What’s Actually Being Proposed?
Typical credit score systems in the USA rely on hard data like the amount of credit you have, your usage of the credit, your number of accounts, and how many times you’ve been late on payments.
Researchers for the IMF are talking about going beyond that. After all, typical credit scoring methods make it hard for people with no credit history to get credit, and more people may become credit risks in a worse economy even if their histories look good.
The researchers describe their proposed solution on the IMF blog:
So, in the future, your online searches, purchase history, and even the browser and device you use to access the internet may be fed to a machine-learning algorithm (what we call “AI”) and used to determine your credit score.
Yes, if you’re using an inexpensive Android phone rather than an iPhone, or if you use Firefox rather than Google Chrome, that might negatively impact your credit score under this proposal.
By the way, this isn’t the first time there were serious proposals to use online activity to determine credit scores. Remember back in 2013 when companies proposed using your Facebook friends to determine your credit score?
It’s worth noting that, as of 2021, this is just a proposal. You can still go view your credit report and you won’t see any browsing history in there. However…
Credit Decisions Are More Than a Single Score
Credit scoring systems are more complicated than many people understand. In the USA, you have three big credit report companies: Experian, Equifax, and Transunion. These reports contain hard data on your credit usage.
There are different ways of “scoring” that data, including different generations of FICO scores. Depending on the type of credit you’re applying for, these models will give different credit score numbers based on the same data. For example, there are different models for mortgages and car loans. Someone might be considered more at risk of defaulting on a car loan than a mortgage, for example.
A bank or company extending credit may run its own credit-scoring model on the data and take into account various factors. Other factors may also be included. For example, LexisNexis offers “Alternative Data” to companies who might want to use that for credit decisions. This includes information like a person’s professional licenses, assets (like owning a home), and “public source data.” It’s pitched as a way for companies to identify credit-worthy people who have thin traditional credit files.
In the USA, the Equal Credit Opportunity Act defines a number of factors that cannot be used for credit decisions:
Credit Scores Used to Include Personal Details, Too
It’s worth noting that credit scores historically included other types of personal information—not just the current “hard” financial details they’re supposed to include—until the system was reformed with laws like 1970’s Fair Credit Reporting Act and 1974’s Equal Credit Opportunity Act.
A Time Magazine article from 1936 describes how the credit reporting system of the day worked. The bolding is ours:
It describes what might happen to a woman who moves across the country:
As you can see, the system included various details about people’s personal lives, which were used in credit decisions.
Of course, the IMF researchers aren’t proposing anything quite like that! They’re just proposing taking into account your online search history and the web browser you use to access the internet. And it will be machine learning algorithms (“AI”) making the decisions.
However, while the system may not have a human banker judging your “personal habits,” AI can still be biased—and is it really right to reject someone’s credit application because they’re using the wrong web browser? (Hey, the researchers are the people who brought up using web browser choice as a metric, not us!)
Bring on the VPNs
In the future, using a VPN might one day be important for maintaining your credit score! Online privacy is incredibly important, but bear in mind that a VPN alone isn’t a silver bullet for protecting your privacy.
