Location Quotient: Descriptive Geography for the Community Reinvestment Act
Geo Info Systems
volume 5, Number 6, June 1995.
by Grant Thrall, Judy Fandrich, and Susan Elshaw-Thrall
Location Quotient: Descriptive Geography for the Community Reinvestment Act
Geo Info Systems
volume 5, Number 6, June 1995.
by Grant Thrall, Judy Fandrich, and Susan Elshaw-Thrall
Susan-Elshaw Thrall is a cofounder of Spatial Decisions and Analysis Inc., and serves as its Vice President. She is a Professor of Computer Science, College of Business, Lake City Community College. She has a Ph.D. in Geographic Information Systems, Computer Sciences and Computer Education from University of Florida.
The Federal Reserve Board in Washington DC, reported in 1991 that home mortgage loan applicants from minority households of every income category were being rejected at levels higher than nonminority applicants (Quint, 1991). The survey data comprised more than five million loan applications made to banks, savings and loans, and other lending institutions.
An earlier study in Boston, Massachusetts, USA, found that the lending patterns by mortgage originations in African-American neighborhoods versus European-descent neighborhoods could not be explained by variables such as income or housing development (Bradbury et al., 1989). Using Federal Reserve Board data, these studies demonstrated that African-American loan applicants with good credit records had been turned down at twice the frequency as European-descent applicants with otherwise similar profiles.
Perhaps the first geographic information systems (GIS) study of spatial patterns of mortgage loans was performed by Essential Information Incorporated (EII, Washington, DC). The company evaluated 1.25 million loan applications made in 1991 and 1992 (EII, 1993). The EII study revealed that minority neighborhoods located within territories of lending institutions had either been excluded or were underserved with respect to mortgage loans. Banking regulations prohibit the act of avoiding and underserving minority neighborhoods, also known as redlining. The lending industry claimed that the EII study was biased (Rockwell, 1993). Related geography literature includes the work of Palm (1985), Wheeler and Dillon (1985), and Warf and Cox (1995).
In this "Shop Talk" column we present our geographic description (see Thrall, 1995) of mortgage patterns for St. Lucie County, Florida, USA. St. Lucie County lies north of Palm Beach and Martin counties in southeast Florida. Data for this study came from the 1992 legal property tax records of the county, which include 132,000 parcels. All parcels were geocoded; however, only owner-occupied, single-family dwellings with outstanding mortgages were used. In 1991 there were 11,480 owner-occupied, single-family dwellings with mortgages in St. Lucie County. Two financial institutions, First Citizens Federal Savings and Loan (Ft. Pierce, Florida) and Harbor Federal Savings and Loan (Ft. Pierce), dominate the county lending market with a total of 39 percent of all mortgages. First Citizens has a 13 percent market share versus 26 percent for Harbor Federal.
HISTORY OF FAIR LENDING LEGISLATION
The earliest regulation prohibiting race as a factor in the housing market was the 1968 Civil Rights Act Title VIII, which prohibited "racially motivated refusal to sell or rent, denial that a unit is available when in fact it is available, racial steering by brokers, and statements of racial preference or exclusion in advertisements" (Lake, 198 1 ). The Title VIII law assumes that the market is free of bias so discrimination is treated as an aberrant practice, The law requires the plaintiff to pay court costs, with damages capped at $1,000 (Lake, 1981).
The Home Mortgage Disclosure Act (HMDA) of 1975 requires lenders to disclose by standard metropolitan statistical area the number and value of mortgage loans that the lender either originated or purchased.
The Federal Fair Housing Act of 1976 prohibits "racially discriminatory lending" (Squires and Velez, 1987). Under the Fair Housing Act "proof of racial discrimination requires a demonstration of discriminatory effect." In comparison, those "cases brought under the Fourteenth Amendment guarantees of equal protection entail demonstrations of discriminatory intent" (Clark, 1981 ). In other words, under the Fourteenth Amendment discriminatory intent has to be shown, while under the Fair Housing Act only discriminatory effect need be demonstrated.
