DNA Database Expansion Failed to Boost Repeat Offender Arrests in Arizona Study

Expanding a DNA database to cover every convicted felon in Arizona did not measurably increase the share of arrested defendants with prior felony records, according to a quasi-experimental study published in the American Journal of Criminal Justice. The finding challenges one of the core assumptions behind decades of DNA collection legislation in the United States: that enrolling more offenders in searchable databases should translate directly into more repeat offenders being caught.

The study, conducted by Stewart J. D’Alessio and Lisa Stolzenberg, exploited a natural experiment created by Arizona law. Effective January 1, 2004, the state required all persons convicted of a felony to submit a DNA sample within 30 days of sentencing, a sharp expansion from earlier rules that covered only select offender categories. Because the reform enlarged the pool of searchable profiles without changing the underlying DNA technology, the researchers argue it isolates the effect of database expansion itself rather than concurrent improvements in laboratory procedures.

If DNA databases work as designed, the logic goes, individuals with prior felony convictions are far more likely than first-time offenders to have profiles on file. Any genuine improvement in forensic identification should therefore register as a growing proportion of arrests involving people with prior felony records. First-time offenders, whose profiles are generally absent from the database, serve as an implicit comparison group because their identification likelihood should remain comparatively stable.

To test this, the researchers analyzed 8,296 felony defendants prosecuted in Maricopa and Pima Counties across eight waves of the State Court Processing Statistics program from 1994 to 2009, alongside 107,639 defendants from 65 comparison counties, for a total analytic sample of 115,935 defendants. The outcome measure was simple: whether a defendant had at least one prior felony arrest. The analysis proceeded through a deliberately graduated sequence of statistical models, each relaxing the assumptions of the last.

The conventional pre-post comparison looked strikingly supportive of database expansion. The odds that an arrested felony defendant had a prior felony arrest were roughly 71 percent higher after the 2004 law than before it, an odds ratio of 1.71. But that interpretation assumes the composition of arrested offenders was stable before the reform, and it was not. When the researchers added a linear time trend, the odds of a defendant having a prior felony arrest were already rising by about 10 percent per year before the statute took effect. Once this trajectory was modeled, the level shift attributable to the expansion became small and statistically indistinguishable from zero. An event-study analysis confirmed the pattern: nearly all of the movement in the series occurred before the statute took effect, and the series actually flattened after 2004 rather than accelerating, the opposite of what the effectiveness hypothesis predicts.

A difference-in-differences design using the 65 comparison counties initially appeared to reinstate support for the expansion, with Arizona’s repeat offender share exceeding the contemporaneous national movement by 5.5 percentage points. Yet the credibility of that estimate depends on the assumption that Arizona and the comparison counties would have followed parallel trends absent the reform. The data rejected it. Relative to the comparison counties, Arizona’s repeat offender share was 8.7, 9.4, and 10.9 percentage points lower in 1994, 1996, and 1998 respectively, with a joint Wald test of the pretreatment coefficients rejected at p = .001. Arizona was converging toward the comparison counties throughout the pretreatment period, so the post-2004 coefficients are more consistent with continuation of a preexisting trajectory than with a discrete policy effect.

Robustness checks further undermined the estimate. Allowing each county to follow its own linear time trend reduced the treatment effect from 5.5 to 1.5 percentage points and rendered it statistically insignificant. Restricting the sample to counties observed in all eight waves, or substituting prior felony conviction for prior felony arrest as the outcome, likewise produced insignificant estimates. Most striking was a historical placebo test: assigning the intervention to the year 2000 and fitting the model only to pre-2004 data produced a larger estimated effect than the actual 2004 reform. In an exploratory county placebo comparison, 22 of 45 comparison counties yielded placebo estimates at least as large in absolute value as Arizona’s, meaning the state’s estimate was not unusual at all.

The study also tested two finer predictions of the database effectiveness perspective, and both failed. The theory predicts that any identification effect should concentrate in offenses that commonly yield biological evidence, such as rape, murder, robbery, assault, and burglary. In Arizona, however, the share of defendants with a prior felony arrest rose by 12 percentage points for evidence-poor offenses and 13 percentage points for evidence-rich offenses, an essentially negligible interaction. The pooled interaction that looked statistically noteworthy was driven by movement among the comparison counties, not by the Arizona reform. A predicted gradient by criminal history intensity also failed to materialize: difference-in-differences coefficients for defendants with one or two prior felony arrests and those with three or more were 0.8 and 4.8 percentage points respectively, both statistically negligible.

The authors situate these results within a broader literature that has generally found DNA databases useful but conditional in their benefits. Prior work has reported that DNA registration in Denmark reduced recidivism within the following year by up to 42 percent, and that DNA processing in an American field experiment increased suspect identifications, arrests, and prosecutions in property crime cases. But reviews of the United Kingdom’s National DNA Database have concluded that its wider security impact remains difficult to establish, because effectiveness depends on crime-scene evidence collection, laboratory capacity, timely processing, and the ability of police to act on database hits rather than on database size alone. The limited-effect perspective advanced in this study does not deny that forensic DNA can generate decisive leads in individual cases; it holds that the conditions required for expansion to shift the aggregate composition of arrestees are rarely satisfied. Biological evidence is absent from the overwhelming majority of offenses that generate felony filings, laboratory analysis is slow and resource-intensive, the most prolific offenders are often incarcerated, and conventional investigative techniques usually identify suspects before forensic analysis becomes relevant.

The researchers acknowledge several limitations. The State Court Processing Statistics capture felony filings only in May of even-numbered years, yielding just five pre-intervention and three post-intervention waves and limiting precision. The comparison counties were themselves exposed to DNA expansions at staggered dates, since the mean year states extended collection to all felons was 2003, which may bias the difference-in-differences estimate toward zero, though the authors note this contamination cannot explain the pretreatment divergence. The study observes the statutory requirement rather than actual sample collection or profile upload, and prior felony arrest is an imperfect proxy for database enrollment. The analysis is also confined to two metropolitan Arizona counties. Even so, the methodological lesson may prove the more consequential one: compositional outcomes in criminal justice exhibit strong secular trends, and the share of U.S. adults carrying felony records was rising sharply during the very decades databases expanded. Evaluations that compare periods before and after such reforms without modeling preexisting trends risk attributing those trends to the policy itself. Expanding enrollment, the study concludes, appears necessary but far from sufficient to improve forensic identification of repeat offenders at the population level.

Subject of Research: Quasi-experimental evaluation of whether expanding a state DNA database to all convicted felons increased the arrest of repeat offenders

Article Title: DNA Databases and the Arrest of Repeat Offenders: A Quasi-Experimental Test of DNA Database Expansion

Article References: D’Alessio, S. J., & Stolzenberg, L. (2026). DNA Databases and the Arrest of Repeat Offenders: A Quasi-Experimental Test of DNA Database Expansion. American Journal of Criminal Justice. https://doi.org/10.1007/s12103-026-09962-8

Image Credits: AI Generated

DOI: 10.1007/s12103-026-09962-8

Keywords: DNA databases, forensic genetics, CODIS, repeat offenders, quasi-experimental design, difference-in-differences, Arizona, criminal justice policy, parallel trends, placebo test, State Court Processing Statistics, offender identification