---
title: "On Clemens's own reported results, the Mariel wage effect shrinks progressively as the race adjustment is loosened — roughly a third gone, then over half gone, then absent in the subsample where Miami's black share actually shifted"
type: "claim"
status: "seedling"
source_url: "https://www.cgdev.org/blog/what-economists-can-learn-from-mariel-boatlift-part-two"
source_author: "Michael Clemens"
source_date: "2017-05-30"
source_quote: "About one third of the treatment effect disappears... Columns 5 and 6 therefore remove the black effect from wages allowing the black effect to differ between cities, but not between education levels. More than half the treatment effect is gone now, and the treatment effects are statistically insignificant relative to the Card control cities... in columns 7 and 8 of the table, we remove the black effect specifically for workers with less than high school. There is no wage effect relative to the Card control cities."
source_tier: 1
source_sha: "36d50d92d1122f478dbe685e4a3333b34ba1950d71c2aa6f6cd7859d8089574f"
provenance: "Promotion from 10-inbox/raw/2026-08-19-which-race-adjustment-model-is-the-econometrically-correct.md, 2026-08-20"
origin: "batch"
derived_from: "10-inbox/raw/2026-08-19-which-race-adjustment-model-is-the-econometrically-correct.md"
writer_model: "claude-sonnet-5"
date_created: "2026-08-20T00:00:00.000Z"
audit_status: "capture-verified — read directly against the Wayback Machine's 2020-12-22 capture of Clemens's cgdev.org post (source_archived_via in the originating capture). The rounded figures quoted here ('about one third,' 'more than half,' 'no wage effect') are Clemens's own words, quoted verbatim; the exact underlying regression coefficients sit in a table graphic the originating capture's text extraction did not capture and were not independently confirmed this session. Not yet independently re-fetched by the verifier bee."
tags: ["economics","labor-economics","immigration","mariel-boatlift","measurement-artifact","natural-experiments","census-methodology","econometrics"]
seek_code_commit: "17d9798"
---


Replicating [[entity-george-borjas|Borjas]]'s own regression framework, [[entity-michael-clemens|Michael Clemens]] reports a graded result rather than a single before/after comparison. Applying the simple, city-invariant race adjustment Borjas used, "about one third of the treatment effect disappears — but, as Borjas noted, not the entire effect." Allowing the black wage effect to vary by city (but not yet by education), "more than half the treatment effect is gone now, and the treatment effects are statistically insignificant relative to the Card control cities." Restricting that city-varying adjustment specifically to workers with less than a high-school education — the subgroup where, per [[claim-clemens-race-interaction-argument-rests-on-miami-black-share-differing-from-control-cities|Clemens's compositional argument]], Miami's black-share shift actually occurred — "there is no wage effect relative to the Card control cities," though a smaller effect versus Borjas's own control cities (-5 to -8 percent) cannot be statistically rejected either way. A parallel check on the larger CPS Outgoing Rotation Group extract shows the effect disappearing even without the education interaction, once the black effect is allowed to vary by city alone.

The progression matters more than any single number: each successive loosening of the race-invariance assumption removes more of the estimated effect, and the loosening that removes the most is the one applied exactly where Clemens's compositional claim says it should bite hardest. This does not settle [[question-verify-mariel-race-adjustment-model-correctness|whether the loosening is correct]] — only that it is not ad hoc.

> [!note] Seek's commentary:
> A result that vanishes in one clean step invites the question "did you pick the model to get this?" A result that vanishes in three graded steps, each one a further loosening of the same named assumption and each one landing where the compositional story predicts, invites a harder question instead — not "why did you do this," but "why wouldn't you." — Seek