In this paper, we use two comprehensive micro-data sets to study how the distribution of mortgage debt evolved during the 2000s housing boom. We show that the allocation of mortgage debt across the income distribution remained stable, as did the allocation of real estate assets. Any theory of the boom must replicate these facts, and a general equilibrium model shows that doing so requires two elements: (1) an exogenous shock that increases expected house price growth or, alternatively, reduces interest rates and (2) financial markets that endogenously relax borrowing constraints in response to the shock. Empirically, the endogenous relaxation of constraints was largely accomplished with subprime lending, which allowed the mortgage debt of low-income households to increase at the same rate as that of high-income households.

Loan modifications can ease borrowers’ financial burdens and mitigate loan losses. However, the threat of future strategic renegotiation may cause lenders to tighten ex-ante credit provision. We evaluate this trade-off in a dynamic model of loan underwriting with frictional renegotiation and calibrate it using loan-level CRE data from banks and CMBS. We find that modification frictions can rationalize a number of empirical facts regarding how CRE loan underwriting and performance differ across lenders. Key to this result, high frictions to modifying CMBS loans reduce renegotiation, increase debt capacity, and cause high-leverage-demand borrowers to select into the CMBS market. Consequently, easing CMBS modification frictions reduces welfare by restricting the menu of LTVs available in the market.