Automated Underwriting: An Art of Science?

May 31, 2004 | Last updated on October 1, 2024
10 min read

Although automated underwriting is only now beginning to surface as a real tool used by property and casualty insurers in determining the pricing and approval of coverage, the use of such methods became highly popular in the mortgage lending and life insurance sectors during the 1990s. So much so, that a significant portion of home mortgages and life insurance policies are now finalized by this means, with service providers offering consumers instant approval or non-approval on applications. In addition to the significant cost savings achieved by eliminating manual processing, the use of automated rating has become a major competitive factor in today’s electronic age of instant and hassle-free service.

The small steps that have been taken by the property and casualty insurance industry over recent years toward automated underwriting have been mainly focused on personal lines business. There are two reasons for this: the large number of policy applications which generate relatively low premiums on an individual basis, and the fact that electronic rating relies on statistical patterns that can only be achieved by having a critical mass of consistent data. The lack of available data and consistency in applying risk ratings have been the biggest stumbling blocks in the advancement of automated underwriting, insurers say.

However, insurers expect that the future of underwriting will become increasingly “electronic-based”, with advancements made by companies playing a significant role in their competitive standing. And, despite growing controversy in the U.S. regarding the use of credit scoring information which has become a common risk-rating denominator (regulators and consumer groups fear this will result in pricing discrimination against minority groups and low income earners), insurers say a move in this direction will actually result in lower premiums for most policyholders in that they would not be subsidizing the cost of the high-risk segment of the risk pool.

U.S. DIRECTION

The objective of every insurance company is to set rates for cover as closely as possible to the actual cost of claims. If a company sets rates too high, it will lose marketshare, and setting rates too low will result in a loss, observes a discussion paper issued by the New York-based Insurance Information Institute (III) on automated underwriting tools. The drive to achieve better risk rating through automated underwriting not only benefits insurers, but the majority of consumers as well in that they are not subsidizing people more likely to file claims, the paper notes.

The rating and pricing of a risk mainly depends on two factors, the propensity of the individual or business to make a claim, and then the value and inherent risk associated with the property itself, the III says. As such, the latest advancements made by the insurance industry in automated underwriting have been “insurance scores” which are based on an individual’s credit history, and the use of shared property claim databases enabling insurers to determine the historical claims experience of the property to be insured.

Independent actuarial studies have clearly identified a correlation between poor credit history and insurance loss experience, the III points out. Research done by EPIC Actuaries (involving 2.7 million records) indicates that, with regard to property damage liability coverage, individuals with the worst insurance “credit” scores had expected losses of 33% above average while policyholders with the best scores were 19% below the average loss experience. “Actuarial studies show that how a person manages his or her financial affairs, which is what an insurance score indicates, is a good predictor of insurance claims. Statistically, people who have a poor insurance score are more likely to file a claim,” the III states.

The other side of the risk evaluation equation is property loss data. The two major property claim databases in the U.S., the III notes, are the “Comprehensive Loss Underwriting Exchange (CLUE) which is operated by an outside data management company and includes claims data from 600 insurers writing homeowners’ cover, and the Automated Property Loss Underwriting System (A-PLUS) which is administered by the industry’s Insurance Services Office (ISO) with about 1,250 contributing companies.

The use of “loss histories” as a primary underwriting tool has been used by insurers for decades, the III points out. The difference today is that technology has greatly reduced manual processing by enabling insurers to electronically tap directly into data sources, thus reducing the cost and turnaround time required in handling policy applications. The “information age” has, however, drawn public attention to how financial services companies use personal information.

CREDIT FRICTION

In this respect, for U.S. insurers, the biggest concern with automated underwriting at present is that both the use of credit scoring and the collection of individual claim information for sharing purposes has attracted negative attention from regulators and consumer groups, the III notes. The National Conference of Insurance Legislators (NCOIL) had adopted a “model law” on the use of credit scoring in underwriting in 2002. The model law requires insurers to inform consumers that a credit report may be used in the risk evaluation process, and that in the event coverage is turned down, that the individual in question be informed that credit was the basis for the adverse action taken (this is also required under the federal Fair Credit Reporting Act). The NCOIL provision also prohibits use of credit information as the sole basis for refusal to insure, non-renew or cancel a policy.

