If you find this article informative and worthwhile, please support my work by donating if you can.

logo    The Use of Credit Scoring by Insurers


Insurers want to be able to base rates on credit scores. The insurers claim that people with poor credit scores file more insurance claims than others and should, therefore, pay higher rates.

There are at least three major problems with this claim. Lets take the simplest first.

Erroneous credit reports

Numerous studies have shown that a large percentage (up to 80% in some studies) of credit reports contain errors. But if credit scores are calculated from erroneous data, the scores calculated are also erroneous. So if insurers are allowed to base rates on credit scores, they are charging higher rates to people whose credit scores do not accurately reflect their true standing.

Secrecy of the formulas used to calculate credit scores

I have read that the various credit bureaus use different formulas to calculate these scores and that the bureaus consider their formulas to be proprietary. But anyone conversant with even rudimentary mathematics knows that if you know a formula, you can make the result come out any way you wish merely by changing the terms on the opposite side of the equality sign. And you can do this as many times and as often as you wish.

For instance, say Credit Bureau A, has a special relationship with the insurance industry. Could be anythinga pure business relationship which it values, ownership of stock in insurance equities, personal friends in the insurance industry, whatever. Now it calculates the credit scores for the people in its database, and examines the results. Someone says, "it would be nice if the results were ten points lower, wouldn't it?" And the person who programmed the formula into the computing device says, "Oh, that's easy," and alters the formula. Then the scores are recalculated.

Now suppose that a year later, circumstances have changed, and someone again says, "it would be nice if the results were ten points higher, wouldn't it?" And again the person who programmed the formula into the computing device says, "Oh, that's easy, and again alters the formula.

Since the formula is proprietary and secret, no one outside the bureau ever knows. So the credit scores calculated could be the result of a mere whim of the people running the bureau. Why should insurers be allowed to use such scores?

Finally, and most importantly, the nature of statistical studies.

When I studied statistics, the nature of a statistical study was described as follows: First the statistician proposes a hypothesis (a null hypothesis in statistical language), say x causes y. Next he accumulates data on both x and y and analyzes it, using statistical techniques to determine if the two sets of data are correlated.

Now here's the clincher. If the data are not correlated, the statistician knows that his hypothesis is wrong and rejects it. However, what if there is a correlation? Can he conclude that x is the cause of y? Not by a long shot!

Why? Because correlation is a many-one relationship. Any one thing has positive correlations to many other things, so no more importance can be assigned to any one correlation than to any of the others without more investigation.

Consider this example:

Smoking can be positively correlated to lung cancer.

Some genetic factors can be positively correlated to lung cancer.

Exposure to industrial pollutants can be positively correlated to lung cancer.

Exposure to some kinds of radiation can be positively correlated to lung cancer.

This is a many-one relationship. The trouble is, it does not tell us which one, if any, is the cause of lung cancer. None may be for all we know. And if someone makes the mistake of attributing causation to any one of them, he is committing the fallacy of confounding.

Confounding is a significant problem in statistics. Because of it, many researchers now replace the word cause with the words contributing factor . But even that assumes too much. Statistical texts are full of examples of published studies that were guilty of confounding. It is not a rare fault.

Are the statistical studies used by the insurance industry guilty of confounding? We don't know, because those studies, too, are proprietary and secret. They have never been subjected to objective analysis. So could these studies, if there even are any, be guilty of confounding? I think so. I can think of something that can reasonably be responsible for both low credit scores and a high number of insurance claims.

Think about itinadequate income.

Certainly people with inadequate income can be expected to have a difficult time paying their bills and, thus, have poor credit scores. But why would they make more insurance claims?

There are two different groups of people who are subject to inadequate incomes. Let's consider them one at a time.

Some people, because of their stations in society, work for inadequate wages all of their lives. They live in rundown neighborhoods where crime is high. If they have home-owners or renters insurance, they can be expected to have more property loss claims. These neighborhoods often consist of houses that were built when older building codes were in force. Their construction may be of poor quality, and the homes may not be in good repair. These houses would be more likely to be damaged by severe weatheranother reason for insurance claims. These houses are more likely to have outdated heating and electrical systems and be more fire pronemore claims. And the vehicles these poor people drive are apt to be older and also in disrepair, more dangerous, and more likely to be involved in accidents.

But not everyone with inadequate income falls into this group. Some people have inadequate income only for short periods, but they still end up with poor credit histories and poor credit scores. What can cause the temporary loss of adequate income? A family tragedy, a severe illness, the temporary loss of a job, being called to active duty from the National Guard or Military Reservesalmost anyone can add items to this list. But these people may not live in run-down, crime-ridden neighborhoods, may not live in older houses in disrepair, may not drive older vehicles. And they may not file many insurance claims.

Nevertheless, if the insurers are allowed to base rates on credit scores, these people get charged higher rates too. Should they be? I don't think so. But shouldn't insurers then be allowed to base rates for the first group on their credit scores? No, I don't think so, for not all people in that group file many insurance claims, and life for those people is already hard enough. We need not make it harder by charging them higher insurance rates just because it makes making profits easier for insurers.

This little essay does not contain any information that is not generally known. So what is truly amazing is why lawmakers haven't banned this practice. Is the reason ignorance or perfidy? (3/2/2005)