The Impact of Credit-Based Insurance Scoring on the Availability and Affordability of Insurance - Part I

By Lawrence S. Powell, Ph.D.
September 24, 2008

Following is the testimony presented by Lawrence S. Powell, PhD before the United States House of Representatives Financial Services Committee Oversight & Investigations Subcommittee on May 21, 2008, by:

Advertisement

Lawrence S. Powell, Ph.D.
Research Fellow - The Independent Institute (http://www.independent.org/), and
Whitbeck-Beyer Chair of Insurance & Financial Services
University of Arkansas-Little Rock
****************************************

Introduction:*
(* Much of this testimony is drawn from a study being written for the Independent Institute.)

Insurance companies face an unusual challenge. They must set prices for the products they sell before they know all of the costs. To meet this challenge, they employ complex pricing methods developed by actuaries using applied economic and statistical techniques. It should then come as no surprise that some aspects of actuarial science and insurance pricing are puzzling to people who have not developed substantial expertise in this field.

Insurance scoring, the use of credit information in insurance underwriting and pricing, is an example of a beneficial practice that is sometimes misunderstood. Insurance scoring benefits consumers in several ways, all of which stem from its accuracy as a predictor of insured losses.

The purpose of my testimony is to present comprehensive information about insurance scoring in a non-technical format. In Section 1, I present a brief conceptual summary of insurance pricing and insurance scoring. In Section 2, drawing from existing studies, I present evidence that insurance scores are powerful and accurate predictors of insurance losses. In Section 3, I conclude with discussion of the appropriateness of insurance scoring.

Section 1: Insurance Pricing and Insurance Scoring

An insurance company facilitates risk pooling, reducing the uncertainty of individual pool members. Uncertainty decreases because the ultimate value of the group's losses is more predictable than that of an individual. Swiss mathematician Jacob Bernoulli first proved this phenomenon, known as the law of large numbers, around 1690. Relying on the law of large numbers, a group of pool participants can each pay the average or expected loss of the group, rather than paying for a much less predictable and potentially larger individual loss on one's own.

Risk pooling is most effective when all members of the pool have the same expected loss. Insurance companies rely on risk classification systems to ensure that groups of insureds pay premiums commensurate with their exposures to risk. When insurers pool exposures with unequal expected losses, the low-risk group must subsidize the high-risk group. This creates an incentive for low-risk pool members to purchase less insurance than high-risk pool members, a scenario called adverse selection. Adverse selection can break down the risk pooling mechanism and, in extreme cases, lead to insolvency of the pool.

Insurance companies use information about applicants for insurance to classify them into groups with very similar expected loss. Of course, no risk classification system is perfect. In addition to other restrictions, insurers can only use rating information if it is cost effective; meaning the cost of obtaining the information is less than the difference in expected loss between groups. For example, assume there are only two types of drivers, low-risk and high-risk. The low-risk group has expected loss of $500 and the high-risk group has expected loss of $700. If it costs more than $100 to classify a driver, it will be more cost effective to simply pool them together and charge both groups $600. However, if an insurer can identify low-risk drivers for, say, $20, it benefits the low-risk drivers to charge them $520, and charge the high-risk drivers $720. On the other hand, insurers could be more precise in risk classification if they hired private investigators to follow each driver for six months before offering an insurance policy. Obviously, this would cost more than $100, and raise privacy concerns. To have enough money in the risk pool to cover expected losses, low-risk drivers would have to pay more than $600. In this case, there is no justification for such an unfair classification.

There are many variables insurers use to classify drivers based on expected loss. These include, but are not limited, to geographic location, age, gender, marital status, miles driven, type of vehicle, use of vehicle, driving record and insurance score. An insurance score is a numerical prediction of propensity for loss estimated using certain information from a driver's credit history. The actuarial literature shows it is one of the most accurate and cost effective loss predictors available (EPIC, 2003).

There are several apparent misconceptions about insurance scores. To understand why insurance scores are beneficial to insurance systems, it is important to start with an accurate description that is free of incorrect assumptions. The variables commonly used to estimate insurance scores include measures of performance on credit obligations, credit-seeking behavior, use of credit, length of credit history, and types of credit used (FTC, 2007). They do not include income, wealth, race, ethnicity, or any prohibited factor.

Insurance scores and credit scores are calculated using some of the same information, but they are not equivalent. The important difference is that credit scores use these variables (and others) to estimate the probability of a borrower defaulting on a financial obligation, while insurance scores estimate the probability of having insured losses.

An important fact often overlooked in the debate about insurance scoring is that the only way including insurance scores in an insurance rating model can result in higher premiums is for the sample population with lower scores to have more insured losses. As I describe in more detail in Section 3, any deviation from using the most accurate, cost effective predictors results in unfair outcomes and damage to the insurance mechanism.

One observed barrier to understanding insurance scoring is manifest in the common criticism that there is not an intuitive link between insurance scores and driving ability. While several studies develop potential causal links between insurance scores and driving, I find it more compelling to recognize an alternative relation. The use of insurance scores does not rely on a link between credit information and "driving ability." Rather, it is a link between insurance scores and insured losses.

There are many factors unrelated to driving ability that increase the likelihood of insured losses. For example, someone who always makes debt payments on time to avoid higher interest rates the next time they borrow may also choose not to file a small insurance claim to prevent an increase in insurance premiums in the future. It may also be the case that insurance scores measure hazards other than lack of driving ability.

(The views expressed in this article/commentary are solely those of the author and do not necessarily represent the views of MyNewMarkets.com, the Insurance Journal or Wells Publishing.)

This is the first of a five-part series. The remaining commentaries can be found at www.mynewmarkets.com under the "Articles" tab.

Credit Scoring Insurance Series

1. The Impact Of Credit-Based Insurance Scoring On The Availability And Affordability Of Insurance - Part I
2. The Impact Of Credit-Based Insurance Scoring On The Availability And Affordability Of Insurance - Part II
3. The Impact Of Credit-Based Insurance Scoring On The Availability And Affordability Of Insurance - Conclusion
4. Credit-Based Insurance Scoring And Measurement Error In The FTC Race Proxy Finding
5. Insurance Credit Scoring And Near Death Experiences