Wednesday, July 2, 2008

Your Creditor is Silently Judging You

You stay below your credit limit, pay your bills on time, and always pay more than the minimum balance due. Yet your creditor mysteriously raises your interest rate and cuts your credit line in half. What gives? Maybe it was because you swiped your card at the bar or at a massage parlor, and the lender's scoring model deemed your purchasing behavior a credit risk.

Lenders, insurers, other financial firms (and, increasingly employers, landlords, cell phone carriers, and the voting public) have always used credit scoring models to make decisions about people. While we have some very general information about the design of the most widely-used models, many lenders use proprietary formulas based on their own evaluative criteria. In the case of CompuCredit's Aspire Visa, these criteria allegedly included individual spending behavior.

The Federal Trade Commission recently filed suit against CompuCredit accusing them of "deceptive" marketing practices because they did not properly disclose to consumers that they monitored spending. The suit alleges that when the cards were used at places like tire and retreading shops, massage parlors, bars, billiard halls, and marriage counselors (really? marriage counselors?), CompuCredit cut their credit lines.

Now keep in mind, the legal issue here is not that a lender monitored purchasing behavior and used it to evaluate the terms of credit. The legal issue is that the lender did not properly communicate the policy to cardholders.

But the relevance to consumers is far broader, because this suit gives us a glimpse into the heretofore secret world of credit scoring. Consumer advocates have long suspected that purchasing behavior is included in scoring models but because the models were protected as intellectual property there was no way of knowing for certain. As Jennifer Silver-Greenberg writes in BusinessWeek:

With competition increasing, databases improving, and technology advancing, companies can include more factors than ever in their models.... The worry is that companies may tweak the credit scoring system in unfair or biased ways, weeding out or limiting borrowers based on race, gender, or sexual orientation.
Personally, and this is my opinion here, I would say of course creditors are monitoring purchasing behavior. If they can do it, they are doing it. It's simple human nature to want to know things about others, and creditors are not any purer than the rest of us. They have the additional incentive of being in business with their customers, so if the mechanism exists for them to gather the information then I think we can safely assume that they will do so.

The concern is not that financial institutions are gathering the information, the concern is in how they are evaluating it. "Scoring model" sounds analytical, but in reality all models that evaluate data are exactly as biased as the person or people developing them.

Models are, in effect, hypotheses. You come up with a hypothesis, you enter the data, and you discover if the data predicts the outcome that you are expecting. If it does then you consider the hypothesis valid. It's reliable if it gives the same valid outcome again and again.

The problem with the credit scoring model is that it's only being selectively tested and only by the same biased people who developed it. In the natural and social sciences we cannot say, "Hey, psychotherapy and antidepressants are more effective in treating depression than psychotherapy alone. Don't ask me how I know this, just buy my medicine." The scientific method dictates that we present all of it -- our hypothesis, subject selection, testing process, data gathering, analysis -- and encourage others to try exactly what we did and produce the same results.

People might not be so outraged about creditors analyzing purchasing behavior if we were allowed to shine the light of public scrutiny on their scoring model.

This particular instance with the Aspire Visa also just smacks of moralism. Why is using your credit card at a bar or marriage counselor so bad? What about people who use a credit card "convenience check" to pay their rent -- that seems more indicative of financial instability than a night out at the pub. We can't ask the lender these questions because their process is protected.

In our complex financial world, selectively directing the quality and availability of credit products has profound social consequences. We've already seen how institutions were able to perpetrate bias (consciously or unconsciously) in the subprime lending market. With credit so interwoven into the social fabric, do we have a right to public oversight of how lenders make their decisions?

2 comments:

Nelson said...

1.) I doubt moral judgment was the basis for the penalty for seeing a marriage counselor. Getting divorced is a negative financial event, and it only makes sense to scale back credit if you have reason to believe that the risk of divorce is present.

2.) Your essay implies that credit scoring models are hypotheses which aren't verified, but that is simply false. Credit card companies have a history of transactions they can mine to see if certain categories of purchases correlate with financial deterioration (e.g., the example above of seeing a marriage counselor). My guess is that in any credit card company which is focused on making money (i.e., every one of them), back testing will be used to verify or disprove any hypothesis an analyst might make concerning the link between certain kinds of purchases and the risk of lending to the corresponding consumers.

Reducing credit lines for no good reason is well known to hurt profits, so normally this will be an unusual phenomenon.

Amanda Clayman said...

Your point about the negative financial repercussions of divorce is a good one. Frankly, I drew a blank about why marriage counseling would even be on the list of risk-predictive expenditures, mostly because I see lots of couples in my counseling practice who are not in any danger of divorce. Is divorce a negative financial event? Definitely. Does the act of engaging in couples counseling mean marital trouble? Not always. (It might cause trouble if attending counseling means that credit lines get cut and that creates additional financial stress in the marriage, but that’s another discussion.)

Obviously lenders want to maximize profit above all other possible agendas. I don’t mean to suggest that CompuCredit’s mission is to push a moral point of view (I just think that’s why the story sounds so sordid). However, I think that it’s a myth that organizations can be truly unbiased, no matter how disinterested they profess to be in anything beyond the scope of profitability. A lender can create a profitable model that is more or less destructive for consumers. That may affect how profitable it is, but profitability is a matter of degrees. I am skeptical that destructive impact is even being measured or that ethical values are being considered, and I think they probably should be.

Changing the terms of credit between lender and consumer can cause great upheaval in the life of the consumer. I would love to know more about how the scoring model is developed and back-tested, otherwise I remain unconvinced that simply pursuing profit is enough to eliminate all other biases from creeping into the system.