«OHIO STATE LAW JOURNAL VOLUME 66, NUMBER 4, 2005 Predatory Lending and the Military: The Law and Geography of “Payday” Loans in Military Towns ...»
To complete our study, we required four types of data: population information, military base locations, bank locations, and payday lender locations. All civilian population information was obtained from the U.S.
Census Bureau.285 The absence of an authoritative reliable source for military population made analysis requiring this information somewhat more problematic. Because military personnel are frequently being deployed, reassigned, trained, and moved, many of the bases we contacted were unable to give us reliable manpower figures. After consulting with representatives from the Department of Defense (DOD), we selected the DOD’s annual Base Structure Report of 2004 as our primary databank.286 Data regarding personnel was cross-referenced with a report published by the DOD’s Statistical Information Analysis Division287 as well as with the data from the Census Bureau.
Data on military base locations in general is widely available. However, the precise boundaries of military bases are sometimes ambiguous. In delineating base boundaries, we primarily relied on maps issued by the United States Geologic Survey (USGS) and published by the Environmental System Research Institute (ESRI). However, we found several instances where USGS maps did not match maps created by either the U.S. Department of Transportation or other private digital map vendors. Discrepancies in base location were resolved via telephone calls to information offices at individual bases. Many bases are large and include multiple parcels of land, sometimes flung over several counties. Where this was the case, the ZIP code region(s) containing the base headquarters and the majority of on-base housing was used to delineate the boundaries of the military installation under consideration.
While bank and bank branch addresses were easily obtained from the Federal Deposit Insurance Corporation (FDIC),288 obtaining reliable data on payday lender locations proved more challenging. We obtained the addresses of 285 See United States Census Bureau, Census 2000 Summary File 3, http://www.census.gov/Press-Release/www/2002/sumfile3.html.
286 See DEPARTMENT OF DEFENSE, BASE STRUCTURE REPORT (2004), http://www.defenselink.mil/pubs/20040910_2004BaseStructureReport.pdf. According to officials in this office, this data was submitted to the DOD by officials on base.
287 Department of Defense, Directorate for Information Operations and Reports, Statistical Information Analysis Division, Distribution of Personnel by State and by Selected Locations, http://web1.whs.osd.mil/mmid/pubs.htm (last visited Oct. 17, 2005). According to officials in this office, this data was collected through payroll records.
288 Federal Deposit Insurance Corporation, Find an Institution, http://www2.fdic.gov/idasp/main.asp (last visited Oct. 17, 2005). The FDIC recognizes several different categories of banks. For our purposes, we included all branch locations irrespective of the FDIC=s categories.
700 OHIO STATE LAW JOURNAL [Vol. 66:653 payday lenders from the state regulatory authority charged with oversight of payday lenders in all but three states included in the study. In most instances, regulatory oversight offices host a website where the addresses of payday lenders can be downloaded; several other states sent lists of payday lenders via electronic mail or as paper copies via U.S. Postal Service. Though we believe the individual licensing agencies are the best source for addresses, we do not believe they are comprehensive. Ample anecdotal evidence suggests that many payday lenders operate without a license from the state. We were able to phone several payday lenders listed in local telephone directories that were not licensed or included on the list of payday lenders provided by various states.
Conversations with state authorities and other industry observers confirmed our observations.289 Though incomplete, we are confident that the lists provided by the states do include businesses engaged in the business of payday lending. To that end, each regulatory authority was contacted in order to ensure that the criteria used to define “payday lender” in our study was consistent from state to state. In three states vital to our survey—New York, North Carolina, and Texas—we could not obtain adequate data from state regulators, and accordingly we used alternative data gathering strategies. Our data collection methods for these three states are elaborated in Part IV alongside discussions of the law and empirical findings in those states.
In terms of mapping technique, we used commercial mapping software to map the addresses of individual payday lender and bank locations onto TIGER centerline files.290 Using these files, we are able to enter a database of addresses into mapping software that places points on street maps indicating the location of each address. For each case study location, a minimum 75% match rate was achieved; but in most cases, especially for payday lenders, match rates of over 90% were realized, giving us reliable sample sizes and excellent statistical confidence.291 Matched addresses were randomly checked for accuracy by 289 Telephone conversations with several state officials and other industry analysts confirmed our suspicion that there are many unregulated payday lending operations in each state. Telephone Interview with Jennifer Delacamp, Lawton Area Supervisor, Consumer Credit Counseling of Oklahoma (Jan. 2005). Independent of the conclusions of this Article, it is troubling that some payday lenders simply have refused to acknowledge the authority of state regulators by openly disregarding state licensing requirements.
290 TIGER Line files are the basis for street and road maps used by many government agencies. See, e.g., KANG-TSUNG CHANG, INTRODUCTION TO GEOGRAPHIC INFORMATION SYSTEMS 308 (2002) (describing TIGER/Line files). Our maps were created using the Geocode function in ESRI=s ArcMap 9.0 software, a common professional geography computer program which allows users to compile, author, analyze, map, and publish geographic information. See Environmental System Research Institute, What is ArcGIS?, http://www.esri.com/software/arcgis/about/overview.html (last visited Oct. 17, 2005).
