Does Underwriting Matter?

by David West

Think of an insurance company as an airplane that has its progress measured by its loss ratio. Up in the cockpit, the pilots use many different instruments to monitor the plane’s progress. If the plane deviates from its course, the pilots make adjustments using their controls. Think of the pilots as the company’s actuaries; the instruments are their key performance indicators, and their controls are pricing.

Now take a step back in the cabin. The folks in the cabin have a general feel for the direction in which the plane is going. They might even get a few indications on the display. But imagine that they are also trying to control the plane — without any direct control. If all of the passengers move to one side of the plane, it might turn a bit. If they all move to the back, it might climb.

Unfortunately, they don’t have the feedback they need to bring the plane to a successful landing. The passengers are your underwriters. Their gyrations are selection rules. They seem to have the desired effect, but they can’t really see where they are going, so they don’t really have any idea whether their movements are truly affecting the flight — or whether it is just the actuaries and their pricing.

I’ve spent many hours in airplanes. I’ve felt the effects of passengers moving around in the plane. Yes, there are some effects, but passengers never realize that. And if a pilot wants to make the plane turn, climb or dive, the passengers can do nothing to prevent it.

Underwriting is quite similar to being a passenger on an airplane. Pricing drives your company. In large part, your changes to selection rules — or your return to discipline or deviation from it — will not have the results you desire. In some cases, your movements can actually do more harm than good. Underwriters lack the analytic tools that the actuary has, but this doesn’t have to be the case. How underwriting got to this point and what underwriters can do about it are the focus of this article.

The evolution of insurance
Insurance has evolved over many centuries in response to the changing needs of society. The first forms of insurance appeared about 3000 B.C. when the Sumerians and Babylonian merchants pooled their resources to protect one another’s wealth from losses due to bandits along their trade routes. The practice was quite common 12 centuries later when King Hammurabi wrote laws that governed risk transfer activities. Six centuries later, the Phoenicians used a form of insurance to protect their ships’ cargoes. Ancient Greeks and Romans created burial insurance to benefit members of “benevolent societies.” Even during the Middle Ages, guilds provided insurance-like benefits to guild members to protect their families against illnesses befalling the male family leaders. Throughout these millennia, insurance existed through informal contracts and gentlemen’s agreements. It wasn’t until the advent of modern insurance in the late 17th century that the practice of underwriting came into existence.

Because of a finite capacity for risk and the recognition of variation in loss potential between different seagoing vessels, underwriting practices came about as a means of maximizing the profits and minimizing the losses of people who backed the contracts with their personal wealth. While the original underwriters were actually the wealthy individuals who would pay for losses out of their own pockets, it didn’t take long for the professional underwriter to emerge as a decision maker rather than a financial backer.

The underwriting professional who emerged from Lloyd’s Coffee House had two basic duties: to select appropriate risks and to ensure that the negotiated price paid by the insured would provide adequate compensation for the risk assumed. This continues to be the primary role of the underwriter today.

Pricing and risk
Actuaries establish modern insurance pricing for personal lines insurance using complex analytic techniques (quite often using SAS). With appropriate methods and adequate historical data, the actuary can calculate an insurance premium for any insurance exposure. The result is that the only bad risk is an underpriced risk.

Underwriters are not responsible for pricing; rather, they are responsible for ensuring that all rating characteristics are correct. This leaves them with one primary responsibility: risk selection. When an insurer receives an application, it is the underwriter’s job to ensure that adequate information is available to assign the correct rate and to determine whether to issue a policy.

To assist underwriters in determining which policies to issue and which policies to reject, insurance companies developed their own unique risk selection rules. If an application violated a rule, the underwriter is generally expected to decline the risk. The art of underwriting appears when an underwriter considers making exceptions to the selection rules.

One of the basic requirements for calculating insurance rates is the need for a homogeneous set of exposures. Risks in any pricing classification should have similar characteristics, and thus they would have similar probabilities of loss. One goal of underwriting rules is to ensure the consistency of the book of business.

Underwriters are trained to consider a variety of risk characteristics. They are taught that such characteristics indicate higher or lower loss potential than the average. Although selection rules address many situations, the underwriter’s training kicks in when exceptions are made.

