Four consumer groups have dismissed the Federal Trade Commission report on credit scoring discussed in my July 20 blog entry as "biased insurance industry propaganda," and has called on Congress to "reject the defective study and ban the use of credit scoring in insurance." Click on to read the full release and respond. (I personally think that given the political calculus in Washington, nothing more will come of this until a Democrat is in the White House and Democrats extend their voting leverage in Congress. What do you think?)
Here is the release:
"Consumer and Civil Rights Groups Reject Federal Report on Insurance Credit Scoring"
"Fatally-Flawed Report Relies on Handpicked Data by Insurance Industry, Fails to Respond to Congressional Mandate"
WASHINGTON, D.C.--July 24, 2007--Representatives of consumer and civil rights organizations today condemned a congressionally-mandated report on insurance credit scoring by the Federal Trade Commission as biased insurance industry propaganda. The groups called for Congress to reject the defective study and ban the use of credit scoring in insurance.
Insurance credit scoring is the use by insurers of consumers’ credit reports for determining insurance eligibility and premiums. Unknown to most consumers, insurers’ use of consumer credit information has spread to almost all insurers and is one of the most important factors in determining how much a consumer pays for auto or homeowners insurance.
Previous studies by the Missouri and Texas Departments of Insurance have found that insurance scoring discriminates against low income and minority consumers because of the racial and economic disparities inherent in scoring. The Missouri study concluded that a consumer’s race was the single most predictive factor determining a consumer’s insurance score and, consequently, the consumer’s insurance premium.
Before the introduction of the credit scoring systems the insurance industry had used other unsupported standards and stereotypes with a racial proxy effect. After the major companies were sued for fair housing violations and were forced to eliminate these practices, the industry introduced a new practice – credit-based insurance scoring – that consumer and civil rights groups see as re-introducing racial and ethnic effects into the pricing of insurance.
The relationship between insurance credit scores and race is so strong that even though the FTC used data handpicked by the industry, it found that credit scoring discriminates against low income and minority consumers, and that insurance scoring was a proxy for race.
Representatives of the Consumer Federation of America, the National Fair Housing Alliance, the National Consumer Law Center, and the Center for Economic Justice said the FTC study is fatally flawed because the insurance industry controlled the data used in the analysis. Instead of requiring the submission of comprehensive policy data by a large number of insurers, the FTC used data handpicked by the insurance industry.
“The FTC’s approach to collecting data for the analysis is like the federal government trying to do a study on the health impacts of tobacco use with data selected by tobacco companies for the study,” said Allen Fishbein of the Consumer Federation of America. “By relying on handpicked data, the insurance industry was unnecessarily given opportunity to control the outcome of the study.”
The FTC study also confirms that, despite growing reliance on credit-based insurance scores, scant evidence exists to prove there is a meaningful connection between a consumer’s score and auto insurance losses. Without the need to demonstrate such a connection, insurers could use any consumer characteristic, such as hair color, to price insurance products.
“Despite finding no explanation for the alleged connection between insurance scores and losses, the FTC report somehow concludes credit scoring is valid and good for consumers. This is not an impartial analysis, but simply advocacy for insurers," said Birny Birnbaum of the Center for Economic Justice. Birnbaum, a former insurance regulator, has studied insurance scoring for over 15 years.
The groups also dismissed the report for failing to respond to the Congressional mandate to examine the impacts of insurance credit scoring on the availability and affordability of auto and homeowners insurance, and for parroting insurance industry propaganda about insurance credit scoring.
Section 215 of the Fair and Accurate Credit Transactions Act of 2003 required the Federal Reserve Board and the FTC to study the impact of credit scoring on the availability and affordability of credit and insurance and to determine whether credit scoring was truly related to insurance losses or simply a proxy for race, income or other factors.
