Sequential Self-Selection of Program Adherence

Conceptual modeling and empirical analysis of individuals’ sequential self-selection of adherence to voluntary treatment or education programs is an ongoing and unsettled area of inquiry. Both the representation of the decision process and the implementation of econometric methods for estimating the unknowns of such models are difficult, especially when the decision process allows repeated exit and reentry to the program. We present and apply a conceptual model that is both consistent with the random utility model structure and capable of representing complex patterns of self-selected adherence while being relatively straightforward to specify and implement empirically. The approach is based on a "distributed error" form of stochastic heterogeneity that exploits the use of singular normal distributions and leads to efficiency in the representation and computation of maximum likelihood estimates of model parameters. The model can be straightforwardly applied to non-normal distributions as well, and to more elaborate nonlinear specifications of the systematic drivers of self-selected adherence decisions.