« Slightly hoarse | Main | Assorted links »
Bad Credit, Bad Driver
Some states ban the use of credit scores to price auto insurance in part because African-Americans and Hispanics tend to have lower (worse) credit scores and thus pay higher auto insurance rates. The brute facts, however, are that credit scores are good predictors of auto claims. Luke Froeb at Management R&D summarizes a recent FTC study on the issue.
- Credit scores effectively predict ... the total cost of [auto insurance] claims.
- Credit scores permit insurers to evaluate risk with greater accuracy, which may make them more willing to offer insurance to higher-risk consumers ... . [note: this is why you can call up GEICO, let them look at your credit report, and get an auto insurance quote over the phone].
- ..as a group, African-Americans and Hispanics tend to have lower scores than non-Hispanic whites and Asians.
- ...scores effectively predict risk of claims within racial and ethnic groups.
- The Commission could not develop an alternative scoring model that would continue to predict risk effectively, yet decrease the differences in scores among racial and ethnic groups.
Thus banning the use of credit scores would at best force good drivers (of all races) to subsidize bad drivers. At worst, if insurance companies cannot price according to risk an adverse selection problem could be created in which good drivers purchase less insurance (to avoid having to pay the subsidy to bad drivers) thus pushing rates even higher and perhaps unraveling the market.
Posted by Alex Tabarrok on July 28, 2007 at 02:12 PM in Economics | Permalink
Comments
Not to mention that enabling bad drivers to drive more cheaply increases the number of accidents. Since in most accidents both parties are somewhat at fault, this increase in the number of accidents should be expected to be more than linear with any increase in the number of bad drivers. In general, we don't need society to be arranged so that everyone can drive, but rather so that everyone can get around without too much inconvenience in some manner. The more of a burden we collectively impose upon someone by kicking them off the road the less willing we are to kick them off the road and the more likely they are to impose an increasingly severe safety burden on everyone.
Posted by: michael vassar at Jul 28, 2007 2:26:20 PM
Insurance companies use lots of other measures to set rates: age, gender, prior accident history, location and brand of car come to mind.
Except for new drivers it would seem that most of these should provide enough of a basis to make an initial quote. Rates are always modified by subsequent experience in any case.
The philosophical question that lies behind:
"if insurance companies cannot price according to risk an adverse selection problem could be created..."
is the foundation of the libertarian/liberal divide. Jared Bernstein calls it YOYO (you're on your own) vs WITT (we're in this together).
Annuity companies used to determine retirement benefits based upon gender. Women sued and said they got a lower payout for the same contributions. This was true in terms of the yearly payout, but not statistically if you considered total lifetime benefits. The courts felt it was sexual discrimination.
In the case where the woman didn't collect the theoretical total payout expected (by dying sooner) she was correct. Fairness demanded that the risks be shared equally even if men now get a smaller monthly payout to equalize the women.
If the libertarians would only get past the idea that maximizing for them personally is the be all and end all of society we would have a lot less class resentment.
Posted by: robertdfeinman at Jul 28, 2007 2:45:53 PM
If the libertarians would only get past the idea that maximizing for them personally is the be all and end all of society we would have a lot less class resentment.
If the liberals would only keep their hands off my wallet (as well as my insurance and annuities), we'd have a lot less theft.
Posted by: Dennis Mangan at Jul 28, 2007 3:00:25 PM
Obviously whatever system is in use is not benefiting any group because good drivers (Asians and Whites ) and bad drivers (Hispanics like me)all seem to be paying higher rates every year.
Posted by: Alex at Jul 28, 2007 3:22:28 PM
If the libertarians would only get past the idea that maximizing for them personally is the be all and end all of society we would have a lot less class resentment.
I'm more liberal than libertarian, but I think this is too harsh. Most libertarians are such for reasons other than selfishness.
How do those who say the use of credit scores is racist explain the lower credit ratings of African-Americans and Hispanics? Do they claim bias in the credit bureaus, or do they look for reasons why African-Americans and Hispanics might genuinely have worse credit?
