Journal of Modern Accounting and Auditing, ISSN 1548-6583 November 2013, Vol. 9, No. 11, 1497-1506

D

DAVID

PUBLISHING

The Demographic Impact on Credit Scores: Evidence From Statistical Methods and Geographic Information Systems (GIS) Mapping∗ Anna E. Newman, Joseph A. Newman Auburn University Montgomery, Montgomery, USA

Average credit scores for people in the United States (US) differ from state to state. Some states have high, and some states have low average credit scores. Since lenders and employers use credit scores to make loan and employment decisions, people living in states where average credit scores are high should experience the benefits of living where credit scores tend to allow more favorable loan and employment decisions. Although credit scores are the direct result of credit histories, credit histories may be impacted by demographic factors. If the demographic factors that impact credit histories are identified, ways to improve credit scores are likely to be discovered and available to people and state government policymakers. This study looks for demographic factors to indirectly explain the average credit scores for people living in each state of the US. The methodology includes statistical analyses and geographic information systems (GIS) mapping. Statistical analyses provide evidence to suggest that state average credit scores are explained by the demographic factors of education, family, income, and health. GIS mapping reveals clusters of states with similar demographics and credit scores. Keywords: credit scores, demographics, geographic information systems (GIS) mapping

Introduction Credit scores are used in a variety of ways. Lenders use credit scores to decide if a loan will be made and the interest rate to charge as discussed in Perry (2008). Healthcare providers use credit scores to determine whether patients will receive free or discounted care as discussed in Bernerth, Taylor, Walker, and Whitman (2012). Insurance companies use credit scores to underwrite both automobile and homeowner’s insurance as discussed in Kabler (2004). Credit scores are being used by employers in background checks for hiring decisions as discussed in Bernerth et al. (2012). Additional use of credit scores has been called for to help investors judge their risk exposure in asset-backed securities as discussed in Stachel (2004), or in allocating accounts receivable in the healthcare industry to collection agencies as discussed in Jackson (2008). In all these uses, higher scores are associated with greater benefits, and lower scores are associated with greater costs. ∗

This paper was presented at the November 2012 Meeting of the Academy of Business Research, Biloxi, Mississippi. Anna E. Newman, graduate student, Department of Sociology, Auburn University Montgomery. Email: [email protected]. Joseph A. Newman, Ph.D., associate professor, Department of Economics and Finance, Auburn University Montgomery. Correspondence concerning this article should be addressed to Joseph A. Newman, Department of Economics and Finance, Auburn University Montgomery, Montgomery, Alabama, USA. Email: [email protected].

