In an earlier column (“ZIP Code Tables and Buggy Whips, Relics of the Past,”), I did more complaining than offering solutions… kind of like a teenager. Because I am way past the acne stage, it is time to repent.

There is no doubt that basing rating territories, or almost anything else, on ZIP codes is like building a house on quicksand, or at best a fault line. ZIP codes constantly change, but you don’t want your rating territory boundaries to. So what’s an underwriter to do?

The answer is to find a polygon that doesn’t change, and use it as a building block for territories. Use as many of these small polygons as necessary to come up with your territories. There’s nothing to it.

My first choice would be to use Census Tracts. They change only every 10 years, then usually by dividing existing tracts. They usually differ in size geographically, but are each somewhat similar in the number and type of people contained within them. They typically contain about 4,000 people and nearly 1,500 households.

However, just as ZIP codes are not made to meet insurer’s needs, neither are Census Tracts. These tracts are created primarily for administering governmental programs. The shape is driven by the desire to have the population within them homogeneous. This helps governmental agencies direct programs to similar occupants within a specific area.

Some insurers feel that this homogeneity is a good reason not to use Census Tracts. They fear that regulators may claim they are redlining. If some tracts can be easily linked to a socio-economic factor, is that not discrimination? I don’t think so.

Rating territories are built, or should be built, based on loss experience. If the losses within a Census Tract justify it, then so be it. Of course, the “macho” attitude has its problems. For instance, in Ohio, insurers must uniformly apply underwriting guidelines within municipal corporations. Because tracts don’t necessarily follow municipal boundaries, Ohio could be immersed in a problem. O.K…. My solution isn’t perfect. It’s just better than anything else being used.

Another option is to create a grid system and analyze loss experience within each cell. The size of the cell can be varied based on population density. Let’s start off with squares that are 100 miles long on each side overlaying the United States. They can then be divided into smaller and smaller cells until each has nearly the same population. The population in the cells can be gauged based on ZIP+4 population statistics.

If the policy distribution is fairly uniform, the loss experience per cell should be consistent. Of course, this assumes a lot. Most insurers don’t have major national databases of loss experience, but most submit loss experience data to an organization that combines the data with everyone else’s and develops rating territories. To date, most insurers have been too protective of their policyholder’s name and address to share accurate loss experience data.

Let’s get real. Every insurance company in existence has access to databases of information on every household in the United States. It may be embarrassing, or even astonishing, but they even know to which magazines you subscribe. I get direct mail pieces from at least two or three insurers each week, and I live in New Jersey where no one wants to write insurance.

What makes insurers think that by not supplying accurate loss experience data they will protect their marketplace?

It is time for insurers to require that organizations that develop loss cost data and design rating territories do so with today’s tools. Figuring out in which rating territory a risk is located is not a matter of simply looking it up in a ZIP code table. That method is archaic.

Today, we have access to tools that help build rating territories that truly reflect loss experience. Then, we turn risk addresses into latitudes and longitudes and let the software determine the rating territory. This can be done accurately, in less than a second.

Now that everyone knows this technology is available, how will you respond to a regulator when he or she asks why you aren’t using it?