The subprime mortgage was developed to accommodate borrowers who would otherwise not have access to more conventional mortgages. Almost by definition, the creation of a subprime mortgage implies a deterioration of underwriting standards. In addition, the phenomenal growth of the subprime mortgage market naturally implies a significant decline of underwriting standards in the overall mortgage market. It is widely believed that a secular deterioration of underwriting standards starting around the latter half of 2004 led to the collapse of the subprime mortgage market. In light of this, some questions arise: How did poor underwriting bring about the collapse of the subprime market? More importantly, how would subprime mortgages perform if underwriting standards did not deteriorate?
The subprime market is largely defined as one meant for borrowers of modest credit quality. Naturally, it is widely believed that in order for this market to grow, it had to lower its standards and serve borrowers of even poorer credit quality. However, a recent study of more than nine million mortgages securitized and sold as subprime has found just the opposite. It reveals that minimum credit quality—as measured by their credit (FICO) scores—on subprime originations actually improved during 2000-2006. In particular, the percentage of loans with origination FICO not greater than 500 dropped from 2.45 percent of total originations during 2000-2002 to 0.31 percent during 2004-2006. But how did the bottom segment of the mortgage market record such high growth and still show an improvement in credit quality? The answer may lie in the fact that more than 70 percent of originations for every year during 2000-2006 were refinances. Over half of these mortgages were cash-out refinances; that is, households refinanced an existing mortgage into a subprime mortgage and in the process cashed out on their home equity.
An important feature of the subprime market during 2000-2006 was the significant growth in the proportion of originations with lower documentation and higher loan-to-value ratios (LTV). There is a clear trend of a decline in underwriting standards along these dimensions. However, a multidimensional view of underwriting reveals a less-known trend: Over the years, lenders increasingly relied on FICO scores to offset other riskier attributes of borrowers. As a result, average FICO scores were significantly higher for originations whose other attributes (such as lower documentation or higher loan-to-value ratios) were riskier. In particular, the percentage of loans with origination FICO less than 500 and LTV greater than 80 percent dropped from 0.8 percent of total originations during 2000-2002 to 0.06 percent during 2004-2006. Moreover, the percentage of low-documentation loans with origination FICO less than 500 dropped from 0.34 percent of total originations during 2000-2002 to 0.07 percent during 2004-2006. These figures seem to suggest that the minimum criteria for obtaining a subprime loan actually tightened over this period.
The patterns of underwriting suggest that lenders placed emphasis on FICO scores not just as an adequate indicator of credit risk, but also as a means to adjust for other riskier attributes on the origination. With the benefit of hindsight, some industry experts have faulted originators on this account arguing that FICO scores failed as predictors of default. In contrast, the recent study finds that the performance of FICO scores as indicators of default did not deteriorate over this period. In particular, they show that on average, the decrease in the probability of default in moving from a lower FICO score to a higher FICO score does not deteriorate over this period. Moreover, after controlling for other attributes on the loan origination, the increase in survival probability for a given improvement in FICO increases over the years. In sum, their results suggest that the overall trend of emphasis on FICO scores at the time of origination was not misplaced.
How would ex post facto default rates change if a “representative” origination of 2005 vintage had been originated in 2001? The study’s conclusion is that representative subprime originations for later vintages (namely, 2005, 2006 and 2007) would perform significantly better in 2001 and 2002 than representative originations of the same vintages (namely, 2001 and 2002). In light of this evidence, it is difficult to conclude that underwriting was central to the collapse of the subprime mortgage market. This non-result is a significant departure from conventional wisdom on the subprime crisis. Still, it is not difficult to see why a discerning reader may not find this result implausible. The argument that a significant deterioration in underwriting after 2004 triggered the collapse of the subprime market implicitly suggests that originations of earlier vintages had relatively robust underwriting. Taken to its logical conclusion, it could also suggest that the underwriting framework for earlier vintages could help provide a sustainable framework for future subprime originations. In contrast, their results do not rule out the possibility that the design of subprime contracts could have been fundamentally flawed since the inception of this market.