Part of it has to do with the modelling we chose to use. We decided to use linear trends for the frequency and the severity of the disaster. We fit the model onto history, and there was a large jump toward the end of the time frame of the model, so the model doesn't capture that as a permanent increase. It's mitigated by the fact that we fit the model over a long period of history.
It's difficult to say what that impact will be in the very long term, but it's largely a result of the fact that these changes are very recent. We don't know if these changes will continue to accelerate rapidly, as they did in the last 10 or 15 years, so that's attenuated downward.