The new business cycle: the impact of the application and production of information technology on U.S. macroeconomic stabilization

Business Economics, Oct, 1997 by Martin Fleming

Because dating conventions have been applied in a rigorous and consistent manner in the postwar period, comparisons within the current period are not subject to arbitrary differences that can arise as a result of the assignment of dates. Given that the U.S. economy is now in its ninth period of expansion since the end of World War II and that the current expansion has proceeded for at least twenty-five quarters, the strong possibility exists that U.S. expansions have lengthened over the past twenty years in comparison to the immediate postwar period.

Table II shows the dates of the nine expansions in the postwar period presented in rank order by length as measured in quarters. Of the four expansion periods subsequent to 1975, three are among the four longest expansions over the entire postwar period with the current expansion still underway.

The average length of the four most recent expansions is twenty quarters, while the average of the five earlier expansions was sixteen quarters. However, because of the very small number of observations in each sample, pre- and post-1975, conventional hypothesis tests are inappropriate. As a result, it has become customary in this literature to calculate the Wilcoxon statistic based on the sum of the ranks during the two periods. For the test statistic to be significant and for the null hypothesis of equal means to be rejected, the ratio of the two sums of ranks must be 2.0 or greater. In this case, the ratio is 1.5 and the null hypothesis of no difference is accepted.

Two expansions have a substantial impact on the calculation of the test statistic. The expansion of the 1960s, which is frequently thought to have been lengthened by the Vietnam War effort, is more than twice as long as the next longest expansion during the early period. In addition, the expansion of 1981 and 1982 is also generally thought to have occurred under unusual circumstances. The recession of 1980 was associated with the quick institution and rapid removal severe credit restraints. Simultaneously, severe imbalances had built up in the economy after a prolonged period of expansion from 1975 to 1980 that were not addressed during the brief recession. Consequently, the entire 1980 to 1982 period could be viewed as a period of recession or adjustment. When these two expansions are removed from their respective periods, the lengths of the average expansion during the two postwar periods are much different, and the significance test rejects the null hypothesis of no difference.

While it is not unreasonable to ignore the brief expansion of 1981-82, it does seem very unreasonable to ignore the entire decade of the 1960s. In reviewing the GDP data, it appears that activity was very weak in the 1966-67 period, and it is possible that, except for the military build-up, a recession might have occurred. Indeed in his work, Watson finds that an alternative set of dating conventions does produce a recession during this period. Assuming such a recession, which produces two expansions of twenty-one quarters and nine quarters instead of one expansion of thirty-five quarters, the lengths of the average expansion during the two postwar periods are much different and the Wilcoxon test of significance again leads one to reject the null hypothesis of no difference.