Implied vs Historical Volatility as Predictors of Future Realized Volatility

Sammanfattning

This thesis investigates whether implied volatility, extracted from option prices usingthe Black–Scholes model, provides more accurate forecasts of future realized volatilitythan historical volatility. The analysis covers three major US equity indices and theindividual stocks of three large-cap companies. Predictive performance is evaluatedthrough univariate and multivariate regressions, best subset model selection, and out-of-sample forecast error metrics, including mean absolute error and root mean squarederror.For all three indices, implied volatility is the strongest predictor of future realizedvolatility, producing the lowest forecast errors and remaining statistically significant inmultivariate models. For individual stocks, the pattern reverses: historical volatilityoutperforms implied volatility across all three companies, and implied volatility isexcluded entirely from the best-fitting model for Procter & Gamble. These findingsare consistent with the semi-strong form of the Efficient Market Hypothesis, whereimplied volatility dominance at the index level reflects efficient information aggregationin liquid option markets, while historical volatility proves more reliable in thinnermarkets where option prices are more volatile.