Executive Summary
California’s AB 372 Domestic Violence Pilot was enacted in 2018 in response to concerns that the state’s long-standing 52-week batterer intervention program (BIP) was not achieving improvements in recidivism. This pilot represents a shift toward a more individualized, risk- and needs-based approach to domestic violence intervention. Grounded in the Risk-Need-Responsivity (RNR) framework, the pilot allows counties to match BIP dosage to risk level and to impose additional requirements to participate in programs and services, such as substance use treatment, to address needs related to recidivism. This study evaluates whether the pilot’s more flexible model improved recidivism outcomes for individuals sentenced to probation for domestic violence offenses.
The analysis leverages a difference-in-differences design comparing outcomes for people convicted of domestic violence and sentenced to probation in six pilot counties to outcomes for people sentenced in non-pilot counties before and after implementation. Supplemental analyses using detailed probation data
from three pilot counties provide additional insight into policy impacts. Across both approaches, findings consistently show that the DV Pilot is associated with meaningful reductions in reconviction outcomes.
Key Findings
- Risk- and needs-based interventions reduced reconvictions: The DV Pilot resulted in a 15% reduction in overall reconvictions and a 20% reduction in domestic violence reconvictions, suggesting that matching BIP dosage to risk level is more effective than a one-size-fits-all intervention model.
- No evidence of an increased public safety risk: The DV Pilot did not increase overall or DV rearrest rates, indicating that shorter or less intensive interventions for lower-risk individuals did not compromise public safety.
- Policy shift toward RNR principles is supported by evidence: The results found through this pilot evaluation reinforce the broader research literature showing that individualized risk- and needs-based interventions outperform one-size-fits-all approaches, supporting continued investment in
assessment-driven supervision.
Policy Context
For several decades, California has relied on a one-size-fits-all approach to interventions for people convicted of domestic violence and sentenced to probation. Under Penal Code §1203.097, courts must impose mandatory probation conditions, including the requirement that people convicted of domestic violence complete a batterer intervention program (BIP) lasting “not less than one year,” typically consisting of weekly group sessions and periodic progress-reporting to the court. In practice, this statutory framework has created a statewide expectation that all people sentenced to probation for domestic violence complete a 52-week BIP.
Traditional BIPs are delivered in group settings led by trained facilitators. Many programs draw on elements of the Duluth model, which emphasizes understanding patterns of power and control in abusive relationships and challenging attitudes that support violence (Shepard & Pence, 1999). Programs may also incorporate cognitive-behavioral strategies intended to promote behavioral change and accountability (Morrel et al., 2003). While this model has served as the foundation of domestic violence programming in many jurisdictions across the United States, it has also faced criticism for relying on a largely uniform program structure that does not differentiate among individuals based on their risk levels or underlying behavioral needs.
A growing body of research has raised questions about the effectiveness of standardized BIPs in reducing recidivism. Meta-analyses of BIP evaluations generally find modest and inconsistent effects on re -offending, with outcomes varying depending on research design, comparison groups, and methodological approaches (Babcock, Green, & Robie, 2004; Feder & Wilson, 2005; Wilson, Feder, & Austin, 2021). At the same time, research has identified substantial heterogeneity among individuals who commit domestic violence, as well as a range of co-occurring factors associated with committing these offenses, including substance use, trauma exposure, and antisocial behavior (Capaldi et al., 2012; Cafferky et al., 2018). These findings have led some researchers and policymakers to question whether uniform intervention models adequately address the underlying drivers of domestic violence for all individuals.
Within the broader field of correctional rehabilitation, there is growing evidence in support of the Risk-Need-Responsivity (RNR) framework as a more effective approach to reducing recidivism. The RNR model emphasizes matching intervention intensity to an individual’s risk level and targeting programs and services at needs — such as substance use treatment or mental health services — most likely to be associated with re-offending (Bonta & Andrews, 2007). Evidence from correctional interventions suggests that matching program dosage to risk level leads to better recidivism outcomes than requiring a high level of programming regardless of the risk of re-offending (Lowenkamp, Latessa, & Holsinger, 2006). Although there is limited research on the application of RNR principals to domestic violence, a recent meta-analysis found domestic violence interventions that incorporated RNR principles improved recidivism outcomes relative to one-size-fits-all program models (Travers et al., 2021).
