A comparative analysis of quantile regression models in predicting stock returns under extreme market conditions

Details

Supervisors

Iania, Leonardo

Faculty

Louvain School of Management

Degree label

Master [120] : ingénieur de gestion, à finalité spécialisée

Abstract

enAbstract: This master thesis examines whether macro-financial indicators contain predictive information for monthly STOXX Europe 600 log-returns, and whether these relationships are asymmetric across market states and unstable over time. Using a monthly dataset spanning January 1999 to August 2025, the study focuses on five widely used predictors implied like volatility (VSTOXX), term-spread shocks, oil-price changes, inflation, and industrial-production growth, to assess return predictability at both the center and the tails of the distribution. Empirically, the analysis proceeds from a mean benchmark to distribution- and time-sensitive frameworks. A heteroskedasticity-robust OLS benchmark suggests predictability is weak on average, with volatility standing out as the most robust mean predictor. In contrast, Quantile Regression (QR) reveals pronounced heterogeneity. Volatility exhibits strong tail asymmetry, inflation matters primarily in the lower tail, and term-spread shocks become relevant in selected quantiles, while oil prices and industrial production display limited stable effects. Time-Varying Quantile Regression (TVQR) further shows that these quantile effects evolve across crises and calmer regimes, indicating that predictability, when present, is regime-dependent rather than constant. A Bayesian Quantile Regression (BQR) robustness check yields wide credible intervals spanning zero across predictors and quantiles, highlighting substantial parameter uncertainty at the monthly frequency. In-sample evaluation confirms that TVQR delivers the lowest pinball loss and near-ideal quantile calibration, outperforming static alternatives. Overall, this study, among the few to focus specifically on the STOXX Europe 600, concludes that macro-financial predictability is it is best characterized as tail-dependent and time-varying and should be interpreted cautiously when uncertainty is fully accounted for.