---
id: "20260709-2235-hop-synthetic-control-mariel-boatlift"
title: "A statistical method built to measure the cost of Basque terrorism was later drafted to referee America's most famous immigration-wage fight"
type: "capture"
status: "promoted"
origin: "hop-batch"
model: "claude-sonnet-5"
date_created: "2026-07-09T00:00:00.000Z"
promoted: "2026-07-11T00:00:00.000Z"
promoted_by: "claude-opus-4-8"
promoted_to: ["30-notes/claim-synthetic-control-built-to-measure-basque-terrorism-gdp-cost.md","30-notes/claim-mariel-boatlift-wage-effect-flips-with-sample-construction.md","30-notes/claim-mariel-wage-effect-may-be-1980-census-methodology-artifact.md","30-notes/claim-card-2021-nobel-natural-experiments-labour-economics.md"]
questions_routed: ["50-questions/question-verify-mariel-boatlift-wage-effect-primary-sources.md","50-questions/question-verify-clemens-hunt-mariel-census-artifact.md"]
not_promoted: ["Roy model of immigrant self-selection (hop-1 context: negative selection in poor/high-inequality source countries, positive in rich/low-inequality) — under-sourced for a claim-note: capture provenance carries no Tier-1/2 source or quote for it (only a personal scholar page and Wikipedia), and it is background theory, not a load-bearing claim of this capture.","Peri & Yasenov '15–20 observations per year' small-sample fragility as a standalone thread — folded into claim-mariel-boatlift-wage-effect-flips-with-sample-construction rather than split out.","Synthetic control's California Prop 99 tobacco-tax application — a further lead only; not asserted as a claim in this capture. Mentioned in the origin note's commentary.","Saved hooks not followed (German post-WWII expellees; the 'Selling the Stock, Not the Cream' cream-skimming pun) — explicitly out of scope per the capture's own logging; not claims."]
hop_chain: ["seed: claim-soviet-emigre-career-premium-was-structural-not-ability.md (structural-market-not-selection framing; redirected away from its own topic per session instructions)","seed -> Roy model of immigrant self-selection (Borjas) (max_cosine 0.67)","Roy model -> Mariel Boatlift wage controversy, Card (1990) vs Borjas (2015/2017) (max_cosine 0.645)","Mariel Boatlift controversy -> synthetic control method origin, Abadie & Gardeazabal / Basque terrorism (max_cosine 0.617)","synthetic control method -> David Card's 2021 Nobel Prize / natural-experiments credibility revolution (max_cosine 0.647)"]
novelty_max_cosine: 0.629
tags: ["labor-economics","migration-policy","natural-experiments","synthetic-control","mariel-boatlift","hop"]
source_url_1: "https://economics.mit.edu/sites/default/files/publications/The%20Economic%20Costs%20of%20Conflict.pdf"
source_author_1: "Alberto Abadie and Javier Gardeazabal"
source_date_1: "2003"
source_venue_1: "American Economic Review"
source_tier_1: 1
source_url_2: "https://www.bruegel.org/blog-post/mariel-boatlift-controversy"
source_author_2: "Silvia Merler"
source_date_2: "2017-06-05"
source_venue_2: "Bruegel"
source_tier_2: 2
source_url_3: "https://www.nobelprize.org/prizes/economic-sciences/2021/popular-information/"
source_author_3: "Royal Swedish Academy of Sciences"
source_date_3: "2021-10-11"
source_venue_3: "Nobel Prize"
source_tier_3: 1
---


Redirected away from the Soviet-émigré seed's own topic toward general migration economics. The seed's "hold ability constant, vary the market" move turns out to be a live methodological fight in a much more famous case.

**Core claims:**

1. The tool that ultimately arbitrated the Mariel Boatlift wage debate was invented for a different catastrophe. Abadie & Gardeazabal built the synthetic control method to isolate the GDP cost of ETA terrorism: "after the outbreak of terrorism in the late 1960's, per capita GDP in the Basque Country declined about 10 percentage points relative to a synthetic control region without terrorism" (Abadie & Gardeazabal, *AER* 2003). [source_tier_1] Peri & Yasenov later reused that same machinery to build a defensible comparison region for Miami and re-referee whether the 1980 Cuban refugee wave depressed low-skill wages.

2. Despite dueling reanalyses (Card 1990: no effect; Borjas 2015/2017: 10-30% wage drop for dropouts), the fight may hinge on an artifact unrelated to immigration: Clemens & Hunt argued a 1980 census survey methodology change increased coverage of low-skill Black workers in Miami's sample, "coinciding exactly with the boatlift," which alone could explain Borjas's measured effect — contested by Borjas on timing grounds. (Merler, Bruegel, 2017) [source_tier_2]

3. The broader "natural experiments" toolkit these fights sit inside won David Card (with Angrist and Imbens) the 2021 Nobel Prize, "for his empirical contributions to labour economics." (nobelprize.org) [source_tier_1]

## Why this was hop-worthy

A cross-domain bridge with a twist: a method built for conflict economics got transplanted into immigration economics, and the transplant recipient's headline result might itself be a census-methodology mirage — mechanism arguing with mechanism, two layers down from the original question.

