Evaluating Policy Priorities under Social Learning and Endogenous Government Behavior

We provide two methodological insights to the challenge of ex ante policy evaluation in macro models of economic development. First, we argue that the problems of parameter instability and lack of behavioral constancy can be overcome by considering learning dynamics. With this aim, we do not define the social constructs that influence agents' decisions as parameters, but as variables determined through stable functional relationships such as social norms. Then, since post-implementation reactions demand adaptive responses from the policymaker, we specify endogenous government behavior. In contrast with Lucas critique's expectation channel, behavioral shifts from social norms are hardly predictable. Second, we show the flexibility of agent computing for policy evaluation through a model of policy prioritization in economic development. We perform Monte Carlo simulations to estimate retrospective and counterfactual policy priorities, and evaluate how different types of budgetary allocations generate gains in efficiency. We find that prescriptions that take into account these learning dynamics are significantly more efficient than those neglecting them. Our results also suggest that policy failures in enhancing economic development have less to do with how disciplined are governments in following them, and more with how unaware are consultants about the systemic and adaptive nature of the process of economic development