May 24, 2021

Re: Executive Summary of Racial Discrimination in Lending in San Diego County

To San Diego County Lenders, Regional policymakers, and Residents,

Financial discrimination on the basis of race is rampant in most San Diego communities. The NAACP San Diego Branch reviewed publicly available federal loan disclosures released under the Home Mortgage Disclosure Act (HMDA) and Community Reinvestment Act (CRA), and found lending patterns for home mortgages and small business loans consistent with racial discrimination. A statistical analysis of loan data shows very troubling patterns of racial discrimination across a wide range of incomes to the disadvantage of people of color, including evidence of racial steering by mortgage originators, and a disparity for business investment in communities of color, including southeast San Diego, Chula Vista, National City, and San Ysidro. To read the full report, please see: Racial Discrimination in Lending in San Diego County

To each of our addressees we have the following specific messages:

To Lenders

We have analyzed and evaluated banking practices and found racially discriminatory lending practices. Please read the full report for more details.

To Regional policymakers and elected officials

Where the federal government leaves off, we are asking you to correct the systemic injustices we highlight in our report by using City and County funds to increase Black homeownership rates in the City and County and help eliminate racial financial discrimination.

To Residents

We want our report exposing the racial inequities in our communities to motivate you to require fair and equitable practices from your bank and your elected officials.

 

Sincerely,

 

Francine Maxwell, President
NAACP San Diego Branch

 

Derrick Luckett, Chair
NAACP San Diego Branch Housing Committee

Michael Gallaspy, Co-chair
NAACP San Diego Branch Housing Committee

PO Box 152086
San Diego CA 92195

(619) 431-1633 Phone/Text
info@sandiegonaacp.org
www.sandiegonaacp.org

Celebrating 102 years of civil rights advocacy in San Diego

Racial Discrimination in Lending in San Diego County

May 24th, 2021

 

To lending institutions in San Diego County, elected officials and policymakers for regional jurisdictions, and residents: financial discrimination on the basis of race is rampant in most San Diego communities. The NAACP San Diego Branch reviewed publicly available federal loan disclosures released under the Home Mortgage Disclosure Act (HMDA) and Community Reinvestment Act (CRA), and found lending patterns for home mortgages and small business loans consistent with racial discrimination.

Where a person lives is a key issue impacting all aspects of their life, yet communities in San Diego County continue to be subject to racial discrimination in mortgage lending. This letter is not our final word on the matter, but rather a declaration that we will hold lenders accountable for their complicity in perpetuating financial racism and we will hold elected officials accountable for their inaction in addressing this problem.

As Dr. Vanessa Gail Perry notes in the online report 2020 State of Housing in Black America:

 

“Homeownership enables families to build wealth, helps stabilize communities, and has been linked to access to educational, employment opportunities, increased safety as well as physical and mental health… Despite the emphasis on this aspect of the ‘American Dream’ in popular culture and among scholars and policymakers, Black and other minority families have faced a number of barriers to homeownership, most of which can be traced to cumulative disadvantage and structural inequalities.”

 

Despite federal laws on the books for decades racial discrimination and disparities in lending persist to this day in both the home mortgage and small business loan credit markets. Academics have long noted racial discrimination in lending nationally (Faber) (Bates and Robb) (Blanchflower et al.) (Quillian et al.). In this letter, we review the data to show discrimination occurs here in San Diego.

This letter is in two parts and addresses two different credit markets: home mortgages and small business loans. These categories were chosen because loan data is publicly available from the Federal Financial Institutions Examination Council (FFIEC). Inspired by similar national analyses, we conducted an original statistical analysis of loans in San Diego County, the first of its kind to our knowledge. For small business loan disclosures, we combined this with census tract-level demographics from the Census Bureau’s American Community Survey (ACS), particularly 5-year estimates for 2019.

To each of our addressees we have the following specific messages:

To Lenders

We have analyzed and evaluated banking practices and found racially discriminatory lending practices.

To Regional policymakers and elected officials

Where the federal government leaves off, we are asking you to correct the systemic injustices we highlight in our report by using City and County funds to increase Black homeownership rates in the City and County and help eliminate racial financial discrimination.

