How Subprime Loans Emerged in Minority Neighborhoods
It is commonly believed that subprime lenders, who lend to subprime borrowers, provide credit to high-risk communities that could not obtain credit from low-risk lenders. Eglė Jakučionytì and Swapnil Singh challenge this view. They show that policy changes introduced in 1995 by US institutions Fannie Mae and Freddie Mac increased securitization costs for loans in minority neighborhoods. Prime lenders moved out, and with less competition, subprime lenders were able to penetrate these minority neighborhoods more easily.
Over the past two decades, subprime lenders – lenders that provide loans to subprime borrowers – have been in the spotlight for several reasons. These reasons range from financial stability concerns to predatory lending behavior towards minority groups. Despite these concerns, while subprime lending has grown in the United States, the prevailing view since the early 1990s has been that subprime lenders provide credit to high-risk communities that could not obtain credit from low-risk lenders (Collins et al ., 2004). In our recent article (Jakucionyte and Singh, 2021), we challenge this view by exploring the origins of subprime lending in minority neighborhoods in the United States in the 1990s. We show that the emergence of subprime lenders in minority neighborhoods was due to advances in lending technology and specific policies endorsed by the US Government Sponsored Enterprises (GSEs), Fannie Mae and Freddie Mac.
Figure 1 illustrates our point. The figure represents at the neighborhood level the share of subprime loans compared to the share of the black population for two periods: 1993-1995 and 1996-2000. During the period 1993-1995, there is no association between the share of subprime loans and the share of black population in the neighborhood. However, the relationship becomes positive during the period 1996-2000.
Figure 1 – Relationship between share of subprime loans and share of black population
Source: Data from the Home Mortgage Disclosure Act 1993–2000, Decennial Census 1990, HUD Subprime Lender List. Description: The figure provides the nonparametric relationship between the share of subprime loans and the share of black population in the census tract for two different time periods: 1993–1995 (solid red line) and 1996–2000 (black dashed line). The share of black population in the census tract is Winsorized at 1 and 99 percentiles. Local polynomial regression with bandwidth equal to 0.1 is used for nonparametric estimation.
The discrepancy between the two periods – 1993-1995 and 1996-2000 – is related to a specific change in 1995. Prior to 1995, mortgage underwriting in the United States was done manually, which was slow, inaccurate and perceived as suffering from the personal prejudices of subscribers. . In 1995, Fannie Mae and Freddie Mac introduced two modifications to solve this problem. First, they took steps to automate mortgage origination. Second, they sent letters to associate lenders encouraging them to use FICO credit scores as an objective and accurate measure of borrower creditworthiness. These institutions have also provided specific thresholds for credit scores and guidance to lenders on how to act when observed credit scores are above or below these thresholds (Foote et al., 2019). For example, if the credit rating is below the specified threshold, the application required further review by the underwriter before selling it to GSEs. Implicitly, GSEs signaled that credit ratings would affect the chances of securitization succeeding.
It is important to note that the approval of credit scores by GSEs and the provision of specific thresholds have only affected certain credit providers – the major lenders. In the 1990s, subprime loans were mainly securitized in the private market (Temkin et al., 2002). This configuration of the securitization market is important in explaining the differences in lending after 1995. As minority neighborhoods have on average lower creditworthiness, the GSE policy has increased securitization costs for loans in minority neighborhoods. This would encourage major lenders to move to non-minority neighborhoods, that is, neighborhoods with a larger pool of low-risk borrowers. Subprime lenders have not been directly affected by the policy, but they could enter minority quarters in response to less competition from primary lenders.
Figure 2 illustrates how the presence of subprime lenders changed in neighborhoods after 1995. Controlling for differences between local mortgage markets across the United States, we can see that the share of subprime lending in black neighborhoods increased by disproportionately after 1995. In a specific regression framework, called the difference-in-differences, we find that relative to the pre-1995 period, the difference between the share of subprime loans in minority and non-minority neighborhoods increased by 5 percentage points.
Figure 2 – Share of Subprime Loans in Census Tracts with Low and High Black Populations
Source: Data from the Home Mortgage Disclosure Act 1993–2000, Decennial Census 1990, HUD Subprime Lender List. Description: The figure shows the evolution of the share of subprime loans for the treatment (red dotted line) and control (solid black line) groups. Treatment group refers to census tracts with the share of black population in the top quartile. The residual share of subprime loans is calculated by separating the county fixed effects.
