The Hit Rate
The hit rate is the fraction of searches that actually turn up contraband or evidence. It is a key disparity metric because it measures whether searches of one group are held to a lower standard of suspicion than searches of another: if searches of Black drivers succeed less often than searches of White drivers, officers are on average searching Black drivers on weaker evidence.
What the hit rate measures
Most disparity metrics on this site describe how often something happens to people who are stopped: how often they are searched, handcuffed, or arrested. Those rates can differ across groups for many reasons, and agencies often respond to disparities by pointing to differences in where and when they patrol, or in the behavior officers encounter.
The hit rate asks a different question: when officers decided a search was justified, how often were they right? A search is a “hit” if the officer found contraband or evidence. Because the hit rate conditions on the officer’s own decision to search, it is an outcome test: it evaluates the quality of the decisions themselves rather than the raw frequency of enforcement.
The logic of interpretation is inverted relative to the other columns in the disparities table. For search, force, and arrest rates, a higher rate for a group is a potential sign of over-enforcement. For the hit rate, a lower rate for a group is the warning sign: it means searches of that group were less likely to be justified by what they found. A group with a hit-rate disparity ratio below 1.0 relative to White individuals is being searched on comparatively weaker evidence.
Where the idea comes from
The outcome test traces back to the economist Gary Becker’s The Economics of Discrimination (University of Chicago Press, 1957), which modeled discrimination as a “taste” that decision-makers indulge at a cost. Becker’s insight was that a prejudiced decision-maker will accept worse outcomes from the group they disfavor: a lender who dislikes some group of borrowers will hold them to a higher bar, so the marginal loans they do make to that group will perform better, not worse. Comparing outcomes across groups can therefore reveal a discriminatory bar even when the decision-making process is unobservable.
Knowles, Persico, and Todd (2001) (“Racial Bias in Motor Vehicle Searches: Theory and Evidence,” Journal of Political Economy 109(1): 203–229) brought this logic to police searches. In their model, officers who care only about finding contraband will allocate searches so that hit rates equalize across groups; persistently lower hit rates for one group indicate that officers are applying a lower threshold of suspicion to that group. Ayres (2002) (“Outcome Tests of Racial Disparities in Police Practices,” Justice Research and Policy 4(1–2): 131–142) generalized the argument and applied it to police data, and Anwar and Fang (2006) (“An Alternative Test of Racial Prejudice in Motor Vehicle Searches,” American Economic Review 96(1): 127–151) developed related tests that are robust to some of the original model’s assumptions.
Recent applications
The hit rate has become a standard tool in large-scale empirical studies of police stops:
- The Stanford Open Policing Project analyzed nearly 100 million traffic stops nationwide (Pierson et al. 2020, “A large-scale analysis of racial disparities in police stops across the United States,” Nature Human Behaviour 4: 736–745) and found that searches of Black and Hispanic drivers turned up contraband at lower rates than searches of White drivers, evidence that the bar for searching those drivers was lower.
- California’s own RIPA Board annual reports have used the metric in every report since the Board began analyzing stop data — see the next section.
- Researchers at Stanford also developed the “threshold test” (Simoiu, Corbett-Davies, and Goel 2017, “The Problem of Infra-marginality in Outcome Tests for Discrimination,” Annals of Applied Statistics 11(3): 1193–1216), a Bayesian refinement that jointly estimates search thresholds and risk distributions rather than comparing average hit rates directly.
How the RIPA Board defines and uses it
The RIPA Board — the state body that publishes the annual reports this data comes from — has relied on this metric from the start, though its name has shifted: “hit rates” (2019), “search yield rates” (2020), and “discovery rates” (2021 onward).
The Board’s 2019 report, which laid out the methodology it planned to apply once data arrived, introduced the outcome test in the academic literature’s own terms:
Outcome tests compare the discrepancies between the percentages of successful searches conducted on stopped individuals. These percentages are also referred to as “hit rates.” For discretionary searches based upon consent, reasonable suspicion, or probable cause, equal hit rates across identity groups may signify a lack of bias, whereas differences may imply differential standards in conducting a search. (p. 25, citing Knowles et al. 2001)
The 2020 report, the first to analyze actual stop data (July–December 2018, from the eight largest agencies), operationalized the metric as the “search yield rate”: “proportion of searched individuals found in possession of contraband or evidence” (p. 36), “a measure of search efficacy” (p. 37). Its executive summary reported that “when officers searched individuals, contraband or evidence was generally found on White individuals at higher rates than individuals from all other groups,” and that for the highest-discretion searches — those based only on consent — “yield rates for racial/ethnic groups of color were lower than for White individuals” (pp. 9–10).
The 2021 report (2019 data) settled on the current name, explaining that the analyses “are also often referred to in research literature as ‘hit rates,’” but that “‘discovery rates’ is a more transparent term” because it matches the “Contraband or Evidence Discovered” data element in the RIPA regulations (p. 48). The same page states the test’s logic in the Board’s words:
One assumption of the test is that if officers are less likely to find contraband after searching people of a particular identity group, then those individuals are objectively less suspicious and may be searched, at least in part, because of their perceived identity. (p. 48)
Its headline finding: “individuals perceived as Black, Hispanic, and Native American had higher search rates despite having lower rates of discovering contraband compared to individuals perceived as White” (p. 11).
Discovery-rate disparities have also driven the Board’s policy recommendations. The 2022 report found that consent-only searches of Black, Hispanic/Latine(x), and Multiracial individuals “resulted in lower rates of discovery of contraband (8.5%, 11.3%, and 13.0% respectively) than searches of all other racial/ethnic groups” (p. 11), and partly on that basis the Board “recommends severely limiting or ending the practice of consent searches” (p. 12).
