Friday, October 7, 2011

The change in payrolls actually decreased from August to September

The payrolls number today was deceiving. Taken at face value, it implies an increase in job creation between August and September, from the (revised) 57k in August to 103k in September.

But there was a strike in August., which involved 45k workers. Employees in strike are not counted in the payrolls by the Bureau of Labor Statistics, because they are not paid while they are in strike. I would argue that those employees should be counted in if they are expected to return to the payrolls in the short term. I'm not criticizing the BLS's methodology, but rather trying to capture the underlying trend of employment. If we add those 45k to the change in payrolls in August we get a change of 102k (57k + 45k), and if we subtract those 45k from the change in September, we get a corrected change of 58k (103k - 45k). Job creation, therefore, weakened from August to September.

In addition, keep in mind that the threshold for statistical significance of the change in payrolls is 96.4k. So the change in payrolls in August was not different from zero, from a statistical standpoint, and the change in payrolls in September was only barely significant.

The three month average of payroll changes, as of September, increased to 96k, from 68k in August. Over the last 12 months, nonfarm payrolls increased 1.01%.