Today we take another look at the WSJ Economic Forecast Survey. This time we focus on non-farm payrolls. In some sense, non-farm payrolls provide a more accurate picture of the labor market than looking at headline unemployment. Typically we look at the change in non-farm payrolls. The graph below presents actual non-farm payrolls (in yellow) and the consensus forecast (in blue).
In general, the forecasters have been more or less on the trend over the past few years, however, they severely underestimated the Great Recession. In fact, the worst any individual thought job losses would be was 400,000, whereas job losses actually reached 750 thousand per month. While forecasters do seem to do well with the overall trend, they do not capture anything close to the month to month variation seen in the realized data.
Payrolls data suffer from major revisions. It is unclear whether forecasters are targeting initial estimates, or revised estimates, however, the picture above would be qualitatively unchanged. In the face of this much volatility it also makes sense for forecasters to focus on the trend rather than the noise. This strategy has served survey participants well, except for the recession. For now, it is unclear why forecasters were systematically over-optimitisic during the great recession.