Understanding credit rating score

CPCU panel probes the mechanics and the rationale for controversial tool

The Society of CPCU's annual meeting in Orlando last October included a panel on credit scoring which featured companies that are active in providing credit data to insurers. The panel consisted of the following persons:

William T. Atkins, CPCU (Moderator), V.P. Personal Lines for Pacific Insurance

Lamont D. Boyd, CPCU of Fair Isaacs, Inc.

Gary Skerl, Progressive Insurance (responsible for building their Credit Base Scoring algorithm)

John B. Wilson, Risk Modeller for Choice Point

Gregg L. Antenen, President, Convergent Data

Seminar moderator William Atkins admitted in his introduction that creditbased scoring is facing backlash. Several experts on credit-based scoring joined to discuss different aspects, with a focus on how to speak to the public about this critical issue. The seminar then opened with each panelist making some comments. A question and answer period followed.

Lamont Boyd, who has a great deal of experience in credit-based scoring through his work at Fair Isaacs, explained that their business purpose is to generate data and analytics to help businesses make more efficient decisions. In his opinion, the development of credit-based scoring falls precisely within that purpose. He emphasized a distinction between credit scores (predictors of creditworthiness) and credit-based insurance scores (predictor of loss propensity). Speaking of the 1996 NAIC White Paper, he pointed out that it referenced the Tower-Tillinghast study that independently confirmed a significant correlation between credit scores and loss ratios. Boyd also noted that nothing is used in credit-based scoring formulas that can negatively affect (unfairly discriminate against) insurance consumers.

Gary Skerl from Progressive Insurance Company said he is a believer in the correlation between credit-based scoring and loss predictability. Progressive uses its own algorithm to create its C-B scores and is developing the following:

1. New Generation Credit Scores-a simplified model that uses nine rather than 16 variables. The algorithm calculates scores by assigning everyone a base score of 100. Then they either deduct or add to the base to create a score ranging between 49-228, with the lower number being more desirable. Their model will be used nationwide and will be filed in those states requiring such plans to be filed.

2. How to deal with consumers who have either no credit history or insufficient credit histories ("no-hits or thin files")-Skerl mentioned that some states are requesting that this class of insureds be treated (grouped) as average or best credit score groups. The elderly no-hits, which is a small segment of no-hits, has better loss experience, so Progressive plans to break out this group vs. younger drivers who do not have established credit.

3. Credit Assistance Team-The team responds to consumers who call a toll-free number with concerns over how they've been affected by use of their credit history within an insurance transaction. The company will use the team to help consumers in the following manner:

a. Personal Insurance Credit Report-Skerl explained that this document is a two-page breakdown that shows an individual's variable score and how it compares with aggregate scores.

b. Progressive will consider extraordinary events that affect scores and also consider past credit history. For example, it will give full consideration to a crisis medical situation or the fact that a person's past history had been quite positive for an extended amount of time, but may have just recently deteriorated.

c. The company's Credit Assistance Team will help with correcting credit history errors by showing insureds/applicants whom to contact in order to take care of credit report problems.

John Wilson of Choice Point explained the development of his company's approach. Rather than using proprietary information, Choice Point chose to use sources for its scores that would allow it to share its model information with agents, regulators, consumers and others. A Power Point presentation explains credit-based underwriting scores to agents. Choice Point has met with regulators to discuss how they created their scoring model (in the hopes of gaining greater regulatory acceptance).

Wilson addressed the possibility that regulators might ban the use of credit-based scoring. In anticipation of that possibility, Choice Point is performing studies on alternatives, such as loss history on other lines of business (e.g., homeowner loss history as a predictor of auto loss) or information on prior coverage history.

Gregg Antenen noted that his company, Convergent Data-which is less than two years old-has a live product that uses sources other than credit to predict the likelihood of losses. Those sources include check writing and sub-prime data. Gregg pointed out that individuals write checks in many instances and their respondent companies collect information on:

* How many times an individual writes a check

* The amount of the check

* The reason for writing the check

* Number of times checks have bounced

Much of the information comes from Telecheck Services. Convergent also makes extensive use of a unique bureau that collects information from various services such as collection agencies, check-cashing services and rental centers. This information can enhance creditbased scoring because these sources are predictive but do not report to credit bureaus.

The data can supplement and refine credit bureau source information, as well as help to flag problem groups and improve rating. Gregg's company is able to develop additional information on "no-hits" and "thin files." Their services help to create actionable scores for persons in these groups.

Moderator Atkins then allowed for questions from the audience.

Question: What is the underlying reason for backlash against the use of credit-based scores?

Boyd: indicated that the problem could be related to persons who have been negatively affected by use of private information. Also, he did not think that the concept of credit-based scoring was explained very effectively. They (Fair Isaacs) and other companies are making more information available through their Web site in order to help the public better understand their scores and also to provide guidance for improving their score.

Skerl: expressed a belief that the industry could do a better job. He cautioned, however, that as companies provide more explanations, the door could be opened to increased numbers of questions about the predictive nature of the information. He emphasized that a bridge needs to be built over the gap in understanding classpredictive behavior.

Question: Will there be a backlash on the use of sub-prime credit to focus on lower economic class segments?

Antenen: replied that his company (Convergence Data) does not focus on lower economic classes, but rather on possible negative use of credit facilities (implying that this should not cause any negative public reaction).

Question: Other than paying bills on time, how can insurance consumers improve their credit scores?

The panelists agreed that it is important to reduce credit balances. It also is important to check credit history for any errors and aggressively correct them. It would also help to seek any information offered by insurers or vendors about how to monitor elements of the scoring models that an individual can control.

Question: How can vendors take frequent inquiries into account? (Note: Credit reports also include information on how many times different parties request a person's credit history. This element often negatively affects a person's score.)

Boyd: said that Fair Isaacs groups inquiries, particularly those related to mortgages, into a single inquiry.

Skerl: admitted that Progressive is playing "catch-up" on this issue, studying how to minimize the effect of inquiry frequency on scoring.

Question: How can insureds deal with the inefficiencies of credit bureaus in handling errors?

Wilson: said he believes that current laws provide bureaus with an incentive to deal with correcting information. He also said that, largely, bureaus accurately report data.

Boyd: pointed out that one recent studyl from the Insurance Research Council revealed that MVRs are not very accurate and credit-based scores help to supplement traditional, accepted underwriting criteria.

Skerl: remarked that there aren't enough studies available to prove the accuracy (of credit bureau information). In his opinion, it is up to individuals to ensure that their credit information is correct.

Question: When do companies that offer credit-based insurance scores have to update their scoring models?

Boyd: stressed that it is important to re-evaluate and revalidate scoring models to make sure that credit management patterns are reflected (in them).