1498

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

Therefore, it is important to understand why average scores differ, and sometimes considerably, from one state to another. For example, according to Stachel (2004), a common benchmark for subprime lending is a credit score of 660, where borrowing costs rise sharply. This suggests that the average borrower in Mississippi pays a sharply higher interest rate, since the average credit score in Mississippi is below that subprime lending benchmark. Many other states have average credit scores that suggest nearly half the population of the state could be paying high interest rates on loans. Credit scores are directly impacted by bill and loan payment histories, as well as outstanding balances. However, there are likely to be indirect impacts on credit scores through factors that influence bill and loan payment histories and outstanding balances. Identifying those indirect factors should therefore identify the underlying factors that indirectly contribute to credit scores. Knowing the underlying factors that indirectly contribute to credit scores provides an opportunity for people and policymakers to manage those factors in order to help raise credit scores. Raising credit scores should increase the financial benefits to both individuals and the states in which they live. The purpose of this study is to identify underlying demographic factors that indirectly explain why average credit scores differ from state to state. The study uses correlation, regression, and geographic information systems (GIS) mapping to test the influence of demographic factors previously found useful to explain credit scores, and uses a new measure of a demographic factor to provide additional insight. Correlation and regression results provide evidence to suggest that some demographic factors found useful in previous studies to explain credit scores were also found useful in this study, but some were not corroborated. Factors found useful in past studies to explain credit scores and supported with evidence from this study are education, family, income, and health. This study finds that education, measured by the percentage of high school graduates in a state, is the strongest single variable associated with higher credit scores. Also important is family, measured in two different ways: single mothers and people living alone. States with more households headed by single mothers tend to have lower credit scores, and states with more households headed by someone living alone tend to have higher credit scores. Income, measured by household income, is higher in states with higher average credit scores. Health, measured by the percentage of the state population reporting some types of physical or mental disability, is a greater percentage of the population in states with lower average credit scores. These variables, and especially the new health measure, disability, may explain why unemployment, age, and race measures, found useful in previous studies, were not useful to explain credit scores in this study. GIS mapping provides additional useful insight by showing that the factors explaining credit scores are concentrated in specific regions of the United States (US), as are differences in average state credit scores. The primary contributions of this study are the new measure of health, the corroboration of earlier findings regarding the influence of specific demographic factors, the questioning of some previous findings, and the GIS mapping insights. These findings are valuable for making private and public policy decisions that could lead to higher credit scores and subsequent benefits to people and states. The paper proceeds with a literature review identifying demographic factors previously found associated with credit scores. A methodology section specifies the hypotheses being tested, the data utilized, and the analytical techniques employed. A results section displays and discusses output from the analytical techniques. The paper ends with a conclusion section that highlights important findings of the study, emphasizes the limitations of the study, and proposes additional avenues for research.

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

1499

Literature Review Credit score research in the past focused on different concerns, but often used similar demographic factors to explain credit scores. Bernerth et al. (2012) focused on the use of credit scores as a screening tool in employment practices, but also found higher credit scores associated with more education. Perry (2008) used survey data to examine the effect of personality on credit scores and also found higher scores related to increases in education, income, and age, but lower scores related to poor health, unemployment, and lower income. The study also describes a quiz administered to respondents that showed higher credit scores corresponding to more financial knowledge and feelings of control over life. Lyons, Rachlis, and Scherpf (2007) used surveys to understand the extent of knowledge people had regarding credit reports, and found less knowledge in people who were less educated, had lower income, were older, and identified themselves as Hispanic. Several studies examined the extent to which low credit scores perpetuate themselves. Spader (2010) theorized that credit scores create a “feedback loop”, because people are assigned credit products by their scores, and these products by their nature determine default risk. An example is a person with a low credit score being offered only subprime loans. However, Agarwal, Skiba, and Tobacman (2009) found evidence to suggest that people are not assigned credit products; instead, they choose them. Their study looked at payday loan borrowers who also had credit cards, and found that most borrowers could have used a credit card rather than a more expensive payday loan. In addition, after using the payday loan, borrowers were much more likely to default on their credit cards. The authors conclude that poor choices, instead of credit product assignment, cause lower credit scores. Those poor choices could be impacted by the level of education. An investigation of the effect of credit score on consumer payment choice was done by Hayashi and Stavins (2012). They provided evidence to suggest that people with lower credit scores tend to use debit cards more than credit cards, but also identified other significant variables. Their study shows that higher credit scores exist for older people with more education and higher income, who are married or widowed, and Asian or white. Since their study used self-reported credit scores, the authors validated the relationships between reported credit scores and demographic variables by obtaining credit scores directly from a credit bureau and comparing credit score averages to census tract data. Their validation lends support to using aggregated demographic variables to explain state average credit scores and inferring useful lessons for individuals, and mitigates the aggregation inference concerns mentioned in Robinson (1950). An attempt at explaining average credit scores with aggregate demographic variables is documented in Kabler (2004). That study used demographic data aggregated by ZIP code in Missouri to explain credit scores and found significant associations to income, education, divorce, age, population density, and race. Race dominated the associations with lower credit scores associated with racial minorities, and the race association could not be eliminated using other variables. Kabler (2004), using data within just one state, provided a starting point to expand the scope of research to all 50 states. In addition to expanding the scope of research to all 50 states, this study expands the depth of research by including both a statistical analysis and a GIS mapping in its methodology.

