Prison Rehabilitation Programs: Efficiency and Targeting - William Arbour Guy Lacroix

Once the evaluator has devised an intervention plan, the case is transferred to another agent for the remaining of the sentence and the inmate decides whether or not to follow the initial recommendations. [...] If they were evaluated, we include a vector of scores for components of the LS/CMI in the regression (si); we set these scores to 0 for individuals not evaluated.14 xi is a vector of other controls, including fixed effects for the sentence year and for the prison, dummy variables for the category of crime, age groups, whether the inmate has dependents, and whether he is part of an indigenous ethni. [...] The idea behind her approach is to compare the change in the estimated coefficient to the change in the R2 after adding additional control variables. [...] In Figure 5, in blue, we plot the distribution of the predicted treatment effects in the sample, and most of the distribution lies on the left of zero as expected. [...] = 0.0115 4 predicted TEs 2 0 −0.3 −0.2 −0.1 0.0 0.1 0.2 Predicted Treatment Effects Figure 5: Density of Predicted Treatment Effects Notes: The red line shows the distribution of the predicted treatment effects corresponding to the sum of the average treatment effect and the simulated zero-mean shocks with the estimated variances of the CATEs.