Bayesian neural networks for macroeconomic analysis
Hauzenberger, Niko and Huber, Florian and Klieber, Karin and Marcellino, Massimilano (2024) Bayesian neural networks for macroeconomic analysis. Journal of Econometrics. 105843. ISSN 0304-4076 (https://doi.org/10.1016/j.jeconom.2024.105843)
Preview |
Text.
Filename: Hauzenberger-etal-JE-2024-Bayesian-neural-networks-for-macroeconomic-analysis.pdf
License: 
Download (1MB)| Preview |
Abstract
Macroeconomic data is characterized by a limited number of observations (small T), many time series (big K) but also by featuring temporal dependence. Neural networks, by contrast, are designed for datasets with millions of observations and covariates. In this paper, we develop Bayesian neural networks (BNNs) that are well-suited for handling datasets commonly used for macroeconomic analysis in policy institutions. Our approach avoids extensive specification searches through a novel mixture specification for the activation function that appropriately selects the form of nonlinearities. Shrinkage priors are used to prune the network and force irrelevant neurons to zero. To cope with heteroskedasticity, the BNN is augmented with a stochastic volatility model for the error term. We illustrate how the model can be used in a policy institution through simulations and by showing that BNNs produce more accurate point and density forecasts compared to other machine learning methods.
ORCID iDs
Hauzenberger, Niko
ORCID: https://orcid.org/0000-0002-2683-8421, Huber, Florian, Klieber, Karin and Marcellino, Massimilano;
-
-
Item type: Article ID code: 90490 Dates: DateEvent5 September 2024Published5 September 2024Published Online19 August 2024AcceptedSubjects: Social Sciences > Finance Department: Strathclyde Business School > Economics Depositing user: Pure Administrator Date deposited: 05 Sep 2024 14:27 Last modified: 11 Nov 2024 14:26 URI: https://strathprints.strath.ac.uk/id/eprint/90490
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)