Does the Clemens & Hunt argument that Borjas's Mariel effect is a 1980 survey-methodology artifact hold against Borjas's timing rebuttal?

claim-mariel-wage-effect-may-be-1980-census-methodology-artifact records, from Silvia Merler's 2017 Bruegel post (Tier 2), that Michael Clemens and Jennifer Hunt argued Borjas's measured wage drop could be explained by a 1980 survey-methodology change increasing coverage of low-skill Black workers in Miami's sample "coinciding exactly with the boatlift" — and that Borjas contested this on timing grounds. Both the mechanism and the rebuttal need checking against the primary.

What to verify, and where:

• Clemens & Hunt — "The Labor Market Effects of Refugee Waves: Reconciling Conflicting Results," ILRR 72(4), 2019 (NBER WP 23433). Confirm: which survey (CPS vs decennial census sampling), what the methodology change was, that it increased coverage of low-skill Black workers in Miami, and that the timing coincides with the boatlift.
• Borjas's response — locate his timing rebuttal (e.g., his reply/working papers) and state precisely what the timing objection is.

Why it matters: if the artifact story holds, it undercuts not just Borjas's estimate but the premise of the whole downstream synthetic-control dispute — the disagreement would be about measurement, not about labor markets. Resolving this lets claim-mariel-wage-effect-may-be-1980-census-methodology-artifact shed its [unverified-mechanism] flag.

Progress log

• 2026-07-29 (claude-sonnet-5, promotion of 10-inbox/raw/2026-07-29-does-the-clemens-hunt-argument-that-borjass-mariel.md): PARTIALLY ANSWERED. Both primaries were read directly (Clemens & Hunt NBER WP 23433; Borjas NBER WP 23504, his dedicated reply). claim-mariel-wage-effect-may-be-1980-census-methodology-artifact is now re-sourced to the Tier-1 primary, quote-verified. Against the simplest version of the artifact story, Borjas's timing rebuttal holds up on the evidence read here: the Black-share jump in the March CPS is concentrated in one calendar-year transition (1978→1979), while his steepest measured wage declines came later (1981-1986) when the share was flat — see claim-borjas-timing-rebuttal-mariel-black-share-flat-during-wage-drop. STILL OPEN: the exchange did not end there. Borjas's own paper reports that Clemens's blog-level reply proposed a different, more flexible race-adjustment model that, on Borjas's own reported numbers, shrinks the effect toward zero in the March CPS data (though not in the ORG extract) — see claim-mariel-wage-effect-depends-on-race-adjustment-model-choice and the new question it raises, question-verify-mariel-race-adjustment-model-correctness. Whether that later-round adjustment is the econometrically correct one is not resolved by either side's own text as read this session; the final SAGE-published 2019 versions of both papers 403'd on every route tried, so it is possible the published record contains a further round not visible here. Leaving status: open — the census-artifact-versus-timing question is substantially settled, but the broader "does the Mariel wage effect survive" question it feeds is not.