Did the CRA cause the mortgage market meltdown?
Two Federal Reserve economists examine whether available data support critics' claims that the Community Reinvestment Act spawned the subprime mortgage crisis.
Published March 1, 2009 | March 2009 issue
As the current financial crisis has unfolded, an argument that the Community Reinvestment Act (CRA) is at its root has gained a foothold. This argument draws on the fact that the CRA encourages commercial banks and savings institutions (collectively known as banking institutions) to help meet the credit needs of lower-income borrowers and borrowers in lower-income neighborhoods.1/ Critics of the CRA contend that the law pushed banking institutions to undertake high-risk mortgage lending.
This article discusses key features of the CRA and presents results from our analysis of several data sources regarding the volume and performance of CRA-related mortgage lending. On balance, the evidence runs counter to the contention that the CRA lies at the root of the current mortgage crisis.
Assessing banks in context
The CRA directs federal banking regulatory agencies, including the Federal Reserve, to use their supervisory authority to encourage banking institutions to help meet the credit needs of all segments of their local communities. These communities, referred to hereafter as CRA assessment areas, are defined as the areas where banking institutions have a physical branch office presence and take deposits, including low- and moderate-income areas. The banking agencies periodically assess the performance of banking institutions in serving their local communities, including their patterns of lending to lower-income households and neighborhoods, and take the assessments into consideration when reviewing the institutions' applications for mergers, acquisitions, and branches.
The CRA emphasizes that banking institutions fulfill their CRA obligations within the framework of safe and sound operation. CRA performance evaluations have become more quantitative since 1995, when regulatory changes were enacted that stress actual performance rather than documented efforts to serve a community's credit needs. However, the CRA does not stipulate minimum targets or even goals for the volume of loans, services, or investments banking institutions must provide. While it is fair to say that the primary focus of CRA evaluations is the number and dollar amount of loans to lower-income borrowers or areas, the agencies instruct examiners to judge banks' performance in light of 1) each institution's capacity to extend credit to lower-income groups and 2) the local economic and market conditions that might affect the income and geographic distribution of lending.
Timing and originations
Before we turn to our analysis of CRA lending data, we have two important points to note regarding the CRA and its possible connection to the current mortgage crisis.
The first point is a matter of timing. The current crisis is rooted in the poor performance of mortgage loans made between 2005 and 2007. If the CRA did indeed spur the recent expansion of the subprime mortgage market and subsequent turmoil, it would be reasonable to assume that some change in the enforcement regime in 2004 or 2005 triggered a relaxation of underwriting standards by CRA-covered lenders for loans originated in the past few years. However, the CRA rules and enforcement process have not changed substantively since 1995.2/ This fact weakens the potential link between the CRA and the current mortgage crisis.
Our second point is a matter of the originating entity. When considering the potential role of the CRA in the current mortgage crisis, it is important to account for the originating party. In particular, independent nonbank lenders, such as mortgage and finance companies and credit unions, originate a substantial share of subprime mortgages, but they are not subject to CRA regulation and, hence, are not directly influenced by CRA obligations. (We explore subprime mortgage originations in further detail below.)
The CRA may directly affect nonbank subsidiaries or affiliates of banking institutions. Banking institutions can elect to have their subsidiary or affiliate lending activity counted in CRA performance evaluations. If the banking institution elects to include affiliate activity, it cannot be done selectively. For example, the institution cannot "cherry pick" loans that would be favorably considered under the law while ignoring loans to middle- or higher-income borrowers.
In the next section, we discuss the data analysis we undertook to assess the merits of the claims that the CRA was a principal cause of the current mortgage market difficulties. The analysis focuses on two basic questions. First, what share of subprime mortgage originations is related to the CRA? Second, how have CRA-related subprime loans performed relative to other loans? We believe the answers to these two questions will shed light on the role of the CRA in the subprime crisis.
CRA-related lending volume and distribution
In analyzing the available data, we consider two distinct metrics of lending activity: loan origination activity and loan performance. With respect to the first question posed above concerning loan originations, we determine which types of lending institutions made higher-priced loans, to whom those loans were made, and in what types of neighborhoods the loans were extended.3/ This analysis therefore depicts the fraction of subprime mortgage lending that could be related to the CRA.
