Combined risk factor models boost euro area tail forecast accuracy
An ESCB expert group has developed a Macro-at-Risk framework using 350 indicators to model tail risks surrounding euro area growth and inflation projections. The study introduces the M@RX MATLAB toolbox and novel copula and parametric tilting techniques for central bank forecasting.
Decoding tail risk drivers
The study evaluates quantile regression models across 350 quarterly macro-financial indicators divided into nine categories.
Results demonstrate that predictive power is highly horizon- and direction-dependent.
Labor market variables, such as unemployment and NAIRU, prove particularly informative for medium-term upside inflation risks, whereas uncertainty, money, and credit aggregates dominate downside inflation risk projections.
For real GDP growth, financial conditions and monetary aggregates reliably drive downside tail risks, while labor market slack gains importance over longer horizons.
Crucially, multi-variable models combining distinct indicator groups systematically outperform single-factor specifications by capturing complementarities across risk categories.
Bridging projections and probability
To make density forecasts operationally relevant for central bank decision-making, the authors introduced technical innovations integrated into the new M@RX MATLAB toolbox.
A parametric tilting technique re-centres model-derived predictive distributions around official Eurosystem point projections without distorting underlying tail asymmetries.
Additionally, a Gaussian copula method aggregates multi-step quarterly forecasts into annual average probability distributions while preserving serial correlation.
Applied to December 2022 projections, the framework accurately anticipated persistent upside inflation risks and persistent downside risks to euro area economic growth.
Essential tools with structural blind spots
The M@RX framework delivers a vital quantitative tool for evaluating macroeconomic uncertainty beyond simple baseline projections.
Yet, its reduced-form architecture cannot explain the structural causes behind extreme tail events.
Without structural grounding, these models offer clear warnings but little guidance on policy mechanics.