Immigrants commit crime at substantially lower rates than U.S.-born citizens, across every offense class studied

Summary

Immigrants — including undocumented immigrants — commit crime at lower rates than U.S.-born citizens, and this holds across every major crime class that has been studied: violent crime, property crime, drug offenses, weapons offenses, and homicide. The gap is largest for weapons offenses (U.S.-born citizens ~10x more likely to be incarcerated for these) and violent offenses (~5x), and smaller but still substantial for property crime (~2x) and drug offenses (~1.8x). The best state-level natural experiment — Texas, the only state that records arrests and convictions by immigration status — shows undocumented immigrants arrested at less than half the native-born rate for violent and drug crimes, and about a quarter the rate for property crimes, including for homicide specifically. This pattern has held for over 150 years of U.S. incarceration data (1870-2020) and has strengthened since 1960. A supplementary check run directly against AskAmerica's own crime and Census data finds no statistically significant relationship between a U.S. state's foreign-born population share and its overall violent or property crime rate (p=0.84/0.97 and p=0.56 respectively) — consistent with, though not the same test as, the offender-level literature.

1. The core offender-level finding, by crime class

The single most direct evidence on this question is the Texas Department of Public Safety data, because Texas is the only U.S. state that records criminal arrests and convictions by immigration status, distinguishing U.S.-born citizens, documented (legal) immigrants, and undocumented immigrants. A federally funded 2020 study by Light, He, and Robey published in PNAS (and summarized by the National Institute of Justice) used this data and found that, relative to U.S.-born citizens, undocumented immigrants were arrested at:

• Less than half the rate for violent crimes and drug crimes
• About one-quarter the rate for property crimes
• Consistently lower rates across assault, sexual assault, robbery, burglary, theft, and arson
• Lower rates for homicide specifically — a 2024 Cato Institute follow-up on Texas homicide data (2013-2022) found a homicide conviction rate of 2.2 per 100,000 undocumented immigrants and 1.2 per 100,000 legal immigrants, versus 3.0 per 100,000 native-born Americans over the same decade

At the national level, a 2021 Justice Department study cited by the Migration Policy Institute found that although prosecutions of immigrants rose 1990-2018, nearly 90% were for immigration-status violations (not the crime classes typically of public concern). Among the remainder, U.S.-born citizens were 10x more likely than immigrants to be incarcerated for weapons offenses, 5x more likely for violent offenses, over 2x more likely for property crimes, and nearly 2x more likely for drug offenses.

2. The long-run incarceration gap (1870-2020)

The most methodologically rigorous long-run study is Abramitzky, Boustan, Jácome, Pérez, and Torres (NBER Working Paper w31440, published in American Economic Review: Insights, Dec. 2024), which built the first nationally representative incarceration series by nativity spanning 150 years. Its findings: immigrants have had lower incarceration rates than the U.S.-born for the entire 1870-2020 period, and the gap has widened since 1960 — immigrants today are about 60% less likely to be incarcerated than U.S.-born citizens overall, and 30% less likely than U.S.-born whites specifically. The authors rule out changes in immigrants' observable characteristics or immigration policy as the explanation; they attribute the widening gap to a broader divergence in outcomes between less-educated immigrants and their U.S.-born counterparts, and to immigrants' relative resilience to economic shocks (fewer are incarcerated during downturns, comparatively).

Cato Institute's companion series (Illegal Immigrant Incarceration Rates, 2010-2023 / 2010-2024) finds a similar pattern specifically for undocumented immigrants: an incarceration rate of roughly 613 per 100,000 in 2023, below the native-born rate, with legal immigrants incarcerated at an even lower rate than undocumented immigrants, who in turn are incarcerated at a lower rate than the native-born.

