Building of a stochastic scenarios generator
An important step in the building of an appropriated DFA or internal probabilistic model is the identification of the key random variables which affect the evolution of the assets and liabilities of the insurer.
This process is of course eased up by the activity of identification and assessment of the risks described earlier.
Once these variables are identified, it has to be decided:
- Which ones will be directly modelled and which other ones will be induced
- How they will be modelled
- What are the interactions between these variables (cascade structure)
Examples of random variables which are typically included in a generator of stochastic scenarios are inflation, interest rates, capital appreciations and dividends for shares, catastrophic claims, patterns of payments of the claims, business cycles, etc.
The objective of the generator is to produce several thousands of scenarios for each of these variables for the next 10, 25 or even 100 years, where the values taken by these variables have to be coherent between them - through functional relations or through a correlation structure - and through time. The starting values generated for the variables also have to be in adequacy with the current economic environment. This is the so-called process of calibration of the stochastic scenarios generator, which is very important for the results obtained with the model to be meaningful.