In 2008, the Judicial Council of California evaluated the State’s approach to addressing domestic violence and found little evidence of effectiveness under the current BIP program model (MacLeod, Smith, & Rose-Goodwin, 2008). The authors of this study recommended the use of risk and needs assessments, along with integration of substance use treatment where appropriate.
AB 372: The Domestic Violence Pilot
In 2018, California’s AB 372 Domestic Violence Pilot was enacted in response to concerns that traditional approaches were not achieving improvements in recidivism and emerging evidence that RNR models could be more effective (Penal Code §1203.099). The legislation authorized a time-limited pilot allowing six counties — Napa, San Luis Obispo, Santa Barbara, Santa Clara, Santa Cruz, and Yolo — to implement alternative domestic violence intervention models for individuals sentenced to probation. Rather than requiring all participants to complete the standard 52-week BIP, the legislation permitted pilot counties the flexibility to tailor BIP dosage to risk level and to require additional programs and services based on assessed risks and needs.
Implementation of the pilot began in July 2019 in some counties, although early efforts were disrupted by the COVID-19 pandemic. Statewide support for implementation and legislative reporting has been coordinated in partnership with the California State Association of Counties (CSAC) and the Chief Probation Officers of California (CPOC), which have helped facilitate data collection and reporting to the legislature. In 2023, the legislature extended the pilot’s original sunset date to July 2026 to allow time for an evaluation of the effects of AB 372 on recidivism outcomes.
A central component of the pilot model is the use of validated risk assessment tools to measure a person’s risk of general recidivism and domestic violence recidivism. The assessment happens when individuals begin probation supervision. As shown in Table 1, counties use different validated instruments to measure general recidivism risk, including the Level of Service-Case Management Inventory (LS-CMI), the Correctional Assessment and Intervention System (CAIS), the Offender Risk Assessment System (ORAS), and the COMPAS assessment.
For domestic violence risk, all pilot counties use the Ontario Domestic Assault Risk Assessment (ODARA), a validated tool designed to estimate the likelihood of future domestic violence incidents. The ODARA relies primarily on criminal history and incident characteristics to generate a risk score. While the assessment produces a standardized score, individual counties define their own risk thresholds corresponding to low-, medium-, and high-risk categories. As a result, individuals with similar ODARA scores may be classified differently across counties depending on local policy decisions.
Following the completion of risk assessments, counties apply a structured decision-making framework to determine supervision levels and BIP dosage. Each county has developed its own decision-making matrix that combines general recidivism risk with domestic violence risk to guide case management decisions. These matrices are designed to align program dosage with risk level, consistent with principles from the RNR framework. People with higher risk levels typically receive more intensive programming, while people with lower risk levels may receive shorter and less intensive interventions.
Table 2 illustrates the variation in BIP dosage across the pilot counties. In contrast to the traditional statewide model requiring a 52-week program for all participants, program dosages for low-risk participants range from approximately 10 to 26 weeks across the pilot counties, while higher-risk participants receive longer interventions ranging from 26 weeks to the traditional 52-week model. This approach is intended to focus program resources on people most likely to re-offend while avoiding overprogramming to people assessed as lower risk and potentially increasing their risk of re-offending (Lowenkamp, Latessa, & Holsinger, 2006).
The DV Pilot also allows counties to require additional programming for people on probation for domestic violence based on their assessed risk level and needs. Research has demonstrated that factors such as substance use, trauma exposure, and mental health challenges may contribute to intimate partner violence (Capaldi et al., 2012; Cafferky et al., 2018; Wilson, Graham, & Taft, 2014). Addressing these needs may also enable more effective participation in BIP interventions. The DV Pilot allows counties to require people on probation for domestic violence to participate in additional programs or services based on their assessed risk level and needs.
