Preliminary Employment Data Might Miss a Recession Onset
As we move further into 2008, concerns are growing about the U.S. economy heading toward recession. The Employment Situation reports released by the Bureau of Labor Statistics have received a lot of attention in recent months, as economists try to determine the extent to which housing troubles may have spilled over to the broader economy. This month’s Employment Situation reported a decline of 63,000 two nonfarm payrolls in February and a revised loss in January, which increased the initial tally of 17,000 job losses to one of 22,000. The last time two consecutive months of decline occurred was in June 2003.
While the timely information provided by preliminary numbers can help us to assess labor market conditions, those numbers are subject to two monthly revisions after they are first released, as well as annual revisions every February. These revisions can be substantial and are sometimes even larger than the payroll changes themselves. The graph below, showing initial releases and revised numbers, demonstrates how significant revisions for any given month can be. January’s report this year, for example, initially reported a gain in December of just 18,000 nonfarm jobs but was revised up in the following report to a gain of 82,000. In August last year a payroll loss was initially reported, but with the revision the net change moved into positive territory.
Historically, payroll numbers usually dip sharply during or prior to recessions. But it important to note that this observation is based on revised numbers. The data initially reported might have shown a different picture at the time. To get an idea of how much this picture might change from initial release to revision, we prepared graphs of both sets of employment numbers around the two most recent recessions.
Around the 2000–2001 recession, both initial and revised data indicate a slowing labor market approaching July, although the initial data show a somewhat steeper descent. Despite the slowing trend in nonfarm employment growth in early 1990, growth continued to average about 200,000 jobs per month over the year. Payroll growth sharply turned negative in July, the official starting point of the recession. However, July’s loss of 219,000 jobs ended up being revised to a loss of just 89,000 later.
Revisions appear to have been more dramatic leading up to the 2001 recession than the 1990–1991 recession. The initial data show slowing, but employment gains looked solid right up to the onset of the recession in March. However, the revised data paint a much less optimistic picture, twice crossing negative territory in the two quarters preceeding the recession. Employment gains of 268,000 in January and 135,000 in Feburary were revised down to –16,000 and 61,000.
As of January 2001, labor indicators such as payroll employment, the unemployment rate, and the employment-to-population ratio all looked to be holding strong. Reports based on the initial releases of early 2001 thus painted a relatively positive picture of the labor market. Even the Cleveland Fed’s January Economic Trends assessed labor markets as “holding steady, albeit with slower job growth than earlier in 2000, despite signs of weakening in the overall economy.” However, with April’s employment report (of March activity), negative change was posted, the unemployment rate edged up 0.1 percent, the employment-to-population ratio decreased 0.1 percent, and the percentage of the civilian labor force unemployed for 15 weeks or longer increased slightly. The author of the Trends article’s commented that, “While variations in these labor market series are common, even during periods of robust economic growth, their recent simultaneous movements seem atypically strong and suggest that first-quarter economic activity slowed considerably.”
In both recessions, payrolls declined in the first month of the recession. While it seems as though payroll numbers might be insightful turning-point indicators, there are some notable exceptions as well. For example, initial releases for July and August 2000 showed respective declines of 108,000 and 105,000. However, these numbers were later revised upward, revealing increases of 163,000 and 3,000 jobs.
Three-Month Moving Average of Employment Changes
Three-month moving averages can remove some of the volatility of preliminary data and provide a more tempered trend of payroll employment. A moving average is useful because it takes into account both the latest preliminary data and past months’ revisions. However, the diluted nature of moving averages also delays their response to turning points in economic activity. In the 2001 graph, for instance, a three-month moving average smoothes out the peaks and troughs of the monthly change data, but it also shifts the start of the decline to after the start of the recession.
The current three-month moving average of payroll change declined 55,000 to 42,000 between December and January.