The metric remains central in the most recent report: the 2026 report, analyzing the 2024 data, found that “consent searches yielded lower discovery rates (20.30%) than non-discretionary searches (26.40%)” and that “[o]fficers reported the lowest discovery rates in consent searches of individuals perceived as Black (16.59%) and Native American (18.29%) and highest for individuals perceived as Pacific Islander (25.26%) and White (24.16%)” (p. 11; detailed analysis at pp. 63–64). Those 2024 figures exclude Terry frisks, a break from earlier years explained in the next section.
The 2024 break: frisks split out of the discovery rate
Starting with the 2024 data, the RIPA form gives officers a separate box for a Terry frisk — a pat-down of outer clothing for weapons — rather than folding it into “search of person.” That schema change forced a methodological break in the Board’s own discovery-rate analyses, and it matters for reading any 2024 hit rate.
The Board did not drop frisks from its search counts. Its actions-during-stop analyses use a combined “Search & Terry Frisk” measure, and appendix Table A38 gives a “Total Search Count” of 607,762 that includes frisk-only records — the same denominator behind the statement that frisks were 14.59% “of all searches” (2026 report, p. 82). But in the discovery-rate analyses the Board takes frisks out, and says so:
In this section, the denominator is not all stops, as it is in most other analyses in this report, but all searches. Also, as of the 2024 RIPA data collection, officers are not required to record a basis for search in Terry frisks, so this analysis only analyzes stops in which a search of person or property occurred. (2026 report, p. 63, n. 76)
The reason is partly mechanical — a frisk has no recorded basis for search, so it cannot be sorted into “consent-only” or “non-discretionary” — and partly substantive: a frisk is a different act with a different purpose (weapons, not evidence), and it turns up contraband at less than half the rate of a full search. Appendix Table A39 reports the two separately:
| Perceived race/ethnicity | Search, no frisk | Frisk, no search |
|---|---|---|
| Asian | 23.51% | 8.22% |
| Black | 28.38% | 13.05% |
| Hispanic/Latine(x) | 25.47% | 13.37% |
| Middle Eastern/South Asian | 21.81% | 6.66% |
| Multiracial | 30.05% | 11.31% |
| Native American | 25.18% | 5.59% |
| Pacific Islander | 27.48% | 5.05% |
| White | 29.69% | 9.58% |
| Statewide | 27.23% | 12.34% |
The Board also added a standalone section on frisk disparities in this report (2026 report, pp. 82–83), reporting frisk-only stops — frisks with no accompanying search — at 3.32% of stops of individuals perceived as Black against 1.04% for those perceived as White. Since a frisk requires only reasonable suspicion that a person is armed — a lower bar than the probable cause or consent behind most searches — who gets frisked is itself a discretion question, and the outcome-test logic applies to frisks on their own terms.
Limitations
The hit rate is informative but not decisive, and it should be read with its known weaknesses in mind:
Infra-marginality. The formal outcome test concerns the marginal search — the one just barely worth conducting — but observed hit rates are averages over all searches. If groups have different underlying distributions of suspicion, average hit rates can differ even when officers apply identical thresholds, and can even point in the wrong direction. This is the central statistical objection to the hit-rate comparison (see Ayres 2002 and Simoiu et al. 2017, whose threshold test was designed specifically to mitigate it).
Equilibrium assumptions. Engel and Tillyer (2008) (“Searching for Equilibrium: The Tenuous Nature of the Outcome Test,” Justice Quarterly 25(1): 54–71) argue that the economic model behind the test — officers as rational hit-rate maximizers, motorists adjusting their behavior in response — is a fragile foundation for real-world enforcement data.
Search types are mixed. Not all searches are discretionary. Searches incident to arrest, inventory searches of impounded vehicles, and searches required by warrant or parole/probation conditions happen regardless of the officer’s suspicion, and lumping them in with discretionary searches dilutes the signal. Terry frisks, separately identifiable from 2024 onward, are a fourth distinct type with their own much lower yield. The RIPA Board itself makes these distinctions, comparing “non-discretionary” searches against consent-only searches (with frisks excluded from both) — and has repeatedly found that consent-only searches of individuals perceived as Black had the lowest discovery rates of any group (2022 report, p. 11; 2026 report, p. 11). The site-wide hit rate on this site makes none of these distinctions.
What counts as a “hit.” The measure treats all contraband equally: a small amount of marijuana and a firearm both count. Studies that weight finds by seriousness can reach different conclusions than raw hit rates.
Small numbers. For smaller agencies, or when filtering to a single year or stop type, hit rates are computed from few searches and can swing wildly. A hit-rate disparity based on a handful of searches is weak evidence either way.
How this site computes it
In the disparities table, the hit rate is searches that found contraband or evidence ÷ all searches, computed per perceived racial/ethnic group from the RIPA contraband-or-evidence-discovered fields. All search types (discretionary and non-discretionary) are included. The “Disp.” column next to the hit rate is the ratio of a group’s hit rate to the White hit rate — and unlike the other disparity columns, values below 1.0 are the potential sign of bias.
For 2024, and unlike the Board’s discovery-rate analyses, the denominator here includes frisk-only records. That keeps the metric comparable across years — before 2024 a frisk was recorded as a search of the person, so every pre-2024 hit rate on this site already has frisks in it — at the cost of departing from the Board’s 2024 figures. The effect on levels is about a point (statewide 2024: 26.13% including frisks, 27.23% excluding them). The effect on the disparity ratios that the table actually highlights is smaller still: the Black/White ratio is 0.94 on the combined measure against 0.96 on the Board’s search-only measure, and the Hispanic/White ratio is 0.86 either way. See Methodology for the exact field definitions.