Just why does one have to unlock a credit report before you can test drive a car you plan to buy with cash? Why does it determine whether you can rent an apartment, get a job, insure your car, or for all I know get a date? At the HFSC hearing on the credit bureaus earlier this week, the reporting companies and many of the lenders who rely on them argued that credit files and the scores assigned to them ensure credit availability and sound underwriting. Do they, or have we all made the deal that the prisoner in an iconic British TV series years ago refused – i.e., we are now numbers, not men and women, at least when it comes to engaging in economic activity?
Credit scoring was intended to be as benign as an infinite source of clean energy. The technique grew up as banks began in the 1980s to fear costly enforcement actions if they were found to discriminate on the basis of race, ethnicity, or gender. Scoring that seemed to take the subjective out of credit underwriting was not only an important innovation, but also a mighty handy defense – i.e., it wasn’t I, just an objective score that made me say no. Year-in, year-out, this back and forth has characterized annual release of the HMDA data and the industry’s defense of persistent credit-availability and pricing disparities for minority borrowers. And, year-in, year-out, the defense has insulated the industry from enforcement, if not criticism.
Is blame for ongoing discrimination, at least of the disparate-impact variety, just being shifted from banks to abstruse models developed by lightly-regulated credit bureaus and scoring companies? Chair Waters clearly thinks so – her bill gives the CFPB authority over scoring models precisely to ensure that they are fully within the ambit of the fair-lending laws. At the least, it’s critical now to ask the same questions of scoring that I raised in a recent paper on algorithms: does the model simply hide the misdeed?
Assume, though, scoring beats borrower-by-borrower underwriting for credit quality and thus is a valuable tool without discriminatory side-effects. Even so, scoring combined with securitization has standardized credit allocation under arcane models often based on erroneous data with far-reaching and often adverse credit-availability impact. With the technical ability to take many loans and then turn them into a single asset came the power to originate loans as long as their risk characteristics were close enough to a desired outcome to achieve a market-ready price. Bit by bit, loans were booked by FICO scores calculated on credit-bureau files even if other risk criteria were better predictors of default. As we noted in an Economic Equality blog post, “subprime” home-purchase mortgages based on credit score were a good deal less risky than speculative loans for high-score borrowers who were leveraged in ways not caught by the scoring model.
Economies of scale with considerable benefits for efficiency and liquidity thus begat systemic risk. Perhaps even more interestingly, credit-score efficiencies combined with the scale of modern finance also begat business models so dependent on scoring that the only small-dollar loans made in any kind of volume without a federal guarantee have to have a credit score. Before scoring, before securitization, and before banks got so big that economies of scale dictated credit decisions, small loans made economic sense on a one-off basis. Now, given the high cost of underwriting a loan regardless of amount absent a simple score, lenders will only devote scarce capital to loans with secondary markets, score-driven cross-subsidies, or high enough balances to make return-on-asset objectives.
If banks got smaller, would portfolio lending get bigger? I doubt it – another unintended effect of credit scores is to spawn the idea that a single risk weighting captures risk across an entire portfolio of loans for the same type of borrower or collateral. Thus, all small business loans have a single risk weighting under the standardized risk-based capital rules and are stress-tested the same way even if the business is different, the borrower has a lot of expertise, or there’s a lot of liquid collateral. No wonder too many start-up small business owners fund their ventures by putting their homes at risk – they’re actually arbitraging the score-dependent mortgage market to raise small-business funding. Scoring for mortgage lending makes the borrower eligible for a second lien or refinancing to fund a new business because nothing in the current model captures loan purpose and thus the real risk a true underwriting looking beyond a simple score might capture with just a single question.
So, should we just throw out credit scores? I doubt it given the inevitability now of big banks, risk-based capital rules, secondary markets, and investor expectations based on economies of scale. The solution seems to be first to confine credit scores to the purpose for which they are designed – credit underwriting – doing so by legislation if no other remedy is possible. At the same time, we need to take a hard look at why portfolio underwriting is seemingly impossible for wide swaths of retail and small-dollar lending in a regulated institution.
What would happen if banks were allowed to make underwritten – not scored – loans up to certain percentages of regulatory capital without application of an automatic risk weighting? I asked a senior bank regulator this a while back and was told that, for all its virtues, this idea added undue complexity. Are banks really so unable to think for themselves?