Seeking A New Analytical Consistency With CRA

Written by Dan Immergluck, Ph.D.
on September 24, 2010 No Comments
Categories : Word On The Street

WORD ON THE STREET: Overall, Community Reinvestment Act (CRA) implementation effectively suffers from covering an ever-decreasing share of mortgage and financial services markets and from inconsistent and undulating enforcement. The portion of the mortgage market that is subject directly to CRA (originated by depositories) declined markedly in recent decades. The share of all home purchase loans made by CRA-regulated institutions fell from 36% to 26% over the 1993 to 2006 period. For refinance loans, the share fell from 45% to 25%.

With respect to the consistency of enforcement, the inconsistencies in the implementation of CRA have allowed for weakened and undulating enforcement of the law. Data from the Federal Financial Institutions Examination Council show that the share of institutions receiving ‘Outstanding’ CRA ratings varied greatly across regulatory agencies, especially in the mid-2000s.

The proportion of institutions regulated by the Office of Thrift Supervision (OTS) receiving Outstanding ratings from 2004 to 2007 ranged from approximately 25% to 35%, while for the Federal Deposit Insurance Corp. (FDIC), the figure fell in the 7% to 15% range. In my own work, in an analysis of the Investment Test results on almost 200 CRA performance evaluations, I found that institutions regulated by the OTS made far smaller levels of investments than those regulated by the other agencies, after controlling for asset size, region of the country and performance on the lending and service tests.

Mortgage markets have changed dramatically since the adoption of CRA. In recent decades, the industry has generally consolidated, with large national lenders accounting for greater and greater market share.

During the 1980s and 1990s, the growth of nonbank mortgage companies meant that more lenders were not subject to CRA. However, the growing dominance of large, nationwide bank-holding companies (BHCs) in the mortgage market, which sometimes occurred, in part due to the acquisition of formerly independent mortgage companies, could have provided an opportunity for improving CRA coverage.

However, the failure to modernize CRA to keep up with the changing structure of the mortgage market has resulted in adverse impacts to CRA coverage. Currently, CRA coverage in the mortgage market is actually quite ambiguous and at least partially determined by the composite desires and choices of the regulated entities themselves. That is, the lending of bank-affiliated enterprises is ‘included’ in CRA performance evaluations largely at the choice of the examined bank.

While there are efforts to limit ‘cherry picking’ by regulated institutions, the examination procedures continue to allow regulated entities to include the loans of their affiliates at their option; this suggests that such loans will be included only if they are expected to improve the banks' CRA test results. This makes little sense. When I give an exam, I do not allow students to instruct me as to what the exam may cover.

To help remedy the problem of declining CRA coverage and to rationalize the CRA process, CRA examinations should be conducted on a BHC level. That is, there should be a single CRA examination for BHC, and the assessment area(s) should be determined based on the lending patterns of the holding company as a whole.

If it is not possible to require that nonbank affiliates be included in the BHC umbrella for CRA evaluation purposes, and a lender is still given the option to include affiliates, it should be required to include all affiliates for all product lines.

Ideally, assessment areas should be based on the lending patterns of all BHC entities, including mortgage company affiliates. It makes little sense to define areas based on only one part of a BHC's business line, simply because it is originated via a depository unit vs. a nondepository unit.

Assessment areas should be developed based on an analysis of market penetration. For metropolitan areas, the market analysis should be conducted on the metropolitan statistical area (MSA) level; for rural areas, it should be conducted on the county level. An MSA should be included in an assessment area if: 1) the BHC originates an appreciable share of loans in that MSA (e.g., 0.05%, as proposed in H.R. 1479 and as recommended by the National Community Reinvestment Coalition); or 2) the MSA is among those MSAs accounting for the bulk of the BHC's lending activities.

This analysis should be done for all major product lines (e.g., mortgages, small business loans). The same process should then be used for determining which nonmetropolitan counties should be included in the assessment area.

Unfortunately, although this sort of analysis can be readily performed for mortgage markets and, with some limitations, for small business lending (at least for large banks), it is not generally possible, at this time, for deposit services given the lack of comprehensive, geographically specific data on deposits. Bank branches can be analyzed, but they are severely limited as proxies for deposit services.

Without such data, developing appropriate assessment areas for the service test will remain difficult. To implement the service test in an adequate fashion, regulators need to promulgate rules for the collection and disclosure of data on basic financial services, including deposit accounts.

Consistency, please

One challenge that the agencies have routinely faced is evaluating a financial institution's CRA performance in the context of the credit and financial services needs of the local communities it serves. This becomes a particularly difficult task in evaluating institutions whose assessment area span across multiple MSAs or states.