To Residents

We want our report exposing the racial inequities in our communities to motivate you to require fair and equitable practices from your bank and your elected officials.

Mortgages

Attitudes about racism have shifted in America over time from being socially acceptable to less so. The forms that racial discrimination takes in mortgage lending have shifted as well. Forms of racial discrimination in mortgage lending include overtly racialized and government-sanctioned redlining (Rothstein) that resulted in Black Americans being denied conventional loans, and steering by real estate professionals to lead prospective buyers away from or towards particular neighborhoods on the basis of race. Although redlining is illegal, many overtly and increasingly subtle racially discriminatory practices persist today (Choi). In practice, the effects of racial discrimination can be measured in mortgage disclosures even when individual acts of discrimination may go undetected or undocumented; Quillian et al. note regarding mortgages that “racial gaps in loan denial have declined only slightly [from the late 1970s to the present], and racial gaps in mortgage cost have not declined at all, suggesting persistent racial discrimination.”

 

Figure 1: Single family home mortgage approval rates in 2018 and 2019. We group people of Hispanic/Latino ethnicity and any race into one category, and consider non-Hispanic Asians, non-Hispanic Whites, and non-Hispanic Blacks separately. People who identify as two or more races, and those who identify as other races are considered separately as well.

 

We analyzed HMDA mortgage applications for San Diego County for 2019 and 2018. We considered applications for single family homes and first liens, which includes over 174,000 applications that were either approved or rejected (withdrawn or incomplete applications are not considered). In San Diego County the racial gap in mortgage approval rates appears at statistically significant levels for all race/ethnicity categories compared to Whites, except for Asians who enjoy similar approval rates to Whites (Figure 1). Whites enjoy an approval rate of 85%, while Blacks have the lowest approval rates at about 74%. Two big factors in mortgage applications are income and credit score; HMDA disclosures do not include credit scores, but do include an applicant’s income; therefore, we may answer the following question: For Blacks in San Diego, is the discrepancy in approval rates explained by differences in income levels? Figure 2 shows that it is not, as Blacks and Whites of similar income levels still exhibit statistically significant differences in approval rates; the same is true for Hispanics.

 

Figure 2: Applicants are split into equal-sized income quintiles; each quintile represents 20% of applicants. At all income levels, the difference between White approval rates and those of Blacks or Hispanics are statistically significant.

 

Finally, we considered whether the approval rates could be explained by where mortgage applicants want to live, asking if the discrepancy in approval rates is explained by location and income jointly? In principle Blacks and Hispanics could be applying for mortgages in highly competitive areas at higher rates than Whites, with the effect of driving down their approval rates. In San Diego County, 32 census tracts showed statistically significant differences in White and Black approval rates at various income levels to the disadvantage of Blacks, with an average difference in approval rate of 34% (figure 3). Using the same criteria, only 2 cases showed statistically significant differences in favor of Blacks. There are 113 census tracts where Whites enjoy statistically significant higher approval rates than Hispanics; the average difference in approval rate is 25%. There are 30 tracts where Hispanics have higher approval rates than Whites (figure 4). In 8 tracts White and Hispanics are alternately advantaged dependent on income levels. For example in tract 20.109 considering applicants making between $73k and $102k annually, Whites have an approval rate of 65% compared to 90% for Hispanics; at $102k to $137k annually however, Whites have an approval rate of 95% compared to 69% for Hispanics. The results suggest racial steering by loan originators.

 

Figure 3: Census tracts in San Diego County where there are statistically significant differences in SFH mortgage approval rates for Whites compared to Blacks, at various income levels. We consider a tract to exhibit a “racial advantage” if for at least one income quintile there is a statistically significant difference in approval rates. (Some such tracts are not depicted here, as they fall outside of the map bounds.)

 

Figure 4: Census tracts in San Diego County where there are statistically significant differences in SFH mortgage approval rates for Whites compared to Hispanics, at various income levels. We consider a tract to exhibit a “racial advantage” if for at least one income quintile there is a statistically significant difference in approval rates. For several tracts there was a statistically significant difference in approval rates that favored Whites and Hispanics differently at different income levels; for such tracts the racial advantage is dependent on income level. (Some tracts are not depicted here, as they fall outside of the map bounds.)

