Fair Housing Testing Study in Connecticut Reveals Discrimination Findings

A Review of A Review of Results of Three Years of Fair Housing Testing in Connecticut Results of Three Years of Fair Housing Testing in Connecticut Who conducted the three year study: The CT Fair Housing Center Who paid for the study: Funded by HUD grant dollars When: 2012 2015 What: Complaint-based and systemic housing discrimination investigations across Connecticut by more than 300 tests to determine proactively if illegal discrimination was influencing various parts of the housing market

This report referred to as 3yrs Tester Report in text below concluded: Race in home sales, rental & lending testing: More than 75%, 55% and 50% of the tests showed differential treatment with the African-American tester receiving less favorable treatment , respectively. National origin rental testing: Latino was treated less favorably than white in 61% of the state and in 73% of the tests in New Haven county Deaf and Hard of Hearing Rental Tests: Differential treatment toward people who are deaf or hard of hearing in 27%of the tests Independent living testing: in 71% of the test, housing providers made it clear that they would not make housing available to a person unless she could prove she had the ability to live independently. Rental transgender testing: less favorable treatment of the transgender tester in 100% of the tests CTFA thus concluded people of color still face significant obstacles to enjoying equal access to housing in Connecticut Words in are a direct citation from CTFH 3yrs Tester Report

CONSEQUENCES:This test report, if found credible, will have a serious impact on Realtors, Housing Providers, the Banking and Insurance Industries In CT, a legal accusation can arise from testing results only, e.g., there doesn t need to be a real victim but the accused will typically need pay tens of thousands dollars or more to settle a claim of discrimination and to defend themselves This fosters an unfavorable environment for business and will significantly worsen the already anemic environment for real estate in CT Discrimination is a serious accusation but everyone belongs to a certain category; race, color, national origin, religion, disability, marital status, age, source of income, gender identity or family status . Questions to Ask: 1.Is this report technically sound and scientifically credible? 2.Who are the real victims of discrimination claims? Words in are a direct citation from CTFH 3yrs Tester Report

3yrs Tester Report violated the most basic design principle of an un-biased and fair survey: Conflict of Interest Settlement income / legal service fees counted for the one of the most important revenue streams in 2013, $517,314 based on CTFH annual report. There is an obvious financial conflict of interest that CTFH and its own staff organize, execute and analyze the discrimination survey. CT law allows allegation based on 100% only tester results. As stated in this report, p13, 46% of the testing cases were filed as claims and CT Fair Housing Center collected more than $1 million dollars! All testers involved in this report are paid an undisclosed amount according to CTFH website Did they get paid a Percentage of Profits and thus have an incentive to fabricate results? Imagine a drug company using their own lab testing to certify their new drug for clinical use. This is what is happening with the Fair Housing Tester Program in Connecticut. Words in are a direct citation from CTFH 3yrs Tester Report

3yrs Tester Report: Problems with test design regarding Tester What is a Tester: People recruited/trained/paid by CTFH. A CTFH staff person, the testing coordinator will assign protected tester , or PT, which is given characteristics slightly more qualified and a characteristic of protected class . The characteristic of protected class is something fictional (e.g., family status), but most cases have to be real (e.g., race, gender, age, nation of original) for an interview type of testing. And a control tester, or CT who does not belong to that protected class. Tester are preconditioned thus test may become highly hypothesis driven.CTFH s recruitment webpage starts by recruiting testers to measure and document differences and determine whether people are being treated differently based on their membership in a protected class . All Protected classes are listed on their recruitment page. It is reasonable to assume more conditioning will happen during actual training. A classical example of preconditioning: Two identical cakes, if priced differently, consumers will overwhelmingly rate the more pricy one as tasting better. Because they are pre-conditioned that higher price = better quality . The CTFH precondition equation goes: protected class = discrimination in Housing . NOTE: in many tests that CTFH conducts, it relies on a narrative from tester after the interaction.

