When I posted a blog entry on Sept. 19 about NU Associate Editor Mark Ruquet's column in the Sept. 18 print edition of National Underwriter, citing on a personal level the drawbacks of allowing insurers to use credit scoring to price coverage, it created quite a stir, drawing a wide range of comments. After carefully considering how readers responded, he's weighing in with some counterpoints of his own. Read on.
Mark Ruquet, NU Associate Editor:
It is great to see that people read my column either here on this blog or in NU, and are eager to respond. Unfortunately, the comments filed here seem to underscore what I wrote.
To the defenders of credit scoring--you missed my point. As an analysis of risk, it may well be an excellent tool in the actuarial process to help measure risk. It knows no race, color or creed and only indicates how responsible an individual is. Theoretically, a person who is responsible with their wallet should be responsible behind the wheel.
Yes, credit scores are popularly used by the business community, and its popularity is growing. But, popularity does not always equate to wisdom, especially for the business community.
Anybody remember junk bonds? A popular investment tool that ended up putting a few people in prison and untold losses among investors.
For those opposed to the use of credit scores--insurers don’t use them as the sole metric for determining risk. There are a number of factors that go into the mix. It is actuarial science. Anyone in this business who relies on one piece of information to determine risk is a fool, and will one day join the cavalcade of insolvency.
However, all of this is irrelevant to my main complaint.
The problem with credit scores is they are not reliable. One individual I spoke to said there is a 10 percent error rate in credit reports. Another study said the error rate is 79 percent, 25 percent of which are serious enough to result in the denial of credit.
Are the numbers right? We can spend all day debating that point among those with narrow political agendas. But there is no denying errors are made. There are credible questions about the reliability of the information that goes into creating the scores. Whether it is 10 percent or 79 percent, you are looking at millions of people. Those are real numbers.
Those supporting credit scores need to take off their blinders and insist of the reporting agencies that the information they provide be accurate. Credit scoring may be useful, and harmless from an actuarial perspective, but if the numbers are bad going in, might not the results be equally poor coming out?
Comments (5)
Mark, YOU miss the point! Credit scoring works well DESPITE THE ERRORS you cite! That means one of several possibilties:
1) The errors are unimportant. That is, errors in one part of the credit record are not enough to offset the correct parts of the record. The fact that credit scoring does NOT rely on a single source is testiment to this. This is the Law of Large Numbers at work, and it's a powerful law.
Bottom line: individual policyholder errors do not seem to prevent them being put in the correct rating bucket.
2) The regression methods used in credit scoring are good at seeing through the noise. This is a variant on 1), but an important difference--the model itself is robust and helps clean up statistical errors.
3) It's all luck. Unlikely.
At best, your point is that the errors undermine confidence in the system. Which was MY point--we are allowing emotion to override the truth.
Nobody is saying ignore data quality (certainly not me--I'm an actuary, and data quality matters). But your argument simply is not as powerful as the fact that credit scoring seems to well address all of your concerns. All that is left is irrationality.
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 www.bernsteinresearch.com
Posted by Todd Bault | September 22, 2006 10:41 AM
Posted on September 22, 2006 10:41
I would definetely support laws to make the credit bureaus and the creditors who report to them responsible for any errors they have caused.
As long as the punishment, i.e. fines, is large enough to make the bureaus and creditors change their processes, that should do the trick--something in the six-figure range for the first offense.
One interesting thing would be if the insurance industry led the fight for these laws, which in turn could lead to more accurate information and possibly help the consumer with better rates.
Considering how the carriers who use credit in some form or another have been made to look bad by most consumer groups, this would be a good way for the industry to help improve its image.
Posted by David Voteau | September 22, 2006 12:30 PM
Posted on September 22, 2006 12:30
A question, to which I have to confess I do not know the answer: How accurate are the other factors used in risk models?
How many errors do MVR reports have?
How about flood zones?
The issue, surely, is not that this is an imprecise source of data (name one that is precise), but whether or not credit data is fundamentally less precise than other sources?
One of the reasons for using multiple factors--and credit-based data rarely accounts for more than about 15% in the models we see--is to manage the risk of poor data.
Credit data must be worse than average for this complaint to have validity. Is it?
Posted by James Taylor | September 22, 2006 12:41 PM
Posted on September 22, 2006 12:41
Having worked at a national credit reporting agency (CRA) and now in the insurance industry helping "defend" the use of credit-based insurance scores, I have to comment.
While I do not want to defend the CRAs, as that is their job, I have a few thoughts on the accuracy issue.
Any database, list, phone directory, etc. will have errors (to which I ask: What is an acceptable error rate for you? For PIRG?).
The key is, are the errors material to a credit decision?
The "studies" cited by "Consumer Reports" were often from biased sources and counted having 'street' instead of 'avenue' as an error.
Also, think of this--there are more than 220 million credit active consumers in the U.S. With even a .001% error rate at any given time (a remarkably low/near impossible rate), how many reports have some type of "error?" 2,200-plus.
That sure makes good fodder for articles, self-styled "consumer advocates," grandstanding politicians, etc., to attack the end users, when clearly a vast majority of consumers benefit from the use of credit scores.
The bottom line is, if it didn't work, be it for creditors, landlords, employers, cell phone providers, the government and insurers, it wouldn't be used.
Please contact me if you wish to discuss further.
Thanks,
Jeff Junkas
American Insurance Association
Posted by Jeffrey Junkas | September 22, 2006 1:18 PM
Posted on September 22, 2006 13:18
While we all would agree that acccuracy in credit reporting is essential to the use of credit scores as one element in insurance pricing, let's not lose sight of a fairly robust rate pursuit industry that helps remind some consumers that they do drive to work 20 miles every day, or that there is a teenager somewhere in the household, or that a car registered in Florida sits in a garage somewhere on Long Island...
Accuracy in reporting is a chronic problem for our industry that the credit score--even with its occassional significant error--has helped alleviate.
So let's push hard to get the credit bureaus and the institutions that report to them to strive for greater accuracy. But let's also be sure to find better ways to solve the broader inaccuracies in the customer profiles and driving history that go into our complex ratemaking models.
Posted by Frank Cacchione | September 22, 2006 3:02 PM
Posted on September 22, 2006 15:02