3. What the broader research literature concludes about place-level crime and immigration

Beyond offender-level studies, a large literature asks whether places that receive more immigrants see more crime. Ousey and Kubrin's 2018 meta-analysis (Annual Review of Criminology) pooled roughly 50 studies and about 550 individual findings from 1994-2014: most findings show no correlation between immigration and area crime rates, and where an effect exists, it is roughly 2.5x more likely to be a negative (crime-reducing) association than a positive one. The American Immigration Council's 2024 analysis of national FBI Uniform Crime Reporting data from 1980-2022 found that as the immigrant population share roughly doubled (6.2% to 13.9%), the total crime rate fell 60.4% (violent crime -34.5%, property crime -63.3%) — an inverse trend, not a causal test. Their state-level beta regression (2017-2022, all 50 states) found no statistically significant relationship between a state's immigrant population share and its total crime rate.

4. AskAmerica warehouse cross-check: does this hold at state grain in 2022?

No table in this corpus (crime.cde_offenses, crime.cde_arrests, or the BJS NCVS/NIBRS tables) records an offender's nativity or immigration status — so the offender-level claim above cannot be recomputed directly from this warehouse. What the warehouse CAN test is the place-level question: does a state's foreign-born population share predict its crime rate? This is a different, weaker, and purely cross-sectional test — it says nothing about who commits crime, only whether places with more immigrants look more or less crime-prone.

Units excluded from the query, and what excluding them changed: crime.cde_offenses carries a synthetic row per offense-month with state_abbr = 'national' holding the pre-aggregated U.S. total; every query in this analysis filtered it out with state_abbr <> 'national' because the analysis is a 51-unit (state + DC) cross-section and the national row is not a state — including it would double-count the country's crime once as a synthetic 52nd "state" and bias the intercept and standard errors of the regression. This exclusion is definitional, not a data-quality judgment call, and it does not affect the reported statistic in any respect the reader could recover by including it: 'national' is a rollup of the same 50 states already in the regression, not an independent data point, so there is no meaningful "with vs. without" comparison to report for it (unlike the DC exclusion below, which IS an independent unit and is reported both ways).

Joining census.acs_nativity (foreign-born share, 2022, state grain) to crime.cde_offenses (violent-crime and property-crime rates, summed across 2022's reported months, state grain) via geo.states and running OLS across all 50 states plus DC:

• Violent crime rate ~ foreign-born share, DC included (n=51): coefficient +79.8, p=0.84, R^2 approx 0.0009
• Violent crime rate ~ foreign-born share, DC excluded as a sensitivity check (n=50): coefficient -12.3, p=0.97, R^2 approx 0.00002
• Property crime rate ~ foreign-born share, DC excluded (n=50): coefficient +765, p=0.56, R^2 approx 0.007

The finding is robust to whether DC — an outlier on both foreign-born share and violent crime — is included: the coefficient stays statistically indistinguishable from zero and even flips sign trivially between the two specifications, which is itself evidence there is no real linear relationship to detect. This null result at state grain in a single recent year is consistent with the Ousey-Kubrin meta-analysis and the American Immigration Council's own state-level regression, though it is a single cross-section with only one predictor and no controls (poverty, urbanization, policing intensity), so it should be read as a consistency check, not independent proof.

Caveats

• The Texas DPS data underlying the most-cited offender-level study is arrests/convictions in one state; it is the best available U.S. natural experiment on this question because Texas is the only state that codes immigration status, but it is not a national dataset.
• crime.cde_offenses draws on FBI data collected during the SRS-to-NIBRS reporting transition; state-year coverage in 2022 ranged from about 63% (Mississippi) to over 90% population coverage, which affects the precision of the state-level cross-check but not the direction of its null finding.
• The AskAmerica regression is uncontrolled and cross-sectional (one year, one predictor) — it cannot separate immigration's effect from correlated factors like urbanization or regional policing differences, and is offered as a consistency check against the literature, not as new causal evidence.
• "Immigrants" in the literature spans legal permanent residents, naturalized citizens, and undocumented immigrants; the studies cited above generally report all three sub-groups separately, and undocumented immigrants show the lowest offending rates of the three in every study that breaks them out.

