Abstract: The aim of this study is to evaluate the performance of structural equation models (SEM) in the estimation of capabilities in a multidimensional development space. We simulate a society using agent-based modelling techniques. As capabilities are hardly observed in reality, it is necessary to perform a simulation exercise for comparing the capability estimates with their simulated “true” values. Rather than using a classical Monte Carlo approach, we opt for the use of a synthetic population generated by an agent-based model. The disconnect between the data generating process of the simulation model and the econometric model makes the evaluation exercise more reliable for real-life situations. Our results suggest that the SEM approach provides good estimates of capabilities. We compare two different specifications of the links among the dimensions and find that the one that accounts for their interdependence produces better results. We conduct a sensitivity analysis which shows that results are relatively stable across different sets of parameter values.