Labor Statistics Mirages Come to a Crashing Halt

Well, here I am minding my own business, working on a boring venture capital post, when here comes the Bureau of Labor Statistics announcement of October jobs data. The BLS has been working toward a Nobel prize for fiction writing all year, but you can only carry that so far. Maybe they figured that everyone is distracted by Ukraine and Israel or just the latest TikTok videos, but their house of cards can only be towered upward so far, and today’s announcement begins the downfall to reality. “Expectations” were for 180,000 new jobs, according to the survey of businesses. Reality was 150,000 and September was revised down by 39,000 jobs and August by 62,000; so 100,000 total fewer. Funny how every month this year has been revised down from initial releases. you might think the BLS would ask itself if there is something wrong with its methods for initial estimates.

Meanwhile the household survey, which gets less attention but may be more accurate, and which this year has shown the widest divergence ever from the survey of business, showed a drop of 348,000 in the number of employed people. The unemployment rate rose to 3.9% and large numbers of people are working multiple jobs to keep up. The BLS did use one of its usual tricks to keep the report from looking even worse. The sketchy model used to estimate how many jobs were added by new businesses versus lost by businesses going bust (the “birth/death” model) added 412,000 jobs, the second highest ever.

[I’ve always found the Household numbers far sketchier than the Establishment (business-reported) data, but YMMV. The wide swings in the Household base data are not at all new, and are products of the polling model that BLS uses to compile the Household data. It’s those wide swings that are evidence of its lack of reliability, and those have gone on for years. I’m not saying that the BLS reports don’t have issues, but most of them are with Table A rather than Table B, likely for the same reasons political polls have gotten less reliable — fewer people bother to cooperate with the pollsters. — Ed]