The focus of this study is to compare matching estimators to conventional logistic regression. Matching estimators can be used for analysis of observational data where assignment to treatment lacks randomization. The individual units of study are matched to units in the opposite treatment groups and then overall treatment effects can be estimated. When the necessary assumptions are met this estimate is unbiased. I will be looking at the relationship between the educational experience of juvenile delinquents and their behavior after release from a correctional facility. I speculate that matching estimators are more effective than the regression modeling strategies at estimating the average treatment effect in an observational study because they impute counterfactuals while controlling for confounding variables (covariates) and do not require a functional form and therefore, less assumptions need to be met.