Methodology The methodology used in this study starts with hypotheses about the relationships of credit scores to demographic factors, and then tests the hypotheses with empirical data using three different techniques.

1500

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

Hypotheses The hypotheses are concerned with the relationships between credit scores and demographic factors. Therefore, the null and alternative hypotheses are as follows: HO = State average credit scores are not impacted by demographic factors. HA = State average credit scores are impacted by demographic factors. Demographic factors found useful in past studies to explain credit scores include: education in Perry (2008), Kabler (2004), Hayshi and Stavins (2012), and Bernerth et al. (2012); family in Kabler (2004) and Hayashi and Stavins (2012); income in Perry (2008), Kabler (2004), and Hayashi and Stavins (2012); health in Perry (2008); unemployment in Perry (2008) and Kabler (2004); age in Perry (2008), Kabler (2004), and Hayashi and Stavins (2012); and race in Kabler (2004) and Hayashi and Stavins (2012). Unemployment, age, and race were initially used to explain credit scores but were dropped from this study, because their influence was apparently captured by other factors. Therefore, the general model for the final analyses used to test the null and alternative hypotheses is as follows:

Credit Score = f ( Education, Family, Income, Health) The expected relationship between credit score and education should be positive, since more educated people should have greater knowledge and understanding of credit scores and their importance. The expected relationship between credit score and family will vary by the variables used to measure family. Households with couples to share expenses should have the best credit scores, and households headed by one person, especially when children are present, should have lower credit scores. The expected relationship between credit score and income should be positive, since greater income should provide more discretionary income. The expected relationship between credit score and health problems should be negative, since people in poor health are likely to have higher medical expenses. The relationships of credit score to the demographic factors are assumed to be monotonic, so a linear functional form is utilized. The variables used in this study to measure credit scores and the demographic factors are described next. Data Data used to measure the hypothesized demographic factors influencing credit scores are from all 50 US states and various sources. The average state credit score measure is labeled Score, and it is the average state credit score reported for each state on the Fair Isaac Corporation website in September 2012. The measure for education is labeled Graduate, and it is the percent of the population 25 years old or older who are high school graduates, taken from surveys done during the years from 2006 to 2010 by the US Census Bureau. Family status has several measures. One measure is labeled Mothers, which is the percentage of households headed by females without a husband present but with children under 18 years of age, taken from the Profile of General Population and Housing Characteristics 2010 carried out by the US Census Bureau. Another measure of family status is labeled Alone, which is the percentage of households with just one person reported by the US Department of Commerce, US Census Bureau, in their Profile of General Population and Housing Characteristics: 2010 Demographic Profile Data. The measure for income is labeled Income, which is median state household income from the US Census Bureau, State Median Income, Median Annual Social and Economic Supplement, and Median Household Income by State⎯Single Year Estimates for 2011. The measure chosen to measure health is labeled Disabled, which is the percentage of the state population ages