The 1977 Community Reinvestment Act (CRA) requires "each appropriate federal financial supervisory agency to use its authority when examining financial institutions, to encourage such institutions to help meet the credit needs of the local communities in which they are chartered consistent with the safe and sound operation of such institutions." CRA also requires banks to "demonstrate that their deposit facilities serve the convenience and needs of the communities in which they are chartered to do business" (U.S. Statutes 91:1147).
In 1989, the passage of the Financial Institutions Reform, Recovery, and Enforcement Act (FIRREA) "strengthened the Community Reinvestment Act of 1977 by requiring the appropriate federal depository institution's regulatory agency to prepare a written evaluation" of how the entire credit needs of a community are being met (Meyer and Ballard, 1990). Although FIRREA was a response to the savings and loan crisis, the act serves to increase the effect of CRA on both the financial industry and home mortgagee.
In 1988 and 1989, congressional hearings were held to review the compliance of banks to CRA. The 1989 hearings by the House of Representatives Subcommittee on General Oversight and Investigations of the Committee on Banking, Finance, and Urban Affairs focused on the issue of mortgage discrimination (House Serial No. 10 1-48). The hearings were held to determine whether federally insured lenders were making loans within their communities and whether federal regulators were enforcing CRA. The hearings held by the U.S. Senate Committee on Banking, Housing, and Urban Affairs led to the insertion of an additional requirement into FIRREA.
This additional requirement was the collection of data about both approved and denied mortgage applications. The regulation also made this data public information. For the first time, information about rejected applicants was collected and made public. Many rejected applicants had already completed an initial quasi-formal, prescreening process, which involved finding a real-estate broker, locating a seller, and signing a contract. In addition, under FIRREA it was now possible to evaluate lending institutions' decision-making processes (Squires, 1992). The quasi-formal process was completed before an applicant applied for a mortgage loan. Analysis of the data collected and reported by the Federal Reserve Board in response to the 1989 FIRREA regulation suggested that racial discrimination in lending practices may have been a common practice nationwide. This conclusion was based on the dramatic differences, by race, in rejection rates at every income level (Quint, 1991).
Collecting data and evaluating lending practices is difficult and tedious, even for researchers who have the resources of the Federal Reserve Board (Bradbury et al., 1989). In addition, until the FIRREA requirement, no data were collected on rejected loan applicants. Because enormous amounts of data are involved, because the data have been difficult to obtain for examination, and because GIS technology has a steep learning curve, information now made available through the FIRREA requirements has not been adequately analyzed.
DATA
A source of data about existing mortgage liens of interest to those evaluating compliance with FIRREA, CRA, and HMDA regulations is available to the public in most areas through the countywide property appraisal data base and associated property-lien data base. In St. Lucie County, these two data bases are combined. The property data for this descriptive illustration contain a total of 1 1,480 owner-occupied, single-family dwellings with mortgage liens for the year 1991. For a detailed discussion of the attribute and geographic components of the data, see Thrall and Elshaw-Thrall (I 991) and Thrall et al. (1993).
The demographic data are from the U.S. Census (U.S. Bureau of the Census, 1990a, 1990b). Because CRA and related regulations address issues of equity toward minority populations, census demographic information used for this study includes African-descent (black) population count, European-descent (white) population count, and total population count. Census data from 1990 were imported to the GIS and then transformed to be at the geographic scale of the census tract. We used census tracts from 1980. For a discussion of this, including conversion from 1990 to 1980 census tracts, see Thrall (1992) and Howenstine (1993).
Information from the county property assessor used in this study includes lender codes. Each financial institution initiating or holding mortgage loans in St. Lucie County is assigned a unique lender-identification code. The appropriate code is included with the record for each parcel in the property assessor data file.
Also included is information about whether the owner of the dwelling qualifies to receive homestead exemption under the Florida Constitution (Article VIII, Section 6). To minimize bias that may result from the inclusion of second homes or rental properties, the observations used for this analysis are limited to only owner occupied, single-family dwellings. Homeowners in Florida are allowed a homestead exemption, which allows homeowners to deduct $25,000 from the assessed value of their principal residence when calculating property taxes. A residence that receives homestead exemption in the State of Florida is, therefore, an owner-occupied dwelling.
