Generative Economic Modeling
with Hanno Kase & Matthias Rottner
Abstract | Draft upon request
We introduce a novel approach for solving quantitative economic models: generative economic modeling. Our method combines neural networks with conventional solution techniques. Specifically, we train neural networks on simplified versions of an economic model to generate approximations of the full model’s dynamic behavior. By relying on these less complex satellite models, we circumvent the curse of dimensionality and are able to employ well-established numerical methods. We demonstrate our approach on models with nonlinear dynamics and heterogeneous agents. Finally, we apply generative economic modeling to solve a high-dimensional HANK model with financial frictions.
Distributional Impact of Asset Price Fluctuations
Abstract | Draft upon request
This paper studies how fluctuations in asset prices, driven by noise traders in segmented financial markets, affect the real economy and households across the income and wealth distribution. I show that these fluctuations help explain empirically observed portfolio choices, a fraction of the observed equity premium, and a small part of real fluctuations. Poorer households avoid risky assets because they rely on liquid and safe assets for self-insurance. Richer households require a premium of 0.8 percent per year to hold risky assets, which accounts for a fraction of the observed equity premium. In quantitative terms, a 10 percent increase in asset prices, which corresponds to a typical quarterly shock, raises aggregate output by 0.8 percent. Asset price shocks are redistributive because they directly benefit wealthy households, who own most assets, and also affect poorer households indirectly through general equilibrium effects such as changes in wages and prices.