“Incredibly, the FTC report downplays its own findings about the racial impact of insurance scoring – the primary question asked by Congress – and emphasizes the allegedly ‘predictive’ nature of credit scoring,” said Chi Chi Wu, staff attorney at the National Consumer Law Center. “It’s outrageous that the FTC says that ‘credit scoring is good for consumers’ when it has a disparate impact on minorities. The FTC appears to believe minorities aren’t ‘consumers’ worth protecting.”
Buried in the report is the fact that the alleged correlation between risk and credit-based insurance scores might be explained by other factors. Instead of pursuing these other factors, the FTC employed subjective and pejorative racial stereotypes to try to support the alleged link between credit-based insurance scores and legitimate risk.
“To add insult to injury, the FTC report mimics the insurance industry blaming-the-victim psychobabble of claiming credit history is related to responsibility and risk management. A look at the actual scoring models shows that socio-economic factors have more impact on the score than loan payment history and that an insurance credit score has little to do with personal responsibility and everything to do with economic and racial status,” said Shanna L. Smith, president and CEO of the National Fair Housing Alliance.
The group calls on Congress to reject this flawed and biased study and to tell the FTC to conduct an objective, independent study. In addition, based on the available evidence of racial discrimination, Congress should ban the use of insurance credit scoring.
Comments (11)
Why would a Democrat in the White House make a difference when a Democrat in the FTC did not?
Commissioner Jon Leibowitz was one of the four in the 4-1 vote to release this report. In his brief concurring statement, he said: "The differences in credit-based insurance scores across racial and ethnic groups are a disturbing reminder that our society is--still--not race-blind and that vestiges of our history of discrimination remain ever-present."
What is he saying here? Don't blame me for my vote, blame society?
The only dissenting vote was Pamela Jones Harbour, Independent.
SAM RESPONDS:
Regardless of Mr. Leibowitz's vote or party affiliation, Democrats are more likely to support the average citizen against Big Business (insurance included) than are Republicans (Sen. Trent Lott, R-Miss., notwithstanding, only because he has a personal axe to grind after battling with his own insurer over a Hurricane Katrina claim.)
Posted by Mary Concannon | July 24, 2007 3:38 PM
Posted on July 24, 2007 15:38
I'm really getting tired of the lies the consumer advocacy industry is spreading. I used to have a lot of respect for these groups, but now I think they are filled with a bunch of hacks.
The efficacy of credit scoring has been REPEATEDLY shown to be statistically significant, by multiple studies. To insinuate the correlation is spurious is either ignorance or a lie.
The press release has one outright lie: the Texas study REJECTED the notion that race mattered. The commissioner even wrote a cover note admitting his initial assumptions were wrong on this point! The Missouri study does show that race is correlated to credit scores. The FTC study is indeterminate.
One very important point is that insurers are NOT allowed to use actual race or income data. Thus, the claim must be that credit scores are surrogates for these variables. So I have a challenge for the consumer advocates. It's very simple.
The key to solving this problem are the multivariate models used for pricing, of which credit scoring is one variable.
If insurers were allowed to collect race and income data (properly "anonymized" so that individual policies could not be identified), they could "control" for these variables, collapsing any disparate impact. That would leave only the component of credit scoring which differs for reasons other than race and income, by definition.
I claim that credit scoring would still have significant predictive power. If this procedure were implemented, would consumer advocates then allow the residual credit scoring information?
Bob Hunter has been known to post here, and he's an actuary, so he should understand exactly what I'm saying. I'd be very interested in his answer.
If the answer is that "insurers can't be trusted with even the anonymous data," then we have our answer: consumer advocates aren't actually interested in helping consumers, they are interested in bashing insurers. At least we'd know.
I fully recognize how inflammatory some of my language is. I stand by all of it. Consumer advocates are being terribly dishonest and will potentially hurt the very people they are trying to help. People need to understand this.
