Sounds familiar…

Ah, yes, the early days of @sociologicalsci.bsky.social!

Thanks! @mikehoutnyu.bsky.social and I estimated the reliability of all "can't actually change" questions. Parental occupation was more reliable than childhood income, but much less than parental edu. Could be driven by coding (as you mentioned), hard to categorize jobs, career changes, and more!

OA Preprint for “Growing up Different(ly than Last Time We Asked): Social Status and Changing Reports of Childhood Income Rank” on @socarxiv.bsky.social: osf.io/preprints/so...

Replication package: github.com/ophastings/c... (data via @kjhealy.co's incredibly convenient gssr package)

(5/5)

So what? Retrospective reports are useful to measure intergenerational mobility and childhood effects. But that's a problem if current circumstances shape reports of the past. Here we don’t have “true” childhood measures, so we can only examine patterns of instability in repeated measurements. (4/5)

We also found these patterns were much stronger for males than females, suggesting recall bias of childhood income rank may be larger for men and more anchored by present-day experiences. (3/5)

Surprisingly (to us, anyway) these changes were *not* associated with corresponding changes in one’s current income. Instead, they were associated with shifts in *subjective* indicators of current social status. (2/5)

I only tested your Bluesky-registered hypothesis. 😉 But agree if there's indeed a null effect (or even if there is something) to publish, would want to try some other religion measures and some subgroup analyses (class, age, race...).

Here's attendance vs looks by gender with all three years. Not a lot of action...