The following is our geographic description (see Thrall, 1995) of mortgage patterns for two lending institutions in St. Lucie County. Our goals were to discover, from the descriptive presentation evidence for spatial patterns of mortgage loans, the appearance of biased - or unbiased lending practices.
The methodology for this illustration and the data are, in our opinion, insufficient to conclude whether or not lending practices are biased. The geographic analysis of mortgage-loan distribution is made possible by using GIS to associate the spatially distributed population data from the U.S. Census with data from the combined St. Lucie County property appraisal and property lien files. GIS technology can be used to evaluate mortgage distribution patterns by population demographics and location. The technology also displays features that may reveal the appearance of redlining, which under the Fair Housing Act is based on effect rather than intent.
METHODS
Summary numerical measures based on point-data information aggregated to census-tract level are derived using GIS point polygon operations (for discussion, see Thrall and Marks, 1993). Each of the properties within the county was assigned a unique latitude-longitude coordinate, which was used to position the property within the appropriate census-tract polygon. Because of its "column aggregate" function, we used GISPlus software (Caliper Corporation, Newton, Massachusetts, USA).
In GISPlus, users choose a geographic data file of polygons. In our example, we chose census tracts. The polygons overlay the geocoded point file - in our example, the property lien file. The column aggregate function counts the number of points (owner-occupied, single-family dwellings) within each census-tract polygon. The resulting summation becomes a new attribute field of the census tract polygon layer. We then use the GIS-derived counts by census tract to calculate a location quotient (LQ) for each census tract for each of the two lending institutions.
LQ is a common geographic mathematical procedure to measure the degree of relative concentration of an activity on a map. LQ - which allows the comparison of an area's share of some activity with the share of some base aggregate - is sometimes referred to as a concentration ratio or a self-sufficiency ratio. For our purposes, LQ is interpreted as a measure of the degree of concentration of mortgage loans within census tracts.
LQ has been applied in location analysis and in regional development. During the 1940s, the National Resources Planning Board (NRPB) calculated LQ for each manufacturing activity for every state (Isard, 1960). Walter Isard, an eminent regional scientist and economic geographer, believed that LQ could produce a rough benchmark in the analysis of localization in an area. Mayer and Pleeter proposed (1975) some theoretical guidelines for calculating the location quotient with respect to import-export analysis.
The LQ is defined as:
LQ = X*/Y*
where:
X* = percentage of a selected subset within a small area.
Y* = percentage of the aggregate base within the small area.
The LQ can be calculated as:
LQ[ij] = (x[ij] /X)/(y[ij] /Y); i=1,....,n;
j = 1,....,m; LQ[ij] <=> 1
where in this illustration:
x[ij] = number of homestead-exempted (owner-occupied) mortgages within census
tract i owned by financial institution j.
X = total number of all homestead-exempted (owner-occupied) mortgages by financial institution j within entire county (across all n census tracts).
y[ij] = total number of homestead exempted (owner-occupied) mortgages within census tract i owned by all m financial institutions.
Y = total number of all homestead exempted (owner-occupied) mortgages by all financial institutions within entire county (across all n census tracts).
If LQ is equal to 1.0, then the participation of a particular financial institution in mortgage loan activity within a given census tract can be said to be in the same proportion as the participation of all other financial institutions in the census tract. If LQ is less than 1.0, then the loan activity is under-represented by the particular financial institution in the given tract. If LQ is greater than 1.0, then the participation by the given financial institution in loan activity within the given census tract is overrepresented relative to other financial institutions operating within the same countywide market.
LQ is calculated for each tract for each of two financial institutions in 22 census tracts. A statistical t-test can be calculated for each measure of LQ to determine if it was statistically significantly equal to I.O.
For CRA purposes, we are interested in identifying those census tracts, if any, with significant numbers of minority households whose LQ is significantly below I.O. This approach is a so-called one-tail test of significance.