3yrs Tester Report: Problems with test design regarding Tester Testers characteristics is far more complex than CTFH parameters. Each conversation can be different; different conversation is insufficient to establish discrimination existed. A Large body of this report detects differential treatment based on the interaction with a agent and/or computer system. Testers, even if we assume went through proper training and calibration (which is not based on this report), are still distinct individuals that their characteristics are more complex than CTFH parameters such as credit , income , debt ratio and membership of a protected class . Did they always ask the exact same questions the exact same way with exact same attitude? Do they have any previous personal experience which will affect their interaction with a housing professional? Did they wear the same attire? Some alleged differential treatment such as agent pointed out the locations of homes where friends live , or given a breakdown of closing cost in writing , or told more about amenities can be a very random act as each conversation can be different. Unless it is mandated that we have to use robots on the receiving side of the test, difference like this will always exist and it is a far-stretch to conclude this is due to the membership of protected class. How different individuals can have very different evaluation of the same object: A child and an adult went to the same amusement park at the same time, took exactly same rides, their description of the park can be completely different. A person had a dog attack (bite) them as a child, will rate the friendliness of the same dog significantly lower than another person that was never bitten by a dog. It is not that the object is different; the observers themselves are different.

3yrs Tester Report: Problems with test design regarding Tester The most severe design flaw is the extremely small number of testers. The report disclosed that in some categories of testing, the number of the testers is extremely small. For instance, for gender identity testing, there is only ONE PT and the conclusion is 100% DISCRIMINATION. Such a shockingly small sample size has close to zero statistical value. Classical examples of proper design sample size: If you want to know whether there is any difference of running speed between boys and girls, you don t ask one boy and one girl to run 100 times, average their time and compare. Instead, you need to have 100 boys and girls run, average their time and then compare. CTFH design is essentially asked one pair of boy/girl to run 10 times, and concluded 100% of the time, one runs faster than the other.

3yrs Tester Report: Problems with data analysis: standard deviation Most of the numerical statements in the report, when validated, make no statistical sense. One such example, a quote from the report, 40% of the tests the African-American tester was told she qualified for a smaller mortgage or a higher interest rate even when she had more income than the white tester . There are multiple problems with this statement to support discrimination happened based on race. 1. At any given time, mortgage rate is a variable fluctuate within some range, as evident in the plot below. In other words, higher or lower may not be different. 2. Does this statement say in 60% of the tests, white testers are given higher rates? Which probably make sense that white tester ends up having a higher chance of given a higher rate, though maybe none are statistically different, since white tester has lower income. So, where is the discrimination?? Classical examples of standard deviation: Mortgage rate went down in 2014 as shown in this plot. However, at any given time, a fluctuation of a 0.1% point up or down is not real difference. A cookie package labels 100g, but it actually weights 99.6g. You can not sue the cookie maker since that is within the range of standard deviation.

3yrs Tester Report: Problems with data analysis: Simply Wrong Interpretation Simply wrong data interpretation. There are many numerical datasets in the report that simply contradict with the statement/conclusion. For example, figure 7 on p18. PT (African-American with slightly higher income/credit) was shown 10% more in area higher in white population than CT census; 7% less in severely segregated in white and 4% less in severely segregated in black. Another fact the report stated is African- American testers were sent nearly twice as many listings (306) as White testers (168). So, if PT was sent twice more listings (probably due to the fact that higher credit/income and other un-controlled variables) (Continued on the next page) Breakdown of Listing sent to PT & CT by % of white Breakdown of House shown to PT & CT by % of white CT census 72% white Less white CT census 72% white Less white More white More white ~ 10% ~ 20% ~ 10% ~ 4% ~ 7%

3yrs Tester Report: Problems with data analysis: Simply Wrong interpretation Simply wrong data interpretation. (cont. from last page), more listing in white area (though difference might be too small to be significant), and was shown significantly higher percentage in white area (again, may due to the higher credit/income PT has), HOW DID CTFH DRAW THE CONCLUSION THAT 75% OF THE TIME, DISCRIMINATION TO AFRICAN AMERICAN HAPPEN? CTFH concluded discrimination occurred! Quote report: if the real estate agents were looking solely at objective characteristics, there would be more similarity in the homes shown to both testers? But did CTFH just said PT and CT are different that PT , generally had slightly more income, a slightly better credit score, and was pre-qualified for a slightly larger mortgage than that of the CT . Classical examples of wrong interpretation; Calling a Stag a Horse : A Chinese folktale says an extremely powerful dictator had a Stag that he shows to everyone but says: Take a look at my Horse, BUT none dare to say to the Dictator; No, this is a stag

Questions to Ask: 1.Is this report technically sound and scientifically credible? NO 2.Who are the real victims of CTFH discrimination investigations?