However, during 2003, 42 states considered legislation to restrict use of credit scoring, resulting in new laws being introduced in 20 of the states, the III adds. None of the states banned credit scoring outright. Maryland applied the most restrictive law in banning the use of credit scoring for homeowners’ insurance entirely while allowing its use for new auto applications with a 40% cap on rate variable based on credit. Texas also limits insurers to a rate variation of no more than 10% based on credit scoring (although there is no restriction on class of business).

The debate on credit scoring continues, with the Missouri Department of Insurance having recently released a study indicating that low-income and minority groups are adversely affected by insurance scoring (the III says the department’s findings are based on “flawed methodologies”). And, in reauthorizing the Fair Credit Reporting Act at the end of 2003, the federal government gave the Federal Trade Commission (FTC) a directive to investigate the use of credit information by financial services companies. The objective of the investigation is to determine the affordability and availability of financial services products to various “demographic groups” relative to the use of credit rating practices. A report providing recommendations by the FTC is expected to be released by mid-2006, the III observes.

Furthermore, the National Association of Insurance Commissioners (NAIC) has indicated an interest in undertaking a credit scoring study, but the individual commissioners have been unable to agree on how to carry it out, the III notes. As a result, the high level of ongoing regulatory attention surrounding credit scoring casts a shadow of uncertainty over the practice’s future – which could present a severe blow to future development of automated underwriting.

DATA PRIVACY

Several states are currently looking into legislation that will affect claim databases, the III observes. At its most recent December 2003 meeting, NCOIL also indicated that it will place CLUE reports on its discussion agenda for this year. “CLUE reports became an issue in several states including California and Utah after realtors complained that deals were falling through because insurers were canceling new policies after examining claims information,” the III adds.

One of the biggest concerns the legislators have with the use of claim databas es in risk rating is that enquiries (without resulting in an actual claim) made by policyholders to their insurers may be counted as a claim. Insurance companies are obligated to open claim files when policyholders make claim inquires, thus the “enquiry” could show up in a database, the III notes. Once again, a critical risk rating tool currently used by insurers in automated underwriting faces the possibility of tight regulative restrictions at some future point.

CANADIAN APPLICATION

“The modern world of [personal lines] insurance is very much a numbers game,” says Saskia Matheson, the personal insurance auto leader at Royal & SunAlliance Insurance Group. She notes that roughly 80% of personal lines policy applications received by Royal & SunAlliance are processed electronically without an underwriter ever looking at them. Of the remaining 20%, about one quarter is “business that you [an insurer] don’t want”, while the other 15% is directed to an underwriter for further analysis, she adds.

Royal & SunAlliance has been particularly progressive in adopting automated underwriting by building “risk filters” into its mainframe system so that when policy applications submitted by brokers come in the “front end”, the information is streamed through this risk sorting process, Matheson explains. The break-through in automated underwriting technology really only occurred about five years ago, she notes, with much of the business before then dealt with manually and the risk determined by the “sense” of the underwriters. “Technology has changed the industry from having to handle individual risks and having to do a lot of ‘guess work’.”

Automated underwriting largely depends on identifying claims patterns, which makes it ideal for high-volume business such as personal lines, Matheson observes. And, with the pressure on insurers to reduce operating costs while business volumes continue to grow, “we [the industry] need more and more sophisticated [technology] tools to do that,” she notes. By better utilizing the information already made available in the policy application for risk rating purposes, insurers can reduce their reliance on outside information sources such as the motor vehicle registration offices. “Motor vehicle ‘abstracts’ are expensive to get,” she points out. As such, an insurer would only want to go to this length in the event of something unusual surfacing in the policy application.

