UC Berkeley

This dissertation consists of three studies which analyze different aspects of risky behaviors and criminal participation.

A longstanding question is whether alcohol and marijuana use by teenagers exerts a "stepping stone" effect, increasing the chances that they will use harder drugs in the future. Empirically, teenagers who use alcohol or marijuana in one period are more likely to use cocaine in the future. This pattern can be explained in one of two ways: by a causal effect of soft drug consumption on future consumption of hard drugs (i.e., a true stepping-stone effect) or by unobserved characteristics that make people more likely to use soft drugs at a relatively young age, and hard drugs at a later age (i.e., correlated unobserved heterogeneity). Distinguishing between these alternatives is highly policy relevant because, to the extent that there is a true stepping stone effect, policies that reduce the use of soft drugs by young people will have lasting impact on the use of hard drugs by adults. In Chapter 1, I use data from the National Longitudinal Study of Youth 1997 (NLSY97) to estimate a dynamic discrete choice model of teenager's use of alcohol, marijuana and cocaine over multiple years, and separately identify the contributions of state dependence and unobserved heterogeneity. I find modest-sized but statistically significant "stepping-stone" effects from softer to harder drugs that are largest among the youngest individuals in my sample. In contrast, I find little evidence of a stepping stone effect from cocaine to alcohol or marijuana. Simulations show that restricting alcohol and marijuana use at young age has a modest impact on reducing later cocaine use.

Chapter 2 examines the role of an increase in alcohol consumption on drug initiation, hard drug consumption, and criminal participation. Using a regression discontinuity research design, I exploit the discontinuous increase in alcohol consumption at age 21 provided by the minimum legal drinking age. Using a survey of respondents during the year after they turned 21, I found that the probability of cocaine initiation decreased by 1.5 percentage points and the share of respondents who consumed cocaine in the last year decreased by 2 percentage points. Self-reported criminal participation, such as drug dealing, property destruction and attacking an individual, remained unchanged at age 21, with the exception of stealing, which decreased by 3 percentage points. These estimates are robust to a variety of specifications.

Between 1993 and 1995, a number of states implemented "Three Strikes and You're Out" laws that enhance the sentence length for repeat violent offenders. Chapter 3 develops a simple dynamic model that suggests that these laws will lead to an increase in the number of criminal cases that go to trial, rather than being settled with a plea bargain, since the threat of higher future sentences increases the cost of a being convicted for a strikeable offense. I use data from the 1990-2006 State Court Processing Statistics database and a state-by-year difference-in-differences research design to compare the change in the likelihood of plea bargaining by violent offenders after the passage of Three Strikes laws, relative to the trend among non-violent offenders. I also separately examine effects for offenders with at least one prior violent conviction, and compare the effects of the Three Strike law in California, which imposes extra sentencing for any third felony (violent or not), versus the eleven other states with Three Strikes laws. The results show that the introduction of Three Strikes laws significantly reduce the number of criminal cases that are settle with a plea bargain, imposing a potentially costly burden on the legal system.