The Daily Caller

The Daily Caller

A new theory of BLS skew

Conn Carroll suggests a fresh theory of how the jobs number in last month’s BLS household survey exploded upwards: it’s not that federal Census workers spontaneously and perhaps subconsciously classified more people as working (my pet theory). It’s that the people they were interviewing spontaneously and perhaps subconsciously classified themselves as working–a “historic [in Carroll's words]  873,000 jump in the number of Americans telling BLS they were employed” (especially that they were employed part-time). [E.A.]

Why? Carroll cites two earlier instances (after 9/11) when the jobs number spontaneously jumped after President George W. Bush told people how important economic growth was to the fight against terrorism. Perhaps a similar rally-round optimism preceded this election. Patriotism–the Ultimate Conspiracy! …

P.S.: Of course the beauty of Carroll”s theory is it isn’t conspiratorial at all. It lets all government workers–for the Census and Labor department–completely off the hook while nevertheless sowing distrust about the October September numbers. …

Hostage to fortune: I’ve posted this, even though I’m going to bed and in a few hours, before I wake up, the news will tell the non-sleeping world if Carroll’s non-paranoid’s paranoid theory has been vindicated or not. But even if it’s not vindicated today–i.e., maybe more Americans told BLS they were employed because more Americans were in fact employed–it might be in the future. …

11/2 Update: Here is the BLS report for October. The establishment survey reports 171,000 new jobs. The household survey reports 410,000 new jobs. Does that disprove Carroll’s theory? I don’t think so. Certainly the household survey didn’t come crashing down into negative territory, the kind of return to “reality” Carroll’s graphs seem to predict. But neither did the establishment survey rise dramatically to validate the gigantic 873,000 jump in last month’s Census household survey of the employed–which would, I think, disprove both Carroll’s theory and the “blame Census workers” theory. The two surveys converged a bit (there were also upward revisions of prior months for the establishment survey) but my impression from this BLS monthly report is that the divergence is still substantial, with the household survey reporting much more rapid job growth. And of course it’s the household report that’s the basis for the headline-grabbing unemployment rate.

If the household report is inaccurately exuberant, why might the inaccuracy have persisted for two months? Well, if it’s related to the election–because the election provokes either a patriotic or political skew–the answer is obvious: the election hasn’t been held yet! You wouldn’t expect a return to cold, hard reality until next month.

That said, “establishment” job growth was positive and the household survey is volatile, according to BLS. The anti-paranoid default CW position–that September’s results reflected real, modest improvement,  plus normal random variation, and not a “skew” (of any sort)–would seem to have been strengthened. Others who know more about the data will, I hope, be publishing their own conclusions.