Criminal specialization in the criminal justice context

The dissertation consists of two studies. Study 1 examines how criminal specialization predicts the sentencing outcomes. Theories of sentencing have pointed out the association between the sentence and the assessment of the defendant's risk and culpability, and one of the most important indicators of an individual's risk is his or her criminal records. Most quantitative studies of sentencing today take criminal records into consideration by controlling for the number of prior criminal justice contacts, and overlook the nature of the prior crimes. The concept criminal specialization refers to the tendency for an individual to repeat the same or a set of related crimes. In the present study, I use four different measures, which model different dimensions of criminal specialization. I model criminal specialization from the records of over 110,000 defendants convicted in New York State between 2010 and 2012, and add the measures of criminal specialization to models explaining incarceration and incarceration length. Study 1 finds that even though the four different measures capture different aspects of specialization, all the measures find that the analytic sample contained a mix of versatile defendants and specialized defendants. However, the four measures perform very differently in predicting the sentence.;Study 2 estimates the magnitude of the plea discount in New York State during three different observation periods. Piehl and Bushway (2007), building on the methodological framework of Smith (1986), introduced an approach to estimate the magnitude of sentence difference due to charge bargaining. However, what is absent in this framework is the estimate of overcharging. Prosecutors may, intentionally or unintentionally, file initial charges that are unlikely to secure a conviction at trial. As a result, the counterfactual sentence estimated from the initial charge may not be a credible threat when the defendant decides between going to trial or pleading guilty. The present study takes overcharging into consideration, and compare the estimates of the plea discount when overcharging is taken into account and when it is not. I obtain the benchmark of the estimates from the prediction of the model "bargaining in the shadow of trial." The study finds that taking overcharging into account brings the estimates of the plea discount closer to the prediction of the "shadow of trial" model.