The Academic Cost of Juvenile Incarceration: Evidence from Regression Discontinuity and Instrumental Variable Analyses
*Names in bold indicate Presenter
Two statistical estimation methods are used – a regression discontinuity design and an instrumental variable model – to estimate the causal effect of incarceration on the educational outcomes of youth. Both designs offer advantages and disadvantages in overcoming the systematic differences between youth who are incarcerated and those who receive community dispositions. The DOP’s use of a risk assessment tool and cutoff points for disposition recommendations creates an ideal opportunity to employ a regression discontinuity design, which compares a narrow band of youth with asset scores on either side of the cutoff point used to recommend placement. Youth on either side of the cutoff points have very similar risk assessment scores, but receive different recommendations from the DOP, providing an estimate of the impact of the recommendation on future educational outcomes.
While the regression discontinuity design offers strong internal validity, the analyses are limited to observations within the narrow bands around the cutoff points, thereby omitting many of the youth in the dataset. The tradeoff is particularly salient in this analysis, when the sample is already limited to those observations with measured outcomes. An alternative estimation strategy leverages the seemingly random assignment of cases to family court judges. Because judges have differing proclivities for sending juveniles to prison, the assignment to a judge gives an individual youth a probability of being incarcerated that is not related to his/her characteristics or case. As a result, the judge’s placement rate is a strong instrument for an incarceration disposition in the model of the impact of incarceration on educational outcomes. Results across both estimators show that juvenile incarceration sentences decrease academic effort and attendance and increase risk of school dropout.