Data & Methodology
The report draws on two key data sources. First, we use statewide data provided by the California Department of Justice (DOJ) through their Automated Criminal History System (ACHS). The ACHS data include individually identified, event-level records of all arrests and convictions for people in California. The ACHS data allow researchers to identify the county of arrest or conviction and to construct demographic characteristics, criminal history characteristics, and recidivism outcomes. Second, we use data provided by county probation departments in Napa, Santa Barbara, and Santa Cruz — the three pilot counties that volunteered to partner with researchers on this evaluation. Importantly, the probation department data provides precise information about the date people start probation supervision and their placement onto domestic violence caseloads. We estimate the effects of the DV Pilot on two-year recidivism rates, measured as the overall rearrest rate, DV rearrest rate, overall reconviction rate, and DV reconviction rate.
Statewide Data
The statewide analysis leverages AB 372 as a natural policy experiment that allows us to estimate the effect of the DV Pilot on recidivism outcomes in the pilot counties relative to the non-pilot counties. This approach allows us to compare people who committed similar offenses across counties with different policy environments. For our statewide analysis, we use the ACHS data to identify people convicted of domestic violence and sentenced to probation in the pre-pilot period (2017–2018) and the post-pilot period (2021–2022). We exclude people sentenced during the implementation period (2019–2020) due to both the extended roll-out of the pilot across counties and the implementation complications brought by the COVID-19 pandemic. Our statewide analysis includes the six pilot counties.
In the statewide data, we do not have access to nuanced information about the circumstances of the conviction offense. We limit the study group to those convicted of offenses that clearly indicate domestic violence. However, there are cases related to domestic violence for which the conviction charges used by prosecutors are more general, such as a charge of assault. Our statewide analysis excludes these cases. One of the advantages of the supplemental county analysis is that we gain access to more detailed information about whether a conviction was associated with domestic violence based on the probation caseload data. A limitation of the statewide data is that we do not have access to precise probation start dates for people across the state. Instead, we predict the start date based on the timing of the conviction and the sentence received. Among those observations we were able to match with county data, we predict the start date exactly for 51% of cases, within three months for 93% of cases, and within six months for 100% of cases. We have no reason to suspect that there would be differences in the quality of our estimates of predicted start dates for people in pilot and non-pilot counties.
The two key strengths of the statewide analysis are that it allows us to include all six pilot counties and to compare outcomes for people sentenced to probation for domestic violence in pilot counties to those in non-pilot counties. To do so, we use a difference-in-differences model, which estimates the differential change in recidivism for people sentenced in the pilot counties relative to any change in recidivism experienced by people sentenced in the non-pilot counties over the same period. This approach allows us to adjust for both differences in the characteristics of these populations and any broader changes affecting recidivism outcomes. The validity of the model requires that we demonstrate parallel trends in recidivism outcomes prior to the implementation of the pilot. These trends are detailed in Appendix A.
Supplemental County Analysis
The supplemental county analysis draws on the same statewide data and time periods discussed above, but we have access to additional details about the probation experience of people sentenced in the three pilot counties that partnered with researchers for this study. Here we use both county and state data to identify people sentenced to probation for domestic violence offenses. We estimate the change in recidivism outcomes for those convicted of domestic violence offenses compared to people who were sentenced to probation for other types of offenses. We exclude people from the comparison group who were eligible for probation term limits under AB 1950 (2021) because this policy change overlaps with the implementation of the DV Pilot and likely had independent effects on recidivism. One disadvantage of this supplemental county analysis relative to the statewide analysis is that we are comparing the outcomes of people with different types of offenses.
As is the case with the statewide analysis, we use a difference-in-differences model with a pre-implementation period of 2017–2018 and a post-implementation period of 2021–2022. Again, this approach requires that we demonstrate parallel trends in recidivism outcomes for the treated and comparison groups — in this case, people sentenced to probation for domestic violence offenses and people sentenced for other (non-AB 1950 eligible) offenses. Given substantial differences in the characteristics of these groups, we may not expect their recidivism outcomes or trends to be similar, particularly when the composition of these groups is also changing over time. We do not observe parallel trends across all recidivism outcomes within the three partner counties, and we note the lack of parallel trends where relevant in the presentation of the findings below. These trends are detailed in Appendix B.
Findings
Statewide Analysis
We begin by presenting the characteristics of people sentenced to probation for domestic violence offenses in the pre- and post-pilot periods across the pilot and non-pilot counties (Table 3).