Underwriting rules are not static. The majority of insurers change their rules according to changes in marketing direction and especially in response to changes in loss results. When loss ratios climb, insurers tighten underwriting rules; as loss ratios drop, they loosen rules. When executives talk about increasing their underwriting discipline, they mean that selection rules will become more restrictive and fewer exceptions will be made.

Over the past two decades, many insurers have built systems that automate insurance pricing as well as the application of selection rules. The promise of such systems stemmed from the perceived ability to eliminate the underwriter from the selection process on all but the most complex cases. In most cases, these systems were designed to emulate the thinking of underwriters and to arrive at the same decision that an underwriter would have made. Some systems were more simplistic and functioned as rules engines that simply applied the selection rules.

Finding the flaws
The typical underwriting practice described above has four flaws. First, changes to selection rules in response to changes in loss results cause the book of business to violate the homogeneity requirement for accurate actuarial calculations. Risks that meet a stricter set of rules will have a different loss cost than risks that meet the older, looser standard. Applying rates based on one set of risk characteristics to risks with a different set of characteristics produces rate inaccuracy. Inaccurate rates result in an increased risk of attrition when the rates are too high. Rate inadequacy and adverse selection result when rates are too high.

The second flaw is that many underwriting selection rules are proxies for rating characteristics. When the rate structure accounts for a characteristic, there should be little reason to consider the characteristic in underwriting. Doing so would magnify the effect of the rate. For example, producing rates that apply surcharges to drivers who have had accidents but then refusing to insure drivers who have had more than one accident magnifies the impact of the rate. If the rate is accurate, then there should be no reason to decline the risk.

The third flaw is in the rules themselves. Underwriting selection rules are based largely on folklore and tradition. They are not based in science. Few companies, if any, perform any statistical validation of the impact of their selection rules on loss performance. Underwriters are taught to consider certain characteristics, but these characteristics may not have any actual bearing on loss performance. In the absence of a thorough statistical analysis of the impact of selection rules on loss performance, the use of underwriting rules may have the same effect on profitability as a study of used tea leaves. Companies simply do not know whether their rules are working.

The final flaw is the automation of underwriting rules. If the rules are not validated, then automating them may simply reduce the cost of underwriting — but it could also result in putting bad business on the books faster. If the systems are designed to emulate an underwriter’s thought processes, but those thought processes are based on training that includes the analysis of ineffective risk characteristics, then the system is emulating a flawed process. Emulating a flawed process only creates a faster flawed process.

At least one company has eliminated the entire problem of underwriting selection rules. This large U.S. auto insurer has essentially eliminated the need for underwriters by creating a rate structure that is so granular it produces an accurate rate for every exposure. Because it has a rate for every exposure, it makes no risk selection decision. The insurer’s system can assign the proper rate. This company is truly governed by the principle that the only bad risk is an underpriced risk.

The question then becomes, “Does underwriting matter?” For companies using selection rules that have not been validated, the answer is not available. Such companies simply do not know. For companies that have highly granular, accurate rate structures, the answer is definitely not.

The resolution to this problem is a two-part initiative. The first part is to use analytics to thoroughly examine the effects of underwriting rules. This will require that companies capture data on the answers to underwriting questions, which many companies discard after they issue a policy. Even conducting a randomly selected sample of business issued over a defined time period and then measuring the loss results for this sample over a multi-year time period will go a long way toward ensuring that the questions underwriting asks and the reaction to the answers will be valid and appropriate.

The second part is a more complex evolution of the organization’s pricing structure. This evolution will take time, but it will pay huge dividends. This will require the application of advanced analytic techniques, such as the creation of generalized linear models and generalized additive models, to build highly granular rate structures. As an organization develops a structure, it is essential that actuaries and underwriters combine forces to identify the gaps in the rate structure; only then can the insurer implement appropriate risk selection rules that will ensure the consistency of the book while avoiding issuing policies to underpriced risks.

The case for the use of analytics in insurance has never been stronger. By using the SAS insurance analytic platform, insurers will gain the ability to create more accurate rates, improve their competitive position, lower their expense ratios and create rational selection rules.