Full Employment

The headline Establishment Survey number disappointed at just 151k in January. The figure does not represent the actual amount of jobs gained, of course, since employment in January is typically about 3 million less than December. What that estimate represents is a combination of imputations about seasonality (comparing the actual change in the data set to the “typical” decline December to January) and at the start a stochastic modeling based on a non-exhaustive sample. That sample size is admittedly large, about 163,000 businesses and government agencies representing 623,000 individual worksites, but that does not mediate the necessity to refer to statistical processes.

The technical notes claim that the confidence interval for the CES is about 115,000 + or -, and that just for a 90% level of significance (1.6 standard deviations). So what the Establishment Survey just reported was in no way job gains of 151k but rather that January 2016 likely would have been between +36k and +266k had January been a “typical” and smooth month for employment. And even that interpretation is slightly misleading, since the confidence interval really only means that if the BLS ran the survey 100 times (polling a different 163,000 businesses each time), only 90 of those 100 are expected to fall in that range.

In short, there is entirely too much focus on the immediate monthly numbers – including when they are disappointing. Overall, the payroll figures remain as they have been;irrelevant for reasons that have to do with how they are constructed in the first place. There are any number of clues as to this statistical situation, but the easiest and most ready signal is basic common sense. Was January 2016 likely to have been a somewhat softer continuation of an otherwise robust employment condition unbroken by all that has occurred dating back almost two years now?

The fact that even the CES has clearly slowed in 2015 and into 2016 suggests that the true state of labor and payrolls did as well; what is left is to argue about is the degree of that slowing and from which point (“best jobs market in decades” or something much less than that) it began.