Most notable is that the pilot counties represent a small share of the total population in the state and, therefore, we have fewer observations for the pilot than non-pilot counties. However, the characteristics of people in these groups are quite similar. Most people sentenced to probation for domestic violence are male and the average age for this group is approximately 35 years old. The largest racial/ethnic group is Hispanic, but the share Hispanic is somewhat larger in the pilot counties than the non-pilot counties (52.9% vs. 49.1% in the pre period).
Similarly, the share of this population identified as Black is smaller in the pilot counties than the non-pilot counties (8.1% v. 17.9% in the pre period). These differences in the racial/ethnic composition are largely explained by differences in the composition of the broader populations of people living in the pilot versus
non-pilot counties.
Table 3 also shows a pre-post decline in the number of people convicted and sentenced to probation for domestic violence in California in both pilot and nonpilot counties. With the reduction in the number of convictions, we also see an increase in the severity of criminal history characteristics. In both the pilot and nonpilot groups, the share of people with a prior felony or prior serious conviction increases, as does the number of prior convictions. These trends suggest a broader movement toward prioritizing correctional resources for more serious offenders.
Before implementing the difference-in-differences model to estimate the causal effects of the DV Pilot on recidivism outcomes, we first descriptively explore how recidivism rates changed over time. Figure 1 summarizes the change in two-year rearrest, DV rearrest, reconviction, and DV reconviction rates between the pre- and post-pilot periods. The two-year rearrest and DV rearrest rates declined for both groups, with slightly larger declines among people sentenced in the pilot counties. These changes in descriptive rearrest rates show the importance of accounting for broader statewide recidivism trends.
The change in two-year reconviction and DV reconviction rates are somewhat different for the pilot versus non-pilot counties. Among those sentenced to probation for domestic violence in the pilot counties, the overall reconviction rate declined by 9.4 percentage points, or 26% relative to the baseline rate. The reconviction rate in the non-pilot counties also declined, but by only 4.4 percentage points, or 13%. These findings suggest there was a broader trend leading to the reduction in overall reconvictions, but that the pilot counties achieved a larger reduction than the non-pilot counties. When we turn to reconvictions for domestic violence offenses, the difference is more pronounced. There was a decline in the DV reconviction rate of 0.6 percentage points (or 5%) in the non-pilot counties, but in the pilot counties reconvictions for DV offenses declined by 2.5 percentage points, or 27% relative to the baseline rate. Taken together, these descriptive changes suggest the pilot counties were able to achieve larger reductions in reconviction rates than the non-pilot counties over time.
As described in the methodology section, the difference-in-differences model adjusts for differences in the characteristics of people in the pilot and non-pilot counties, as well as changes over time in the characteristics of people sentenced to probation for domestic violence offenses. The model also “differences out” broader trends in recidivism that may show up in both the pilot and non-pilot counties to estimate the differential changes in outcomes in the pilot counties. In other words, our goal is to isolate the effect of the pilot on recidivism outcomes, excluding any other potential drivers of changes in recidivism rates.
The findings from this causal model parallel the descriptive analysis detailed above. We do not find evidence of a differential change in the two-year overall rearrest or DV rearrest rates. Figure 2 shows our estimates of the effect of the pilot on rearrest are near zero and non-significant. In other words, rearrests declined at similar rates in both the pilot and non-pilot counties and, therefore, we do not attribute these declines to the DV Pilot. It is important to note that California is a “mandatory arrest” state. As a result, when law enforcement responds to a domestic violence incident, they are required to make an arrest regardless of the severity of the incident.
When we turn to the analysis of two-year reconviction rates, we find evidence that the DV Pilot reduced recidivism. We estimate a significant reduction in the two-year overall reconviction rate of 5.5 percentage points, representing a 15% reduction relative to the baseline rate for the pilot group. We also estimate a significant reduction in the two-year DV reconviction rate of 1.8 percentage points, or a 20% reduction relative to the baseline rate. In comparison to impacts typically found in the literature, these changes represent large reductions in recidivism for people convicted of domestic violence (Babcock, Green, & Robie, 2004; Travers et al., 2021). Full regression results for the statewide analysis are presented in Appendix A.