Todd R. Bault, FCAS
Sanford C. Bernstein, LLC
1345 Avenue of the Americas
15th Floor
New York, NY 10105
Ph: 212 756-1857
Fx: 212 848-2370
todd.bault@bernstein.com
The comments herein are part of a larger body of investment analysis. For our research reports, which contain information that may be used to support investment decisions and appropriate disclosures, please see our website at http://www.bernsteinresearch.com
Posted by Todd Bault | July 25, 2007 8:59 AM
Posted on July 25, 2007 08:59
In one way, the FTC report helps prove credit scoring is not a valid class. the FTC reports that it cannot determine what credit score correlation is measuring (i.e., why, if a person is laid off because of outsourcing or illness or something, and falls behind with some bills, they are worse car drivers).
In other words there is no thesis being tested by the correlation, just a correlation of loss ratio with some data.
Years ago, California's DMV found a statistically significant correlation between accident record and hair color. Should that be used? Credit scoring is like hair color...no logic behind it.
In the old days, when actuaries rather than data miners set classes, there was a thesis that was tested. For example, we think people with lots of accidents and tickets in the recent past will continue to be worse drivers in the coming year. Collect the data and it is true.
With credit scoring there is no thesis like that, and the data is not available for people to review (but the safe driver plan data was available for review).
A black box with no thesis is shifting sands upon which to build a class system.
Posted by Bob Hunter | July 25, 2007 11:29 AM
Posted on July 25, 2007 11:29
Tremendous post, Todd. The reality is that consumer groups are opposed to risk- based pricing. This would seem to be contrary to their stated mission, but they have repeatedly demonstrated this.
Question: Where do most consumer groups get their funding? Who pays Bob Hunter's salary?
Posted by Anonymous Actuary | July 25, 2007 11:47 AM
Posted on July 25, 2007 11:47
I have seen credit scoring accurately reflect the actual loss ratio of a company's book of business. Anyone who has seen that demonstration will have a hard time dismissing the accuracy of credit scoring as a predictive underwriting tool.
I was under the impression that even though the factors making up the credit score were in a "black box," they had to be approved by the various state insurance departments before a company was allowed to use them. Or the insurance department had the right to order the company to make changes after the fact.
Presumably, those folks would be inclined to follow the mandates of law that preclude the use of race, religion, etc., as factors.
Our society has decided that it is best for all if decisions regarding people in general are racially neutral. This leaves me to wondering, can factors that are solely based on allowable, non-racially discriminatory conditions come up with actuarially sound results that follow racial lines? Do the people in the insurance departments really look at the factors that make up the credit scoring model? If not, they are letting us down.
If there are racial factors in the credit scoring model, it is time to open the "black box" and make those factors known, and then eliminate them.
But what if there are no racial factors, or they are given only minimal weight?
What does that tell us?
Posted by Susan Maloney | July 25, 2007 12:37 PM
Posted on July 25, 2007 12:37
"Anonymous" boldly requests info on who pays my salary. Fine. The answer is no one does.
I have been doing my consumer advocacy pro-bono since I began the National Insurance Consumer Organization in 1980. That is over 25 years of pro-bono service so far (when you subtract the time I did get paid, when I was Texas Insurance Commissioner). I do it as a service.
I also work on tribal reconciliation as a ministry in Africa, pro-bono.
I do work as a consultant to governments and consumers for money, enough to limit my practice to less than one-third of my work hours.
Now a question for you, anonymous. Who are you, and who pays your salary?
Posted by Bob Hunter | July 25, 2007 2:51 PM
Posted on July 25, 2007 14:51
Bob Hunter said:
In one way, the FTC report helps prove credit scoring is not a valid class. The FTC reports that it cannot determine what credit score correlation is measuring...In other words there is no thesis being tested by the correlation, just a correlation of loss ratio with some data.
Bob, there is no other way to put this: You are either willfully ignorant or lying. I think it's the latter, and I never make such an accusation lightly. But I have just had it.
I cannot believe that you are unaware of the many other studies out there which say exactly what is the mechanism of credit scoring, which is why I must assume you are lying.
Lower credit-scoring customers have higher frequency: they make more claims. In fact, there appears to be no severity issue (best I recall--could be wrong here).
