The first chapter explores the determinants of people's decisions to take 401(k) loans. We model factors that rationally would induce people to borrow from their pension plans, and we explain why people do not often use 401(k) loans to replace their more expensive credit card debt. Next we test our hypotheses using a rich dataset and show that people who are liquidity-constrained are more likely to have plan loans, while the better- off take larger loans when they do borrow. Plan characteristics such as the number of loans allowed also influence borrowing and loan size in interesting ways, while loan interest rates have only a small impact. The second chapter investigates the determinants of defaults on 401(k) loans, using a rich dataset of over 100,000 participants who terminate employment with a plan loan outstanding. Overall, one in ten plan loans results in a default, and eight of ten workers who leave a job with a plan loan outstanding then default on that loan. Explanations relate to employee characteristics and plan design features: those who are liquidity-constrained are more likely to default than repay theft loans at job termination. Moreover, borrowers with several smaller loans are more likely to default than are participants with a single loan of the same total size, perhaps due to heterogeneity in credit demand or lack of self-control. Local economic conditions have little impact on 401(k) loan defaults during the period we analyze. The third chapter explores whether social interactions influence 401(k) investors' decisions to hold equity and allocate their portfolios. We provide empirical evidence that participants are influenced by their coworkers when they make equity investment decisions. Specifically, we show that individuals are likely to increase their risky share if their peers earned higher equity returns in the past period relative to average returns; they are also likely to decrease their risky share when past peer equity returns are strongly negative. These results are consistent with the limited attention hypothesis that people are more likely to pay attention to significant outcomes.