For general marketing information, however, a so-called two-tail test of significance would be more appropriate. For that test, tracts having LQs that are greater than, equal to, or less than 1.0 would be identified. If, for the financial institution, an LQ measure falls outside the confidence interval and is greater than 1.0, then overrepresentation of mortgage loans by the financial institution can be said to occur. For a particular census tract, if the LQ measure falls outside the confidence interval for the t-test and is less than 1.0, then underrepresentation can be said to occur. A two-tailed t-test therefore provides three statistically significant areas of evaluation. The left tail is the area of underrepresentation. The right tail is the area of overrepresentation. The area between the two tails is where the particular financial institution is statistically not different in spatially concentrated lending patterns than other financial institutions in the county. For our illustration we use a two-tailed approach.
A third alternative is to concentrate on variation in individual institutional behavior. We are not as concerned with what is typical for all financial institutions in a county as we are with individual institutional behavior and whether there is the appearance of redlining by a particular financial institution. Instead of testing for an LQ equal to 1.0, we may instead test for how the LQ for a particular census tract is similar to or differs from the average LQ for a particular financial institution across all census tracts in that institution's market area. Variation from the financial institution's average LQ can identify where there is greater or lower concentration than normally expected for the particular lender. For this illustration we test on the basis of the financial institution's average LQ, but the general procedure followed in this illustration can be followed for each of the above alternative scenarios.
RESULTS OF ILLUSTRATION
St. Lucie County has a population of about 150,142, of which 24,664 are black and 122,133 are white. Roughly 3 percent of the total population are of Asian or Native American Indian descent. Three-fourths (18,626) of the black population reside in four adjacent census tracts (1,2,3,9) [Figure I ].
Figures 2 and 3 show the category that the location quotient for each of the two financial institutions falls within, at a 95 percent level of confidence; the breaking points for the categories relate to the statistical t-test described above. Figure 2 depicts the results for First Citizens and Figure 3 for Harbor Federal.
Table I describes the participation of the two financial institutions in the four minority-dominated census tracts and also includes the numerical value for the LQ for each financial institution in each of the four census tracts. Note that the confidence interval for each financial institution is different. The range normal is defined as the LQ value not falling in either the upper or lower tail of the confidence interval. Table 1, therefore, shows that what is calculated as normal for First Citizens is different than what is normal for Harbor Federal.
It should be noted that LQ is a relative measure of spatial concentration, not an absolute measure of spatial concentration. Hence, if few total observations occur in any census tract, then over or under-representation can appear as a numerical artifact. That explains the pattern in the eastern coastal-barrier island census tracts, which contain few single-family dwellings and instead are dominated by multiplefamily dwellings.
RESULTS
Both lending institutions had mortgages outstanding in each census tract of the county. Hence, it is reasonable to conclude that the entire county is the market area for each lending institution.
First Citizens, the second-largest mortgage holder in St. Lucie County, is heavily represented in the mortgage market in the new, rapidly growing, southern portion of the county; its main office is in the northcentral portion of the county, and it has market presence throughout the populated areas of the county. Harbor Federal also has its main office in the historic central- northern core of the county; yet, it is underrepresented in mortgages within the census tracts that surround its main office.
Using data current through 1991, this process demonstrates that the two lending institutions are underrepresented relative to other financial institutions in the four African American-dominated census tracts.
We could conclude that, for the two financial institutions illustrated here, the.,. is the appearance of redlining; however, it should be emphasized that we cannot infer the cause or intent of the appearance of redlining. Moreover, the census tracts that have a high proportion of African American minority population also contain a low proportion of single-family dwellings that qualify for homestead exemption. Therefore, as in the affluent barrier-island census tracts discussed above, the location quotient can mathematically demonstrate concentration when that concentration is actually spurious.
Nevertheless, for practical purposes in terms of compliance with federal regulations, the two financial institutions both appear to have low concentration of mortgages in minority-dominant census tracts. The burden of proof falls on the financial institutions to demonstrate that it is not their intent to discriminate; that does not translate into a requirement to make bad loans. Rather, our interpretation of federal regulations is that the financial institutions must be able to document that they expend reasonable effort to market their products in minority-dominant census tracts.
CONCLUSION
Our purpose was twofold. First, to present an overview of CRA and related federal regulations on lending institutions, which have become a significant component of applied GIS regional literature and GIS commerce. Second, to demonstrate the use of the geographer's location quotient. We have shown how one version of the location quotient can be calculated and used to describe the spatial pattern of mortgage lending institutions.
