A0800
Title: The quantile regression forest reloaded: The distributional random forest
Authors: Joan Paredes - European Central Bank (Germany)
Jeffrey Naef - University of Geneva (Switzerland) [presenting]
Abstract: Assessing the joint risks of inflation and economic activity, so-called stagflation risks, is a key priority for central banks with medium-term-oriented mandates. A novel approach is introduced based on multivariate evolution of the quantile regression forest, the distributional random forests (DRF), which allows for the flexible modeling of the joint distribution of inflation and activity, conditional on a rich set of macroeconomic predictors. The method extends traditional random forest techniques by jointly estimating the conditional distribution of multiple outcomes, thereby capturing complex, potentially nonlinear interactions and co-dependencies relevant for stagflation dynamics. Drawing inspiration from the Barro-Gordon framework, the focus is on the probability of scenarios where inflation remains elevated while output contracts, events that are particularly challenging for monetary policy.