Small Business Loans

The Community Reinvestment Act (CRA) is a federal law that encourages depository institutions to meet the credit needs of the communities where they operate. The law mandates the collection of data on loans and assigns compliance grades that influence how an institution’s application for deposit facilities is perceived. All institutions regulated by the OCC (Office of the Comptroller of the Currency), Federal Reserve System, and the FDIC that meet the asset size threshold are subject to data collection and reporting requirements; for 2019, agencies with $1.284 billion in assets need to report. Not all banks are evaluated every year; in 2019 grades were assigned to 1,483 banks, with the results summarized in table 1. Remarkably, only 1% of institutions nationwide received the grade Needs to Improve while none received a grade of Substantial Noncompliance; all others received passing grades of Outstanding or Satisfactory. These passing grades are deceptive because federal regulators are not using race as one of the factors when grading banks.

CRA loan disclosures track small business loans of less than $1 million by census tract and include the median income of the tract as a percentage of the area median income, a useful economic indicator. However in contrast to HMDA disclosures they do not include demographic information about loan applicants, therefore we combined the loan disclosures with tract-level demographic information from the Census Bureau. As the stated aim of the CRA is to encourage investment in the community, we ask whether banks in San Diego County are investing in tracts in a racially equitable manner? In order to answer this question, we analyzed CRA loan disclosures from 2019. We drew maps of the region and colored census tracts with color scales indicating number of loans and racial/ethnic demographics of tracts. These maps illustrate where loans are going and where they are not, and starkly show that many communities of color are not the beneficiaries of these investments (figures 5, 6, 7 and 8).

Having identified gross trends using these maps, we again compare tracts at similar income levels but with different racial demographics in detail, to answer the question: does income alone explain the observed racial disparities? In order to control for population, instead of the number of loans or the total loan amount per census tract, we consider a derived value by dividing the total loan amount for a census tract by the population. We call this value the per capita loan amount (PCLA); it represents the number of dollars per person invested in a community through small business loans.

Moreover half of all the tracts we looked at in San Diego County have a non-Hispanic Black population of less than 2.8%; for simplicity we compare tracts having a Black population above or below this threshold at different income levels as measured by percentage of median family income (MFI). At many income levels there is a discrepancy to the disadvantage of Blacks; the discrepancy is statistically significant at two different income levels (figure 9). For other income levels the statistical test was inconclusive, but we again caution that this does not mean there was no discrimination; see our statistical note at the end of this letter for more detail.

 

Table 1: CRA grades for banks in 2019.

CRA grade in 2019

Number of institutions

% of total institutions

Satisfactory

1303

88%

Outstanding

160

11%

Needs to Improve

20

1%

Substantial Noncompliance

0

0%

 

Figure 5: Map centered on southeastern San Diego. Census tracts are colored according to the number of loans originated for businesses in those tracts. Most areas in southeastern San Diego have around 50 loans or fewer.

 

Figure 6: Map centered on southeastern San Diego. Census tracts are colored according to the percentage of residents who identify as non-Hispanic Black. Compare to figure 5; those areas with higher percentage of Black residents are the same areas receiving the fewest number of loans.

 

Figure 7: Map centered on south San Diego County. Census tracts are colored according to the number of loans originated for businesses in those tracts.

 

Figure 8: Map centered on south San Diego County. Census tracts are colored according to the percentage of residents who identify as Hispanic or Latino. Compare to figure 7; those areas with a higher percentage of Hispanic or Latino residents are the same areas receiving the fewest number of loans.

 

Figure 9: Median per capita loan amounts (PCLA) for census tracts at various income levels. The median percent of non-Hispanic Black residents for tracts considered is 2.8%. We split all the tracts we considered into two equal-sized categories: above and below the median percent of Black residents. The graph shows that tracts with a greater percentage of Black residents (blue bars) typically have lower PCLAs than tracts with a lower percentage of Black residents (orange bars). In other words, fewer dollars per person are invested in those communities through small business loans. This discrepancy is statistically significant (p < 5%) for 100% to 110% MFI and 130% to 140% MFI income levels.

