In Chapter 1, I explore the digital divide in homeownership outcomes as better financial opportunities increasingly move beyond the technology frontier. Low-income households derive significantly less savings from mortgage refinancing than their wealthy counterparts. I document that the rise of refinancing inequality in the United States can be partially explained by the gap in access to modern information and communications technology. Using granular spatial variation of a large-scale broadband subsidy program, I show that high-speed internet facilitates refinancing activity and reduces monthly mortgage payments. These effects are large and persistent, corresponding to a 5 percent increase in disposable income and up to $18,000 in total savings for low-income households. The growth of refinancing is pronounced in underserved areas with low access to bank branches and among populations that are likely to have low financial and digital literacy.

In Chapter 2 (with Andrea L. Eisfeldt and Dimitris Papanikolaou), we show that the recent underperformance of value investing strategies can be attributed to the mismeasurement of intangible assets. We propose a simple improvement to the classic Fama and French value factor that incorporates intangibles and addresses differences in accounting practices across industries. Our intangible value factor prices assets as well as or better than the traditional value factor but yields substantially higher returns. This outperformance holds over the entire sample period, including in more recent decades during which value has underperformed. We also find evidence that the adjustment better identifies firms with superior fundamentals as measured by productivity, profitability, and financial soundness.

In Chapter 3 (with Marcelo Rezende), we study whether regulatory costs induce banks to search for yield by holding riskier assets. We test this hypothesis by analyzing a cost specifically imposed on the size and composition of a bank's balance sheet -- deposit insurance premiums charged by the Federal Deposit Insurance Corporation (FDIC). Using supervisory data and a sharp cutoff in the schedule of deposit insurance premiums, we show that higher balance sheet costs indeed cause banks to substitute excess reserves (a liquid asset with no credit risk) for short-term interbank loans (a less liquid asset with credit risk). We argue that optimal deposit insurance pricing should account for this potential feedback effect.

In the first chapter of this dissertation, I study the asset pricing implications of heterogeneity in the financial intermediary sector. To examine the impact of massive balance sheet adjustments within the intermediary sector during the Great Recession, I propose a dynamic model with heterogeneous intermediaries and financial frictions. The model implies that a significant fraction of risk premia variation can be attributed to the composition of the intermediary sector. Asset reallocations, comparable in magnitude to the one observed during the recent financial crisis, lead to a substantial increase in the risk premia and volatility. An empirical measure of the composition of the financial sector strongly forecasts future excess returns with significant additional predictive power beyond many established forecasting variables in the literature. Moreover, shocks to wealth distribution among intermediaries have strong explanatory power for the cross-section of assets and are priced with a positive price of risk.

In the second chapter of this dissertation (with William Mann), we estimate small marginal costs and large markups at private colleges in the United States. For identification, we exploit a tightening of credit standards in the PLUS loan program, which decreased enrollment, revenues, and expenditures at private colleges with low-income students. We estimate that markups represented more than half of charges for students disqualified by the change.

Markups were higher at for-profit schools, and in states with fewer public schools and lower education spending. Our results contrast prior estimates of small markups in higher education. We conclude that the positive relationship between federal student aid and rising tuition costs, the so-called Bennett Hypothesis, arises due to imperfect competition between colleges.

In Chapter 1 of this dissertation, I study the informational content of GSE Credit Risk Transfer (CRT) bonds. CRT bonds amount to catastrophe bonds on underlying mortgage collateral, and are informative about the market price of credit risk for conforming mortgage loans. I analyze the information content of this new asset class, which holds effectively half of the default risk of the \$12 trillion US mortgage market. I build a pricing model and extract default probabilities from a comprehensive hand-collected dataset on CRT bond issuances using TRACE pricing data. I find that while the guaranty-fee (g-fee) implied by early CRT issuances was high and near the level charged by the GSEs (30 bp), the market has stabilized in recent years and implied g-fees have fallen significantly to around 10-20 basis points. I discuss the implications of these findings for several facets of the mortgage market.

In Chapter 2, I build and estimate a top-down portfolio credit model matched to a sample of GSE Credit Risk Transfer bond (CRT) tranche market prices. Two state variables, modeled as exogenous Poisson intensities, represent two sources of default risk affecting mortgages, which intuitively map into a level of routine mortgage defaults and the risk of catastrophic losses. The market views the (risk-neutral) probability of a catastrophic event, on the order of the 2008 housing market crisis, as occurring roughly once every 25 years. The implied guaranty-fee (g-fee) is on the order of 15 basis points over the sample, and both risk factors contribute roughly equally to this spread. I also analyze the level of credit-protection offered by the current CRT design in a simulation study, and discuss the conditions under which ``credit risk transfer transfers credit risk."