Monday, December 01, 2008

The NBER Votes for December, 2007 as the Cyclical Peak

It is now official. A recession began in 2007. The National Bureau of Economic Research's Dating Committee met by phone Friday and made that decision. Looking at monthly indicators and cross checking them against various ways of aggregating quarterly GDP, they judged the recession started in December of last year.

But the Dating Committee should beware. Its analysis is vulnerable to revisions. The data as currently reported shows payroll employment peaked in December, 2008. The Committee looked at various other significant indicators peaked between November, 2007 and June, 2008. Our experience from the 2001 recession was that the payroll employment data got revised down dramatically well after the fact. Since the Bureau of Labor Statistics (BLS) adjusts the data each month for firm births and deaths, it overestimates job growth during recessions. I expect the same will happen again. That would move the payroll employment peak back. A downward revision in payroll employment would also cause personal income minus transfers and GDP to be revised downward.

I agree the economy turned down in 2007. I would argue for an earlier date than December: perhaps one of the summer months. The household survey data indicate an earlier downturn. Employment divided by population peaked in December, 2006 (i.e. a year earlier) and the unemployment rate bottomed out in March, 2007 (nine months earlier.) I rely on the ratios because they avoid difficulties associated with the time series of the household employment aggregates. The household data should do a better job of picking up employment weakness tied to the sectors at the center of the drama: mortgage brokers, real estate agents, and construction workers. Housing starts peaked in January, 2006 (a month or two earlier if you look at the moving averages) and had fallen about 40 percent by summer, 2007.