Document Type
Honors Project - Open Access
Abstract
Household debt plays a significant role in shaping the macroeconomy. The relationship between growth and household debt is nonlinear; however, most research regarding household debt and GDP utilizes linear techniques. Using a Deep Vector Autoregression (Deep VAR) model, a recursive neural network-based approach that has been shown to outperform linear alternatives, I analyze the relationship between household debt and GDP using quarterly data from an unbalanced panel of 39 countries over the past 40 years. The Deep VAR outperforms linear alternatives, further providing evidence that the relationship between household debt and GDP is non-linear. Using impulse response functions (IRF), I find evidence supporting Minsky’s financial instability hypothesis, highlighting the effects of relaxed borrowing conditions. Moreover, I identify a cyclical pattern in household responses to GDP shocks, where households deleverage during periods of economic growth and increase leverage during contractions.
Recommended Citation
Nutt, Bram, "Capturing Non-linearity between Household Debt and GDP: A Deep VAR Approach" (2025). Economics Honors Projects. 130.
https://digitalcommons.macalester.edu/economics_honors_projects/130
Included in
© Copyright is owned by author of this document
