The Market Conditions chip on the VIX forecast is a 3-state Markov-switching model. It assumes volatility is always in one of three hidden states — calm, normal or stress — each with its own mean level and volatility of log(VIX), plus a probability of switching between them.
A single mean-reverting model (like the log-OU baseline) treats a sleepy summer tape and a full-blown panic with the same math. But volatility plainly behaves differently inside a stress episode than in a calm market: in stress it is higher, jumpier and slower to settle; in calm it drifts quietly near its floor. The Market Conditions model captures that by letting each state carry its own mean and volatility, and by estimating how likely VIX is to stay put or switch. The forecast then blends the states by how probable each one is right now.
Flexibility isn't free. Three states mean more quantities to estimate from a limited history, which raises the risk of reading patterns into noise. Whether that flexibility pays off on current data is exactly what the "Model performance" table under the VIX forecast is for: it shows the Market Conditions model's coverage, MAPE, RMSE, Brier and hit-rate next to the simpler log-OU and AR(2) baselines, so you can see whether the extra states earn their complexity.
Calm, normal and stress. The model treats these as hidden states log(VIX) switches between, each with its own mean level and volatility, and estimates how likely it is to stay in one or jump to another.
Because volatility behaves differently inside a stress episode than in a calm market — higher, jumpier and slower to settle in stress; quiet near its floor in calm. One fixed equation can't express that; three states with their own dynamics can.
More states mean more quantities to estimate from limited history, so there's a higher risk of fitting noise. Check the 'Model performance' table under the VIX forecast to see whether the Market Conditions model actually beats the simpler baselines on current data.
Browse all S&P 500 tickers to see this metric applied to individual companies.
Educational research only — not investment advice.