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

1501

21-64 reporting a visual, hearing, ambulatory, cognitive, self-care, or independent living disability, and summarized by state in the “2010 Disability Status Report” published by the Cornell University Employment and Disability Institute, Cornell University. Another, less satisfactory measure for health, the percentage prevalence of self-reported obesity among adults in each state from the Centers for Disease Control and Prevention as of 2011, was originally used in the analysis and found to be a duplicate and inferior measure of health compared with Disabled, so it was removed from the analysis. Summary statistics are provided in Table 1 for each variable used in the final analyses. Table 1 has mean, standard deviation, minimum, median, maximum, skewness, and kurtosis values for each variable. A cursory examination of the data comparing means to medians shows slight differences between the two statistics, supporting the view that symmetry exists in the variable distributions. In addition, skewness measures that might suggest outliers and kurtosis measures of the distribution peaks generally support normal distributions. However, the Alone variable has a somewhat larger negative skewness and a higher positive kurtosis than the other variables, suggesting a possible left-tail outlier and a distribution with a higher peak than normal. Therefore, a more thorough test for influential observations, the Cook’s distance score, was calculated and reported for each regression output. Table 1 Descriptive Statistics Standard Minimum Median Maximum Skewness Kurtosis deviation Score 693 17.2 654 696 720 -0.4 -0.8 Graduate 86.5 3.5 79.6 87.4 91.3 -0.4 -1.1 Mother 6.9 1.0 5.2 6.9 10.0 0.7 0.7 Alone 27.1 2.1 18.7 27.4 31.5 -1.5 4.8 Income 50.6 7.5 39.9 49.4 68.9 0.6 -0.4 Disabled 10.9 2.5 7.3 10.5 17.7 0.8 0.1 Notes. Descriptive statistics are from data for all 50 US states. Score is the average state credit score reported by Fair Isaac Corporation in September 2012. Graduate is the percent of the population 25 years old and older who are high school graduates, taken from surveys done during the years from 2006 to 2010 by the US Census Bureau. Mother is the percentage of households headed by females without a husband present and with children under 18 years of age, taken from the Profile of General Population and Housing Characteristics 2010, US Census Bureau. Alone is the percentage of households with one person from the US Department of Commerce, US Census Bureau, Profile of General Population and Housing Characteristics: 2010 Demographic Profile Data. Income is median state household income from the US Census Bureau, State Median Income, Median Annual Social and Economic Supplement, and Median Household Income by State⎯Single Year Estimates for 2011. Disabled is the percentage of the state population ages 21-64 reporting a visual, hearing, ambulatory, cognitive, self-care, or independent living disability, and summarized by state in the “2010 Disability Status Report” published by the Cornell University Employment and Disability Institute, Cornell University. Variable

Mean

Techniques The techniques used to look for relationships between state average credit scores and state demographic variables include correlation, regression, and mapping. The Pearson product moment correlation is calculated for each pair of variables to measure their linear relationships and to test for significance. Regressions investigate the relationship between the dependent variable, Score, and the other demographic independent variable predictors. Since the dependent variable is continuous, least squares regression is employed. Mapping is done using GIS software to see regional similarities and differences on a map of the contiguous US. States are shaded lighter and darker corresponding to their levels of credit score and demographic factor variable.

1502

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

Results Results were obtained by calculating Pearson correlations and ordinary least square regressions and by doing GIS mapping. Correlations The correlations for variables used in the final analyses are in Table 2. Correlations between the state average credit scores (Score) and the independent variables used to explain those scores are in the first column labeled Score. The variable representing the percentage of adults that are high school graduates (Graduate) has the highest correlation to state average credit scores. Average credit scores tend to be high in states with high percentages of high school graduates. The variable representing the percentage of households headed by a female with children but no husband (Mother) has a high negative correlation to state average credit scores. Average credit scores tend to be low in states with high percentages of households headed by a female with children but no husband. The variable representing the percentage of households with just one person (Alone) has a relatively small positive correlation to state average credit scores. Average credit scores tend to be high in states with a high percentage of people living alone. The variable representing average household income (Income) has a fairly high positive correlation to state average credit scores. Average credit scores tend to be high in states with high average household incomes. The variable representing the percentage of people in the state that are disabled (Disabled) has a fairly high negative correlation to state average credit scores. Average credit scores tend to be low in states with a high percentage of people reporting a disability. P-values are in parentheses below the correlations and indicate that all correlations in the Score column are significant at the level of 1% except Score and Alone, which are significant at the level of 10%. Table 2 Correlations Variable Graduate

