Profitwise News and Views

September 2003 Issue

Seeds of Growth Conference Summary

Seeds of Growth Sustainable Community Development: What Works, What Doesn't, and Why

Session Five - Public Policy - CRA
It has been more than 25 years since Congress passed the Community Reinvestment Act (CRA). Since enactment of the CRA, policymakers and community development practitioners have sought to measure the impact of the CRA. The papers presented in this session shed light on the effects of the CRA on lending and credit access.

The paper by Federal Reserve Board economists Robert B. Avery, Paul S. Calem, and Glenn B. Canner, The Effects of the Community Reinvestment Act on Local Communities, addresses questions about the effects of the CRA on local communities. To estimate the marginal effects of the CRA, the analysis compares economic circumstances in census tracts with incomes just above and just below the low-income threshold designated by the act. The analysis measures changes in neighborhood outcomes between 1990 and 2000.

Since its inception, the CRA has focused on measuring the activity of lending institutions in the lower-income census tracts of their market areas. Lower-income census tracts are defined by regulators as those with median family income below 80 percent of the median income of a region, such as the Metropolitan Statistical Area (MSA) or the nonmetropolitan portion of a state. Importantly, the classification of the census tract by its relative income only occurs once per decade, following the release of census data. At that point, all parties have a clear understanding of which census tracts are going to be the focus of attention when CRA assessments are made. Canner et al. designed their study to rely on this long-standing focus of the CRA.

Canner discussed four additional points regarding the CRA that were directly relevant to the paper's line of research. First, the CRA regulations were overhauled in 1995 to make the process more quantitative and performance-oriented. There are now three specific performance tests for large banking institutions (lending, investment, and service). Second, the regulations specify four categories for performance ratings (outstanding, satisfactory, needs to improve, and substantial noncompliance) and the analysis makes use of these ratings. Third, since 1995 the regulations broadened the focus of CRA assessments to include lower-income borrowers, not just lower-income neighborhoods, and the research accounts for this broader focus. Finally, the research focuses on home lending activity as a primary measure of CRA activity.

Canner noted previous research on the effects of the CRA primarily focused on performance and profitability of CRA-related lending activities and changes in lending volume as a consequence of the law. The weight of the evidence suggests that CRA-related lending is generally profitable. There also is some limited evidence that the CRA has resulted in the extension of additional credit. Despite previous findings, Canner argued that there has been very little research conducted on the effects of the CRA on local communities.

When the researchers explored ways to test the potential effects of the CRA on local communities, they devised three. The first view is that banking institutions may satisfy CRA provisions through their ordinary course of business. In this case, there is neither a necessary expectation for any change in lending activities associated with the CRA per se nor any effects of the law on local communities. The second view is that banking institutions may respond with intensive marketing and outreach activities to try to acquire more borrowers in lower-income neighborhoods, but not change the pricing or underwriting standards. Here, higher market share may be achieved and the institution's performance may appear better for CRA evaluation purposes; however, these activities will not necessarily result in higher lending levels or have an effect on the community if the institution hasn't lowered prices or used more flexible underwriting. The third view is that banking institutions may respond by reducing prices and/or adopting more flexible underwriting. Lending increases in targeted neighborhoods and perhaps some positive changes in the neighborhood's circumstances, such as homeownership rates, median home values, and vacancy rates, would be expected.

When designing the tests to determine the effects of the CRA on local communities, Canner et al. faced significant challenges. The first was to assess the marginal effect of the law, which was difficult because there is no way to observe what would have happened in the absence of the law. The second challenge was to distinguish any effects of the CRA on neighborhood outcomes from other market and regulatory driven changes. In particular, technological advances over the last decade affect the ability of lower-income families to obtain credit. There has been a substantial growth in sub-prime lending opening up new opportunities for borrowers. Another potential effect on lower-income communities has been the establishment of affordable housing goals for both Fannie Mae and Freddie Mac. And finally, a greater focus on fair lending may result in regulatory overlap to the extent that predominantly minority neighborhoods tend to be lower-income neighborhoods.

To address these challenges, Canner et al. conducted two empirical tests. Both of the approaches make use of the CRA income threshold definition for lower-income census tracts (i.e., those tracts where the median family income falls below 80 percent of MSA median income). The first test, referred to as the "changes analysis," provided evidence on the potential impact of the CRA on changes in neighborhood outcomes during the 1990s. For this test, the paper compares changes in outcomes for two cohorts of census tracts. The first group, the CRA eligible cohort, is just below the CRA eligibility threshold. The second group, the not-CRA eligible cohort, is just above the threshold. The paper asserts that its test is quite close to a natural experiment, and co-author Avery employs the metaphor of a horse race to put the paper's empirical tests in perspective. The horse race starts with the initial conditions in each group of census tracts in 1990, and the winner is the cohort that comes out ahead, in terms of changes in neighborhood outcomes, by 2000. For this test, census tracts were placed in one of two groups, based on their 1990 relative income, and this categorization was fixed for the period of the study (i.e., 1990-2000). Lending in any of the census tracts below the income threshold counts toward meeting CRA obligations while lending in the tracts above does not. By design, the authors selected two groups of census tracts that as of 1990 should have been similar by most measures. However, because initial conditions in the two groups of census tracts may have differed slightly, regression techniques were used to adjust for initial differences. The final step in the changes analysis was to determine if any outcome differences observed were related to several different measures of CRA activity employed during the test.

