Summary
Butcher and Piehl document that male immigrants aged 18–40 have institutionalization (incarceration) rates about one-fifth those of the native born by 2000, despite having demographic characteristics — low education, low wages, young, male, disproportionately Hispanic — that strongly predict high incarceration among natives. Applying native institutionalization propensities to immigrants' characteristics would predict a rate double the native average; the actual immigrant rate is one-tenth that prediction. The gap widened from 1980 to 2000. Using a cross-Census design they rule out deportation as the primary mechanism and find evidence instead for multidimensional positive self-selection into immigration (migrants have lower criminal propensity or higher incentive-responsiveness) and a general deterrence effect that operates more strongly on migrants than on native nonmovers.
Key Claims
- Scale of the puzzle: In 2000, immigrants' institutionalization rate is ≈0.7% vs. ≈3.5% for native-born men 18–40. Demographic characteristics (education, race/ethnicity, age) predict an immigrant rate of 7.3% — 10× the observed level.
- Widening gap: More recently arrived cohorts have successively lower relative institutionalization rates; the gap between immigrants and natives widened from 1980 to 2000.
- Metropolitan-area (MA)-level corroboration: Metropolitan areas (MAs) with the largest increases in immigrant share had the largest decreases in crime rates 1990–2000 — ruling out the "immigrants escape detection" alternative that would inflate area crime.
- Deportation ruled out: Naturalized citizens — who cannot be deported — reduced their institutionalization rates relative to natives as much as or more than non-citizen immigrants. Therefore, the mechanical removal of criminal aliens through deportation cannot be the primary explanation.
- Deterrence confirmed: Naturalized citizens' improvement is consistent with general deterrence from the harsher criminal penalties of the 1980s–90s. Critically, native-born movers (those living outside their state of birth) also reduced their institutionalization relative to nonmovers over the same period. This identifies a general mechanism: migration (domestic or international) selects for individuals who are more responsive to criminal incentives.
- Positive selection confirmed: The most recently arrived immigrant cohort in 2000 was 1.1 percentage points (pp) less likely to be institutionalized than the most recently arrived cohort in 1990, holding time-in-country constant. However, there is no discrete break at the 1996–2000 cohort (whose entry decision came after the 1994–96 criminal alien legislation), suggesting gradual multidimensional selection shifts rather than legislation-driven changes.
- Roy model interpretation: The standard Borjas-Roy model predicts migrants' social outcomes from a single wage-relevant skill dimension. Butcher and Piehl argue migration selects on multiple dimensions simultaneously; criminal propensity or incentive-responsiveness can differ from wage-relevant skills. Immigrants who have low wages and low criminal propensity are consistent with a multidimensional selection process where social network migration costs are negatively correlated with criminal-justice outcomes.
Data and Method
- Data: 5% Public Use Microsamples, U.S. Census 1980, 1990, 2000. Sample: men aged 18–40. Institutionalization (a Census group-quarters category) used as proxy for incarceration; in 1980, 70% of institutionalized men 18–40 are incarcerated; limiting to incarcerated in 1980 does not change results.
- Baseline model: Logit for institutionalization with Census-year dummies, age dummies, cohort-of-arrival dummies, and time-in-country controls; reports marginal effects at sample means.
- Cohort decomposition: Year-specific logits to estimate how each immigrant arrival cohort's institutionalization rate changed between Census years, netting out age and demographic composition.
- Deportation test: Restrict to naturalized citizens (not subject to deportation) and re-estimate the cohort trend; deportation is not the mechanism if naturalized citizens show the same pattern.
- Deterrence / selection test: Compare immigrants to native-born movers (those living outside their state of birth) to identify whether the pattern reflects general migration-selection or deportation/immigration-specific factors.
Concepts Introduced or Extended
Entities Mentioned
Quotes
"Immigrants have very low rates of institutionalization compared with the native born. What's more, immigrants' relative rates of institutionalization have fallen over the last three decades."
"The process of migration selects individuals who either have lower criminal propensities or are more responsive to deterrent effects than the average native."
"Native-born movers and immigrants appear to have responded to the harsher criminal penalties in the 1980s and 1990s... suggesting a general deterrence effect."
My Take
A clean piece of applied microeconomics that resolves what looks like a paradox (poor demographics + low crime) by decomposing three competing mechanisms with a natural comparison group logic. The deportation test (using naturalized citizens as a control group) is particularly clever: if deportation were the mechanism, naturalized citizens shouldn't show improvement, but they do. The native-born mover comparison is also insightful — it rephrases the question from "why do immigrants have low crime?" to "what does migration select for?" and grounds the answer in a general mobility-selection story.
Limitations: institutionalization rather than true incarceration (post-1980 Census limitation); inability to distinguish documented from undocumented immigrants, who may have additional reasons to avoid law enforcement contact; the analysis is descriptive/decomposition rather than formally causal — no instrument for immigration flows. The paper predates the large literature using geographic variation in immigrant enclaves as instruments (Altonji-Card 1991, Card 2001) and does not attempt a structural instrumental-variables (IV) estimate.