A Credit to Humanity?

As anyone who has recently applied for a mortgage has discovered, credit reports are no longer evaluated by humans, and instead are numerically scored by computer software (the most common: “FICO” scores). Similar systems are used to underwrite all other consumer loans, though not as comprehensively as in the mortgage business, where the larger sums get special attention.

Using computers to do the work makes good sense: the native hard-headedness of the machines is more than offset by their supernatural capacity to recognize patterns in complex masses of data.

However, new technology has a way of… morphing… spreading.

I don’t mean to make anyone uneasy, but it’s time you knew that the cost and availability of your auto and home insurance policies are soon to be (or already are) determined in large part by your credit record.

Insurance companies are not studying your credit to see if you will make the premium payments on time. No, the companies — or rather, their computers — are studying your credit to see what kind of insurance risk you will be.

George Orwell or Aldous Huxley or Robert Heinlein would get some dark satisfaction from the following sentence: The worse your credit, the more likely you are to generate an insurance loss; the better your credit, the better the insurance risk you are.

Your credit record is an extraordinarily accurate predictor of the probability that your home will burn down, or that you will have a car accident — in fact, a better predictor of the chance you will have an automobile accident than is your driving record!

The best monograph in the public domain on this subject (“The Impact of Personal Credit History on Loss Performance in Personal Lines”, James Monaghan; the Casualty Actuarial Society www.casact.org/pubs/forum/00wforum) should be read in daylight. You will feel as though someone or something is not just looking over your shoulder, but looking into you… through you… measuring something in you that you don’t understand.

The correlation mathematics are as clear as can be: whether measured by “Amounts Past Due” (credit jargon for the total number of all late payments in your record), “Derogatory Public Records” (judgments, liens, bankruptcies, foreclosures), “Collection Records” (past or present accounts in collection), “Status of Trade Lines” (late payments outstanding), “Age of Oldest Trade Line” (your credit is judged better the longer you have had accounts open — don’t close idle accounts), “Non-Promotional Inquiry Count” (if you ask a lender to check your credit, it hurts; if a credit card company scans you before sending a promotional mailing to you, no harm done), “Leverage Ratio on Revolving Accounts” (balance outstanding at any moment versus credit limit — keep your balances below 50% of limit, even transient, intra-month balances!), or “Revolving Account Limits” (get your limits raised any time you can)… if you have poor credit by any ONE of these standards, you are a higher insurance risk.

This list of credit criteria, with slightly different weighting, is used to compute your FICO credit score. You can assume that your FICO score is a reasonable proxy for your attractiveness as an insurance risk: above 680 (into the 800’s), a good risk; below 620 (into the 400’s), a poor risk.

What, exactly, are the insurance machines measuring here? As Mr. Monaghan says near the end of his analysis, “An outstanding issue that will likely remain outstanding is causality.”

Causality. The credit vs. insurance risk correlation is clear and unambiguous, but nobody knows why. Monaghan recites the leading theories, but all fail rigorous tests of cause and effect. One might suspect that those with poor credit rely more on insurance, and are quicker to file claims, or are more likely to file fraudulent claims; but these hypotheses do not prove out. Some suggest that insurance risk may be the result of low income, or some differentiation by age, or residence (urban, rural, or suburban), but none of these correlate with insurance loss as well as predicted by credit rating. Only two groups present some evidence for cause of both poor credit and high insurance risk: those under stress, and those who are self-employed.

You may have been amused to discover that Amazon.com could quickly predict from your buying habits which books you would next like to buy. Perhaps you felt a little uneasy at the ability of “data-mining” software to identify nearly anything you want to buy or do.

Orwell, Huxley, Heinlein… under the scrutiny of sophisticated pattern-recognition, our credit and insurance ratings may be aspects of a larger quotient, some measure of relative caution in the way we live our lives, or competence at it, or responsibility for it, or… we may not be able to comprehend the cypher in us at all, no matter how transparent our behavior is to the machines.

You do have options: the data show that if you set out to improve your credit, and do so, you will have less chance of a car accident or burglary in the future.

Why, we don’t know. But you will, predictably.

Inhumanly.