The second test was referred to as the "levels analysis," intended to reveal the cumulative effect of the CRA on conditions in local communities. This test was undertaken out of concern that the CRA may have caused banking institutions to change their behavior in the period shortly after enactment and that the changes analysis, which focuses only on the decade of the 1990s, may not reveal such an effect. The levels analysis used regression techniques to relate the neighborhood outcomes in 2000 to indicators of CRA eligibility for individual tracts as of 1980 or 1990. Here, the sample is census tracts in the 70 percent to 90 percent relative income range as of 2000. This group of census tracts has considerable homogeneity with respect to prior years CRA eligibility.

The measures used for neighborhood outcomes for the changes analysis were: changes in the homeownership rate, number of owner-occupied units, vacancy rate, median home value, and crime index. Canner stated that if the CRA leads to an increase in lending, the first three measures are the most likely to be favorably affected by such lending. However, expectations for positive changes in median home value and crime rates are less clear, and any changes would be considered a secondary effect. The levels analysis employed the same measures, except for the number of owner-occupied units.

The paper concludes that the results of the two tests for the effects of the CRA are mixed and difficult to interpret. On the one hand, CRA eligible tracts had more favorable changes for the decade of the 1990s in terms of homeownership, owner occupied units, and vacancies, than would have been predicted using the not-CRA eligible cohort. Moreover, the market share of institutions rated outstanding was higher in the CRA eligible cohort than in others â€“ suggesting a CRA effect. The paper's findings regarding homeownership and growth in the number of owner-occupied units are not conclusive. Moreover, CRA-eligible tracts fared worse than would have been predicted with regard to home values and crime.

Canner pointed out that much more study on the effects of the CRA is planned, as it may be that some of the paper's results reflect the nature of the specific tests employed by the study. For example, it is quite possible that banks don't focus only on tracts below 80 percent of the median income, but rather target their activities to a broader area. And this broader area may have some tracts both below and above the 80 percent threshold. Moreover, banks may focus more heavily on lower-income borrowers to facilitate program design and implementation.

Canner stated that several aspects of the paper's research need expansion, as there are different channels through which the CRA may influence communities. Although the current research focused on home lending, further study should assess the effects of the CRA on small business, community development, and purchase mortgage lending.

In What Makes CRA Agreements Work? A Study of Lender Responses to CRA Agreements, Raphael W. Bostic and Breck L. Robinson examine whether lenders actually change their behavior after entering into CRA agreements. Bostic began by stating that the CRA has two enforcement mechanisms that provide incentives for lenders to pay attention to the act. The first is the public release of institutions' CRA ratings, and the second is the regulatory application (a formal procedure allowing banks to undertake mergers, expansion, new branch establishment and a variety of other actions) process wherein the regulator must consider a lender's CRA rating. If lenders don't consider the CRA, it can be particularly costly - mostly in terms of lost time regarding applications for mergers. Bankers have a motivation to respond to the incentives and objectives set forth in the CRA. Because of this motivation, bankers have taken affirmative steps to demonstrate their commitment to CRA principles. This research looks at one of these steps, which is the CRA agreement.

Bostic defined a CRA agreement as a pledge that a bank makes to extend financial services to a targeted, typically under-invested community. Usually, financial institutions will enter into a formal agreement with a particular community-based organization (CBO). However, the recent trend has been for banks to proactively and voluntarily extend pledges to lend at certain levels to communities. Here, Bostic referenced Wells Fargo's pledge to invest $100 billion over the next ten years in low-income communities in California. Although these are types of CRA agreements, they are less formal. The typical agreement contains a commitment not only for lending, but also to provide additional services, such as mortgage counseling, small business technical assistance, or a pledge to open additional bank branches or extend hours. Further, other types of commitments can include a minority hiring pledge or the establishment of a review committee within a target community. These agreements began showing up almost immediately after the CRA was passed, but real growth in the number of agreements did not come until the late 1980s and early 1990s. Bostic stated that there have been well over 300 agreements since the CRA's enactment.

This research is part of a broader agenda that attempts to measure and quantify the effects of CRA agreements upon lending markets. This paper asks the direct question of whether CRA agreements change the lending behavior of the institutions that actually enter into them. A secondary question is: if changes are observed in lending behavior, what characteristics of the agreements are capable of having the most impact? To answer these two questions, the authors identified institutions that entered into agreements and then tracked the results of those agreements to determine if lending increased or decreased.

In order to conduct the analysis, the authors needed to obtain three types of information. First, the bank's structure needed to be identified and followed through subsequent mergers (as 90 percent of the subject banks underwent at least one merger) to accurately attribute the lending to the source of the agreement. Second, the research required specifics of each agreement, such as commencement dates, time in force, and geography covered. Finally, information on lending activity was also required.