Posted by: Nick Tarleton at Jul 28, 2007 3:30:48 PM
Nick,
Re: "Thus banning the use of credit scores would at best force good drivers (of all races) to subsidize bad drivers."
Simply untrue. Why? If credit scores were the only indicator, it would be true. But they are not - and no need to go into detail of other indicators (but some are accident record, tickets received, age, sex, and type of car).
What you are suggesting is that states have a choice of allowing insurance companies to discriminate again blacks and Hispanics and at the same time provide rates that are more closely keyed to accident likelihood, or not.
You are also suggesting that states have a choice of allowing insurance companies to discriminate again blacks and Hispanics in order to maximize profits. Given that insurance companies are private for-profit entities, obviously their desire is to maximize profits and that it the reason they want to use credit scores.
Your approval for this is repugnant.
Posted by: Dave M at Jul 28, 2007 4:03:17 PM
Dave M,
It is your ability to think and argue rationally that is repugnant and further leads to morally repugnant outcomes.
The first thing that should be clear is in the absence of competition an individual insurance company will only enhance profitability to the extent that the credit scores give it additional information. Thus if credit scores did not enhance the ability of a firm to predict accident rate then they would not use them as they would actually hurt profitability.
The second thing is that insurance is a competitive industry. Once all firms can use this one piece of information, their ability to extract "free" profits from it is removed. The "profit" then moves onto those who have better credit scores. That is, those who this study shows are statistically better drivers
Thus by not allowing the use of credit score you are arguing you would like cheaper insurance policies for bad drivers. By doing this you encourage more bad drivers to be on the road. You also discourage them from changing their behavior in ways which would improve their driving ability. Thus you are encouraging more of the outcomes of bad driving. That is more accidents. Including more fatal accidents involving innocent pedestrians.
More innocents dead in the name of stopping evil “for-profit” corporations. Congratulations. Another victory for the church of modern liberalism.
Posted by: at Jul 28, 2007 4:32:03 PM
Alex, you said in part due to its impact on hispanic and african american drivers. What's the other part of the reason some states are banning the use of credit scores to price auto insurance?
Posted by: Hopefully Anonymous at Jul 28, 2007 4:37:25 PM
What if minorities are on average younger than whites? Is it then wrong to use age as a
factor determining insurance premiums? Some insurance companies give a discount to young
drivers who are still in school and have good grades. What if on average that favors
whites? Is that racial discrimination?
For those that don't know how the statistics works, it isn't that insurance companies want
to throw out all the other data on age, gender, and driving history, they just want to add
this other measure. We all know that some people have 3 wrecks at 30 because they are bad
drivers and others have 3 wrecks because they were unlucky. Is it possible credit score
could shed some light on this? It doesn't seem that big of a leap that if a person is
responsible with his money and credit that he/she is also responsible with a vehicle.
Posted by: Charlie at Jul 28, 2007 5:06:10 PM
This issue is contaminated by the fact that in most places auto insurance is required by the government. If the government requires something, then an extra degree of fairness should be required in the provision of that requirement.
Also, since Blacks and Hispanics tend to have lower credit scores, and people with lower scores tend to be worse drivers, why not just cut out the credit-score middleman and advocate direct discrimination against Blacks and Hispanics without even checking the credit scores?
Posted by: TY at Jul 28, 2007 5:08:25 PM
Nice post.
If one is looking for a general defense of credit reporting, let me suggest my article in The Independent Review:
http://www.independent.org/pdf/tir/tir_05_3_klein.pdf
Credit reporting is really one of the great social accountability mechanisms, bringing tremendous benefits to all.
Posted by: Daniel Klein at Jul 28, 2007 5:18:27 PM
Dave, Tyler said that, not me.
