Modelling real workforce choices accurately via Agent Based Models and System Dynamics requires input data on the actual preferences of individual agents. Often lack of data means that analysts can have an understanding of how agents move through the system, but not why, and when. Hybrid models incorporating discrete choice experiments (DCE) solve this. Unlike simplistic neoclassical economic models, DCEs build on 50 years of well-tested consumer theory that decomposes the utility (benefit) derived from the agent’s preferred choice into that associated with its constituent parts, but also allows agents different degrees of certainty in their discrete choices (heteroscedasticity on the latent scale). We use DCE data in populating a System Dynamics/Agent Based Model – one of choices of optometrists and their employers. It shows that low overall predictive power conceals heterogeneity in agents’ preferences. Incorporating such preferences in our hybrid approach improves the model’s explanatory power and accuracy.
Figure 1: Conceptual picture of the project