Civil Rights in Credit Scores? A Report

Preliminary findings indicate a strong relationship between credit scores and claims

experience. Poor credit scores are associated with increased claims activity. Furthermore, the study

found that Black, Hispanic, young, and low-to-moderate income policyholders tend to have worse credit

scores than White, Asian, older, and high income policyholders. [Texas Department of Insurance: Credit Scoring Study (Dec. 30, 2004)]

Excerpts below under Read More.

Race & Ethnicity

The

Legislature directed the Department to address whether the use of credit
information has “any

disproportionate impact on any class of individuals, including
classes based on income, race or

ethnicity.” As described in detail below, the
Department has determined that in the individual

policyholder data, there are consistent patterns reflecting differences in credit scores, most notably,

between different
racial/ethnic classes. Other classes also present patterns, as discussed below

(pp. 10-11.)

For each data set, the Department compared the average and median credit

scores by
race and found a consistent pattern across all models. Whites and Asians, as a

group,
tend to have better credit scores than Blacks and Hispanics. In general, Blacks have

an
average credit score that is roughly 10% to 35% worse than the credit scores for
Whites.

Hispanics have an average credit score that is roughly 5% to 25% worse than
those for Whites. Asians

have average credit scores that are about the same or slightly
worse than those for Whites (p.

13).

Chart 4 shows that Blacks and Hispanics make up an increasing percentage of

the
individuals in a given credit score range as the credit scores get worse while Whites
make up

an increasing percentage of the individuals in a given credit score range as the
credit scores get

better. For example, the bar with the best credit scores (+40 to +45%)
shows that Whites make up

about 90% of the drivers. In the far left bar with the worst
credit scores (-25% and less), Whites

make up about 35% of the drivers. On the other
hand, Blacks make up about 2% of the policies in the

best credit score range and about
33% in the worst credit score range. In a pattern similar to

Blacks, Hispanics make up
about 5% of the drivers in the best credit score range and 28% of the

drivers in the
worst credit score range (p. 14).

Credit Score &

Risk

For personal auto insurance, the relationship between pure premium and

credit score
was examined. Chart 7 is characteristic of the data sets analyzed. It shows that

as
credit scores improve, the pure premium or average loss per vehicle decreases.
Conversely, as

the credit scores worsen, the average loss per vehicle increases (p. 18).

For homeowners

insurance, the data did not readily lend itself to a pure premium
approach given the wide range of

differences in insured values. Therefore, the
relationship between loss ratio and credit score was

examined. In this analysis, the loss
ratio was calculated using the premiums adjusted to the level

they would have been
prior to the use of credit scores. Chart 8 shows the average adjusted loss

ratio for each
decile of credit scores. (Charts for all available data sets can be found in the

Appendix.)
Like the personal auto data analysis, the homeowners data shows that as the

credit
scores improve the loss ratios improve (p. 19).

Frequency Not

Severity

The Department also looked at claim frequency7 and claim severity8. The

data shows
that credit score has a stronger relationship to frequency than severity. That is, as

the
credit scores improve, the frequency decreases, i.e. people have fewer accidents or
claims.

Severity may decrease as well, but not at the same rate as the frequency. For
some data sets,

severity is nearly flat (p. 20).

More Study

Scheduled

Charts 7 through 10 are based on univariate analysis; they consider

the relationship
between claims experience and a single variable (credit score). In reality there

are many
other variables that impact claim costs, including type of vehicle, ZIP code and age

of
driver. Further, many of these variables are plausibly related to credit score directly
(e.g.,

age of driver) or indirectly via another variable (e.g., high traffic congestion via
territory). For

example, high claims experience for younger drivers may reasonably be
explained by fewer years of

driving experience rather than their low credit scores.
Similarly, as an example of an indirect

relationship, high claims experience in certain
areas of the state may reasonably be explained by

factors such as high traffic
congestion and crime rates rather than low credit scores. Thus, the

issue is not whether
credit scoring is related to claims experience, but rather, whether credit

scoring provides
additional information, over and above traditional or existing rating variables,

which can
enable an insurer to more accurately predict losses. Additionally, it should be

ascertained whether the impact of credit scoring (both positive and negative) is
lessened due to

other explanatory variables.
To answer these questions, it is necessary to augment the univariate

analysis
discussed above by incorporating a multitude of other variables known to impact

claims.

The Department is in the process of conducting such a multivariate analysis using

the
individual policyholder data and will report its results by January 31, 2005 (pp. 22-

23).