How do researchers estimate homicide rates by immigration status and what are the main data sources?
Executive summary
Researchers estimate homicide rates by immigration status by combining three moving parts: a reliable count of homicide events or offending individuals (the numerator), an estimate of the population of people in each immigration-status category (the denominator), and statistical adjustments to account for missing or biased data; because few national criminal datasets record immigration status, most robust work relies on special sources (notably Texas criminal-history records and the National Violent Death Reporting System) and demographic estimates from the American Community Survey or specialist groups like the Center for Migration Studies (CMS) and Pew [1] [2] [3] [4]. Methodological choices — which events are counted (victim deaths versus arrests versus convictions), how the undocumented population is estimated, and whether researchers adjust for undercounting or use multilevel/fixed-effects models — materially change reported rates and explain much of the public disagreement [2] [1] [4].
1. Numerator: Counting homicides — victims, arrests, convictions
Different studies treat the numerator differently: some count homicide victims (decedents) using the National Violent Death Reporting System (NVDRS), which aggregates death records, medical examiner reports and law-enforcement narratives and was used to calculate foreign-born homicide victimization at 3.28 per 100,000 versus 5.60 for U.S.-born in a 2017 NVDRS sample (32 states + D.C.) [2]. Other high-profile analyses use arrest or conviction records — notably the Texas Computerized Criminal History (CCH) database, which uniquely records immigration status and has been the backbone of studies comparing undocumented, legal immigrant, and native-born offending [1]. Arrest/conviction numerators can be cleaner about who was charged or convicted but miss unsolved homicides and can be shaped by enforcement practices and reporting differences [5] [3].
2. Denominator: Estimating the immigrant and undocumented populations
The denominator problem is the core hurdle: undocumented populations are not fully counted in standard censuses, so researchers rely on modeled estimates from CMS, Pew, or the ACS’s multi-year estimates to create per‑100,000 rates [1] [2] [4]. For Texas-focused work, analysts have used state-level CMS figures or DHS estimates to derive counts of undocumented residents and showed that small differences in those population estimates change homicide rates meaningfully [6] [1]. National-level denominators are more fraught because most crime systems (UCR, NIBRS, NCVS) do not record immigration status, forcing analysts either to limit study to special datasets or to extrapolate from states like Texas — a step that introduces uncertainty and potential bias [1] [4].
3. Statistical choices: Adjustments, modeling, and comparability
Researchers use multilevel logistic regression, fixed-effects panel models, and other adjustments to account for state differences, time trends, and covariates like socioeconomic context or race/ethnicity; for example, the NVDRS study used multilevel mixed-effects models to examine foreign-born decedents nested within states [2], while longer-run state-panel work on undocumented immigration and violent crime has used fixed-effects regressions with CMS/Pew population series [4]. These choices help control confounders but do not erase data gaps: unsolved homicides, differential reporting by immigrant communities, and variations in law-enforcement recording of immigration status remain persistent threats to validity [2] [4] [3].
4. Key findings and divergent narratives
The empirical consensus in recent peer-reviewed and government-funded work is that immigrants — including undocumented immigrants, in the contexts where status can be observed — have equal or lower homicide and violent-crime rates than U.S.-born populations, a finding replicated in Texas CCH analyses, NIJ-funded studies, and national syntheses [1] [3] [7]. Nevertheless, high-profile claims that extrapolate from limited state evidence or from administrative counts without transparent denominators (for example, extrapolating Texas figures to national totals) produce much higher estimates and conflicting headlines; fact-checkers and researchers warn that such extrapolations can be methodologically unsound because of data and contextual differences across states [8] [9] [6].
5. Limits, caveats, and what good evidence looks like
Good estimates transparently state which numerator (victim vs. arrest vs. conviction) and which denominator (ACS, CMS, Pew, DHS) are used, adjust for undercounting and local legal practices, and avoid naïve national extrapolation from idiosyncratic state datasets; studies that meet these criteria (e.g., Texas CCH analyses and NVDRS-based work) provide the strongest evidence while still acknowledging uncertainty in undocumented population estimates and potential reporting biases [1] [2] [4]. Where the sources are silent on certain claims, reporting must be honest about limitations rather than asserting definitive national rates when national immigration-status data are absent [1] [3].