How did researchers use Texas DPS immigration-status arrest data to estimate offending rates for undocumented immigrants?

1. The data source Texas provided: arrest records with immigration-status flags

The core empirical input was the Texas Department of Public Safety’s arrest and case-processing records, which include immigration-status information because DPS submits arrestee biometrics to DHS/IDENT and receives status responses that researchers used to classify arrestees as native-born, legal immigrants, or undocumented ^{[2] [1] [4]}.

2. How arrestees were classified into three groups

Researchers combined the DPS status flags with citizenship and place-of-birth fields from the Criminal History (CCH) extract and DHS responses to assign each felony arrestee into one of three mutually exclusive categories—native-born U.S. citizens, legal immigrants, and undocumented immigrants—enabling comparisons that most prior studies could not make ^{[1] [5] [2]}.

3. Counting crimes: incident-based arrest charges, years covered, and scope

The team treated each arrest charge as a separate crime incident—standard incident-based reporting—and focused on felonies recorded from 2012 (the first full year DPS recorded immigration info) through 2018, noting that most Texas arrests (about 83%) list only one charge so charge-counting closely approximates incident counts ^{[1] [2]}.

4. The denominator problem: population estimates for undocumented people

Because undocumented-population counts aren’t directly observed, the study used independent demographic estimates (the Center for Migration Studies and other peer-reviewed models cited by the authors) to produce the denominator for per-capita offending rates—an approach the authors and government summaries describe as standard but inherently reliant on the chosen population model ^{[3] [1]}.

5. Main findings the method produced

Applying arrest counts over the estimated undocumented population yielded lower felony arrest rates for undocumented immigrants than for legal immigrants and native-born citizens—less than half the rate for violent and drug crimes and roughly one-quarter the rate for property crimes during 2012–2018, according to the study and NIJ summaries ^{[1] [3]}.

6. Explicit methodological caveats the researchers flagged

Authors and funders note that arrest data are an imperfect proxy for actual offending because arrests reflect law enforcement activity as well as behavior, and because immigration-status measurement and population estimates contain uncertainty; the study therefore frames its results cautiously as arrest-rate comparisons, not definitive measures of all offending ^{[6] [2]}.

7. Critiques and alternative interpretations: undercounting and classification delay

Republican-leaning and enforcement-focused critics argue the DPS data undercount undocumented offenders because DHS identification can lag or occur after initial arrest—meaning some undocumented people may be misclassified as “other/unknown” or legal at the time researchers pulled records—so the undocumented-arrestee numerator could be understated and rates biased downward ^{[7] [8] [4]}.

8. Corroboration and broader context from other analysts

Migration Policy, the NIJ summary, and other reviews cite the Texas-based result as consistent with broader research finding immigrants commit fewer crimes than U.S.-born populations, while policy analysts and think tanks have used Texas DPS extracts for alternative counts and narrower outcome checks ^{[9] [3] [10]}.

9. Journalistic judgment: what the method buys and what it doesn’t

Using Texas DPS arrest-status data plus external population estimates provides a rare, transparent route to compare arrest rates by documentation status and produces robust descriptive comparisons for 2012–2018; however, the method cannot eliminate biases from delayed DHS identification, policing patterns, or population-estimate choices—so the numerical gap reported is real in the DPS-derived arrest frame but remains contingent on those known measurement limits ^{[1] [7] [3]}.