In some cases, Royal & SunAlliance also uses an external information service, being CGI’s motor vehicle registration program which includes various other rating data including credit reports. The objective is to determine whether the applicant has provided accurate information, Matheson says.

In addition, Royal & SunAlliance has built “back end” risk assessment filters into its mainframe system. This enables the insurer to evaluate the type of business being sent by any particular broker. Again, the objective is to look for patterns, Matheson notes, “to determine if this is the type of business that we want. We can then approach the broker and ask whether they are sending new drivers [higher risk] evenly among the carriers they represent. We see this [the back end risk analysis] as being a business planning tool to better work with brokers.”

Overall, Matheson expects automated underwriting will increasingly become an area of technology development by insurance companies. “It’s an insight to how better to use the information we already have.” As such, she believes applying technology to underwriting is already a competitive factor in how companies are differentiating themselves in the marketplace. “This [automated risk rating] is definitely an area which insurers will continue to develop.”

Kevin Dobson, commercial auto product line leader at Royal & SunAlliance, says automated underwriting has less impact in rating commercial risks than personal lines. However, Royal & SunAlliance has applied electronic risk rating filters to its small business coverages.

The problem with applying automated underwriting to medium to large commercial risks, Dobson explains, is the significantly lower volume of applications involved, while the potential losses that could arise from a claim is much greater. However, he adds, “we are always looking at new options and tools” in evaluating commercial risks. The insurer’s commercial risk division does, however, utilize outside risk assessment software in scoring cover applications. But, in this sense, the technology is more of an aid to the underwriter than a form of processing, he adds.

Al Parro, property product manager at Swiss Reinsurance Co. Canada, concurs that, outside of the direct handling of high-volume personal lines business, the application of risk modeling and rating technology is primarily an underwriting aid, and not the actual process. However, the application of technology to underwriting has resulted in more accurate and specific loss experience data, Parro says. The more information, and the more companies are becoming reliant on electronic risk modeling, the more intense the job of the underwriter seems to have become, he jokes. Because of the aid of technology, underwriters are spending a lot more time going through pricing exercises, he notes. This, however, has enabled insurers to better price risk accordingly and therefore be more competitive in the marketplace.

EARLY STAGES

However, many Canadian insurers still seem to be at the “starting gate” of applying automated underwriting as a process and generally using technology for risk rating purposes. This, however, is not necessarily a bad situation, insurers observe, considering the uncertainty surrounding automated underwriting in the U.S.

Wawanesa Mutual Insurance Co. currently uses its legacy-based IT system for policy renewals, but handles new business applications manually, says Keith Hartry, the company’s habitational and farm underwriting manager. The insurer is planning on building in automated risk filters to its new mainframe technology platform which is currently in development, he notes. “We’re trying to get our [risk rating] rules straight,” he adds with regard to the technology switchover. Hartry believes automated underwriting will play a more prominent role in the industry’s future, however, it is unlikely to ever replace the presence of the underwriter. “You can’t take the underwriter out of the picture, to remove all judgment. The real benefit of technology is to eliminate the manual processing to allow the underwriters to work on what is needed.”

Susan Vella, senior vice president of personal insurance at Chubb Insurance Co. of Canada, says the insurer hopes to import the electronic risk rating and automated underwriting system currently used by the group’s U.S. subsidiary. At this point, however, Chubb Canada mainly uses manual underwriting on its personal lines business. While automated underwriting represents “big changes to the future”, Vella notes that one of the more challenging issues associated with electronic risk rating is identifying which factors, in terms of extraction of data from policy applications, would be pertinent. And, more importantly, she observes, is how those factors are combined to produce an effective risk rating pattern. “Insurance companies have a lot of [policy-based] information, but the challenge is putting it together.”

Vella also points out that building risk rating tools into a company’s mainframe system requires a substantial monetary investment. The cost involved relates not only to development, but keeping the system up to date with the latest technology, she notes. In this respect, and bearing in mind the controversy that has developed around the automated underwriting practices applied by insurers in the U.S., she adds, “sometimes it’s not a bad thing to be behind the U.S. by a few years”.