The main purpose of the study was to learn how well credit scoring worked and whether it was discriminating against certain groups (perhaps because key factors were strongly related to sex or race). The cool part from a statistics viewpoint is that they made up their own credit scoring model in order to evaluate those offered by companies for bias. Of 312 credit-related factors, they got down to nineteen that appear to make a noticeable difference to predictions - and for most people, only half of those really matter.
So what was found? Here's the executive summary:
In general, credit scores work well to identify the risk of default across most subjects, though they don't seem to have disproportionately benefited groups with low levels of credit as some expected - in part because members of these groups appear to be denied credit (or dissuaded from applying for it) for practical reasons.
Recent immigrants and people below 30 get slightly dinged on credit beyond the amount justified by their performance. The length of credit history matters a lot, to the extent that although it hurts these groups, it would be silly to remove it from the model.
Applying for lots of new credit if you already have a long history can hurt your future credit score, whether or not you receive it, as it reduces your average account age - so if you have a history but want more credit, get the cards after you get the mortgage (or better, raise your limit on the existing accounts).
If you have lots of retail store accounts, you're probably not an ideal risk, but finance company accounts don't have much effect. (Note, however, that FICO's 2008 revision may ding you for them.)
Sex overall doesn't seem to be an issue - women score slightly better, but not after controlling for factors like age. However, the use of certain types of credit appears to be somewhat correlated with sex (consider: retail/department store credit).
Race per se isn't much of an issue; however, on average, Hispanics and especially blacks do have lower scores than non-Hispanic whites and Asians. This appears to be because have less of a history and perform worse - have a higher chance to default - most likely because their income is less than 60% of that of non-Hispanic whites and Asians, while their net worth is less than 15%. Indeed, their scores appear to be slightly higher than they should be based on performance.
Due to age of account issues, it's almost impossible to have the best scores if you're below 30, and unlikely below 40. Asians tend to have pretty good scores (and do actually get lower interest rates), but non-Hispanic whites have more of the best scores - perhaps because there are fewer old Asians in the record.
If you have more income, you probably have a higher score - but there are plenty of people with high incomes, but average or even poor credit. Married people also tend to have higher scores.
Perhaps more interestingly - and bearing in mind that the federal model isn't the same as those used in the real world - what factors predict success? What's likely to get you a good credit score?
The goal of the model was to identify a person's likelihood of 90-day delinquency or default on all kinds of new or current accounts. Like others, the federal model divides people into groups, each with a unique set of scorecard factors (see the full scorecard details). Due to the size of their sample, they only have three groups (most commercial offerings have 10-15, and providers have a range of models for specific purposes).
If you don't have much credit history (0 or 1 accounts recorded; ~10% of users, ~34% become delinquent/default, KS [measure of accuracy]: 72):
* Having any record of delinquency (including defaults and civil judgments) is terrible, but especially if you still owe money or the delinquency is less than three months old.
* Keep your average balance on open revolving accounts (e.g. credit cards) below 30% if at all possible. Above that, and you're more likely to default. It gets steadily and severely worse if you're use more than 65%; going over 90% is a great predictor of default.
* If the above information isn't available (say, you don't have a credit card), the risk is about the same as if you had an average of 70% - one reason it's hard to get a good card straight off
* If your average balance on installment accounts is above ~60% of your total credit for installment accounts over the past year, it's not good. If it's over 100% (rare), it gets a lot worse
* Having an account in good standing opened at least 18 months ago is a good sign (this might be a good point at which to apply for more credit).
* Having just a $135,000+ mortgage is great (if you're the 1% in this group who does); failing that, having an account with a limit of even a couple of thousand is pretty good
* Multiple credit inquiries are a slight predictor of default. Having more than 1, 3 or 12 credit inquiries hurt you more.
* If your only account hasn't been reported for more than a few months (perhaps you stopped using it), the chance of default rises significantly
If you have a clean file (no derogatory accounts/collections/records; ~56%, ~7.5% become delinquent/default, KS: 54):
* It's best never to have been delinquent. If you been delinquent in the past, the longer you've been clean, the better. It takes over three years to fully dig yourself out of the hole, but most of the benefit comes in the first six to nine months.
* It's best to use as little of your overall revolving credit limits as possible, but average usage up to about 50% doesn't have much effect on default. Above ~60% is worse, above ~80% it gets really bad, and above ~90% is a major warning sign.
* If you were ever more than 60 days late, it's about as bad as being at 70% of your credit limit vs. 0% - but still a lot better than ever being fully delinquent.
* Inquiries have a small but cumulative effect. If you have fewer than 8, you probably won't notice much. It takes 14 to get to the "70% of credit limit" point.
* Personal finance installment accounts ("make 12 easy payments of . . .") are a bad sign, especially if you've had more than one in the last 12 months (which is as bad going to "70% of limit").
* Using over 50% of the limit on a non-installment account in the last year is bad, but only really significant if you have more than two accounts in that situation.
* The higher your total credit limit across all open accounts in the last year, the better - but the effect is moderate, and there's little difference between $3,000 and $30,000. (Presumably having a mortgage or other high-value loan implies stability.)
* The average age of accounts matters a lot. Above three years is good, above five is great, and there's almost no difference above eight.
If you have a major derogatory mark on your account (e.g. collection or default; ~31%, 65% become delinquent/default, KS: 62):
* Having even one public record or derogatory account with more than $100 still owed on it is bad; more than that and you're at a high risk of default
* The longer you've gone since being delinquent, the better. The first six months is as important as the next five years.
* The more accounts with no late payments reported, the better.
* Try to keep balances on bank cards below 40%. Above 70% is a major warning sign.
* Your risk of default steadily improves as derogatory public records age, but there's little change due to this factor until about two years after the most recent one. It takes seven years to recover (at which point it would probably be cleared from your account by law anyway, and you'd be on a different scorecard).
* The average age of accounts on your reports matters, but not as much as for people with clean credit. Up to four years, it makes no difference, but after that it steadily improves, tailing off around eight years.
* Having more than one account currently or previously 30 or more days due within the past 24 months is a pretty bad sign.
* Having an account that's less than 120 days past due (but presumably not an account in good standing) in the past two months is bad; few people will want to loan money to someone about to default on another account.
* It helps to still be using credit. If you have no accounts reported in the past 12 months, that's bad.
* Try to pay down accounts to zero. Having a high percentage of your accounts be active and open with a balance greater than $0 in the past year is a small indication of default risk (about half of this group have over 85% of their accounts in this situation).
* Having more than four different credit issuers on your record is a good sign.