The share of subprime loans increases after 1995 because primary lenders reduce loans in minority neighborhoods compared to non-minority neighborhoods. Figure 3 succinctly illustrates this point. Subprime lenders increased their lending in minority neighborhoods relative to non-minority neighborhoods, however, the reduction in prime lending in minority neighborhoods preceded the increase in subprime lending and was larger, suggesting that prime lenders came first.
Figure 3 – Effect of credit score endorsement on total loans
Source: Data from the Home Mortgage Disclosure Act 1993–2000, Decennial Census 1990, HUD Subprime Lender List. Description: The figure shows the estimation results of the difference-in-differences model. The dependent variables are the logarithm of the total loan amount issued by prime lenders (left panel) and subprime lenders (right panel). All estimates are for the base year 1995, which is omitted. The regression includes year, census tract, and county-year fixed effects. The sample is constructed using only accepted applications and restricted to census tracts matched by propensity score matching. The regression is weighted by the total number of loans granted at the census tract year level. In the final sample, census tracts with less than five loans per year are excluded. Black whiskers indicate 95% confidence intervals. Standard errors are clustered at the state-year level.
Finally, to further tie our conclusion to credit rating endorsement, we examine the lending behavior of primary lenders with a weak or strong relationship with Fannie Mae and Freddie Mac and illustrate this with Figure 4. As primary lenders who securitize more loans with GSEs would be more affected by the new policy, they would leave minority neighborhoods to a greater extent than prime lenders who tend to securitize a small portion of their mortgages. Figure 4 confirms this prediction. It shows that before 1995, top lenders with a stronger relationship with GSEs had a similar share of loans in both groups of neighborhoods, but after 1995 the share of loans by top lenders with a stronger relationship with GSE decreased in minority neighborhoods compared to non-minority neighborhoods. Major lenders with a weaker relationship barely changed their share of loans made in neighborhoods.
Figure 4 – Effect of credit score endorsement on share of subprime loans, by relationship to GSEs
Source: Data from the Home Mortgage Disclosure Act 1993–2000, Decennial Census 1990, HUD Subprime Lender List. Description: The figure shows the estimation results of the difference-in-differences model for three dependent variables: the share of subprime loans (magenta, diamond) and the share of primary lenders with low securitization (red, square) and high securitization. primary lender (black, circle). Prime lenders are categorized into low or high securitization groups based on the average share of conforming real estate loans securitized over the period 1993–1995. All estimates are for the base year 1995, which is omitted. The regression includes year, census tract, and county-year fixed effects. The sample is constructed using only accepted applications and restricted to census tracts matched using propensity score matching. The regression is weighted by the total number of loans granted at the census tract year level. In the final sample, census tracts with less than five loans per year are excluded. Black whiskers indicate 95% confidence intervals. Standard errors are clustered at the state-year level.
Essentially, we show that the approval of credit ratings along with specific guidelines had an unintended consequence. Approval of credit ratings in mortgage underwriting has contributed to the emergence of subprime lenders in minority neighborhoods and the flight of prime lenders. These results suggest that borrowing conditions may also have changed. For example, as minorities became more exposed to subprime lenders, they might have become more likely to obtain high-cost loans, even if they had similar credit scores to non-minority borrowers. Further analysis of the implications of lender screening would provide valuable insights into the financial stability of minority borrowers and neighborhood inequalities.
Please read our comment policy before commenting.
To note: The post office gives the point of view of its authors, not the position USAPP– American Politics and Policy, nor from the London School of Economics.
Shortened URL for this post: https://bit.ly/3FK46Ed
About the authors
Eglė Jakučionytė – CEFER
Eglė Jakučionytė is a senior economist at CEFER, the research center of the Bank of Lithuania, and a junior researcher at Vilnius University. His research interests include macrofinance, real estate economics, and international macroeconomics, with a focus on household and corporate debt and borrower protection. She holds a doctorate in economics from the University of Amsterdam.
Swapnil Singh – CEFER
Swapnil Singh is a senior research economist at CEFER, the research center of the Bank of Lithuania, and a senior researcher at Kaunas University of Technology. His research interests focus on quantitative macroeconomics, household finance and development economics. He holds a doctorate in economics from the University of Amsterdam.