The paper's findings paint a cautiously optimistic picture suggesting that lenders increase their targeted lending when agreements come into force and that the increased lending persists even after the term of the agreement. Additionally, the authors find that mortgage counseling and technical assistance are key components of effective agreements. The authors include a caveat that their results are preliminary and the analysis currently lacks the incorporation of time, which is a particularly important consideration given that overall lending increased during the time period covered by the study.

Jonathan Zinman, a research economist from the Federal Reserve Bank of New York, was the discussant. In addition to formerly having been a community development practitioner, Zinman also has written on the impact of the CRA, focusing on the effects of small business lending.

Zinman highlighted three themes in his comments. First, he felt the research in this session asks the right questions. The marginal value of research on the CRA, financial markets, and access to capital is extraordinarily high. This is true, both from a pure academic as well as a policy and practitioner perspective, simply because empirically very little is known about these markets, how they really work, and what optimal policies and optimal interventions really look like. Second, the social sciences, and economics in particular, are young sciences and practitioners are just beginning to appreciate and learn to apply methodologies to estimate the causal relationships that are the focus of the research presented. If there is only one take away point for practitioners, it is that they need to adopt a "buyer beware" approach when trying to employ academic or applied policy research to their own operations. Finally, Zinman argued that economists are just beginning to appreciate the most fundamentally important and interesting questions in this area.

Commenting on the papers, Zinman focused on their respective models and methodology - the means to identify the causal effects of the CRA on outcomes of interest - than their specific results. He emphasized the need to identify and establish the causal relationships asserted by the papers as being accurate.

In commenting on the paper by Bostic and Robinson, Zinman noted that it precisely documents an association between CRA agreements and increased CRA lending, for a subset of agreements. What the paper does not do, at this stage, is establish a causal link between CRA agreements and the types of lending that the CRA seeks to advance. Zinman's comments focused generally on the role of agreements and trying to estimate the causal link, as opposed to the specific best practices. Zinman also offered comments relative to specific refinements to the model that he believes would produce estimates of the causal role of CRA agreements on the outcomes of interest.

What needs to be known and estimated is whether CRA lending actually responds to CRA agreements per se. What is a concern, as noted by the paper, is that CRA agreements are not randomly assigned and can't be studied in the fashion of a clinical trial. Rather, they occur in the real world and are endogenous - certain banks select or decide to have CRA agreements and what needs to be considered is whether the banks that decide to be a party to an agreement are different from banks that don't elect to sign CRA agreements. A specific item of concern is that banks only make commitments when it is relatively inexpensive for them to do so. If it is believed that banks are profit maximizing and savvy - ideas that economists hold dear - then banks are going to be more likely to enter into CRA agreements for lending that they were going to do anyway. Therefore, the challenge here is to test and control for whether an observed association between CRA agreements and lending is actually the result of the agreements or just lending the banks would have done anyway.

To deal with this selection problem, two different approaches could be taken. First, a researcher could attempt to identify some other reason for forming an agreement, something peculiar to observed communities and banks that makes it more likely to have agreements and where that likelihood is going to be plausibly not related to any underlying conditions of credit supply or demand. The second approach would be to control for the lending that banks that enter into agreements would have done anyway.

Zinman suggested strategies that Bostic and Robinson could follow for identifying causal effects. One is to use statistical matching techniques to compare ex-ante (pre-agreement) lending with post-agreement lending. The other strategy is to measure community capacity to explain the likelihood of there being an agreement in the first place. This approach involves determining why some communities have denser, richer, more capable networks of community groups that are more likely to effectively broker and monitor an agreement than other communities. Zinman asserted that the entire explanation for differences between communities relates to underlying conditions of supply and demand within credit markets.

Regarding the paper by Canner et al., Zinman characterized the changes analysis as very sensible, comparing outcomes in tracts that are essentially identical, except for falling just above or below the relevant income threshold, but felt the levels analysis needed further thought.

Zinman opined that there is tremendous potential to make strides in analyzing the CRA. In particular, there is much variation and institutional richness in the CRA and complexity breeds opportunity in academic research. Table 1 (below) illustrates the different types of CRA incentives and policy changes in institutions that Zinman argued should be taken into account.

Zinman closed by underscoring the importance of understanding the underlying economics and microstructure of these targeted markets and by providing two cautionary tales from the papers.

First, mitigating market failures does not necessarily imply that bank profits will increase. The CRA could improve the functioning of LMI markets and the welfare of society as a whole, without necessarily improving conditions for banks.

Second, lending behavior and even outcomes can change even if the CRA doesn't induce banks to change the way they do business. If this is true, then contrary to Canner et al., the CRA might well work without banks needing to change the way that they do business. If banks' underwriting policies depend in part on property values or expectations about property values, and on the perceived probability that other lenders will lend in a given neighborhood, then banks could increase their lending without changing policies or practices.