Statewide Analysis
While we observe declines in the overall and domestic violence rearrest rates in both pilot and non-pilot counties, we do not find evidence of a differential decline in rearrest rates in the pilot counties. Given the evidence of a differential decline in reconviction rates, this finding requires further investigation, so we explore additional measures to better understand how rearrest rates are changing in the pilot counties relative to the non-pilot counties.
We do not see a change in the rearrest rate for people sentenced to probation in the pilot counties relative to the change for people in non-pilot counties, but we do find evidence of a relative change in the composition of those rearrests. For the pilot counties, we estimate a significant reduction of 2.0 percentage points in the felony rearrest rate and a corresponding increase of 1.7 percentage points in the misdemeanor rearrest rate, although the estimate for the change in the misdemeanor rate is not statistically significant (Figure 3). These findings indicate that although the rearrest rate did not decrease in the pilot counties relative to the non-pilot counties, there was a significant relative reduction in the severity of rearrest offenses in the pilot counties.
We also examine shifts in the composition of reconvictions (felonies versus misdemeanors) for people sentenced to probation in the pilot counties relative to those in the non-pilot counties. We find the decrease in the reconviction rate in the pilot counties was driven by a reduction in reconvictions for misdemeanor offenses. We estimate a 0.4 percentage point decrease in the felony reconviction rate, but this estimate is not significant. For misdemeanor reconvictions, we estimate a significant, 5.1 percentage point reduction in the pilot counties relative to the non-pilot counties (Figure 3). See Appendix A for full regression results.
In addition to felony and misdemeanor recidivism, we explore two additional measures: the number of days to rearrest and the number of days to reconviction for those who do recidivate within the two-year recidivism window. We find no evidence of a differential change in the time to rearrest for people rearrested in the pilot counties relative to the non-pilot counties. However, we estimate a 39-day increase in the time to reconviction, showing that people sentenced to probation for domestic violence in the pilot counties spent more time in the community before being reconvicted. Time to recidivism is one additional measure that can indicate some improvement in outcomes even in cases where recidivism does occur. See Appendix A for full regression results.
Supplemental County Analysis: Three Counties
We supplement the statewide analysis using data from three counties. As described above, the county probation department data allow us to identify a broader group of people on probation for domestic violence offenses and to use their precise probation start date to construct their recidivism windows. However, this approach also necessitates a comparison of outcomes for people on probation for DV offenses with those on probation for non-DV, non-AB 1950 offenses. We do not expect these groups to have similar demographic or criminal history characteristics.
Table 4 shows the pre- and post-pilot period characteristics of those on probation for DV and non-DV offenses. Demographically, we see the DV group is more likely to be male (85.9% v. 75.0%) and more likely to be identified as Black (5.5% v. 3.2%) or Hispanic (54.3% v. 46.7%) in the pre period. These differences between the DV and non-DV groups persisted over the period, but we see an increase in the share identified as Hispanic and a corresponding decrease in the share identified as White in both groups.
Table 4 also shows the criminal history of the DV group is more serious than the comparison group. People in the DV group are more likely to have a prior felony (32.9% v. 13.9%) or prior serious conviction (8.1% v. 3.1%) in the pre period. The DV group also has a higher count of prior convictions for any offense (3.4 v. 1.4) on average in the pre period, and a higher count of prior arrests or convictions for DV offenses. In contrast with the non-DV group, the criminal history of the DV group appears to increase in severity in the post period, with increases in the share with a prior serious conviction and an increase in the total number of prior convictions.
Figure 4 shows descriptive recidivism rates for these two groups. Recidivism rates are higher for the DV group than the comparison group across all measures. The DV group is more than twice as likely to be rearrested or reconvicted than the non-DV group in the pre- and post-pilot periods.
As we might expect, the DV group has much higher — more than 10 times as high — rearrest and reconviction rates for new DV offenses than the comparison group. Both groups experience substantial reductions in overall rearrest and reconviction rates over this period. However, the DV group also experiences substantial reductions in the DV rearrest rate (3.7 percentage points) and DV reconviction rate (3.2 percentage points). The comparison group experiences a small increase in the DV rearrest rate (0.4 percentage points) and small decrease in the DV reconviction rate (0.3 percentage points). However, it is important to note the very low baseline DV reconviction rate of less than one percent for the comparison group.