One promising suggestion that I recommend expanding upon is to develop well-researched, interagency community development needs analyses for a set of 50 large MSAs and for the remaining balance of each state. Such an approach makes a good deal of sense but should be expanded beyond assessing performance under [proposed] community development (or investment) tests.

This approach could be used to provide more thorough, rigorous and consistent information to be shared by all examiners conducting exams in these areas. I would also recommend expanding the list of MSAs to something more like the top 100 MSAs rather than simply the top 50. This would not only provide more localized knowledge for the midsized and smaller MSAs, but would result in the balance-of-state analyses to more heavily consider needs in rural or small city areas. Such a rationalization of resources should provide for more consistency across examinations and regulatory agencies, and provide for deeper and more meaningful assessments of credit and banking needs in various communities.

Some commentators have suggested that CRA evaluations have become too ‘numbers-driven,’ resulting in lenders being too heavily rewarded for amassing large numbers or shares of low-impact loans, investments or services while receiving insufficient credit for more complex and innovative activities or for activities that are particularly effective at serving a community's credit or financial services needs.

Such perspectives raise some valid concerns. Efforts to quantify results in almost any evaluative context typically run into what researchers refer to as ‘validity-reliability trade-offs,’ in which attempts to develop consistent, reliable and accurate indicators (typically involving quantitative data and tools) inevitably yield indicators that do not fully capture the phenomena of interest.

For any complex phenomenon, no single list of a few quantitative measures will present an entirely valid picture of the phenomenon. This is why evaluators and researchers frequently employ mixed methods of evaluation and assessment: They seek qualitative information to complement the quantitative data.

The answer to the imperfect validity of quantitative measures, however, is not to eschew quantitative indicators. Avoiding quantitative measures is likely to result in greater problems of reliability in assessments. Moreover, reducing the use of quantitative measures may imperil accountability among regulators and institutions and lead to an overall leveling down in the rigor of exams.

Finally, while it is important to give appropriate weight to community development activities and to reward institutions that go further in their efforts to meet community credit and financial services needs, avoiding quantitative measures of mainstream retail loans and services is not the best approach for recognizing such differences.

A better approach is to rationalize and improve the quantitative indicators as much as possible and to combine qualitative and quantitative methods, especially in the analysis of community development activities or investments.

One flaw in CRA implementation during the 1990s and 2000s was the failure of regulators to consider the quality of institutions' lending activities and to make a determination as to whether some portions of lending were, in fact, having detrimental impacts on local communities and households.

Although the bulk of subprime lending was not under the purview of CRA, if some of the changes recommended here are adopted – including expanding assessment areas and examining all lending of BHCs, regardless of the particular channel – it may become more important to consider variations in the affordability, quality and responsibility of lending and financial services products across different communities.

The challenge here is substantial; I am in no way suggesting that considering the quality of retail lending is a trivial task. Promulgating standards for practices or products that are deemed as not beneficial or potentially harmful to local communities is not without likely controversy, and bright lines are not always possible.

Nonetheless, methods and approaches can be adapted from fair-lending and other compliance examination procedures (and can be applied on a geographic basis). For example, regulators should pull random samples of loans from different channels or units of a lender (or BHC) and identify any differences in terms and pricing. Combining this work with analyses of where and to whom different units originate loans should then be used to evaluate how well the BHC, as a whole, serves lower-income communities versus other communities.

The same sorts of analyses should be routinely done with consumer financial services companies. One new tool that could be used to measure loan quality in the mortgage market is the identification of ‘qualified mortgages,’ to be implemented under the Dodd-Frank Act. Examiners should give greater weight to the origination of qualified mortgages in low- and moderate-income communities than to the origination of nonqualified mortgages.

Of particular interest should be the delinquency and default rates of loans originated by the BHC, including all channels. Institutions should maintain reports on the delinquency and default rates of originated loans, regardless of whether the loans remain in portfolio, and be able to identify default rates across different geographies (e.g., low-, moderate-, middle- and upper-income census tracts).

Such analysis should also be broken out by origination channel (e.g., wholesale, correspondent, retail) and by lending unit. Institutions with default rates substantially above industry norms should not receive CRA credit for the corresponding product line, and high default rates should result in lower CRA ratings.

Dan Immergluck, Ph.D., is an associate professor at Georgia Institute of Technology's School of City and Regional Planning. This article is adapted from testimony presented at the Aug. 12 public hearing on CRA regulations in Chicago.

Register here to receive our Latest Headlines email newsletter




Leave a Comment