A note on the statistical tests used

This section is for those who want to better understand the meaning of the phrase “statistical significance”, or who have a good understanding of statistics and want to see the details of our claims. We briefly provide some background on probability and statistics and summarize the numerical results.

A binomial random variable models any situation that has only two outcomes (e.g. flipping a coin to get heads or tails). Any fair coin has a 50% chance of producing heads or tails, but an unfair coin will produce heads or tails more than 50% of the time. This difference between a fair coin and an unfair coin can be observed by flipping it many times and seeing whether you get close to 50% heads and 50% tails; if it deviates significantly from this then most likely the coin is unfair. Coins that are more unfair (e.g. a coin that produces tails 90% of the time) are easier to detect, but unfair coins with less bias are harder to detect and require more trials to make a conclusive determination.

All of this applies to our example in the following way: we are assuming that getting a loan versus not getting a loan is similar to flipping a coin, and by observing many different outcomes we can determine what the odds for loan approval or rejection are, similar to observing the odds of heads and tails when flipping a coin. Banks may use different “coins” for different populations; it’s legal to use a different “coin” based on an applicant’s income, but illegal to use a different “coin” based on protected characteristics like race. If the banks were racially impartial, then the observed odds should be nearly the same when considering Blacks and Whites with similar incomes. As we have seen, this is not the case at many different income levels; there are enough trials that we can say with a high degree of confidence that the metaphorical “coins” used for Whites and Blacks have different odds for mortgage approval. In some cases there weren’t enough mortgage applications to conclusively say they were different, but the presence of even one “unfair coin” casts significant doubt on the impartiality of banks in general.

Formally speaking, for mortgages we modeled mortgage approval rates as a binomial random variable and estimated by maximum likelihood the probability of being approved for a mortgage for the various classes discussed. One may then ask, what is the probability that the observed outcomes for one class would arise from the distribution implied by another class? For example, Whites in San Diego County have an approval rate of about 85%; Blacks applied for 4005 loans of which 2958 were approved (74%). The probability of observing this outcome for Blacks if we assume they have the same chance of approval implied for Whites is vanishingly small; far less than one in a million or even one in a billion, the chance is so small that most people don’t have the vocabulary to even name the number (p-value of 7.173×10-69). Hence we say there is a statistically significant difference. We use a two-sided exact binomial test, with a significance level of 5%. Table 2 summarizes the results of some of the statistical tests for mortgage approval rates discussed in this letter; others are omitted for brevity.

CRA disclosures for small business loans do not include data on rejected applications, therefore we can’t model this as a binomial random variable as with mortgage applications. We must employ a more sophisticated statistical technique. Roughly speaking, our model is closer to a die roll for census tracts that determines the PCLA; most dice have six possible outcomes, and while PCLA has many more possible outcomes it is basically analogous. With enough trials we can determine whether banks are using the same “dice” for tracts that are more or less Black, but at the same income levels. Again, if the banks were racially impartial then we would see similar outcomes when comparing tracts that have similar income levels, regardless of race. As we have seen, this is not the case, which casts more doubt on the racial impartiality of banks.

For census tracts at various income levels, we consider the two distributions of PCLA implied by the data for two classes of tracts; those above and those below the median percentage of Black residents (2.8%). We then apply the two-sided Mann-Whitney U test to determine if the distributions are the same, with a significance level of 5%. The results are summarized in table 3. For small business loans, we excluded from our analysis 35 (5.57%) of 628 census tracts that were exceptional compared to a typical tract; these tracts obscure the general trends that we are interested in exploring. In particular we excluded tracts with a PCLA of more than $3055/person. Figure 10 shows the distribution of PCLA for all census tracts in San Diego County.

 

Figure 10: Histogram of per capita loan amount (PCLA) for all census tracts in San Diego County. The vertical red line indicates the PCLA threshold ($3055/person) above which we exclude tracts from our statistical analysis.

 

 

Table 2: Some statistics on mortgage approval rates discussed in this letter. The p-values reported correspond to an exact two-sided test of the given samples arising from the maximum likelihood binomial distribution implied for Whites at comparable income levels.

