Are there publicly available data to demonstrate the predictive power of credit scores?

Question

Credit reference agencies assign and distribute a credit score for individuals based on their financial behavior. The implication is that this score has predictive power, such that an individual with a higher score is less likely to default on a debt than one with a lower score.

Are there any publicly available data where this relationship can be examined? Have the credit reference agencies (or anyone) published any peer reviewed literature where this relationship has been quantified? Does anyone claim to have quantified this relationship in any verifiable numerical way? The best would be data that allows the individual features to be examined (eg. compare the effect of defaulting on a loan to not having had a loan), but any data that is available would be great.

My personal interest in this is from the point of view of someone who has lived within their means for much of their life and not borrowed money, and has a lower credit score than my peers who have utilised consumer credit to fund their lifestyle.

Answer 1

Probably the most accessible source of this sort of data (at least in the US) would be to look at the various reports that the government-sponsored enterprises (GSEs) that hold a large fraction of American mortgages publish. For example, if you open this FHA Single Family Loan Performance Trends PDF and go to page 4 (page 5 of the PDF including the cover page), you'll see a chart that lists the different delinquency rates of loans held by FHA broken down by credit score as of April 2021. The HUD site has a similar report for each month so you can see how the trend has changed over time if you want.

Of course, this is only information about mortgage loans which we generally expect to be the last thing someone would fail to pay. Default rates on other types of consumer credit will be different (though it is almost certain that people with higher credit scores, on average, default less often).

Note, though, that companies really only use credit scores in wide ranges. FHA, for example, buckets together people with credit scores between 720 and 850 when they report information. No lender cares about small differences in credit scores-- they're going to give the same rate to someone with a 760 score as an 850 score. If your "lower credit score" is still in the excellent range, that's effectively the same as someone with a "perfect" score.

And note as well that credit scores are intended to be predictive for groups not for individuals. If you have a lower score because the credit bureaus literally lack enough information to be able to assess your risk-- you've never had a credit card or a car loan or a mortgage or a student loan-- that lack of information is going to look risky to the credit bureau who has no insight into your finances if no creditor is reporting any information. Someone looking at your assets, bank account, etc. may be able to see that you're quite unlikely to default even if the credit bureaus can't see that. It's similar to the way that body-mass index (BMI) makes sense when applied to groups (BMI > 30 is obese, for example) but falls short when applied to certain individuals (Dwayne "The Rock" Johnson is incredibly fit and incredibly not obese but has a BMI of ~35).