Also, the higher frequency is quite concentrated among the lowest credit scores, so there is a clear behavioral issue in play. There never has been a causality requirement in underwriting: indeed, there cannot be, as that would require unethical social experimentation.
All class ratemaking is by correlation. Your hair color crack is insulting to those of us who actually know what we're talking about.
If you are truly unaware of all of this, then I gladly take back my "lying" accusation, sincerely apologize, but then am forced to substitute "incompetent." Just let us know which one applies.
As sources of pay have been brought up, everyone knows I am a stock analyst covering insurance. In that capacity, it makes not one bit of difference whether or not insurance scoring exists--all I have to care about is what happens to the stocks, up, down or sideways.
I care about this issue because insurance scoring has been a huge pricing innovation that makes more capacity available owing to the decreased uncertainty. In other words, I care about the consumers. I walk the walk, even if I don't wear a light-up "Consumer Advocate" hat.
I have great sympathy to affordability issues, and there are ways to compensate people who cannot afford needed insurance. But credit scoring is a positive boon to the availability problem.
Anyone who has been around the insurance industry for a long time--that means you, Bob--knows that messing with market pricing always results in an availability problem. Look at Exhibit A: Florida.
Finally, you don't even acknowledge my direct challenge to you! Unbelievable. I have to assume that you refuse to answer my question--would insurance scoring be okay if we controlled on race and/or income and kept the residual impact--because I got the answer right to begin with.
You just want to bash insurers, not help consumers. I think we now know where you stand.
Todd R. Bault, FCAS
Sanford C. Bernstein, LLC
1345 Avenue of the Americas
15th Floor
New York, NY 10105
Ph: 212 756-1857
Fx: 212 848-2370
todd.bault@bernstein.com
The comments herein are part of a larger body of investment analysis. For our research reports, which contain information that may be used to support investment decisions and appropriate disclosures, please see our website at http://www.bernsteinresearch.com
Posted by Todd Bault | July 26, 2007 7:38 AM
Posted on July 26, 2007 07:38
I believe the latest comments need to be toned down. They are becoming personal diatribes.
I do have a couple of questions I would like to ask of Todd and those who concur with his analysis of credit scoring.
If there is a direct correlation between credit scoring and higher frequency of claims, why then aren't all P&C carriers using this underwriting methodology?
As a 30-plus-year producer who works with many carriers in the personal lines market, I know for a fact that not all do.
Secondly, prior to credit scoring being used, most carriers relied heavily on a driver's motor vehicle record (MVR) and claims history (CLUE) in determining eligibility criteria and premium rating.
Past history was used to predict future claims experience.
Are you saying that past credit history is a better predictor of future frequency of claims than historical driving experience?
If so, then it needs to be demonstrated and in an objective manner.
If not, or if it's just as good, then we need to exclude a system that puts an emphasis on socioeconomic factors. Because the perception is is that credit/insurance scoring is discriminatory, and, as we know, perception is reality.
I await your response.
Posted by Steve Daroff | July 27, 2007 1:22 PM
Posted on July 27, 2007 13:22
Steve Daroff asked many good questions and made several good points:
Steve said, "I believe the latest comments need to be toned down. They are becoming personal diatribes."
No, they are not. Read carefully what I wrote. There is nothing personal about them.
There is no reason for people in positions of power to abuse that power through lies and deception. It's bad when insurers do this (and they do), but it's also bad when consumer advocates do it. It's worse when consumer advocates are advocating policies that don't help consumers. If you object to my using the word "lie," so be it.
Steve also said: "If there is a direct correlation between credit scoring and higher frequency of claims, why then aren't all P&C carriers using this underwriting methodology?"
Most personal line writers ARE using this method, for auto and homeowners. Variations of it are being tested in small commercial. It wouldn't be expected to work much further, because it requires high frequency of claims with relatively low severity. Thus, this method is not used for catastrophe pricing in homeowners, or in most of commercial lines.
