Discrimination in Loan Applications
What determines whose loan applications are accepted or denied? I’ve taken a look at a dataset released by Lending Club, an American marketplace for loans, to find out.
In this dataset, rejected applications include the applicant’s zip area; issued loans include both the applicant’s zip area & the loan status (fully paid or delinquent). With some counting we can show which area you live in affects both the likelihood your loan application will be successful - and if accepted, the likelihood you’ll fully repay the loan.
Which demographics live in the zip areas where loan applications are regularly denied? Are they the same demographics that fail to repay their loans? Because if they aren’t, something fishy is going on.
I cross-referenced the loan data with census data (age/race/gender by county), and for each demographic compared the average bias (distance from the trend line) for counties with a relatively low saturation (the bottom 25%) of that demographic to counties with a high saturation (top 25%).
Within this dataset there are no statistically significant biases affecting Whites, Blacks, Native Americans or anyone younger than 20. However if you live in a county with a larger than average concentration of Asians, Hispanics or Biracial people and apply for a loan you are 2% more likely to have your application accepted (p < 0.000003). Age also has an effect on acceptance rates.
It’s impossible for me to explain why these biases exist. Perhaps Lending Club has poor judgement, uses predatory sales tactics, or is actively helping some disadvantaged populations - I simply don’t know. Additionally an applicant’s county is an indirect proxy for age/race/gender, perhaps the presence of a certain demographic correlates with someone’s level of education & Lending Club is discriminating against, for example, school dropouts.
Your zip code tells a bigger story than you might expect - and the same way I used zip codes to measure bias large corporations use proxies to help determine your worth, and financial options. A friend recently told me about her own loan application - which was denied because other borrowers in her area were behind on their payments (she asked the broker). It sickens me that institutions judge us without knowing who we are, on the basis of indirect measures, and then hide behind flawed statistics to justify them.
Making the world a better place doesn’t matter unless you give everyone a chance to enjoy it.
For more see this post’s repository and/or read Weapons of Math Destruction by Cathy O’Neil.