We use a causal model to adjust for differences in the demographic and criminal history characteristics of the DV and non-DV groups, as well as broader changes over time in recidivism outcomes. This model estimates the effects of the DV pilot on recidivism outcomes for people on probation for DV offenses relative to those on probation for non-DV offenses in three of the six pilot counties. We estimate differential declines in rearrest and reconviction rates (see Appendix B for regression results). However, we are cautious about the interpretation of these findings because our examination of recidivism trends suggests recidivism was declining for the DV group prior to the pilot and continued through the pilot period, driven by pre-existing declines in Napa County. While declines in recidivism prior to and following the implementation of the pilot are a positive outcome for Napa, this pre-existing trend complicates the causal identification strategy. Therefore, we exclude Napa for our final analysis.
Figure 5 summarizes the estimated changes in recidivism for people on probation for DV offenses relative to changes in recidivism over the same period for people on probation for non-DV offenses. Consistent with the statewide analysis, we do not find any evidence of increases in recidivism rates associated with the implementation of the DV Pilot. Here we estimate a 1.2 percentage point reduction in the rearrest rate, but this estimate is not significant. However, we find evidence of a significant reduction of 2.9 percentage points in the DV rearrest rate. This finding suggests the subset of counties included in this supplemental analysis may have been able to achieve larger reductions in DV rearrest rates than the full pilot group.
Consistent with the statewide findings, we estimate a 4.4 percentage point reduction in the overall reconviction rate and a 2.3 percentage-point reduction in the DV reconviction rate. When we examine pre-period trends in recidivism for this group, we find some evidence of a pre-existing downward trend in the overall reconviction rate for the DV group relative to the comparison group (see Appendix B). Therefore, we caution that we may be over-estimating the effect of the pilot on DV reconviction rates for the subset of counties included here. Although more challenging from a casual evaluation perspective, the supplemental county analysis reinforces the findings from the statewide analysis.
Conclusions & Policy Implications
This study provides evidence that California’s AB 372 Domestic Violence Pilot — designed to replace a uniform, one-size-fits-all BIP model with a risk- and needs-based approach — can produce meaningful reductions in recidivism. Across both statewide and supplemental county analyses, we found evidence that the pilot is associated with reductions in overall and domestic violence offense reconviction outcomes. We did not find evidence that the DV Pilot decreased rearrest rates, but we did find evidence of a shift toward lower-severity arrests in the pilot counties relative to the non-pilot counties. These findings suggest that allowing counties the flexibility to tailor domestic violence program dosage to individuals’ assessed risks and needs led to reductions in some measures of recidivism.
At the same time, this evaluation highlights important considerations for interpretation and future policy development. The statewide analysis benefits from a strong quasi-experimental design and supports a causal interpretation of the findings. However, the supplemental county analysis proves more challenging
due to differences in the pre-pilot recidivism trends of the within-county comparison group. These issues are more likely to arise when the characteristics of the comparison group are very different from the treatment group. In addition, variation in implementation across pilot counties — including differences in risk thresholds, BIP dosage, and additional requirements for programs and services — suggests that local context and implementation decisions may influence outcomes.
Taken together, these findings have several potential implications for policy and practice. First, they are broadly consistent with continuing to offer counties the flexibility to use an assessment-driven RNR model as an alternative to a one-size-fits all approach. Second, the results indicate that shorter, less intensive BIP interventions can be used for lower-risk individuals without increasing recidivism. Third, requiring additional interventions aligned with needs — such as substance use and mental health — may be an important component of addressing the underlying drivers of domestic violence. Continued data collection and evaluation will be critical to understanding longer-term impacts of this policy and identifying which program design components are most effective.
About this research
This research was made possible through a partnership between the California Policy Lab at the University of California and the Committee on Revision of the Penal Code, a state agency that studies and makes recommendations to improve California’s criminal legal system.