Class

Applications

% Approved

p-value

White, overall

82207

84.6

Black, overall

4005

73.9

0.000*

Hispanic, overall

20079

78.9

0.000*

White
$1k to $73k annual income

10208

70.8

Black
$1k to $73k annual income

690

60.1

0.000**

Hispanic
$1k to $73k annual income

4061

68.1

0.000**

White
$199k and above annual income

17239

88.3

Black
$199k and above annual income

248

82.1

0.002***

Hispanic
$199k and above annual income

1372

82.5

0.000***

* Compared to Whites overall

** Compared to Whites at $1k to $73k annual income

*** Compared to Whites at $199k and above annual income

 

Table 3: Summary of per capita loan amount (PCLA) values for census tracts in San Diego County at various income levels. At each income level we consider two classes of tracts above and below the median percentage of Black residents (2.8%) and apply the two-sided Mann-Whitney U test. The U-statistic and p-values are reported.

Income level

Number of tracts <2.8% Black residents

(class 1)

Number of tracts >=2.8% Black residents

(class 2)

Median PCLA for class 1

($/person)

Median PCLA for class 2

($/person)

U-statistic

p-value

40% to 50% MFI

7

16

351

249

40

0.300

50% to 60% MFI

11

23

145

263

163

0.185

60% to 70% MFI

12

29

297

265

161

0.720

70% to 80% MFI

12

37

379

249

199

0.601

80% to 90% MFI

22

26

253

208

268

0.717

90% to 100% MFI

15

36

307

259

208

0.204

100% to 110% MFI

18

26

625

315

151

0.049*

110% to 120% MFI

20

26

566

566

248

0.799

120% to 130% MFI

26

23

504

285

220

0.116

130% to 140% MFI

117

54

640

430

2229

0.002*

* Statistically significant results.

Authors

Michael Gallaspy,* William Rudebusch, Carol Spong,* Rachel Forester*

*NAACP San Diego Branch Housing Committee members

Acknowledgements

Thanks to the NAACP San Diego Branch Housing Committee and its chair Mr. Derrick Luckett.


Bibliography

Bates, T., and A. Robb. “Has the Community Reinvestment Act increased loan availability among small businesses operating in minority neighbourhoods?” Urban Studies, vol. 52, no. 9, 2015, pp. 1702-1721.

Blanchflower, David G., et al. “Discrimination in the Small-Business Credit Market.” The Review of Economics and Statistics, vol. 85, no. 4, 2003, pp. 930–943.

Choi, A., et al. Long Island divided. Newsday, 2019, https://projects.newsday.com/long-island/real-estate-agents-investigation/. Accessed 13 5 2021.

Faber, Jacob William. “Segregation and the Geography of Creditworthiness: Racial Inequality in a Recovered Mortgage Market.” Housing Policy Debate, vol. 28, no. 2, 2018, pp. 215-247.

Perry, V., et al. “2020 state of housing in Black America.” 2020, https://shiba2020.com/. Accessed 13 5 2021.

Quillian, L., et al. “Racial Discrimination in the U.S. Housing and Mortgage Lending Markets: A Quantitative Review of Trends, 1976–2016.” Race Soc Probl, vol. 12, 2020, pp. 13–28.

Rothstein, Richard. The Color of Law: A Forgotten History of How Our Government Segregated America. First edition ed., New York ; London, Liveright Publishing Corporation, a division of W.W. Norton & Company, 2017.

 
County Department of Homeless Solutions and Equitable Communities

As of July 1st, the Health and Human Services Agency has established a new department and four offices. The Department of Homeless Solutions and Equitable Communities will focus on achieving better coordination of existing and new County homeless and equitable community efforts and serve as a central point of collaboration for outside partners to ensure equity among all San Diegans and reduce homelessness in the region.

Read More »
Reject Amendments to the Housing Element

The authors of the appendix evidently hope that by asserting this figure is “conservative” they will avoid genuine scrutiny. No historical evidence has been supplied to justify this claim. No estimate of the current rate of redevelopment is provided. No compelling argument has been made that the 90% capacity will be fulfilled by 2029.

Read More »