| Sumario: | Using data from criminal cases in the State of S\~ao Paulo, Brazil, I analyze whether alternative sentences -- e.g., fines or community services -- impact recidivism. To do so, I define my treatment variable using each case's final rulings and leverage the random assignment of judges within a court district to identify the effect of alternative sentences and find a small and statistically insignificant effect. Moreover, I show that the usual identification strategy, that uses only the trial judge's sentence, fails to identify the correct treatment effect parameter because the trial judge's decision may misclassify the final sentence due to the appeals process. By using each case's final ruling, I also show that this misclassification bias may be as large as 10% of the correctly estimated treatment effect parameter. To correct this measurement error problem, I propose a novel partial identification strategy to identify the marginal treatment effect (MTE) with a misclassified treatment. This method explores restrictions on the relationship between the misclassified treatment and the correctly measured treatment, allowing for dependence between the instrument and the misclassification decision. Using each case's final ruling, I show that this partial identification strategy works appropriately in my empirical application.Comment: Version 2 includes the empirical result
|
|---|