Using loan origination data obtained pursuant to the Home Mortgage Disclosure Act (HMDA), we find that in 2005 and 2006, independent nonbank institutions—institutions not covered by the CRA—accounted for about half of all subprime originations. (See Table 1.) Also, about 60 percent of higher-priced loan originations went to middle- or higher-income borrowers or neighborhoods, populations not targeted by the CRA. (See Table 2.) In addition, independent nonbank institutions originated nearly half of the higher-priced loans extended to lower-income borrowers or borrowers in lower-income areas (share derived from Table 2).
In total, of all the higher-priced loans, only 6 percent were extended by CRA-regulated lenders (and their affiliates) to either lower-income borrowers or neighborhoods in the lenders' CRA assessment areas, which are the local geographies that are the primary focus for CRA evaluation purposes. The small share of subprime lending in 2005 and 2006 that can be linked to the CRA suggests it is very unlikely the CRA could have played a substantial role in the subprime crisis.
To the extent that banking institutions chose not to include their affiliates' lending in their CRA examinations, the 6 percent share overstates the volume of higher-priced, lower-income lending that CRA examiners would have counted.4/ It is possible, however, the examiners might have considered at least some of the lower-income lending outside of CRA assessment areas if institutions asked that it be considered in their CRA performance evaluations. No data are available to assess this possibility; however, the majority of the higher-priced loans made outside of assessment areas were to middle- or higher-income borrowers. In our view, this suggests it is unlikely that the CRA was a motivating factor for such higher-priced lending. Rather, it is likely that higher-priced lending was primarily motivated by its apparent profitability.
It is also possible that the remaining share of higher-priced, lower-income lending may be indirectly attributable to the CRA due to the incentives under the CRA investment test. Specifically, examiners may have given banks "CRA credit" for their purchases of lower-income loans or mortgage-backed securities containing loans to lower-income populations, which could subsequently affect the supply of mortgage credit.
Although we lack definitive information on banks' CRA-induced secondary market activity, the HMDA data provide information on the types of institutions to which mortgages are sold. The data suggest that the link between independent mortgage companies and banks through direct secondary market transactions is weak, especially for lower-income loans. (See Table 3.) In 2006, only about 9 percent of independent mortgage company loan sales were to banking institutions. (Figure not shown in table.) And among these transactions, only 15 percent involved higher-priced loans to lower-income borrowers or neighborhoods. In other words, less than 2 percent of the mortgage originations sold by independent mortgage companies in 2006 were higher-priced, CRA-credit-eligible, and purchased by CRA-covered banking institutions.
Analyzing loan performance
To assess the relative performance of CRA-related, higher-priced loans, we use data from First American LoanPerformance (LP) on subprime and alt-A mortgage securitizations to compare delinquency rates for subprime and alt-A loans in lower-income neighborhoods relative to those in middle- and higher-income neighborhoods. The LP data do not provide information on borrower income or the type of originating institution, but do indicate the ZIP Code of the property, which we use to group loans into neighborhood income categories.5/ The results indicate that the 90-days-or-more delinquency rate as of August 2008 for subprime and alt-A loans originated between January 2006 and April 2008 is high regardless of neighborhood income, with delinquency rates comparable across neighborhood income categories. (See Table 4.)6/
In order to gauge more precisely the possible effects of the CRA, we use the LP data again and focus attention on the subset of ZIP Codes that are similar, in principle, except for their relationship to the CRA. Specifically, we focus only on ZIP Codes right above and right below the CRA eligibility threshold. (A neighborhood meets the CRA threshold if it has a median family income equivalent to 80 percent or less of the median family income of the broader area.) As such, the only major difference between these two sets of neighborhoods should be that the CRA focuses on one group and not the other. This analysis indicates that subprime loans in ZIP Codes that are the focus of the CRA (those just below the threshold) have performed virtually the same as loans in the areas right above the threshold.7/ (See Table 5.)
To gain further insight into the risks of lending to lower-income borrowers or areas, we also compared the performance of first mortgages originated and held in portfolio under the nationwide affordable lending programs operated by the NeighborWorks® America (NWA) partners to the performance of loans of various types as reported by the Mortgage Bankers Association of America. Many loans originated through NWA programs are done in conjunction with banking institutions subject to the CRA, so the performance of these loans provides another basis to address the relationship between the CRA and the subprime crisis. Along any measure of the severity of loan delinquency or the incidence of foreclosure, the loans originated under the NWA program have performed better than subprime loans.8/ (See Table 6.) Although the performance of loans in the NWA portfolio provides one benchmark to compare the performance of CRA-related loans with other loans, it is only one portfolio of such loans; further research of this type could provide a stronger base from which to draw conclusions.
