Paul Donovan
@pdonovan-econ.bsky.social
1.7K followers 800 following 650 posts
Economist. Part time farmer. Amateur boxer. Occasional author. Egotist. Advocate of economics without jargon. Inflation, environmental economics, the economics of prejudice, structural change. ubs.com/pauldonovan
Posts Media Videos Starter Packs
pdonovan-econ.bsky.social
The market also skews to valuing official data a lot more - roughly three times as many economists make forecasts for payrolls as for ADP.
pdonovan-econ.bsky.social
Which is not to say that those biases cannot be corrected for when ADP converts its sample into its final number (the final data is about five and a half times its sample). But it needs official data to try and correct the biases by reweighting, and the sample still has to be presumed representative
pdonovan-econ.bsky.social
But ADP does not include local, state, or federal government, self-employment (a growing trend), has an underrepresentative sample for small businesses, and must be presumed to suffer from selection bias - most firms do not use ADP for their payrolls, so what motivates those companies that do?
pdonovan-econ.bsky.social
Exactly. A company whose recent applicant numbers have gone up less than 100% might want to examine their poor relative attractiveness as an employer. Especially if in a sector where applicants are likely to be disproportionately tech literate younger people
pdonovan-econ.bsky.social
Any company that does NOT have surging applications is in serious trouble. Any labour market NOT experiencing falling externally advertised job vacancies must be dangerously unstable. And anecdotal comparisons to the recent past make for dramatic journalism but are not helpful economic analysis.
pdonovan-econ.bsky.social
4. Employment participation amongst the younger population has been increasing sharply, which is not especially suggestive of a dramatic weakening of the labour market in these sectors
pdonovan-econ.bsky.social
2. Other measures of job vacancy data, especially online vacancies, remain high in these sectors.

3. AI distorts job application data dramatically, making comparisons over time useless. Younger applicants will be happy to use AI to make multiple applications
pdonovan-econ.bsky.social
Poor quality official data unfortunately elevates anecdotal evidence.

1. "Job vacancy" data ONLY measures externally advertised job vacancies. A surge in labour market churn followed by a normalisation can create a rise and then a fall in the numbers without changing the total vacancies.