Changes in Buyer Composition and the Expansion of Credit During the Boom
Duke University – Fuqua School of Business
Massachusetts Institute of Technology (MIT) – Sloan School of Management; National Bureau of Economic Research (NBER)
Dartmouth College – Tuck School of Business
Earlier research has suggested that distortions in the supply of mortgage credit during the run up to the 2008 financial crisis, in particular a decoupling of credit flow from income growth, may have been responsible for the rise in house prices and the subsequent collapse of the housing market. Focusing on individual mortgage transactions rather than whole zip codes, we show that the apparent decoupling of credit from income shown in previous research was driven by changes in buyer composition. In fact, the relationship between individual mortgage size and income growth during the housing boom was very similar to previous periods, independent of how we measure income. Zip codes that had large house price increases experienced significant changes in the composition of buyers, i.e. home buyers (mortgage applicants) had increasingly higher income than the average residents in an area. Poorer areas saw an expansion of credit mostly through the extensive margin, i.e. a larger numbers of mortgages originated, but at DTI levels in line with borrower income. When we break out the volume of mortgage origination from 2002 to 2006 by income deciles across the US population, we see that the distribution of mortgage debt is concentrated in middle and high income borrowers, not the poor. Middle and high income borrowers also contributed most significantly to the increase in defaults after 2007. These results are consistent with an interpretation where house price expectations led lenders and buyers to buy into an unfolding bubble based on inflated asset values, rather than a change in the lending technology.
Changes in Buyer Composition and the Expansion of Credit During the Boom – Introduction
Understanding the origins of the housing crisis of 2007 to 2009 has been an enduring challenge for financial economists and policy makers alike. One of the predominant narratives that has emerged in the literature is that fundamental changes in the origination technology of lenders significantly contributed to unsustainable levels of borrowing and ultimately caused an acceleration of house prices. This interpretation builds on a key finding by Mian and Sufi (2009) that growth in mortgage credit at the zip code level became negatively correlated with income growth in the run-up to the financial crisis, suggesting that lending was decoupled from income, especially in areas with strong house price growth. As a result, there has been a significant emphasis on understanding the role of the financial industry in providing credit to low-income borrowers, which is often referred to as the credit supply side view of the housing crisis.2
In this paper we use mortgage and income data on individual borrowers (rather than the zip code level analysis that has been previously used) to shed new light on the dynamics of loan origination in the run-up to the financial crisis. Our results show that using zip codes as the unit of observation to explain the relationship between growth in lending and income confounds important compositional changes within zip codes. By focusing on individual borrowers, we can decompose the growth in total mortgage debt due to the intensive margin (change in the average size of individual loans) from the extensive margin (the number of new loans that are originated in a zip code). Additionally, we distinguish the income of borrowers from the average household income of the residents in an area.
We provide three main new findings about the relationship between credit origination and income in the run-up to the financial crisis. First, when we relate individual mortgage size to income, measured either using borrower income from mortgage applications or average household income from the IRS, we see that the growth in individual mortgage size is strongly positively related to income growth throughout the pre-crisis period. This means that there was never a decoupling of mortgage growth and income growth at the individual level, the relevant measure for lending decisions. Second, we show that there was an expansion of credit along the extensive margin: poorer neighborhoods have an increase in the number of loans being originated, with modest changes in individual DTI that are similar to those in high income zip codes. This happens because new home buyers had increasingly higher income levels than the average household living in these areas.
At the same time, neighborhoods that experienced strong house price growth see a rise in average mortgage size, but again at DTI levels close to previous periods, since the average income of these buyers also went up significantly. Third, we document how aggregate mortgage origination in the U.S. was distributed by borrower income levels. The large majority of mortgage dollars originated between 2002 and 2006 are obtained by middle income and high income borrowers (not the poor). While there was a rapid expansion in overall mortgage origination during this time period, the fraction of new mortgage dollars going to each income group was stable. In other words, the poor did not represent a higher fraction of the mortgage loans originated over the period. In addition, borrowers in the middle and top of the distribution are the ones that contributed most significantly to the increase in mortgages in default after 2007. Taken together, the evidence in the paper suggests that there was no decoupling of mortgage growth from income growth where unsustainable credit was flowing disproportionally to poor people.
Using data on individual mortgage applications from the Home Mortgage Disclosure Act (HMDA) between 2002 and 2006 we distinguish the growth in average household income at the zip code level (from the IRS) and the income of individual borrowers (from HMDA). Following the earlier literature, we first look at the relationship between the growth in total mortgage credit at the zip code level (the sum of the mortgage balance of all mortgages originated in a given year) and the two measures of income. We find a strong positive relationship between total credit growth and the individual borrower income growth, but a negative relationship with the growth in average household income at the zip code level (the result highlighted in Mian and Sufi, 2009). This suggests that there was no decoupling of total credit growth and borrower fundamentals during this time period. In addition, when we look at a longer time period between 1996 and 2007, we confirm that there was neither a reversal of the sign nor a change in the slope between credit flows and income growth using individual borrower income.
See full PDF below.