Another way to measure the relationship between the CRA and the subprime crisis is to examine foreclosure activity across neighborhoods that are classified by income. Data made available by RealtyTrac on foreclosure filings from January 2006 through August 2008 indicate that most foreclosure filings (e.g., about 70 percent in 2006) have taken place in middle- or higher-income neighborhoods. More important, foreclosure filings have increased at a faster pace in middle- or higher-income areas than in lower-income areas that are the focus of the CRA.9/ (See Table 7.)
Two basic points emerge from our analysis of the available data. First, only a small portion of subprime mortgage originations is related to the CRA. Second, CRA-related loans appear to perform comparably to other types of subprime loans. Taken together, the available evidence seems to run counter to the contention that the CRA contributed in any substantive way to the current mortgage crisis.
Neil Bhutta and Glenn B. Canner are economists in the Division of Research and Statistics at the Board of Governors of the Federal Reserve System. The views expressed are those of the authors and do not necessarily reflect those of the Board of Governors or members of its staff.
1/ Lower-income households are determined by comparing the income of the household to the median family income of the metropolitan statistical area (MSA) or statewide non-MSA in which the property being purchased or refinanced is located. "Lower" is less than 80 percent of the median, "middle" is 80 to 119 percent, and "higher" is 120 percent or more. Lower-income neighborhoods are determined by comparing the median family income of the census tract where the property being purchased or refinanced is located to the MSA or statewide non-MSA median family income. Income categories for census tract classification have the same numerical thresholds as those applied for households.
2/ The change in the CRA rules in 2005 focused primarily on reducing burden for smaller lenders and expanding the focus of the CRA to include some middle-income census tracts in distressed rural areas. No changes were made that encouraged lenders to relax their underwriting standards.
3/ A higher-priced loan is defined as a loan where the spread between the annual percentage rate on the loan and the rate on Treasury securities of comparable maturity is above designated thresholds. For first-lien loans, the focus of attention in this article, the designated threshold is 3 percentage points. For junior-lien loans, the threshold is 5 percentage points. The definition was adopted as part of Regulation C (the regulation that implements the Home Mortgage Disclosure Act) and was intended to identify loans that fell in the subprime portion of the mortgage market.
4/ About one-fifth of the higher-priced loans extended in the banking institutions' local communities were extended by their affiliates.
5/ We classify ZIP Code-based delinquency data by relative income in two different ways. First, we use information published by the U.S. Census Bureau on income at the ZIP Code Tabulation Area (ZCTA) level of geography. Because the ZCTA data provide an income estimate for each ZIP Code, delinquency rates can be calculated directly from the LP data based on the ZIP Code location of the properties securing the loans (see www.census.gov/geo/ZCTA/zcta.html). Second, we calculate delinquency rates for each relative income group (lower, middle, and higher) as the weighted sum of delinquencies divided by the weighted sum of mortgages, where the weights equal each ZIP Code's share of population in census tracts of the particular relative income group. Relative income is based on the 2000 census and is calculated as the median family income of the census tract divided by the median family income of its MSA or a nonmetropolitan portion of the state. The two approaches yield virtually identical results.
6/ A virtually identical relationship across neighborhood income groups is found if the pool of loans evaluated is expanded to cover those originated between January 2004 and April 2008. The only material difference is that the levels of delinquency are lower for both subprime and alt-A loans for the larger sample of loans. Such a relationship is expected, since loans that are relatively long-lived tend to perform well over time.
7/ See footnote 6.
8/ No information was available on the geographic distribution of the NWA loans. The geographic pattern of lending can matter, as certain areas of the country are experiencing much more difficult housing conditions than other areas. Also, no information was available on the age of the loans, which can have an important effect on performance.
9/ These data are reported at the ZIP Code level. We calculate the statistics by relative income group in Table 7 as before; see footnote 6. Foreclosure filings have been consolidated at the property level, so separate filings on first- and subordinate-lien loans on the same property are counted as a single filing.