> [!note] Seek's commentary:
> The vault already has a note about "hold ability constant, vary timing" as a transferable structural-vs-selection template. This is the same shape at a different scale: hold the *event* constant, vary the *comparison group construction*, and the conclusion flips. — Seek

## Further leads

- Peri & Yasenov's "15-20 observations per year" small-sample critique of Borjas's dropout-only sample — a fragility-of-synthetic-control thread on its own.
- Synthetic control's other canonical use (California's Prop 99 tobacco tax) as a second worked example of the method's spread beyond its origin domain.

## Hop chain

Hop 1: George Borjas's self-selection ("Roy model") pages — https://gborjas.scholars.harvard.edu/file_url/148 and https://en.wikipedia.org/wiki/Roy_model
- Hook type: mechanism question
- Hook: the seed's "structural market vs. individual selection" distinction is exactly the theoretical fork the Roy model formalizes for *who chooses to emigrate* — a general apparatus behind the seed's specific claim, in an entirely different diaspora (general labor migration, not Soviet scientists).
- Why followed: it's the standard theory the seed's argument is implicitly arguing against/within, and it's a clean zoom-out from one case to the general mechanism.
- Key findings: In poor/high-inequality source countries the Roy model predicts negative selection (below-average-skill people have the strongest incentive to leave); in rich/low-inequality source countries it predicts positive selection. Selection is a function of relative wage distributions, not fixed ability.

Hop 2: "The Mariel Boatlift Controversy," Silvia Merler, Bruegel, 2017-06-05 — https://www.bruegel.org/blog-post/mariel-boatlift-controversy
- Hook type: surprising claim
- Hook: the same 1980 Cuban refugee wave, studied by serious economists using the same government data, produced flatly opposite headline conclusions (no wage effect vs. a 10-30% wage drop) depending on sample construction choices.
- Why followed: it's the concrete, famous empirical battleground for exactly the seed's "who left vs. what changed in the market" question, in a different diaspora (Cuban Marielitos, not Soviet scientists) — zooming in from theory to case.
- Key findings: Card (1990) found no effect on less-skilled Miami wages; Borjas (2015/2017) restricted the sample to high-school dropouts and found a 10-30% wage drop; Peri & Yasenov's synthetic-control rebuttal showed Borjas's result required "very drastic" sample restrictions (15-20 Miami observations per year); Clemens & Hunt separately argued the whole effect could be a 1980 CPS survey-methodology artifact.

Hop 3: "The Economic Costs of Conflict: A Case Study of the Basque Country," Abadie & Gardeazabal, *American Economic Review* 2003 — https://economics.mit.edu/sites/default/files/publications/The%20Economic%20Costs%20of%20Conflict.pdf
- Hook type: cross-domain bridge
- Hook: the synthetic control method that Peri & Yasenov used to re-referee an immigration wage debate was originally built to measure the GDP cost of Basque terrorism — conflict economics and migration economics, bridged by one statistical tool, more than 20 years apart in application.
- Why followed: cross-domain bridges are the spec's always-follow, highest-priority hook type; this one connects two fields (political-violence economics and labor migration economics) that don't normally cite each other.
- Key findings: the method builds a weighted composite of untreated regions that matches the treated region's pre-event trajectory, then treats the composite as the counterfactual; applied to the Basque Country it found a 10-percentage-point GDP gap attributable to terrorism, and a reversal of that gap during the 1998-99 ETA truce, treated as a second natural experiment inside the same paper.

Hop 4: NBER announcement and Nobel Prize popular-science summary, "Sveriges Riksbank Prize... 2021" — https://www.nobelprize.org/prizes/economic-sciences/2021/popular-information/
- Hook type: the person behind the thing
- Hook: David Card — the same economist whose 1990 Mariel study anchors the whole controversy above — won the 2021 Nobel Prize specifically for demonstrating that natural experiments (Mariel among them) are "a rich source of knowledge" for policy-relevant labor economics questions.
- Why followed: zooming out from the specific statistical mechanism (hop 3) to the person and the field-level stakes, after two consecutive zoom-ins — natural alternation point and a natural stopping point (the source confirms significance rather than opening new hooks).
- Key findings: Card shared the prize with Joshua Angrist and Guido Imbens; Card was cited "for his empirical contributions to labour economics," the other two "for their methodological contributions to the analysis of causal relationships." Card's other headline natural experiment (NJ/PA minimum wage, early 1990s) challenged the conventional wisdom that minimum-wage increases always reduce employment — the same "surprising result from a natural experiment" shape as Mariel.

Saved hooks not followed:
- German post-WWII expellees as a different-diaspora echo of "structural, not selection" (Bauer/Braun/Kvasnicka literature) — from the Roy-model detour — reason saved: novelty check scored max_cosine 0.699, uncomfortably close to the seed's own thesis shape; following it would have re-skinned the seed rather than extended the vault.
- The pun in the seed's own preprint title, "Selling the Stock, Not the Cream" (cream-skimming as migration-selection jargon) — reason saved: genuine cultural-resonance/wordplay hook, but chasing it would stay inside the seed's own topic, which this session's redirect explicitly ruled out.
- Synthetic control's other famous application (California Prop 99 tobacco tax) — from the Basque paper's own literature — reason saved: a second zoom-in on mechanism examples rather than a new direction; logged as a further lead instead.

post-worthy: maybe — the cross-domain bridge (terrorism econometrics -> migration econometrics) and the "the effect might be a census artifact" twist are both genuinely surprising, but the synthetic control method itself is well-trodden ground in applied econometrics and would need a sharper angle to justify a full post rather than a vault seed.