Our calculations suggest that each of the two lending institutions have different geographic market strategies. Both example lending institutions have been found to be underrepresented in mortgages in several census tracts that are dominated by minorities; this finding does not mean that either of the two banks are redlining or discriminating in any manner. The findings suggest only that there is an appearance of redlining in the year the data were current.
REFERENCES
Bradbury, K.L., K.E. Case, and C. Dunham. 1989. "Geographic Patterns in Boston, 1982-1987." New England Economic Review , September-October: 3-30.
Clark, T.A. 1981. "Race, Class, and Suburban Housing Discrimination: Alternative Judicial Standards of Proof and Relief." Urban Geography 2(4):327-338.
Essential Information, Inc., 1993. "Computer Mapping Reveals Evidence of Widespread Discrimination Against Minority Neighborhoods," Washington DC: Essential Information, Inc.
Howenstine, E. 1993. "Measuring Demographic Change: The Split Tract Problem." Professional Geographer 45(4): 425-430.
Isard, W. 1960. Methods of Regional Analysis . New York: John Wiley & Sons.
Lake, R.L. 1981. "The Fair Housing Act in a Discriminatory Market." Journal of the American Planning Association 47(1):48-58.
Meyer, D.A., and S.A. Ballard. 1990. "Issues in Lending - A Practical Guide to FIRREA." The Journal of Commercial Bank Lending 1: 11-26.
Mayer, W., and S. Pleeter. 1975. "A Theoretical Justification for the Use of Location Quotients." Regional Science and Urban Economics 5: 343-355.
Palm, R. 1985. "Home Mortgage Lenders and Earthquake Hazards." Papers of the Regional Science Association 57:139-153.
Quint, M. 1991. "Racial Gap Found on Mortgages." New York Times October 22, pages CI and C2.
Rockwell, L. 1993. "Facts." Forbes September 27, page 86.
Squires, G. 1992. "From Redlining to Reinvestment." Philadelphia: Temple University Press.
Squires, G.D., and W. Velez. 1987. "Insurance Redlining and the Transformation of an Urban Metropolis." Urban Affairs Quarterly 23(l):63-83.
Thrall, G.I. 1992. Using the JOIN Function to Compare Census Tract Populations Between census Years," Geo Info Systems 2(5):78-8 1.
Thrall, G.I. 1995. "Stages of GIS Reasoning." Geo Info Systems 5(2):46-5 1.
Thrall, G.I., and S. Elshaw-Thrall. 1991. "Reducing Investor Risk: A GIS Design For Real Estate Analysis." Geo Info Systems 10):40-46.
Thrall, G.I., and A.P. Marks. 1993. "Functional Requirements of a Geographic Information System for Performing Real Estate Research and Analysis." Journal of Real Estate Literature l(l):49-61.
Thrall, G.I., M. Ruiz, C. Sidman, and S. Elshaw-Thrall. 1993. "Using GIS Tools to Analyze and Visualize Spatial Phenomena." Geo Info Systems 3(5):59-65.
Warf, B., and J. C. Cox. 1995. "US Bank Failures and Regional Economic Structure." Professional Geographer 47(l): 3-16.
Wheeler, J., and P Dillon, 1985. "The Wealth of the Nation: Spatial Dimensions of U.S. Metropolitan Commercial Banking, 1970-1980." Urban Geography 6:297-315.
U.S. Bureau of the Census, 1990a. "Census of Population and Housing, Summary," Tape File I A (STF- IA), U.S. Government Printing Office, Washington, DC, 20233-8300.
U.S, Bureau of the Census, 1990b. "Census of Population and Housing," Summary Tape File 3A (STF-3A). US Government Printing Office, Washington, DC 20233-8300.
U.S. Congress. House Committee on Banking, Finance, and Urban Affairs. Subcommittee on General Oversight and Investigations. 1989. Field Hearing on Mortgage Lending Patterns and Related Issues in the South Queens, NY, Area. 101st Congress. First session. Serial 101-48.
