Nothing says "agriculture" more than a barn, and that is why the emphasis on nonfarm payrolls is so appropriate: Economists can't hit the side of a barn with their estimates for the first release of the nonfarm payroll number each month. With this Friday's employment situation report incorporating the massive distortions of the hurricanes, a tough job is about to get a whole lot tougher. It's quite possible that the number will come in plus or minus several hundred thousand from the loss of 169,000 jobs forecast at the time I am writing. Let's look at these errors and their origins, and then conclude by asking whether we should care come Friday. If we map the Wall Street consensus against the first estimate of nonfarm payroll additions for the past two decades and overlay the forecast error, we can see some patterns emerge. The first is a surprisingly small bias for the consensus forecasts to overshoot the actual number. You might think this was a consistent and consistently large error, but it's not: Only 127 of the past 246 monthly forecast errors, just 51.6%, were negative, and the average error of estimate relative to the nation's labor force was a mere -.004%. Wall Street may cheerlead earnings, but it really tries to get the payroll numbers right.
Bet the Under
But as the cluster of recent forecast errors highlighted in the oval indicates, the recent performance of economists has deteriorated. Since the December 2002 report, those who "bet the under" would have won 19 out of 32 times, or 59.4%. The average error relative to the size of the labor force over this period was a far more sizeable -.023%. As some officials might note, there's a "divergification" we need to explain.