TY, direct racial discrimination would only be justifiable if minorities were worse drivers even after correcting for all other factors - in other words, if being Black or Hispanic made you a worse driver directly rather than being correlated e.g. through credit scores. That probably isn't the case. Although this does raise the significance of my previous question of why race and credit score are correlated.
Posted by: Nick Tarleton at Jul 28, 2007 5:21:37 PM
Actually, switching to a provider that doesn't use just credit as a pricing mechanism can lead to dramatically lower rates. I cut my premiums in half when I switched FROM Geico to a competitor who paid attention to my driving history (never had a moving violation or been in an accident) instead of my credit (abyssmal)
Posted by: Flynn at Jul 28, 2007 6:25:31 PM
"Also, since Blacks and Hispanics tend to have lower credit scores, and people with lower scores tend to be worse drivers, why not just cut out the credit-score middleman and advocate direct discrimination against Blacks and Hispanics without even checking the credit scores?"
It is a good question, but one answered by the study, "As a proxy for race and ethnicity in statistical models of insurance, scores have a 1.1 percent and 0.7 percent effect for African-Americans and Hispanics, respectively. This means that most of their predictive power is not as a substitute for membership in racial or ethnic groups. In addition, scores effectively predict risk of claims within racial and ethnic groups."
As the study points out, the credit score is not the same as just adding a race proxy. Also, the credit score does a good job of predicting risk within racial and ethnic groups. That is, show me two drivers with the same race, age, and driving history and I can guess how often they'll get in an accident. But if you add credit score, then I can make a better guess.
Posted by: Charlie at Jul 28, 2007 7:01:05 PM
Flynn has pointed to the key of the matter. In a free merket, people will have a large variety of life situations. Some companies will provide a good situation for some, and worse situation for others. People will be rewerded for choosing wisely, or not for not.
Posted by: Russ Nelson at Jul 28, 2007 9:55:02 PM
Daniel Klein's "general defense of credit reporting" is so tendentious that it discredits itself. For some examples (just a few of many issues): Klein first asserts (p. 327) that it would be very difficult for someone to obtain a credit report on a neighbor-- an assertion which was untrue in 2001 and laughable today; then (p.333 et seq.) the paper uses only the overheated formula "exorbitant damages" ("exorbitant" means "out of this world," or "greatly excessive") to discuss (or rather, to avoid discussing) the notion of permitting people damaged by false reports from credit-reporting agencies (known as libel when anyone else does it) to hold them to account. Klein sets up then burns down a huge strawman of "exorbitant damages" without ever treating the real question of reasonable sanctions.
Given a chance to explain how useful credit-reporting agencies (CRA's) are-- to all market participants-- Klein blows it by launching a tirade against people who just want a fair chance for the mice who risk being unfairly trampled by elephants.
Klein even makes the elementary economic error of supposing that credit-reporting agencies would prefer Type 2 errors to Type 1. If that were true, why would CRA's need Federal preëmption of state libel laws? In fact, CRA's customers are risk-averse (that's why they purchase credit reports) repeat players who (in the immediate sense) pay for the whole business. They don't much mind Type 1 errors, because at worst those impose a small opportunity cost on report-users, whereas Type 2 errors pose the risk of quantifiable losses. (Corporate middle managers almost never get fired for missing opportunities (dogs that don't bark)-- they get fired for running up visible losses (loans that go sour).) Since CRA customers prefer Type 1 to Type 2 errors, so do CRA's (they are in a competitive business). The problem is, Type 1 errors cause diffuse, uncertain harm to CRA's and their customers, but focused, specific harm to consumers. The only way to get CRA's to care about Type 1 errors would be to hold them accountable for them, but the CRA's have lobbied Congress to do precisely the opposite: Congress has immunized them to liability.
So, the costs of Type 1 errors are externalities to CRA's (which, by the way, explains why Klein's suggestion that consumers pay CRA's extra for accurate reports is a red herring. Consumers already pay CRA's for accurate reports, by financing the businesses which purchase those reports). Only if forced to internalize the costs of Type 1 errors will CRA's take steps to minimize them. In fact, only such a rule could push the Type 1 error rate down to the economically-optimal level! So long as CRA's can exclude the costs of Type 1 errors, they will "overconsume" them, and consumers will suffer.