Steve said: "As a 30-plus-year producer who works with many carriers in the personal lines market, I know for a fact that not all do."
If a customer wants to go to a small regional company that says it will not use credit scoring, they should do it and you should help them. If more consumers do this, insurers might stop using credit scoring, though I doubt this given its predictive power.
This method requires lots of data and initial setup time, so it may not be feasible for smaller companies to use it. But that's competition for you.
Steve also said: "Secondly, prior to credit scoring being used, most carriers relied heavily on a driver's motor vehicle record (MVR) and claims history (CLUE) in determining eligibility criteria and premium rating."
They still do. Driving record has not been dropped, credit scoring has been added. The two together are extremely powerful.
Steve said: "Past history was used to predict future claims experience. Are you saying that past credit history is a better predictor of future frequency of claims than historical driving experience?"
Yes, in a key way. Credit scoring better predicts the class of claimants most likely to make a large number of claims. This is what makes it different from driving record, and points to a non-driving source of claims.
This is particularly true in homeowners, which suggests using insurance as a maintenance policy rather than an indeminty policy.
Steve said: "If so, then it needs to be demonstrated and in an objective manner."
It has been--see this paper: http://www.casact.org/pubs/forum/00wforum/00wf079.pdf .
This is part of why mr. Hunter's last post was so dishonest. Of COURSE this has been demonstrated to be valid, many times, yet he muddies the water like a politician (which is he, essentially) when he CERTAINLY knows of this paper, as well as many others.
It's old now, but it is one of the best for laying out the core issues in an understandable way, yet maintains enough complexity to be useful.
Steve said: "If not, or if it's just as good, then we need to exclude a system that puts an emphasis on socioeconomic factors. Because the perception is is that credit/insurance scoring is discriminatory, and, as we know, perception is reality."
Perception is reality only if we allow our emotions to overwhelm intellect. We do this all the time, of course, to our great shame, as it damages people we claim to be helping. This is why my rhetoric seems so heated to you.
I'm more of a consumer advocate than the CFA is. I actually want to do things that help consumers, regardless of its impact to insurers, help or hurt. The CFA seems to require that consumer advocacy bashes insurers. That's wrong, and I'm obliged to say so.
As I noted, the actual data on unfair discrimination is mixed at best. Also, as I am now repeating for the third time, the models used CAN SOLVE THIS PROBLEM IF IT EXISTS.
Race and income data could be held by state regulators, attached to an insurance rating plan without the company's intervention, and the state could make the final calculation to control on race and income, leaving only that piece of credit scoring not related to race or income.
Here I've added a way to keep the data out of insurer's hands, which is an important point--insurers DON'T have this data!
I would be willing to bet real money that if we did this, credit scoring would still be very significant. So perhaps you will say what you think about this idea, Steve, as Mr. Hunter clearly won't.
Again, great questions, Steve. Thanks for asking them. If you read that paper, I think you will be very surprised.
Todd R. Bault, FCAS
Sanford C. Bernstein, LLC
1345 Avenue of the Americas
15th Floor
New York, NY 10105
Ph: 212 756-1857
Fx: 212 848-2370
todd.bault@bernstein.com
The comments herein are part of a larger body of investment analysis. For our research reports, which contain information that may be used to support investment decisions and appropriate disclosures, please see our website at http://www.bernsteinresearch.com
Posted by Todd Bault | August 1, 2007 8:43 AM
Posted on August 1, 2007 08:43
When I was the Federal Insurance Administrator, about 30 years ago, I appeared at a Senate hearing on medical malpractice. Along with the then Department of Health, Education and Welfare, we had issued a report stating that the explosion in rates was not due to an explosion in claims, as best we could tell from the then limited closed claim data.
One Senator was outraged at our findings and bitterly attacked us. I started to respond to his rather weak arguments when my co-testifier, HEW Secretary Casper (“Cap”) Weinberger, passed me a note that read, “Bob. A word of advice. Never match wits with nitwits.”
