Expanding the scope of microsimulation by modeling and simulating inter-individual interaction Seminar

To realistically describe individual behavior, demographic microsimulation has to consider the effect of inter-individual interaction. For example, in the majority of cases the decision to have children depends not only on the woman, but also on the potential father, and presumably additionally on the social network of both. Individual interaction patterns are usually ignored within demographic microsimulations. The only type of inter-individual linkage that has been considered in microsimulation models so far is demographic kinship. However, when modeling kin at this level mostly very simplifying and therefore distorting assumptions are put. For example, in nearly all cases characteristics of members of kin groups are stated to be uncorrelated, which is an assumption that is simply wrong. No general modeling or simulation technique exists that accounts for correlation between linked lives in a microsimulation. To give an example: many microsimulation models incorporate a marriage market to perform mate-matching. However, after a couple has been matched, the subsequent modeling and simulation of the individuals forming the couple is not clear: in the literature neither a stochastic model of the joint behavior of paired individuals is described, nor related simulation algorithms are discussed.

In this talk I propose a way to extend a continuous-time microsimulation model to tackle this shortcoming. I introduce a simulation model and related simulation techniques to account for interdependencies between the life-courses of spouses of (heterosexual) partnerships. That is, I restrict my consideration to binary linkages. The construction of interaction networks comprising more than two individuals is straightforward from a pure technical point of view: binary links between all interacting individuals make up a complete graph mapping an interaction network. I propose to model couple behavior by considering both spouses of a pair jointly. The capability of the novel microsimulation is illustrated using a simplified example that aims at forecasting the contemporary Dutch population. Here I analyze partnership stability and the individual propensity to quit smoking, considering dual- and single-smoker couples.