This dissertation presents three independent research projects. The first studies the effect of prison crowding on violent misconduct, the second presents a model of prison gangs as profit-maximizing suppliers of illicit goods, and the third is an impact analysis of eyewitness identification protocols.

Chapter 1 provides a brief overview of each research project.

The first research project, Chapter 2, seeks to estimate the causal relationship between prison crowding and violent behavior. This study exploits exogenous variation in California prison populations, resulting from a Supreme Court mandate to reduce prison crowding, to estimate the effect on violence. Using both difference-in-differences and instrumental variables identification strategies, a significant positive relationship is identified that is robust to a variety of model specifications. These are the first empirical estimates showing a causal link between crowding and violence, suggesting that reducing prison crowding by 10 percentage points leads to a reduction in the rate of assault and battery of approximately 15%. In addition, differential reductions in the rates of violence between population types is presented as evidence of a compositional effect associated with shocks to prison crowding, which poses a threat to the validity of empirical estimates of the link between crowding and violence.

Chapter 3 synthesizes existing research on prison gangs into an explicit modeling framework that treats gangs as profit-maximizing suppliers and sources of informal governance in an illicit marketplace. The model offers broad policy implications for prison enforcement and highlights the futility of certain policy approaches that don't account for the profit motive underlying gang activity.

In Chapter 4, we test for the presence of an identifiable impact on police clearance rates from the implementation of statewide reforms that adopt the sequential lineup process. We find insufficient evidence to identify an average effect for all reform states, but evidence this is the result of heterogeneous effects. We are also able to bound an the possible effect on clearance rates to rule out concerns that reforms lead to large reductions in positive identifications.