VIX forecast — Ornstein-Uhlenbeck mean reversion
Three selectable mean-reverting models, each fit in log-space on CBOE VIX daily closes from FRED (VIXCLS) with bootstrap-resampled residual shocks so fat tails survive: Ornstein-Uhlenbeck (AR(1), default), a second-order autoregression (AR(2)) that lets recent momentum influence the near-term path, and a regime-switching model that fits separate long-run means for calm/normal/stress states. A deliberately different methodology family from the index-level and Fear & Greed forecasts, since the VIX reverts to a long-run level rather than trending.
Current VIX: 16.59 (Normal). Fitted long-run mean level: 18.7. Reversion half-life: 8.5 days. Figures reflect the default Ornstein-Uhlenbeck fit at the 3-month horizon; the interactive page also lets you switch horizon (3/6/12 months) and model.
Frequently asked questions
- How is the VIX forecast built?
- A log-space Ornstein-Uhlenbeck (mean-reverting AR(1)) model fit on CBOE VIX daily closes from FRED (VIXCLS), with bootstrap-resampled residual shocks so fat tails survive. The fit yields a long-run mean level and a reversion half-life in days. Educational research; not investment advice.
- Why mean reversion instead of a trend model for the VIX?
- The VIX is empirically bounded and reverts to a long-run average level after spikes — it does not trend the way an equity index can. A mean-reverting process fits that behaviour far better than a drift-based model.
- Can I choose a different VIX forecasting model?
- Yes — the forecast page offers three selectable mean-reverting models: Ornstein-Uhlenbeck (default), a second-order autoregression (AR(2)) that lets recent momentum influence the near-term path, and a regime-switching model with separate long-run means for calm/normal/stress volatility states.
Educational research only — not investment advice.