When an insurance company or a reinsurance company is trying to determine how much it should charge for its product, it is basically trying to predict the future. It bases its prediction on past information and applies this information to certain models. If a company has a great deal of information for the type of risks for which it is trying to predict the future, its models will do a better job than if there is a paucity of information. For most companies that have been writing a particular line of business for some time, they will use a mix of their own data and industry data.

First and foremost, a model is a model is a model. It is not real. Perhaps, using a trip where you have to drive four or five hundred miles away and you have to arrive by a particular time would be a good example. You know where you are starting out and where you want to end your trip. You consult maps or online mapping services and a trip is laid out that includes the duration. What we have done here is to try and predict the future in terms of the route and the length of time for the trip.

Now, most times, everything goes smoothly but we all have had the inevitable “road is closed take this detour” type of trip. It is not the map maker’s fault. When the maps were drawn, the road wasn’t closed. So, what do we do? If we are experienced drivers, we give ourselves a margin of error in the time estimate. If the trip were five miles to the train station, we would most likely give ourselves an extra five or ten minutes. The further we have to go the more time we would give ourselves. If our four hundred mile trip was supposed to take nine hours, we add another one to two hours to the trip. However, if we absolutely had to be there, we might even plan on arriving the night before.

When we have a great deal of information,i.e. the local trip to the train station, we don’t have to give ourselves a large margin of error. However, as in the example above, the less information that we have, the greater a margin of error we give ourselves. My poor attempt with this analogy is that, when we are pricing an excess reinsurance product, the more losses that the reinsurer’s actuary has (like the trip to the train station) the less margin of error that will be used.

We are fortunate that there are not usually too many large losses. This condition is a good news/bad news situation. When there is a small amount of information (like the road conditions 400 miles away), we build into our model a greater margin of error.

We should note that when the reinsurer is pricing the agreement, it is doing so with old information. The newest data will be more than 1 year old when the term of the agreement begins. The end of the agreement will frequently be fifteen months after the reinsurer develops its rate.

Going back to my trip planning analogy, it is like planning your trip with road conditions that are a year old and the trip won’t be finished for another year. How frequently do you think that you’ll get it right without the “are we there yet?”