« Back to Results

Advances in Computational Economics

Paper Session

Saturday, Jan. 3, 2026 8:00 AM - 10:00 AM (EST)

Philadelphia Marriott Downtown, Room 412
Hosted By: Society for Computational Economics
  • Chair: Jesús Fernández-Villaverde, University of Pennsylvania

Macroeconomic Effects of Production Networks: A Deep Learning Approach

Jesús Fernández-Villaverde
,
University of Pennsylvania
Christian Matthes
,
University of Notre Dame
Hikaru Saijo
,
University of California-Santa Cruz
Felipe Schwartzman
,
Federal Reserve Bank of Richmond

Abstract

This paper develops a global deep learning solution for a high-dimensional, multi-sector economic model featuring both production and investment networks. We solve the model by training a neural network to minimize residuals from the key equilibrium conditions. Our first focus is on the divergence between the deep learning solution and the perturbation methods. The first-order solution generates large residuals, particularly for the resource constraint, and overstates volatility in aggregate variables. We analyze how the network structure, the granularity, and heterogeneities across sectors affect the performance of the global and local solution methods. We then apply our method to address substantive macroeconomic issues to demonstrate the relevance of our approach.

Financial and Macroeconomic Data Through the Lens of a Nonlinear Dynamic Factor Model

Pablo Guerron
,
Boston College
Alexey Khazanov
,
Hebrew University
Molin Zhong
,
Federal Reserve Board

Abstract

Through the lens of a nonlinear dynamic factor model, we study the role of exogenous shocks and internal propagation forces in driving the fluctuations of macroeconomic and financial data. The proposed model allows for nonlinear dynamics in the state and measurement equations; can generate asymmetric, state-dependent, and size-dependent responses of observables to shocks; and can produce time-varying volatility and asymmetric tail risks in predictive distributions. We find evidence in favor of nonlinear dynamics in two important U.S. applications. The first approach utilizes interest rate data to extract a factor that allows for an effective lower bound and nonlinear dynamics. Our estimated factor coheres well with the historical narrative of monetary policy. We find that allowing for an effective lower bound constraint is crucial. The second recovers a credit cycle. The nonlinear component of the factor boosts credit growth in boom times while hinders its recovery post-crisis. Shocks in a credit crunch period are more amplified and persist for longer compared with shocks during a credit boom.

Equilibrium Multiplicity in Aiyagari and Krusell Smith

Kieran James Walsh
,
ETH Zurich
Eric Young
,
University of Virginia

Abstract

Repeatedly solving the Aiyagari (1994) model with random parameters, we construct hundreds of examples with multiple stationary equilibria. We never find multiplicity with risk aversion less than ≈ 1.49, depreciation less than ≈ 0.19, or income persistence less than ≈ 0.47, and multiplicity requires a disaster state for income. In cases with multiplicity, the lowest rental rate occurs near depreciation times the capital share. It is possible for the economy, without a change in fundamentals, to transition rationally from a higher-rate equilibrium to one with a lower rental rate, lower inequality, and lower welfare (for most agents). We also construct the first Krusell and Smith (1998) examples with multiple recursive competitive equilibria.

Solving Models of Economic Dynamics with Ridgeless Kernel Regressions

Mahdi Ebrahimi Kahou
,
Bowdoin College
Jesse Perla
,
University British Columbia
Rachel Childers
,
Bowdoin College
Geoff Pleiss
,
University British Columbia

Abstract

This paper proposes a ridgeless kernel method for solving infinite-horizon, deterministic, continuous-time models in economic dynamics, formulated as systems of differential-algebraic equations with asymptotic boundary conditions (e.g., transversality). Traditional shooting methods enforce the asymptotic boundary conditions by targeting a known steady state, which is numerically unstable, hard to tune, and unable to address cases with steady-state multiplicity. Instead, our approach solves the underdetermined problem without imposing the asymptotic boundary condition, using regularization to select the unique solution fulfilling transversality among admissible trajectories. In particular, ridgeless kernel methods recover this path by selecting the minimum norm solution, coinciding with the non-explosive trajectory. We provide theoretical guarantees showing that kernel solutions satisfy asymptotic boundary conditions without imposing them directly, and we establish a consistency result ensuring convergence within the solution concept of differential-algebraic equations. Finally, we illustrate the method in canonical models and demonstrate its ability to handle problems with multiple steady states.

Discussant(s)
Matias Covarrubias
,
Banco de España
Dongho Song
,
Johns Hopkins University Carey
William Jungerman
,
University of North Carolina
Grey Gordon
,
Federal Reserve Bank of Richmond
JEL Classifications
  • C0 - General