A thread on credit reports, gender, and race. Some examples of how bias in 1 lending decision can persist and even exponentially impact a person throughout their entire life. In November, @dhh, a leading tech voice, shared an experience he and his wife had. /1

When @dhh and his wife both applied for the Apple card, he received 20x the credit limit as his wife. The only difference was he was a man and she was a woman. Others were quick to share similar experiences. /2

Because of my background, my brain starting churning. I spent the early part of my career at large banks predicting loan defaults (when borrowers don’t pay back their loan). In theory, data is *supposed* to remove bias because it’s blind to gender or race. /3

The way you predict payment default is by taking *seemingly* unbiased data points and using statistics to predict who is more likely to default on their loan. Banks use these predictions to assign interest rates, credit limits, and even to approve or deny loan applications. /4

An example: Customer A has a higher risk of default and receives a 6% interest rate. Customer B has a lower risk of default and receives a 4% interest rate. /5

The bank is trying to offset the higher risk of default with a higher interest rate (more revenue) to cover the costs of the customer never paying back the loan. /6

The Fair Housing Act (FHA) and the Equal Credit Opportunity Act (ECOA) protect consumers by prohibiting unfair and discriminatory practices. Banks can’t have one treatment for women and one for men. The same applies for race and other factors. /7

Banks also cannot explicitly use these factors in any algorithms they create that determine interest rates, credit limits, or loan approval. I started to think about the information that is available to banks and how they use that information to develop these algorithms. /8

How might this information be biased against protected classes. How might previous decisions by OTHER institutions reverberate (even exponentially) through a consumer’s credit life. Let’s start with the credit limit issue that @dhh had. /9

But let’s assume that happened in the year 2010 instead of last year. Say he received a credit limit of $20,000 and his wife received a credit limit of $1,000 because an organization DID explicitly discriminate. /10

If they both use the card to buy a $1,000 computer, they have the same balance and the same impact to their credit, right? /11

Wrong. One of the biggest factors in determining a credit score is “utilization”. Utilization is your balance divided by your credit limit. @dhh would have a 5% utilization with his credit limit ($1,000/$20,000) but his wife would have a 100% utilization ($1,000/$1,000)! /12

This will impact her credit score negatively. So now if they both apply for another loan in 2014, she will have a higher interest rate and a higher payment. This will impact her the amount she can save and the amount that she can use as a down payment on a house. /13

Her lower down payment now creates another higher interest rate event. You can see how these things can start to add up quickly and can become a chain of events. /14

That’s the IMPACT of the decision by Apple. The cause of the decision is unknown but there is one explanation I can think of. I know banks look at your highest existing credit card limit and will match it for marketing purposes. /15

This is overriding the blind data driven credit decision with a data driven marketing/revenue decision and can result in bias which contributes to unintended negative credit consequences. BTW the point is not to guess the cause. /16

These issues can even persist over multiple generations. I was born in 1981. I’m a white guy. My parents bought a home in 1980. Now assume my parents have the exact same salary and credit profile of another couple who also had a son, John, in 1981. /17

The only difference is John’s parents were black. When John’s parents also tried to buy a home in 1980, a bank discriminated and declined their mortgage application. They were not able to buy a home. They decided to rent instead. /18

Fast forward 30 years. My parents were able to build equity in their home, and were not subject to rent increases, etc. John’s family was not able to build equity and save as much because of rent increases. 30 years would have been 2011, the middle of the housing crisis. /19

If John and I (born in the same year, with the same family financial situation when we were born) now both had the same job/same income, but because of the recession both lost our job. What might happen? /20

I’d want to do what I could to save MY house. After I exhaust all my options, maybe I ask my parents for money to get me over the hump and pay my mortgage for 3 months. And they have it! Because they were able to build equity in their home and save. /21

John does the same thing. His parents give them what they have but its not enough the cover his mortgage for 3 months. He falls behind and loses his house and the equity he had. /22

When John and I apply for a credit card in 2020, the bank pulls both our credit reports. They see *unbiased* data because they don’t know I’m white and John is black. But they ALSO don’t see that our are situations were affected by PREVIOUS bias. /23

There are multiple ways these issues can materialize into credit reports including discrimination in income, layoffs, hiring, loan approval, and credit limits. NO credit data is independent of bias when you look at it like this. /24