Inaccurate Territories and Assignments Hurt Consumers, Agents and Companies
Territories offer tremendous opportunities for insurance companies to better segment their book of business, but private passenger auto and homeowners insurance companies, in general, are not using territories as effectively as they should for two reasons. First, insurers are often not as accurate with the assignment of territories as they intend. Second, companies generally do not keep their rating territories and relativities as up-to-date as they could, particularly with changing demographics.
The accuracy with which an individual risk is assigned to a rating territory is often not as great as companies believe. So what is the problem with a few miscoded risks? From an agent’s perspective, if the risk is overpriced, it may be harder to write because everything else being equal, competitors will have a lower price. If the risk is underpriced, it will affect the agent’s loss ratio.
For consumers, it means paying an incorrect premium. If that premium is too low, the customer probably does not care. If it is too high, not only might the customer be due a refund, they also may decide to leave for a competitor that assigns the customer to the correct territory and has a lower premium.
Some companies may rely almost exclusively on the agent to assign the risk to the correct territory. This responsibility can be especially daunting for independent agents, who work with several different companies that have different sets of territorial definitions. Other companies may have more sophisticated methods, such as a mechanical system based on a particular zip code or street and town address. However, while this is a better system, there are still several factors that can, and do, go wrong.
Figure 1 shows a Cincinnati mailing address for a location that is actually outside the city. Because the insurer relied on the postal address, the risk is rated in the wrong territory, and the insurer’s statistics will be distorted.
Zip codes are another common way insurance companies define territories because they are convenient, perceived to be a better basis than larger geographic designations, and well-understood by the public. However, it is not always the case that they will do a better job. Zip codes were never designed with insurance company territories in mind, but for more efficient mail delivery. Zip codes often transcend county, city and/or township boundaries. Thus a zip code may be considered part of two different territories, and create miscoding of applications. As an example, zip code 23234 includes part of Richmond, Virginia, and part of the surrounding suburbs (Figure 2).
In other cases, a zip code may contain a number of cities within its boundary. This works fine if all the cities within the zip code display the same basic demographics and loss experience. However, if they do not, and the insurer uses different territories, the likelihood of territorial mis-assignments increases.
The U.S. Postal Service routinely changes, reassigns and eliminates zip codes. Hundreds of five-digit zip codes change every year. Remarkably, over 20 percent of zip codes can change on an annual basis. It is very difficult for any individual insurance company to keep up with all these changes, and it is virtually impossible for an individual agent to do so.
Considering all the possible ways to incorrectly assign a rating territory, one should not be surprised that the error ratio is much greater than was previously believed. According to Howard Botts, executive vice president for Proxix Solutions, the error rate for the assignment of customer policies for a large number of companies was consistently in excess of 20 percent. “The highest rates occur in companies that use postal geography, zip codes, to assign policies to territories,” he added.
The second problem with rating territories lies with the actual territory definition and rate relativities companies use. In many states, companies are still using boundaries that were established years ago, without any recognition for how the demographics may have changed. Further, many of these definitions include wide-spread geographic areas. Since territories should include generally homogeneous groups, how homogeneous are these territories if they haven’t been changed for several years, or if they contain wide spread geographic areas? This problem is exacerbated by the huge building blocks often used by companies to define territories. Frequently companies try to lump entire counties, or groups of counties together. Obviously, most counties and even zip codes contain several different demographic make-ups.
The other concern with territorial definitions is that they may have been created for reasons other than trying to find the most homogeneous group or the best statistical definition. Anecdotal evidence suggests that some territorial changes made have been made to accommodate an agent who generated a large volume of business and favorable loss ratios.
Another concern with territories is the rate relativities assigned to individual territories once they are created. Historically, insurers have procedures for calculating territory rate relativities that include one-way loss ratio or pure premium analyses. These procedures do not take into account the relationship, or correlation, between territories and other rating variables, nor do they consider the different mix of business that exists within each territory. As an example, for private passenger automobile, not every territory will have the same percentage of vehicle types, young drivers, or drivers with accidents and violations.
Ways for Improvement
Fortunately, insurers and agents do not have to be stuck with antiquated and inaccurate territories. Better tools are available to improve overall accuracy of territory assignments. Companies can create digitized maps for their territorial definitions. Individual risks can then be geocoded and the assignment of territories is possible of reaching accuracy levels of 98-99 percent. Further, this technology is available such that agents can have correct territories assigned automatically once they input the correct address into their computers.
Improvements can be made in several ways. First, and maybe easiest, is to update actual definitions more frequently. There is no excuse for thinking that 70-year-old territories are doing a good or even an adequate job.
Second, “building blocks” better than cities, zip codes, or other large geographic area are available. Block groups, which can be obtained from the U.S. Census Bureau, are the best currently available because they provide smaller and more consistent geographic units.
A block group (BG) is a cluster of census blocks that share the same first digit of a four-digit identifying number within the census tract. Block groups are the smallest unit for which the U.S. Census releases detailed demographic information. Most BGs contain 600 to 2,000 people, with an optimum size of 1,500. Consequently, these block groups are much more likely to contain a homogeneous group than a larger area such as a zip code, township or county.
Finally, with respect to territorial definitions, it is no longer necessary to combine all coverages or major perils to determine rates. When rates were calculated manually, rating plan simplicity necessitated a combined approach. However, with computers now available to everyone, it is no longer necessary to apply the antiquated rules of 20 or 30 years ago.
For example, companies may consider using the same territories for private passenger auto property damage and collision coverages. However, using the same territory definition for bodily injury, comprehensive, or uninsured motorists’ coverage might not be the best approach. Additional considerations – including medical costs, theft rates and the uninsured motorist population – will have major effects on rates for these coverages.
Insurers in general need to do a better job of recognizing the differences in distributions among territories, and the relationship they have with other rating variables. The use of multi-variate techniques, such as Generalized Linear Modeling (GLM), ensures that territorial definitions and relativities account for only that portion of the risk which is not already being accounted for elsewhere in the rating plan.
Territories are one of the most critical elements in rating private passenger auto or homeowners insurance. A few years ago, a California Actuarial Advisory Committee Study found that territories explained 49.7 percent of the risk for bodily injury coverage. Insurers would benefit from increased attention to the application of the correct territory. Insurers also need to pay as much attention to territorial definitions and rate relativities as they do to other rating variables. New tools and methodologies are available to address these issues and need to be added to many companies’ ways of doing business.
Kucera is senior consultant for Pinnacle Actuarial Resources, Inc. Pinnacle Actuaries, Inc., with offices in Bloomington, Ill., Chicago, New York, New Jersey and San Francisco, is one of the 10 largest property/casualty actuarial consulting firms in the United States and specializes in insurance pricing, loss reserving, alternative markets and financial risk modeling.