Nonparametric VAR with nonlinear factors for macro analysis
This paper proposes a vector autoregression augmented with nonlinear factors, modeled nonparametrically using regression trees. It offers four key advantages for macroeconomic forecasting and structural economic analysis.
Four advantages for complex time series
The model's primary advantages include nonparametric modeling of nonlinearities, reducing misspecification risks.
It employs factor methods for parsimonious modeling, exhibiting functional pooling where a few nonlinear factors capture common nonlinearities across variables.
Bayesian computation via MCMC is efficient, even in high-dimensional settings, avoiding bottlenecks common in alternatives like time-varying parameter VARs.
Furthermore, existing methods for identifying structural economic shocks in linear factor models can be readily adapted for this nonlinear framework.
Exercises with artificial and macroeconomic data demonstrate the model's properties and its utility for forecasting and structural analysis, including uncovering significant asymmetries in responses to financial and monetary policy shocks.
Navigating high-dimensional challenges
Macroeconomic and financial research frequently deals with high-dimensional time series, where traditional VARs and factor models face significant challenges.
Incorporating nonlinearities into parametric VARs often leads to over-parameterized and computationally demanding models, such as time-varying parameter VARs with dozens of variables.
This complexity, coupled with the risk of misspecification when adopting specific parametric forms, highlights the need for more flexible and efficient approaches.
The proposed model addresses these issues by balancing over-parameterization, misspecification, and computational burdens, offering a robust alternative for complex datasets.
A defensive strategy for robust models
This model offers a robust solution to long-standing issues in macroeconomic modeling, particularly for high-dimensional, nonlinear data.
Its nonparametric approach significantly reduces misspecification risk and computational load, making complex analysis more feasible.
While technically advanced, its implications for more reliable forecasting and structural analysis are broadly significant for policy-makers.