Essays in macroeconomics and education

I use a canonical model similar to that of Golosov, Tsyvinski, and Werning (2006) to highlight salient properties of the standard tax system used to decentralize constrained efficient allocations. I first show that while labor and wealth taxes are mainly used for efficiency purposes, risk-sharing is achieved through lump-sum transfers. I then show through various examples that ignoring parts of the optimal tax system---a recurring theme in the literature---can lead to sizable welfare losses.

In order to evaluate the causal effect of student loans on labor market outcomes, we exploit the eligibility for the need-based Stafford loan program (subsidized Stafford loans) to implement a regression kink design. Using a nationally representative sample of students graduating with a bachelor's degree in 1993, we establish that student loan debt leads to lower earnings after graduation. Estimates show that an additional hundred dollars of Stafford loan reduces 1994 annual income by about 0.1%. Extrapolating this result, earnings for an individual with the mean level of borrowing are 5% lower than those of an individual with no debt. The impact of an additional hundred dollars of student debt decreases income by 0.4% in 1997, and the impact of debt on earnings vanishes by 2003. Economic theory shows that there exists a simple mechanism consistent with the empirical finding, whereby more debt leads individuals to quickly find employment rather than wait for an ideal job. Crucial for this result is that student debt is not dischargeable in bankruptcy. Indeed, when debt is dischargeable, higher debt can cause individuals to search for higher paying jobs that are harder to find.

Absent government or otherwise imposed restrictions, competitive loan markets are incompatible with identifiable subsets of the population subsidizing others. We use that insight to identify and measure inefficiencies in government student loan programs using linked survey and administrative data from the Beginning Postsecondary Student (BPS) longitudinal surveys. We use loan repayment histories to compute realized returns for borrowers. We then estimate considerable heterogeneity in expected returns based on ex ante observable characteristics, which suggests inefficiencies that private lenders might exploit by cream-skimming more profitable borrowers. We explore the potential for cream-skimming under different assumptions about the costs of capital for private lenders and study its implications for the pool of government student loan borrowers and expected returns to the government. Finally, we analyze the extent to which student loan limits can be modified based on borrower characteristics to equate expected returns and alleviate concerns about cream-skimming.