Score Graduate Mother Alone Income 0.83 (0.00) Mother -0.79 -0.76 (0.00) (0.00) Alone 0.26 0.12 -0.06 (0.07) (0.41) (0.69) Income 0.62 0.58 -0.42 -0.19 (0.00) (0.00) (0.00) (0.20) Disabled -0.61 -0.57 0.43 0.17 -0.70 (0.00) (0.00) (0.00) (0.23) (0.00) Notes. Correlations are from data for all 50 US states. Score is the average state credit score reported by Fair Isaac Corporation in September 2012. Graduate is the percent of the population 25 years old and older who are high school graduates, taken from surveys done during the years from 2006 to 2010 by the US Census Bureau. Mother is the percentage of households headed by females without a husband present and with children under 18 years of age, taken from the Profile of General Population and Housing Characteristics 2010, US Census Bureau. Alone is the percentage of households with one person from the US Department of Commerce, US Census Bureau, Profile of General Population and Housing Characteristics: 2010 Demographic Profile Data. Income is median state household income from the US Census Bureau, State Median Income, Median Annual Social and Economic Supplement, Median Household Income by State⎯Single Year Estimates for 2011. Disabled is the percentage of the state population ages 21-64 reporting a visual, hearing, ambulatory, cognitive, self-care, or independent living disability, and summarized by state in the “2010 Disability Status Report” published by the Cornell University Employment and Disability Institute, Cornell University.

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

1503

Correlations between pairs of independent variables are in the columns of Table 2 labeled Graduate, Mother, Alone, and Income. The Graduate variable has significant negative correlations to Mother and Disabled and a significant positive correlation to Income. The Mother variable has a significant negative correlation to Income and a significant positive correlation to Disabled. There is no significant correlation between Alone and any other independent variable. In the last column of Table 2, Income is shown as having a significant negative correlation to Disabled. All significant independent variable correlations are at the level of 1%. The numerous significant correlations among independent variables justify using a stepwise regression procedure in addition to a full model regression to see the extent to which regression coefficients are impacted by the inclusion of other independent variables. Regressions Regression output using Score as the dependent variable for a full model and a stepwise procedure using data from all 50 states is in the columns of Table 3. Output is in the full model column and shows that all independent variables are significant at the level of 5% or lower. All of the independent variables have signs consistent with their correlations: Graduate, Alone, and Income have positive coefficients, while Mother and Disabled have negative coefficients. The F-statistic shows significance of the overall model at the level of 1%. The adjusted R-square shows that the independent variables in the full model explain 83.9% of the variation in state average credit scores. Outliers do not appear to significantly influence coefficients, since the highest Cook’s distance measure is 0.24, well below the 0.80 threshold. Table 3 Regressions Independent variable

Full model

Step 2 2.69 (0.00) -6.32 (0.00)

Stepwise procedure Step 3 2.53 (0.00) -6.55 (0.00) 1.51 (0.00)

Step 4 Step 5 1.25 1.51 1.25 (0.02) (0.01) (0.02) Mother -6.83 -6.86 -6.83 (0.00) (0.00) (0.00) Alone 2.37 2.19 2.37 (0.00) (0.00) (0.00) Income 0.49 0.72 0.49 (0.02) (0.00) (0.00) Disabled -1.30 -1.30 (0.03) (0.03) F-statistic 52.2 106.1 70.4 54.7 58.8 52.2 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Adjusted R2 83.9% 68.2% 73.9% 76.7% 82.5% 83.9% Cook’s high 0.24 0.27 0.23 0.48 0.34 0.24 Notes. Regressions use data for all 50 US states to explain state average credit scores. The dependent variable, Score, is the state average credit score reported by Fair Isaac Corporation in September 2012. Graduate is the percent of the population 25 years old and older who are high school graduates, from surveys taken from 2006 to 2010 by the US Census Bureau. Mother is the percentage of households headed by females without a husband present and with children under 18 years of age, from the Profile of General Population and Housing Characteristics 2010, US Census Bureau. Alone is the percentage of households with one person from the US Department of Commerce, US Census Bureau, Profile of General Population and Housing Characteristics: 2010 Demographic Profile Data. Income is median state household income from the US Census Bureau, State Median Income, Median Annual Social and Economic Supplement, and Median Household Income by State⎯Single Year Estimates for 2011. Disabled is the percentage of the state population ages 21-64 reporting a visual, hearing, ambulatory, cognitive, self-care, or independent living disability, and summarized by state in the “2010 Disability Status Report” published by the Cornell University Employment and Disability Institute, Cornell University. The variable coefficients and F-statistics have their p-values below them in parentheses. Cook’s high is the highest Cook’s distance measure for any observation in each regression and must be above 0.80 to indicate an outlier. Graduate

