Underwriting Insight: Out with the Old

Revised: 09/22/2006

Underwriting Insight: Out with the Old
With its widespread use, credit scoring no longer delivers a competitive edge.

August 2004, by Chuck Boucek, Best's Review

One of the major technological innovations in insurance over the past 20 years was the introduction of credit scoring. Beginning with private passenger automobile insurance and branching into other lines, credit scoring has given insurers greater insight into the profitability of individual policyholders. Insurers gain advantage from this insight by making more effective underwriting and pricing decisions.

The fact that a vast majority of companies now use credit scoring means it has become a defensive strategy rather than a strategy to gain competitive advantage. Thus, despite its tremendous impact on the industry, credit scoring is losing its appeal. As Harvard economist Michael E. Porter describes in his book, Competitive Advantage, technological innovations pass through the following four phases (and credit scoring is no exception):

-- Introduction, in which few companies are using the technology;

-- Growth, with companies rapidly accepting the technology and changes focusing on product innovations;

-- Maturity, when product innovation slows and process innovation takes over, resulting in mass production of the technology; and

-- Decline, with investment in the technology reduced due to diminishing returns.

In auto insurance, credit scoring is at maturity, with most companies using the scores of a small number of providers.

Arguably, credit scoring is beginning to move into the decline stage since companies no longer gain significant competitive advantage from its implementation. Companies will question the cost for implementing it in the near future. So is this the end of credit scoring or is innovation on the horizon?

Porter's framework leaves room for rebirth. He describes how a flexible technology can give rise to future product differentiation and prevent a technology from going into decline. Predictive modeling, the broader umbrella from which credit scoring was born, has quietly been evolving and has the potential to generate this rebirth.

Predictive modeling works by appending an external data source -- for example, credit characteristics -- into a company's internal data. A series of analyses and modeling methodologies are then applied to determine the individual credit characteristics that are most predictive of a policyholder's profitability. These credit characteristics are used to produce a model that estimates a policyholder's loss ratio.

Predictive modeling can and has been applied to study real world problems beyond the use of credit data to improve the quality of underwriting decisions. Examples include claim fraud detection, customer retention and response rates to marketing campaigns. One example that holds tremendous potential is the application of predictive modeling to better understand the agencies that are most likely to become profitable and deliver high growth.

To date, companies have a fairly basic tool kit to use in assessing with which agencies to align themselves. Currently, carriers perform personality testing to determine an individual's ability to sell. Also, in the case of an independent agent, the company can look at performance with other companies in the agency's office.

However, there are much richer data sources for assessing future success. Combining data from the U.S. Census Bureau on the growth rate of the community with data regarding agency density can produce a very powerful predictor of the ability to expand in an area.

Carriers that deal with independent agencies have an optimal number of companies that they would like these agencies to represent. But what is the optimal number and does the presence of certain competitors in a company's agencies have a particularly negative effect on profitability? Publicly available data sources exist to analyze this phenomenon and understand how it affects performance.

Finally, the credit characteristics of the agency itself can be a performance predictor -- yet few companies take an in-depth review of the value of this predictor.

The value in modeling agency performance does not lie in confirming some of the above relationships that already are known qualitatively. Instead it lies in understanding the individual factors that are most predictive of agency performance, uncovering new factors that are currently not employed, understanding how those factors interrelate and being able to quantify the impact that each is likely to have on overall performance.

Predictive modeling, which spawned the once innovative field of credit scoring, has the flexibility Porter identifies as crucial to reversing a technology decline -- once again delivering competitive advantage to the industry. Companies that embrace it, especially in the early stages, will reap the rewards.

(Chuck Boucek is a senior manager in the Chicago office of Ernst & Young's Insurance and Actuarial Advisory Services Practice. He can be reached at insight@bestreview.com.)

by Chuck Boucek

An online information service of the United Farmers Agents Association, Inc.