Volatility Forecasting: How Horizon and Benchmark Design Shape Model Comparisons
AI-assisted research by PyInvesting. Sources and limitations are provided below. Educational content; not personalized investment advice.
AI-assisted article drafted and checked with AI against the sources below. Educational information, not personal investment advice.
How can forecast horizons change the apparent leader?
The Federal Reserve study reports that the strongest forecasting model differs by horizon, with Markov-switching HAR leading at short horizons and ARFIMA generally leading at the monthly horizon. [S7]
The same study says machine-learning models sometimes improve on HAR but do not consistently outperform the wider group of econometric models. [S7]
What does a unified trading framework add?
The commodity-futures study compares several seasonal methods within a unified trading framework rather than presenting each method in isolation. [S4]
Its evidence does not establish robust benchmark outperformance and shows that model performance varies materially across market subperiods. [S4]
What this means
Questions for the reader: Is the forecast horizon clearly defined? Is the benchmark appropriate for the strategy being assessed? Are transaction costs and evaluation measures disclosed? Does the comparison distinguish average results from median results, drawdowns, and statistical tests? Does the evidence show consistent performance across subperiods, or only a result for one sample?
Limitations
This article uses the supplied excerpts rather than the full papers. The Federal Reserve excerpt concerns realized-volatility forecasting for the S&P 500 and 40 U.S. equities, while the arXiv excerpt concerns seasonal trading in 15 liquid commodity futures. The excerpts support a framework for reading model comparisons, but they do not establish that either study's results transfer to other markets, horizons, or investment products.
read model comparisons across volatility horizons
Sources
- Linear and Nonlinear Econometric Models versus Machine-Learning Models: Evidence from Realized-Volatility Forecasting — 2026-09-02T15:30:00+00:00
- Seasonal Trading in Commodity Futures: Evidence from Regression and Singular Spectrum Signals — 2026-09-14T04:00:00+00:00