendency for increased borrowing to reduce saving below the 1972 level would be more than offset by increased predetermined saving and therefore saving would rise. It was suggested, however, that the increase in saving would be so slight in the first half that the saving rate would be down from the 1972 calendar year rate. The data now show the 1972 calendar year level of saving was $49.7-billion and $51.0-billion. The 1972 calendar year saving rate was 6.2 percent. The saving rate for the first two quarters was down slightly at 5.9 percent each quarter.

It may be useful to contrast this to some other well-known forecasts. The Council of Economic Advisors, in their 1973 Report, predicted that the saving rate in the first half would be very high—due to the rapid increase in income induced by the tax refunds. On the basis of University of Michigan sample survey data released in late April, 1973 Report, predicted that the saving rate in the first half would be very high—due to the rapid increase in income induced by the tax refunds. A reconciliation of the CEA analysis of the response of consumers to a rapid increase in income and the forecast based on the new model can be made readily in terms of assumed values of the SR-MPC when income is rising rapidly. But a reconciliation of sample survey data suggesting most consumers were saving their tax refunds and the approach offered in this paper is not so obvious. It was suggested by the author (at an April 19 forecasting conference where the two views were discussed) that there was no inconsistency [Burress, April 19,. 1973]. Very simply, it was suggested that it was quite conceivable that even 80 percent of the taxpayers might save their entire tax refunds but that the refund, when taken as a whole, would induce net dissaving. That is, if the other 20 percent used their tax refunds as down payments to purchase large ticket items, the induced dissaving of the 20 percent could be expected to more than offset the induced saving of the 80 percent with the result that the MPC from the aggregate tax refund would exceed unity.