Step 1 4.12 (0.00)

1504

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

Regression output from the stepwise procedure is in the five columns beneath the stepwise procedure label in Table 3. The first independent variable to enter the procedure is Graduate, and although its coefficient is lower in subsequent regressions, it remains significant in the last regression, Step 5. The second variable to enter the procedure is Mother, and its coefficient gets larger through every subsequent regression step. The third variable to enter is Alone and also strengthens in subsequent regressions. The fourth variable to enter is Income, and although its coefficient is smaller in Step 5, it remains significant. The fifth and last variable to enter the stepwise procedure is Disabled. The fifth stepwise regression is the same as the full model. All coefficients in the stepwise regressions are significant at the level of 5% or lower. The F-statistics indicate that all stepwise procedure regressions are significant at the level of 1%. The adjusted R-squares of the stepwise regressions rise monotonically as variables are added. The Cook’s distance measures for the stepwise regressions are all below the 0.80 threshold and provide evidence that outliers are not distorting any stepwise regression coefficients. Mapping A GIS mapping with three attribute queries to depict three new variables created from existing variables appears in Figure 1. These attribute queries give a visual representation of average credit scores and the percentage of adults who are high school graduates in relation to national averages for each state in the 48 contiguous states. Alaska and Hawaii are excluded from the mapping due to their non-contiguous nature, since the visual representation is designed to provide a look at contiguous regions in which credit scores and the demographics coincide. In GIS, attribute queries provide a visual representation of data with the aid of a color shade. The three attribute queries on this map work to provide a color shade distribution of the data. On this map, all three attribute queries are shown, producing a color shade distribution for the three new variables. The three new variables are created by filtering out states which are applicable to each category by using a GIS code formula. These attribute queries classify state average credit scores and the percentages of the adult population in a state with high school graduates in three ways. The first attribute query classification, which provides the first new variable, uses dark gray to shade states with average credit scores and the percentage of high school graduates that are both higher than national averages. The second attribute query classification, which provides the second new variable, uses medium gray to shade states with average credit scores and the percentage of high school graduates that are both lower than national averages. The third attribute query classification, which provides the third new variable, uses light gray to shade states that do not have average credit scores and a percentage of high school graduates that are both above or below national averages. The percentage of high school graduates was used as the paired demographic factor to credit scores, because in the statistical methods section, it had the highest correlation to state average credit scores and was the first variable to enter in the stepwise procedure. The GIS map in Figure 1 that results from using the three new variables illustrates regional differences in the associations between state average credit scores and the percentages of high school graduates by showing clusters of states shaded the same in the upper and lower portions of the map. The upper portion of the map has several clusters of states that show regions where average credit scores and the percentage of high school graduates are both higher than national averages. The lower portion of the map has an unbroken cluster of states where average credit scores and the percentages of high school graduates are both lower than national averages. Only a few states that appear in the northeast quadrant of the map are neither above nor below national averages for both average credit scores and the percentage of high school graduates. Although not

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

1505

shown on the map, Alaska and Hawaii would be shaded dark gray, with each state having average credit scores and percentages of high school graduates above national averages.

Figure 1. Attribute queries showing states by credit score and high school graduate percentages. Attribute queries show states by average credit score and the percentage of adults with a high school diploma. States with above average credit score and high school graduation percentages are shaded dark gray. States with below average credit scores and high school graduation percentages are shaded medium gray. States with neither above nor below average credit scores and high school graduation percentages are shaded light gray. Source: Map by the authors.

