Modeling the Variety of Decision Making

Static or deterministic actuarial models work with a single set of assumptions for all of the variables in the model. Most commonly, those assumptions might be selected to be the most likely values, or to values that produce an outcome for the model calculations that is likely to be achieved “most of the time” (i.e. most likely assumptions with provisions for adverse deviations). Deterministic models are also used in risk management related stress and scenario testing, where the assumptions are chosen to be static but at unlikely and adverse values. It has become common in risk management related calculations to use dynamic or stochastic models. One or several of the key assumptions are set to vary under a set of rules and other assumptions as well as the ultimate calculations should vary along with the stochastic variables. And yet one assumption that is often held to be a static table of actions are the things that management has discretionary power to do in the future. Often called simply “management actions.” Examples that relate to risk models include choices regarding execution of ALM and hedging programs, structure and extent of reinsurance purchases as well as underwriting standards and choice of the products/markets to do business. These management actions are usually the most critical in the lead up to and immediately following an experience of acute adverse experience. This report presents the argument that (a) management actions cannot be definitively determined from past actions or current intentions; (b) there are a number of potential actions that management might take in and around times of adverse experience, and which of those actions that will be taken will be determined by the future management at that time based upon their vision at that time of the most likely potential future; and (c) an actuary can build this additional variability into a stochastic model to provide a more realistic range of future experience among the stochastic scenarios