Simulation-Based Estimation of General Structural Network Models

Structural estimation to fit an economic model on a single graph is an open problem in economics and adjacent fields. While there are techniques to estimate structural models on a collection of independent and identically distributed graphs, there is little literature on a general-purpose algorithm to fit a structural model on a single network--at least work the author is aware of. This paper provides an algorithm borrowed from deep learning that is adapted to network. This approach theoretically allows for Bayesian and likelihood estimation of arbitrary structural models. The main condition is that the graph data is in the support of the networks simulated by the structural model. Tests on simulated data as well as an empirical application demonstrate the validity and the accuracy of this approach.

Fast Simulation-based Bayesian Estimation of Dynamic Models

I show that one can perform Bayesian and MLE estimation of dynamic models without the use of a likelihood function, particle, or Kalman filter. Compared to conventional methods, this allows solution methods like projection and value function iteration to be estimated as well as models with large latent spaces like HANK models. The simulation-based technique scales better than conventional Metropolis-Hastings producing more accurate posteriors with less computational effort, especially for the largest problems like the Smets-Wouters 2007 model and a HANK model.

Improving External Validity of Machine Learning, Reduced Form, and Structural Macroeconomic Models using Panel Data

We show that adding countries as a panel dimension to macroeconomic data can statistically significantly improve external validity of structural and reduced form models, as well as allow machine learning methods to outperform these and other macroeconomic forecasting models.