Every query behind this report

6 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT DISTINCT offense_code FROM crime.cde_offenses ORDER BY 1

SELECT state_abbr, "year", AVG(population_coverage_pct) AS cov FROM crime.cde_offenses WHERE "year"='2022' AND offense_code='violent-crime' AND state_abbr <> 'national' GROUP BY state_abbr, "year" ORDER BY cov ASC LIMIT 10

WITH crime_agg AS (
  SELECT state_abbr, offense_code, SUM(offense_rate) AS annual_rate
  FROM crime.cde_offenses
  WHERE "year"='2022' AND offense_code IN ('violent-crime','property-crime') AND state_abbr <> 'national'
  GROUP BY state_abbr, offense_code
),
nativity AS (
  SELECT state, foreign_born, total_population, foreign_born::double / total_population AS fb_share
  FROM census.acs_nativity
  WHERE "year"='2022' AND geography = 'state'
)
SELECT s.state_abbr, n.fb_share,
  MAX(CASE WHEN c.offense_code='violent-crime' THEN c.annual_rate END) AS violent_rate,
  MAX(CASE WHEN c.offense_code='property-crime' THEN c.annual_rate END) AS property_rate
FROM crime_agg c
JOIN geo.states s ON s.state_abbr = c.state_abbr
JOIN nativity n ON n.state = s.state_fips
GROUP BY s.state_abbr, n.fb_share
ORDER BY s.state_abbr

WITH crime_agg AS (
  SELECT state_abbr, offense_code, SUM(offense_rate) AS annual_rate
  FROM crime.cde_offenses
  WHERE "year"='2022' AND offense_code IN ('violent-crime','property-crime') AND state_abbr <> 'national'
  GROUP BY state_abbr, offense_code
),
nativity AS (
  SELECT state, foreign_born::double / total_population AS fb_share
  FROM census.acs_nativity
  WHERE "year"='2022' AND geography = 'state'
)
SELECT s.state_abbr, n.fb_share,
  MAX(CASE WHEN c.offense_code='violent-crime' THEN c.annual_rate END) AS violent_rate,
  MAX(CASE WHEN c.offense_code='property-crime' THEN c.annual_rate END) AS property_rate
FROM crime_agg c
JOIN geo.states s ON s.state_abbr = c.state_abbr
JOIN nativity n ON n.state = s.state_fips
WHERE s.state_abbr <> 'DC'
GROUP BY s.state_abbr, n.fb_share

WITH crime_agg AS (
  SELECT state_abbr, offense_code, SUM(offense_rate) AS annual_rate
  FROM crime.cde_offenses
  WHERE "year"='2022' AND offense_code IN ('violent-crime','property-crime') AND state_abbr <> 'national'
  GROUP BY state_abbr, offense_code
),
nativity AS (
  SELECT state, foreign_born::double / total_population AS fb_share
  FROM census.acs_nativity
  WHERE "year"='2022' AND geography = 'state'
)
SELECT s.state_abbr, n.fb_share,
  MAX(CASE WHEN c.offense_code='violent-crime' THEN c.annual_rate END) AS violent_rate,
  MAX(CASE WHEN c.offense_code='property-crime' THEN c.annual_rate END) AS property_rate
FROM crime_agg c
JOIN geo.states s ON s.state_abbr = c.state_abbr
JOIN nativity n ON n.state = s.state_fips
WHERE s.state_abbr <> 'DC'
GROUP BY s.state_abbr, n.fb_share

WITH crime_agg AS (
  SELECT state_abbr, offense_code, SUM(offense_rate) AS annual_rate
  FROM crime.cde_offenses
  WHERE "year"='2022' AND offense_code IN ('violent-crime','property-crime') AND state_abbr <> 'national'
  GROUP BY state_abbr, offense_code
),
nativity AS (
  SELECT state, foreign_born::double / total_population AS fb_share
  FROM census.acs_nativity
  WHERE "year"='2022' AND geography = 'state'
)
SELECT s.state_abbr, n.fb_share,
  MAX(CASE WHEN c.offense_code='violent-crime' THEN c.annual_rate END) AS violent_rate
FROM crime_agg c
JOIN geo.states s ON s.state_abbr = c.state_abbr
JOIN nativity n ON n.state = s.state_fips
GROUP BY s.state_abbr, n.fb_share