Posted by: Mark Seecof at Jul 28, 2007 9:58:57 PM
robertdfeinman, how can it be the case that "we're in this together" if I never decided to join up with you and become "we"? If I actually was given the choice to join or not, I might actually do so and then we really would be in it together. However, I am not given any such choice, but simply forced into it and told to do "my part".
Posted by: TGGP at Jul 28, 2007 10:08:00 PM
What's the explanation for the link between credit scores and accident risk? Obviously the link has been proven, but are there any explanations for the common factor, e.g. impulsiveness?
Posted by: Tom Davies at Jul 28, 2007 10:16:15 PM
The problem I have with credit reports is that good people can, for whatever reason, wind up with bad credit reports. This is probably not a concern to the industry. If 1% of people who are actually good credit risks have bad credit scores, the credit reports are still very useful, but very unfair to that unfortunate 1%.
It's bad enough that people with bad scores can't get credit. But when they can't get a job and can't get car insurance, then they are screwed.
Posted by: Half Sigma at Jul 28, 2007 11:15:51 PM
I think that maybe it's best to turn this question around: is it racist to use credit ratings to determine the interest rate and amounts of a loan? This is their specified intention, and there is (I imagine) a good correlation between past credit performance and future payment of loans. However, this has *exactly* the same racial bias as the case above--people of certain races have low credit scores, and credit scores negatively correlate with both risk of default and risk of accident.
By the same logic as above, if African Americans and Hispanics are on average worse drivers, then practically any actual predictor of risk of accident is "racist", including a history of accidents. It makes some sense to look at what might constitute real racism. The key is finding 4 above "...scores effectively predict risk of claims within racial and ethnic groups." Credit scores and accident history both predict whether one African American driver (A) is likely to be a worse driver than another African American driver (B). Skin tone is presumably not a good predictor of whether driver A is better than driver B. Since credit scores are "race-blind" in this sense, using them is not racist. Skin tone is racist, because it treats all African Americans the same based solely on their race.
Posted by: Lucas at Jul 28, 2007 11:32:14 PM
Thus banning the use of credit scores would at best force good drivers (of all races) to subsidize bad drivers.
Good drivers (of all races) subsidize bad drivers no matter what the scenario.
Posted by: fustercluck at Jul 29, 2007 12:01:56 AM
Irresponsible people tend to display their irresponsibility in many ways, not just one or two.
I've personally been in a situation where I had damaged my credit, but had a flawless driving record, was married with an older sedan, et cetera, and still paid higher than normal rates.
In other words, I was one of the statistically few that had only one indicator of irresponsible behavior with no other corresponding traits. That I still paid relatively high rates, as compared to other people I had talked to of a similar demographic tells me that the credit scored was weighted very heavily.
A perfect driving record for years, and a non-risk related car i.e. not a sports car or some other high-end vehicle is much more relevant than the credit score, even though the credit score is applicable in the larger picture.
In my opinion, from my own experience the insurance companies are taking all negative indicators and weighing them equally, which of course is not even close to being accurate.
So it's true that irresponsible behavior in one area of a person's life is most likely indicative of irresponsible behavior in other areas. However, the insurance companies are not treating all of the information as it ought to be treated, and thus they are weakening their own case in the long run.
What will also come out in this debate will be the use of zip codes. Minorities often stay in or close to the older neighborhoods even after they can afford to move on to the burbs. So there's probably a substantial number of minorities who have good driving records, avg credit scores, and are paying still higher rates simply because of the street address.
So perhaps the insurance companies are just going to get burned for their less than scrupulous treatment of their beloved actuarial tables. I
Posted by: Ray G at Jul 29, 2007 1:16:45 AM
In an era of data fraud and identity theft I think credit ratings are next to useless for anything since the accuracy of the information can be easily screwed around with.
