Zero, naught, and nothing: How it can cost you more on your insurance

What’s the difference?

We all know what zero is. It’s a number, and yet it’s not a number. If we are counting, it means “none” of something. Zero is one less than one. Zero is the amount we want to owe on our taxes on April 15. Zero is the balance we want to see on our credit cards. Zero is not the number of gallons of gas we want left in our gas tank when we are 10 miles from the next exit on the Interstate.

So the numeral “0” is nothing, right? When we begin to use the power of statistics, not exactly.

The power of this difference came to my attention when I received an insurance renewal form a few years ago. I read over the entire document, and found a small notice at the end. It stated that I had not been given the best rate because of the results of my credit history. This was of concern to me, because my credit should have been excellent. I had just checked my credit history with all three bureaus and everything was in order—no late payments, several credit cards in good order, paid off every month. What was the problem?

After receiving three of these notices, from different companies, I called the contact person. What was it about my credit history that was not perfect? The answer given was that a sophisticated formula was used to calculate a score based on a number of factors in my credit history. So… what was it that prevented me from getting the best rate? Well, that was private information. I persisted; if they had to give me a notice regarding my credit report, then I felt they should tell me how I was falling short. We went back and forth for quite some time, over several phone calls, even going up the ladder a couple times.

I knew what the problem was. These credit companies are able to use powerful regression formulas to crunch enormous databases, and find small, seemingly insignificant variables that, when put together, could predict some outcome better than chance. In this case, the statistician was using hundreds of variables collected in the vast data pools to predict insurance losses. I wondered what those variables were. Their prediction formula, they said, was derived at great expense, and revealing it could only compromise the company’s use of the data.

My persistence paid off. I was able to find out that the variables that were being used to place me in the more expensive insurance pool were (drum roll): I did not have credit cards when I was in my 20s, I had never had a car loan, I had never had a mortgage. Wrong, wrong, wrong, wrong. If I hadn’t persisted, I would never have found out the problem.

I explained to the woman that when I was 20 I took a job as a bank teller at Virginia National Bank. At that time, I was approved for two credit cards, one a VISA and one a MasterCard, each with a credit limit of $600. I had them for many years, but my best guess is that they were not reported to a credit bureau, because I worked for the bank when they were issued. I applied for a car loan in my 20s as well to buy my first (and only) new car, but it was with a credit union. That was paid off early. But the credit union may not have reported it to a credit bureau because there were no problems with the account. And I wasn’t sure why they mortgage hadn’t been reported to them; it had been paid off without any late payments years before.

How does that relate to zero, naught, and nothing? To a statistician running a regression formula, apparently not enough. To you and me, it could be the difference between paying hundreds of dollars of unnecessary charges over our lifetimes. In running the statistics, the answer to the field “number of credit cards in his 20s” should have been treated as a blank, as in “no information available.” This is “naught.”” The statistician or programmer assumed that naught meant zero, as in “zero credit cards in his 20s.” This assumed that the database was a complete record of all credit events in my history.

I have termed this problem The Car Fax Effect, as I will discuss in my next entry.

The followup to this story is telling. I called the credit bureau who reported the data, and asked to correct my credit report. I was asked what item on my report was incorrect; I replied that it was an item that was not on my report. Unfortunately, I cannot correct an item that is not reported. Ah, so they are not liable for errors that are not reported, but they can fail to report data that they have not collected, thus lowering my insurance score. No one is responsible. What an interesting paradox!

This entry was posted in assumptions, credit history, quantitative research, regression, statistics. Bookmark the permalink.

Leave a Reply