Mon, Feb 13, 2017, noon:
Daniel Almirall, "Getting SMART about adaptive interventions"
a PSC Research Project
Intellectual Merit: A persistent question in criminology and corrections policy is how incarcerating convicted offenders affects their future behavior and social and economic outcomes. From a policy perspective, perhaps the most important question is whether incarceration increases or decreases the likelihood of reoffending, particularly for low-level offenders ?on the margin? for whom probation, jail, or other sanctions are potentially appropriate alternatives to incarceration. Given the dramatic increase in the number of people incarcerated in the U.S. over the last three decades and the high public cost of incarceration compared to other forms of punishment, it is important to understand how incarceration affects criminal offending, as well as offenders? employment prospects when they return from jail or prison. There have been considerable attempts to study the social and economic consequences of incarceration by comparing the outcomes of offenders who received custodial and non-custodial sanctions, but a fundamental problem with such observational studies is that differences between these groups could be due to pre-existing and unmeasured differences that affect both the sentencing decision (e.g., incarceration vs. probation) and the future outcomes (e.g., reoffending, employment). For example, a judge?s assessment of how likely an offender is to reoffend ? and thus the sentencing decision ? can be influenced by information available to them that does not get recorded in administrative data (e.g., statements from witnesses), resulting in omitted variable bias/unobserved confounding. Such problems with causal inferences about incarceration led one recent review of this literature to conclude that ?existing research is not nearly sufficient for making firm evidence-based conclusions for either science or public policy.?
The proposed study would advance this research by exploiting quasi-experimental conditions created by the combination of sentencing guidelines and the random assignment of cases to judges. We would employ two strategies. First, we would use random judge assignments as ?instruments? that
The causal effects of incarceration are difficult to estimate because the differences in outcomes between offenders sentenced to prison/jail vs. probation, even for the same offense, may be due to pre-existing and unmeasured differences between these groups that affect both the sentencing decision and the outcome under study. Recently, several studies have attempted to overcome this problem by using the random assignment of cases to judges as an ?instrument? that generates cleaner counterfactual comparisons between offenders who received custodial vs. non-custodial sanctions, such that inferences about the effects of incarceration on future reoffending [2,3] and employment/earnings [4,5] are less vulnerable to such residual confounding. The logic is that if (a) there is substantial variation across judges in how they would sentence the same offender and (b) the determination of the judge is random (or at least unrelated to pre-existing differences between groups of offenders), then the instrument will isolate the variation in the treatment (e.g., incarceration) that is unrelated to pre-existing differences. Such studies advance the literature on the ef?fects of incarceration and warrant replication on larger and more representative samples. However, the strategy of using judges as instruments for studying the effects of incarceration also has limitations, some of which can be addressed through more powerful research designs. Our pro?posed study would replicate and extend quasi-experimental research on the effects of custodial sanctions by accomplishing the following objectives:
1. Collecting/ archiving a unique data set: We will construct and place in the public domain data on all felons convicted in the state of Michigan from 2003-06 (N=319,212), including their background characteristics, sentences, and incidents o
|Funding:||National Science Foundation (SES 1061018)|
Funding Period: 06/01/2011 to 05/31/2014
Country of Focus: USA