This was selected as Best Answer

As there are not current models, as you already mentioned, I will treat this question as more of a brainstorming exercise; especially since we don’t seem to be getting much in the form of responses.

To start off, I think it would be good to break down the issue into 4 parts:

1.The structure of current CAT models.
2.Theoretically, what kind of changes can be made to accommodate the change from property damage to group life?
3.Practically, can the changes in point 2 be made? What is needed?
4.Any shortcuts?

The current structure, as I understand, is fairly standardized. The CAT model is divided into 3 modules; the Hazard Module, the Damage Module and the Insurance/Financial Module. I’m sure you are more familiar with the exact working than I am, so will not discuss this further.

Theoretically, to convert the PML from property to life, you can use one of 2 approaches:

Approach1: Just modify the Insurance Module. So the part of the model that generates the storms stays the same (of course) and I understand RMS and others factor in the expected increase in frequency and severity there. The Damage module CAN stay the same as well. They are already modeling in the factors for changes in construction and types of buildings and safety codes etc., so it may be advantageous to use that information. What actually changes is the Insurance module, where instead of converting damage into insured loss, you convert damage into loss of life, and then multiply with the expected death benefits.

Approach2: Another approach could be to make changes at the second stage, the Damage module. You may be able to change the definition of ‘damage’ from that of ‘property damage’ to ‘loss of life’, and then the Insurance module would work more or less the same way it does for property damage.

In my opinion the Approach1 is preferable for 2 reasons. Firstly, since improvements in construction and construction density will affect the number of lives lost, it may be useful to leave the Damage module the same as for property damage. And secondly, if you are going to use outputs from the model to calculate the PML internally, it would be easier to manipulate the results of just one stage, instead of two stages.

Practically, of course the best would be if you can convince the vendor that there is enough demand that they should invest in providing a version with the changed Insurance module. However, that may not be a possibility. In that case, depending on your in-house capabilities, you may be able to take the output from the Damage module and fit loss-of-life data onto it. The advantage of using a stochastic model is that you can play around with different exposure levels and come up with different scenarios to calibrate it to any data you may already have.

Of course, data will be the major issue here. I am not sure what level of detail is available in Mexico, but converting building damage to loss-of-life may initially require you to hypothesize based on experience, more so than anything else.

Lastly, if all of the above is completely impossible due to capacity constraints or some other reason, it may be useful to just go back to the basics. Without CAT models, property coverage is priced based on historical data split into CAT and non-CAT losses, with a rate calculated based on non-CAT losses and then a loading added to it. However, those rates have proven to be inadequate, so you may need to add a further loading to your estimate. This loading can be a judgmentally modified version of the [base (property) rate for a given year based on CAT model] / [base (property) rate for the same year without use of CAT models].

Of course there would a lot of considerations to be factored in before such adjustments can be made, but this brain-dump is all I can come up with at the moment. Hope it helps.

All the best in your endeavors and be sure to let us know of your eventual solution to the issue!

posted 11 months ago

Joan:

I have been doing some research on this subject, specially as I work with some mass marketing initiatives, in which modeling of life losses due to natural catastrohpes will be key to allocate the right prices to these exposures.

One of the most interesting issues here, I think, is that as far as loss of life due to natural catastrophes is concerned, the industry will have to gradually shift away from standard actuarial models, whatever these might be. All actuarial models I know, consider past experiences are key to predict future ones. This will only be true if future weather conditions remain the same as in the past. We all know this will no longer hold true, as social infrastructure and climate change will negatively impact (increase) the frequency and severity of human life loss.

Population is growing in very different rates all over the world, large areas which were not populated as recently as 20 years ago, are now full of activity. The Asian Tsunami tought us all a great lesson.

Hence, I think the industry will have to move to a more inductive set of models to create rates. Complex models that will have to consider additional sets of variables, such as as global warming effects, population density projections, etc.

Some group life portfolio models might be available now. The key issue here, is for how long their results will be reliable, or within acceptable tolerance intervals, as current events and conditions move rapidly away from expected projections.

Good luck!

posted 12 months ago