Acknowledgements
Support for this research was generously provided by the Blue Shield of California Foundation, the California State Association of Counties, and the Committee on Revision of the Penal Code. The research was also made possible through partnerships between the California Policy Lab and the following three counties participating in the Domestic Violence Pilot: Napa, Santa Barbara, and Santa Cruz. These counties shared data with researchers and contributed significant staff time and insight to the project, and we appreciate their support.
We also thank other supporters of the California Policy Lab, including the University of California Office of the President Multicampus Research Programs and Initiatives, M21PR3278, The James Irvine Foundation, and the Woven Foundation for their generous support. The views expressed are those of the authors and do not necessarily reflect the views of our funders. All errors should be attributed to the authors.
The California Policy Lab generates research insights for government impact. We are an independent, nonpartisan research institute at the University of California with sites in Berkeley, Los Angeles, and Sacramento.
This research publication reflects the views of the authors and not necessarily the views of our funders, our staff, our advisory boards, the California State Association of Counties, the California Committee on the Revision of the Penal Code, the State of California Department of Justice, or the Regents of the University of California.
Suggested Citation: Bird, M., Nguyen, V., Skog, A., Lacoe, J., Raphael, S., Yang, E. (2026). The Effects of California’s Domestic Violence Pilot on Recidivism Outcomes. California Policy Lab, University of California. https://capolicylab.org/the-effects-of-californias-domestic-violence-pilot-on-recidivism-outcomes/
References
Assembly Bill 372 (Stone), Chapter 290, Statutes of 2018.
Babcock, J. C., Green, C. E., & Robie, C. (2004). Does batterers’ treatment work? A meta-analytic review of domesticviolence treatment. Clinical Psychology Review, 23(8), 1023–1053.
Bonta, J., & Andrews, D. A. (2007). Risk-Need-Responsivity model for offender assessment and rehabilitation. Public Safety Canada.
Cafferky, B. M., Mendez, M., Anderson, J. R., & Stith, S. M. (2018). Substance use and intimate partner violence: A meta-analytic review. Psychology of Violence, 8(1), 110–131.
California Penal Code §1203.097.
California Penal Code §1203.099.
Capaldi, D. M., Knoble, N. B., Shortt, J. W., & Kim, H. K. (2012). A systematic review of risk factors for intimate partner violence. Partner Abuse, 3(2), 231–280.
California State Association of Counties. (2025). AB 372 Domestic Violence Pilot Program: Legislative Report: Year 5.
Feder, L., & Wilson, D. B. (2005). A meta-analytic review of court-mandated batterer intervention programs: Can courts affect abusers’ behavior? Journal of Experimental Criminology, 1(2), 239–262.
Lowenkamp, C. T., Latessa, E. J., & Holsinger, A. M. (2006). The risk principle in action: What have we learned from 13,676 offenders and 97 correctional programs? Crime & Delinquency, 52(1), 77–93.
MacLeod, D., Pi, R., Smith, D., & Rose-Goodwin, L. (2008). Batterer intervention systems in California: An evaluation. San Francisco: Judicial Council of California/Administrative Office of the Courts.
Morrel, T. M., Elliott, J. D., Murphy, C. M., & Taft, C. T. (2003). Cognitive behavioral and supportive group treatments for partner-violent men. Journal of Consulting and Clinical Psychology, 71(3), 517–528.
Shepard, M. F., & Pence, E. L. (Eds.). (1999). Coordinating Community Responses to Domestic Violence: Lessons from Duluth and Beyond. Sage Publications, Thousand Oaks, CA.
Travers, Á., McDonagh, T., Elklit, A., & Shevlin, M. (2021). The effectiveness of interventions for intimate partner violence perpetrators: A meta-analytic review examining the role of risk-need-responsivity principles. Clinical Psychology Review, 86, 102029.
Wilson, D. B., Feder, L., & Austin, S. (2021). Court-mandated interventions for individuals convicted of domestic violence: A systematic review. Campbell Systematic Reviews, 17(3), e1181.
Wilson, I. M., Graham, K., & Taft, A. (2014). Alcohol interventions, alcohol policy and intimate partner violence: A systematic review. BMC Public Health, 14, 1–11.