Posted by: T.W at Jul 29, 2007 1:40:21 AM
I agree with the most of the economics of this post but I disagree with the policy.
The cost associated with the invasion of privacy is not factored in. Lower insurance rates are not sufficient compensation for this type of intrusion into privacy. Access to consumers' personal information puts consumers at an unfair disadvantage. Consumers are not privy to all the underwriting criteria used by insurers and consumers in return, should not be forced to disclose so much to insurers. Anyone who thinks that insurance companies will translate all the cost benefits of near-complete information directly to the consumer is being somewhat naive.
Moreover, we would need a legal framework that would guarantee affordable coverage to the highest risk consumers. Otherwise insurance companies would simply not cover the high-risk group or charge unreasonably high rates which would create a class of uninsured/underinsured and even more dangerous drivers (see "health insurance"). In creating such a safety net, we are inviting more regulation into the insurance industry by way of price ceilings and guaranteed coverage. So we are back to an inefficient system except now our fundamental right to privacy is violated.
Posted by: Chairman Mao at Jul 29, 2007 1:45:06 AM
Above, Mark Seecof comments critically on my paper on credit reporting. The main issue here is whether credit reporting agencies should be liable for "errors" in credit reports. Seecof seems to favor restrictions on free speech.
He might be right that I understated the ease with which one could view another's credit report ca 2000 when I wrote the paper. And I did overuse "exorbitant" in the expression "exorbitant damages." But those are minor.
On the main points, I don't think Seecof's reasoning works at all.
First, I don't see why he says I say credit reporting agencies would prefer the error of excluding accurate information to that of including inaccurate information. I say that there is a trade-off, and that nobody gains from either kind of error. I also point out that the later type are self-correcting.
But more importantly, his logic of "externality" is no good.
From my paper:
"By focusing exclusively on one type of error, critics engage in asymmetric thinking. If bureaus were made to pay exorbitant damages to consumers harmed by the
inclusion of inaccurate information, should bureaus not also be made to pay damages to businesses harmed by the exclusion of accurate information? Imposing symmetrical
responsibility would put the credit bureaus in a no-win situation and might strangle the goose that lays the golden eggs.
"There is another asymmetry in the complaints against credit bureaus: given that credit bureaus make possible much of the opportunity that consumers enjoy, why should consumers be able to sue credit bureaus when reports have errors, but not have to pay rewards to credit bureaus when reports do not have errors? Critics
are asserting, in essence, that consumers are entitled to have their credit report maintained without serious inaccuracies, but, symmetrically, one could assert that
credit bureaus are entitled to rewards from consumers when reports are accurate. It is wiser to drop the entitlement mentality and think in terms of contractual obligations."
There are interesting questions of the meaning of "error."
If a reporter says, "Smith said Jones stole the car," and Jones did not steal the car, and Smith did in fact say he did, the reporter is not erring. If TransUnion says Visa says Jones did not pay his bill, but Jones did and Visa said he didn't, is it an error? Arguably, no. If you read a credit report as saying "the creditors says the following", then most of the "errors" are not credit-reporting-agency errors at all.
As I discuss on p. 327, inaccuracies and misunderstandings arise from many sources other than the credit reporting agency, but they show up in the credit report. Shoot the messenger, and we will have less and worse delivery.
The government and courts do not have any knowledge about where inaccuracies actually originate. How about all the troubles that come from USPS failing to forward mail appropriately? Want to make them liable? If not, why the double standard?
Free speech has given rise to the great social accountability mechanism of credit reporting. Best to keep it as free as we can.
And, yes, I do oppose laws against libel and slander. The best rules are property and contract. Reputation should not be treated as a form of property (discussed pp. 341-342).
Again, the link to my paper is:
http://www.independent.org/pdf/tir/tir_05_3_klein.pdf
Posted by: Daniel Klein at Jul 29, 2007 3:39:26 AM