Conclusions The demographic impact on credit scores appears to exist. Most of the variation in state average credit scores is explained by just four demographic factors: education, family, income, and health. Individuals and policymakers should be able to take steps to increase state average credit scores if they are able to improve these demographic factors. For example, average credit scores might increase if a state could raise its percentage of high school graduates. Credit scores might also increase if a state could lower its percentage of families headed by single mothers. Another way a state could increase its average credit score is to create opportunities for people to earn higher incomes. Finally, average credit scores might also increase if a state helped with health expenses by offering subsidized disability insurance. To the extent that these demographic factors improve in a state, average credit scores might also improve. When state average credit scores improve, those states should enjoy the benefits of a population that has access to more credit at a lower cost, pays lower auto and homeowner insurance premiums, and has greater employment opportunities. In addition, people with better credit scores should need less free healthcare or discounted services, putting less financial pressure on healthcare providers. Finally, investors who see better average credit scores in mortgage-backed securities will provide funds at lower interest rates, thereby providing mortgage funds at a lower cost to borrowers.

1506

THE DEMOGRAPHIC IMPACT ON CREDIT SCORES

Regional credit score differences are evident. States in the southern part of the contiguous 48 states have credit scores lower than national averages. States in northern sections of the contiguous 48 states have credit scores higher than national averages. A major factor associated with the regional differences is the percentage of adults in them who have high school diplomas. It appears reasonable to suggest that if a state raises its percentage of adults with high school diplomas, it should indirectly be contributing to higher state average credit scores. Higher state average credit scores should be especially helpful to states located in the southern part of the contiguous 48 states. More research is needed. For example, other factors that might influence state average credit scores are wealth characteristics such as savings and investments. In addition, studies are needed at the metropolitan area level. The ideal research would use individual credit scores along with their demographic and wealth data. This study is limited by the use of state average variables. The ecological paradox may apply to this analysis, so caution should be used in using state average relationships to explain the credit scores of individuals. However, previous research by Kabler (2004) did show a high correlation in credit score relationships using both averages and individual data. Therefore, the results from this study should not be entirely dismissed. This study provides some evidence that there is a demographic impact on credit scores, based on results from statistical methods and a visual inspection from GIS mapping.

References Agarwal, S., Skiba, P. M., & Tobacman, J. (2009). Payday loans and credit cards: New liquidity and credit scoring puzzles? American Economic Review, 99(2), 412-417. Bernerth, J. B., Taylor, S. G., Walker, H. J., & Whitman, D. S. (2012). An empirical investigation of dispositional antecedents and performance-related outcomes of credit scores. Journal of Applied Psychology, 97(2), 469-478. Hayashi, F., & Stavins, J. (2012). Effects of credit scores on consumer payment choice. Public Policy Discussion Papers, Federal Reserve Bank of Boston, No. 12-1, 1-47. Jackson, G. (2008). How credit scores can make a difference for your revenue cycle? Healthcare Financial Management, 62(2), 34-37. Kabler, B. (2004). Insurance-based credit scores: Impact on minority and low-income populations. Journal of Insurance Regulation, 22(3), 77-90. Lyons, A. C., Rachlis, M., & Scherpf, E. (2007). What’s in a score? Differences in consumer’s credit knowledge using OLS and quantile regressions. The Journal of Consumer Affairs, 41(2), 223-249. Perry, V. G. (2008). Giving credit where credit is due: The psychology of credit ratings. The Journal of Behavioral Finance, 9(1), 15-21. Robinson, W. S. (1950). Ecological correlation and the behavior of individuals. American Sociological Review, 15(3), 351-357. Spader, J. S. (2010). Beyond disparate impact: Risk-based pricing and disparity in consumer credit history scores. Review of Black Political Economy, 37(2), 61-78. Stachel, D. W. (2004). SEC regulation AB: What an ABS investor wants for Christmas. Journal of Structural Finance, 10(3), 24-28.