Were the jobs numbers fudged... Hussman's take.

As usual, John Hussman takes an issue and dissects it partially and by the numbers.....

On the economic front, Friday’s employment report was interesting in that total non-farm payrolls (the “establishment survey” figure most widely quoted in news reports) came in slightly below expectations, but total civilian employment (the “household survey” figure used to compute the unemployment rate) jumped enough to produce a drop in the unemployment rate to 7.8%. While the difference was certainly an outlier in terms of typical correlations between establishment and household figures, it wasn’t the sort of outlier that would justify the suggestions of political conspiracy that were bandied about over the weekend. 

The fact is that on a month-to-month basis, there is only a 50% correlation between the establishment and household employment figures, rising to about 90% correlation for year-over-year changes. The household data is notably more volatile, but the establishment figure makes up for the lower volatility with significant after-the-fact revisions, particularly around economic turning points. The month-to-month changes above and below the 12-month average are about 50% larger in each direction for the household survey than for the establishment survey. What’s interesting is that these changes are often matched by changes in the reported size of the labor force, which is why they don’t usually result in large changes in the unemployment rate from month-to-month. For example, in January 2000, the household figure jumped by over 2 million jobs, while the establishment figure increased by only 248,000 jobs. But the unemployment rate held steady at 4% because the reported labor force also increased by over 2 million workers. 

From that perspective, the unusual feature of last month’s report was that the increase in reported household data exceeded the increase in the reported labor force by an amount that statistically occurs only about 6% of the time. Yes, it was an outlier, but it wasn’t even a two standard deviation event. I would expect some give-back in that “excess” household survey growth, and based on the extent of revisions to establishment survey data around economic turning points, I also expect that the September establishment survey figure will ultimately be revised to show a net loss of jobs on the month. So as a whole, my impression is certainly that the September report presents a healthier picture of the employment situation than will survive later revisions, but there isn’t evidence to suggest any manipulation of the data. 

The entire piece can be read here.