291 See GARETH SHAW & DENNIS WHEELER, STATISTICAL TECHNIQUES IN2005] PREDATORY LENDING AND THE MILITARY 701 cross-referencing matched locations with several widely available on-line address-matching services.292
3. Statistical Analysis of Payday Lender Location Density Maps were analyzed using simple, widely-understood statistical measures in hopes that the findings would be transparent to the widest possible audience. At the county and ZIP code levels, three basic measures of payday lending were employed. The first was the total number of payday lenders per geographic region. The second was payday lenders per capita, generally expressed in terms of payday lenders per 100,000 persons. The third measure we used is a measure of payday lending density relative to banking density. Professional geographers have a variety of commonly accepted methods for measuring relative location density of two business types. Most geographers typically use a standard business density formula known as a “location quotient.”293 In calculating payday lender density relative to banks, we used statistically acceptable variations on the standard location quotient formula tailored to capture subtle differences in payday lender and bank density for our county and ZIP code level analyses.294 GEOGRAPHICAL ANALYSIS 48–53 (2d ed. 1994) (describing statistical significance in mapping match rates).
292 See, e.g., Environmental System Research Institute, ArcWeb Showcase: Map Viewer Application, http://arcweb.esri.com/sc/viewer/index.html (last visited Oct. 17, 2005);
Google Maps, http://maps.google.com (last visited Oct. 17, 2005); Mapquest, Mapping Service, http://www.mapquest.com/ (last visited Oct. 17, 2005); Yahoo Maps, http://maps.yahoo.com (last visited Oct. 17, 2005).
293 Location quotient is the most frequently used statistic to determine a region’s share
of some business activity. One standard location quotient formula is:
Xi ∑X LQ = Yi ∑Y where LQ is the location quotient, X and Y are the businesses in question, and i is the geographic location, such as a ZIP code or a county. SHAW & WHEELER, supra note 291, at
313. However, an in-depth discussion of analytic statistical geography is beyond the scope of this Article. For an excellent introduction to this topic, see generally JAMES E. BURT & GERALD M. BARBER, ELEMENTARY STATISTICS FOR GEOGRAPHERS (2d ed. 1996).
294 The standard location quotient formula is not appropriate for this study, given the data limitations inherent in tracking payday lending locations. Because there are many ZIP codes with no payday lenders, the standard formula is not suited to measuring this industry.
Modifying this formula allows us to use the data we have available to include those areas without payday lenders, instead of tossing them aside, and to see subtle differences between two areas with identical ratios of banks to payday lenders but with different numbers (volume) of banks and payday lenders. In the alternative, we conducted experiments with 702 OHIO STATE LAW JOURNAL [Vol. 66:653 Next, we ranked each of these three statistical measures against their intrastate counterparts, with the lowest rank (first) in each category assigned to the county or ZIP code with the highest score on each variable. So, for example, the county with the highest total number of payday lenders would therefore receive a rank of first in that category. Similarly, the ZIP code region with the highest relative density of payday lenders in comparison to banks would receive the first place ranking for that category. Finally, the ranks for all three categories were averaged together to produce a composite index for each scale level. Because the composite index is a function of our three measured categories, the lowest ranked counties and ZIP code regions will generally feature a relatively large number of payday lenders, a relatively high density of payday lenders per capita, and a relatively high ratio of payday lenders to banks.
These composite index scores were also assigned ranks with the highest composite index score again receiving the first place ranking. Importantly, our composite index scores create an opportunity to express the proximity of the payday lending industry as a whole in any given county or ZIP code to military bases with a single, easily comparable number.
In order to give us some perspective on the per capita density of payday lenders in any unit of analysis, such as a ZIP code, we calculated the statewide average for payday lenders per 100,000 people. By multiplying the statewide average by the population in smaller-area units, such as a ZIP code, we were able to predict the number of payday lenders that should be in that unit of analysis, if it were to conform to the statewide average.295 Finally, we compared numerous formulaic variations and produced nearly identical results. We selected a very
simple county level ratio:
⎛X⎞ LQ = ⎜ ⎟ × 100 ⎝Y⎠ where LQ is the location quotient, X are payday lenders, and Y are banks. For ZIP code regions, our relative measurement of payday lender to bank density needed additional refinement to account for the great number of ZIP codes without banks, payday lenders, or either. Once again, after numerous experiments, we selected the following formula which distinguishes ZIP code regions that have identical ratios payday lenders and banks, but have
different absolute numbers of bank and payday lenders. Our ZIP code region formula is:
⎡ ⎤ X + (X − Y ) LQ = ⎢ ( X + Y ) × 100 ⎥ ⎣ ⎦ where, once again, LQ is the location quotient, X are payday lenders, and Y are banks. We believe these formulas provide the best opportunity to see subtle differences in the density of payday lending (relative to banks) among counties and ZIP codes in each state. Moreover, they are well within traditionally accepted geographic methodology. SHAW & WHEELER, supra note 291, at 313–15.
295 The formula we used to determine the expected number of payday lenders is:
2005] PREDATORY LENDING AND THE MILITARY 703 our prediction, or “expected” number, of payday lenders against the actual number of payday lenders observed in each geographic unit. This allowed us to accurately characterize the actual number of payday lenders as being in excess of, equal to, or below the statewide per capita average for any given regional population.