Abstract
Investors and supervisors increasingly monitor oil, carbon, green asset and cryptocurrency markets as sources of equity risk. This study tests whether these markets improve forecasts of the realised volatility of ten STOXX Europe 600 sectors beyond a heterogeneous autoregressive (HAR) benchmark that already includes market volatility and the VIX. Using weekly data from 2020 to 2026 and 181 out-of-sample weeks, it compares HAR models augmented with fossil energy, carbon, green and digital asset predictors, forecast combinations, penalised regressions and tree ensembles. The external predictors are jointly significant in sample for five sectors, but out of sample, the fossil, carbon and green blocks do not improve one-week forecasts on average. The digital asset block yields small gains (mean out-of-sample R2 of 0.33%) that survive false discovery rate control for five sectors but for only one when weeks end on Friday. Penalised regressions gain 2.0–2.6%, yet versions restricted to the benchmark predictors achieve 1.3–2.1%, so most of the gain comes from regularisation. With retransformation-corrected variance forecasts, model confidence sets retain almost all models, and volatility timing strategies underperform buy and hold after feasible execution. Robust incremental gains from climate transition information are not established; digital asset information offers small, fragile, sector-specific gains.

This publication has 50 references indexed in Scilit: