24/7 Sobriety
Greg Midgette on the effects of the 24/7 Sobriety program (podcast)
Reference: “Criminal Deterrence: Evidence from an Individual‐Level Analysis of 24/7 Sobriety” by Beau Kilmer and Greg Midgette.
Background and Aims The US state of South Dakota’s 24/7 Sobriety Program (24/7) requires individuals charged or convicted of alcohol-involved offenses to avoid alcohol and submit to twice-daily or continuous alcohol testing. We evaluated the impact of the 24/7 program in the US state of Montana.
Methods Using data from everyone in Montana who was convicted of their second driving under the influence (DUI) offense from 2009 through mid-2014, we described program violations among 24/7 participants and then estimated the effect of 24/7 participation on the probability of DUI re-arrest. To address potential selection issues related to individual-level 24/7 participation, we used an instrumental variables approach that exploits county-level variation in program adoption.
Results Among 2,768 persons arrested for a second DUI in our analytic sample, 356 24/7 participants were monitored for an average of 173 days (median=112 days). Among the 332 participants monitored by breath test, 95.5% of scheduled alcohol breath tests were completed and did not lead to a program violation. After controlling for individual- and community-level covariates as well as year and Co. fixed effects, our instrumental variable models suggest that participation in 24/7 reduced the 1-year DUI re-arrest probability by at least 80% (Preferred model: 86% decrease; 8.9 percentage points) compared with a counterfactual group of persons arrested for a second DUI over the same period but not assigned to the program.
Conclusions South Dakota USA’s 24/7 Sobriety Program appears to work in Montana as well. Certain delivery of immediate but modest sanctions for repeat driving under the influence (DUI) arrestees who violate alcohol abstinence orders appear to be able to reduce future DUI arrests.
Evaluating 24/7 Sobriety Program (North Dakota) Participant Reoffense Risk
• Over 90% of ND 24/7 Sobriety Program participants did not reoffend within three years.
• Reoffense was most likely during the 12-month period after program participation ended.
• Age was a significant factor in survival as drivers over age 35 had a six times greater survival risk than the younger cohort ages 18–35.
• Further investigation into a finding that drivers with prior program history may be warranted.
Classical rational choice is predicated on the concept of expected utility theory (Rubinstein et al. 2007), which theorizes individual decisions to be the output of a rational calculus weighing the expected costs and benefits of potential actions. However, much of what we know about the influence of sanction certainty on offender decision-making comports with the behavioral science perspective that departures from rational behavior observed among individuals often manifest in predictable ways (Loughran 2019; Pogarsky, Roche, and Pickett 2018). In particular, scholars have demonstrated that misjudgments about probabilities, specifically in terms of how humans judge and act on them, are numerous (Camerer 1998; Kahneman and Tversky 1979). Cook (2016:1159) argued this point explicitly in consideration of the mixed evidence generated by Hawaii’s Opportunity Probation with Enforcement (HOPE) and subsequent Bureau of Justice Assistance-funded replications (Hawken & Kleiman 2009; Lattimore et al. 2016), which are programs built to deter recidivism in probation through certain sanctions: “a relatively high (but far from certain) punishment…will be the more effective deterrent when assessed by the traditional expected utility framing but that that conclusion is less obvious under the findings of modern behavioral science.”
This study focuses on how justice-involved individuals process information about—and more importantly—act on perceived uncertainty through the behavioral economics concept of ambiguity, or the degree of confidence one holds regarding an unknown probability such as the chance of detection or punishment. In this study, we integrate theoretical insights from criminology and behavioral economics to develop and test novel predictions about the key role of ambiguity in offender decision-making. The study of ambiguity in perceptions of risk in criminal deterrence is relatively nascent, but its importance has been demonstrated in the offender decision-making process (Loughran et al. 2011; Pickett, Loughran, and Bushway 2016; Pogarsky et al. 2018). We contribute to the understanding of ambiguity and certainty in the study of offender decision-making by estimating an ambiguity parameter that to this point has only been described in theory and approximated empirically in contrived settings. Leveraging novel data from a “real-world” community supervision program, we test the impact of introducing ambiguity around the certainty of detection into a decision process in which an individual faces a real threat of incarceration for program violations.
Using data from two states, we directly consider the role of ambiguity in sanction certainty in the context of 24/7 Sobriety, a program that has demonstrated important reductions in drunk driving for community supervision participants (National Institute of Justice 2015). Specifically, we leverage a single difference in the implementation of 24/7 Sobriety that leads to consistent but ambiguous sanction certainty in one state (Montana) compared to another (South Dakota), thereby creating a credible counterfactual comparison. The 24/7 program in Montana mirrors South Dakota, except that they differ on one important dimension: the breathalyzer-determined blood-alcohol concentration (BAC) level that denotes a program violation is zero in South Dakota, and 0.02 mg/l in Montana. Central to our quasi-experimental design, these violation thresholds are explicitly provided to participants. However, the choice to drink (or not) is made with ambiguity surrounding the probability of violation in Montana that does not exist in South Dakota. Evidence of drinking may not result in a violation in Montana, but it certainly will in South Dakota. These conditions provide the opportunity to make reasonable comparisons between a program that incorporates strict sanction certainty and one that, in an attempt to be more flexible, operates with more ambiguity, thus allowing us to test the theoretical prediction that individuals will be ambiguity-seeking in the presence of a near-certain loss.
***
The evidence we present suggests, first, that the basic design of 24/7 may be an effective strategy to deter risky behaviors in community corrections, more so when ambiguity is minimized. This sheds light on the potential theoretical mechanisms at play in Swift, Certain, Fair (SCF) programs such as 24/7 and HOPE. Evaluations of SCF-type programs in multiple states have demonstrated promising crime and substance use reduction effects on participants (e.g., Hamilton et al. 2016; Hawken and Kleiman 2009; Taxman et al. 2003). However, recent research has called into question the universal effectiveness of strategies based on the SCF paradigm (Lattimore et al. 2016; O’Connell, Brent, and Visher 2016), as well as the utility of this certainty-based supervision approach and leading some to argue for the outright abandonment of these types of interventions (Clear and Frost 2014; Cullen, Pratt, and Turanovic 2016). The potential sources of ambiguity in SCF programs that fundamentally rely on certainty to affect behavior are numerous, from minor implementation choices that soften hard and fast rules, to the clarity with which instructions and expectations are conveyed, to core program components such as drug testing. In such programs, ambiguity may result in adverse outcomes.