Distributional modeling and forecasting of natural gas prices

We examine the problem of modeling and forecasting European day-ahead month-ahead natural gas prices. For this, we propose two distinct probabilistic models that can be utilized in risk portfolio management. use daily pricing data ranging from 2011 to 2020. Extensive descriptive analysis shows both time series feature heavy tails conditional heteroscedasticity show asymmetric behavior their differences. state-space under skewed, heavy-tailed distributions capture all stylized facts data. They include impact autocorrelation, seasonality, premia, temperature, storage levels, price Emission Allowances, related fuel prices oil, coal, electricity. provide rigorous model diagnostics interpret components detail. Additionally, conduct a study with significance tests compare predictive performance against literature benchmarks. The proposed (month-ahead) leads 13% (9%) reduction out-of-sample continuous ranked probability score (CRPS) compared best performing benchmark model, mainly due adequate volatility tails.