I have tried to keep this wise counsel, but maybe I will stray from it just a bit today.
I did, in my first post in this series, answer Todd’s question. He--albeit being the only one, apparently, who “actually know(s) what we’re talking about”--was too immature in his thinking to understand that I had answered the question.
The answer was, obviously, “no,” because there was no thesis being tested by even a residual correlation.
All-knowing Todd apparently does not understand the difference between a thesis and causality, either. Here are the Merriam-Webster definitions of interest:
“THESIS: a position or proposition that a person (as a candidate for scholastic honors) advances and offers to maintain by argument; a proposition to be proved or one advanced without proof.”
“CAUSALITY: the relation between a cause and its effect or between regularly correlated events or phenomena.”
Get the difference, Todd? Thesis is a logical underpinning and causality is a cause-and-effect relationship. You can have a class based on logic AND have no cause-and-effect relationship. In fact, that is what all classes I know about--other than credit scoring--are.
I do not believe that a class should be approved that does not have a logical underpinning--i.e., a proposition to be proved. A correlation then makes sense by proving or disproving the idea proposed. A correlation is necessary but not sufficient by itself. A point I tried to make clear with the hair color example.
By the way, California voters understood this when they passed Proposition 103’s three prime auto insurance classes of driving record, miles driven and experience. California has not allowed use of credit scoring in insurance, to its credit. Some states have followed and more will follow.
In his later post--the one another writer labeled a “personal diatribe”--Todd is unable to figure out my obvious “no” to his earlier question and continues to attack me as “…either willfully ignorant or lying.” He thinks I am a liar because I do not agree with the studies, “which say exactly what is the mechanism of credit scoring.”
Hey, Todd, lighten up… to paraphrase a Bill Clinton campaign slogan, it is not the “mechanism” of a correlation, it is the lack of a logical base, stupid.
As we watch the wheels come off of the FTC report on credit scoring--which we rightly pointed out when the study began was using flawed data--I recommend that wise supporters of credit scoring refrain from using the FTC findings. If that is the best you got, you are in deep trouble!
Just as Cap Weinberger offered me advice, I will offer this piece of wisdom to Todd:
Todd, if you are going to blog “personal diatribes” and rant and rave, and call people names, and be illogically in rage, don’t put an advertisement for your employer as the last paragraph of each post. That is not a smart way to make points with your boss.
Posted by Bob Hunter | August 2, 2007 8:00 AM
Posted on August 2, 2007 08:00
I read your column frequently and thought I might respond to your latest.
As an agent for over 40 years we don't get much credit, and are always caught in the middle for things like credit scoring.
Back in the 1980s we had to deal with CLUE reports and started a landslide of paperwork for agents.
Some of the information that was received by the companies on behalf of their insureds was incorrect, but we as agents still had to deal with it at, of course, no cost.
Now we have credit scoring. No one can really explain it because it is another underwriting idea far from perfect.
There is one simple question that even the insurance departments can't answer, and can't show us the loss runs that show people with lower scores have more claims.
All this boils down to is rate increases. It's the same way health insurance works.
If you work for a large company with several hundred employees, your individual rate is lower.
If the same person changes to a smaller company or is self-employed, his rate is off the charts.
That person's risk of loss has not changed, but is penalized for being small. That why we need loss-cost health insurance.
Each state knows what their citizens are paying for health care. You divide that by X number of people, age, ZIP code and so on, that number would be the rate plus the company's expense factor.
This would be the rate reqardless of the size group you are in, or all the in- and out-of-network games companies are playing.
Who pays the premium should be left up the employer and employee.
The concept of the law of large numbers would work because we would be all in the same qroup (working people).
This is what has been done in the fire and causalty rate in our state, including workers' compensation, and the rates have gone down and there is more competition.
Posted by Don Lafferty | August 17, 2007 3:52 PM
Posted on August 17, 2007 15:52