New signals of recession and recovery - Center for International Business Cycle Research

Business Economics, Oct, 1995 by Geoffrey H. Moore, John P. Cullity, Beth W. Taubman

The National Bureau of Economic Research (NBER) has established a widely used chronology of business cycle peaks and troughs. The latest trough came in March 1991, and this is commonly accepted as the starting point for the current expansion. The CIBCR publishes a Coincident Index, so called because its highs and lows have roughly coincided with the business cycle turns. It includes six components: industrial production, nonfarm employment, real personal income, total business sales in constant prices, gross domestic product in constant prices, and the unemployment rate.

The CIBCR also publishes a long-leading index, which on average leads the business cycle and the coincident index by more than a year at peaks and six months at troughs. Its components include: bond prices, building permits, the ratio of price to unit labor costs, deflated money supply, and changes in productivity and services prices. The Center, in addition, publishes a short-leading index. This index includes: the average workweek in manufacturing, initial claims for jobless benefits, net business formation, new orders for consumer goods and materials, contracts and orders for plant and equipment, change in industrial materials prices, stock prices, surveys of changes in business inventories, domestic non financial debt in constant dollars, the layoff rate, and vendor performance.

The short-leading index reaches its turning points before the turns in the business cycle as well as the coincident index. However, its turning points occur later than those in the long-leading index.

All of these indexes have declined during every one of the nine recessions since 1948 and have seldom undergone sustained declines at other times. The short-leading index did decline on two occasions when no recession occurred, in 1950-51 and 1966, but then the coincident index merely hesitated a little. For this reason, it is advisable to look at all the indexes rather than just one. All of the indexes are also relatively free of erratic movements, at least in comparison with their individual components. However, they are not perfectly smooth, and the first figures to be published are also subject to revisions. Hence, one must allow for the possibility of frequent reversals.

With these considerations in mind, we first convert the indexes into growth rates using a special formula to do so. The rate is calculated by obtaining the ratio of the current month's index to the average level of the twelve preceding months and expressing this result as an annual percentage rate. The advantage of this method compared with the ordinary month-to-month change is that a twelve-month average as a base is less subject to erratic movements than any single month.

During a recession we would expect these growth rates to be negative because the indexes themselves would decline. During recoveries they should be positive. The points at which they turn positive may therefore signal a recovery. At the same time, because of erratic movements or revisions, the growth rate may turn negative again. To meet this problem we adopted a simple rule to be applied to the initial signals (P1 and T1) based on the long-leading index: the peak signals require the growth rates in this index to be below zero for at least two consecutive months and the trough signals require the rates to be positive for two consecutive months. The second and third signals of recession (P2 and P3) occur when the growth rates of the short-leading and coincident indexes become negative. The second and third signals of recovery (T2 and T3) occur when their growth rates become positive. However, we require that the preceding signal be reached in the same or a previous month.

In June 1991, the long-leading index growth rate, which had been negative until April 1991, posted its second consecutive positive growth rate (see Table 1). This constituted the first signal of recovery (T1) from the 1990-91 recession. The second signal of recovery [TABULAR DATA FOR TABLE 1 OMITTED] (T2) was also transmitted in June 1991, and the third (T3) occurred in February 1992.

The accompanying diagram puts the whole system together. To repeat, the growth rates of the P1 and T1 signals must go below or above zero for two consecutive months. For the second and third signals, the growth rates of the short-leading and coincident indexes need to go below or above zero for only one month, provided that the preceding signal has been reached.

LOOKING AT THE RECORD

How do the signal dates compare with the National Bureau's peak and trough dates? Table 1 lists them all and from this information one can tell how far the signals led (-) or lagged ( ) behind at peaks and troughs. More specifically, at troughs T1 has always occurred within three months of the turn, and often before it. Its average lead was two months. T2 has usually occurred two or three months after the turn, and T3 about four to six months after.

At peaks, the P1 signal has led seven of the eight peaks between 1953 and 1990. Its average lead was nine months. P2 signals led the business cycle peaks about half the time, with an average lead of four months, although it is increased by one exceptional case in which the signal went off eighteen months before the 1957 peak. P3 signals typically were set off around one to three months after the peak.