Research Methodology and Data Availability in Key CDFI Product Lines

To identify prospects for developing some reasonable measures of CDFI impacts within product lines, I evaluate types and quality of the data available, or that might be made available at reasonable cost, for some major financial product groups.

Single-Family Mortgages

Within this group there are different product lines, including senior and junior home purchase loans, home improvement loans, and refinance loans. Some CDFIs, especially those with a major homeownership focus, offer more than one product line. Perhaps the best-known examples are the NeighborWorks affiliates around the country. Because many programs focusing on single-family lending may include community development outcomes among their primary objectives, and because there is significant evidence that home improvement (Ding, Simons, and Baku 2000) and homeownership (Haurin, Dietz, and Weinberg 2003) result in positive neighborhood spillover effects, they may lend themselves well to geographic impact analysis.

One evaluation approach would use a geographic experiment. After identifying potential census tracts that meet the criteria for a new CDFI single-family lending program, researchers would randomly divide tracts into treatment and control groups. It would be critical for the selection process to be truly random and not corrupted by political or other criteria.

Quasi-experimental methods may be required. In analyzing the impact of New York City investments in housing on property values, Schill, Ellen, Schwartz, and Voicu (2002) employed a model moving neighborhood location and time of property sale to control for the fact that properties in different neighborhoods will likely be subject to different trajectories in property values. In this way, they controlled for selection bias. Galster, Tatian, and Accordino (2006) used property sales data for Richmond, Virginia, to determine if efforts and those of the area's Local Initiatives Support Corporation to concentrate community development investments strategically in certain neighborhoods had any positive effect on property values. They utilized AITS to measure the impact of nearby community development investments on the level and trajectory of property values.

It should be acknowledged that the data requirements for such property-value impact models, especially ones that do a good job addressing the issue of preintervention trajectories and selection bias, can be quite demanding. Not only are sales transaction data over time generally required, researchers may need information on property attributes – typically from a property tax assessor or a multiple listing service when available.

Multifamily Real Estate Loans

Estimating the impact of CDFI multifamily lending activity on local outcomes is conceptually feasible. However, one critical issue is determining what, if any, crosscutting outcomes might be identified in the case of multifamily lending activity. Some CDFIs may view community development outcomes – perhaps measured by neighborhood stability indicators such as property values – as a desired product of multifamily lending. Such would be the case for programs focusing on the repair and improvement of dilapidated multifamily properties. When small-area community development goals are the focus, analyses looking at local property value effects, such as the single-family lending model described earlier, may be feasible.

However, many CDFIs making multifamily loans focus more on the aggregate production of affordable rental housing, with less regard to neighborhood impacts. Moreover, because little good data – other than the decennial census – arc available on vacancy rates or affordability measures at the small-area level, analyses at the neighborhood level are difficult.

Small Business Term Loans and Lines of Credit

CDFI lending to small businesses appears to be a weaker candidate for geographic impact evaluation. Many CDFIs that do small business lending do not target their programs at small-area levels, nor do they generally seek small-area impacts. At larger geographies, the lack of relative density is unlikely to allow for the discernment of geographic impacts. Hollister (2004) and Caskey and Hollister (2001) reviewed some attempts to evaluate the impact of what they call "business development financial institutions." They found existing efforts lacking and are less than optimistic in suggestions for additional research in this area.

In terms of quasi-experimental approaches, there at least two possible routes, both of which appear quite challenging given existing data sources. First is a model that would attempt to measure the structural effects on small business lending markets on an intermetropolitan level. The goal would be to determine whether increased CDFI small business lending is related to increased small business lending by conventional lenders. While CRA small business lending data may prove sufficient for this task, the data are not without problems.[1] The second approach would be to attempt to estimate neighborhood-level impacts of small business lending by CDFIs on neighborhood outcomes. Again, the problem is the weakness of small business lending data, as compared to mortgage data. Another problem is that many of the more active CDFI small business lenders do not target their loans spatially, so these programs are likely to achieve fairly low levels of relative density at the neighborhood level.

  • [1] Thanks to Michael Stegman for making this suggestion.
< Prev   CONTENTS   Next >