This paper investigates the ability of corridor implied variances (CIV) with different corridors to forecast conditional volatility of DJIA index returns, and compares their performance with a CBOE volatility index, VXD, by employing several GARCH models in a model-based out-of-sample context. Besides, it explores the reasons behind the differences in the forecasting ability of CIVs and VXD through a decomposition of the model-free implied volatility. In addition, it addresses the economic difference among aforementioned implied volatility measures in a simulated options market. We find that narrow-corridor CIVs outperform wide-corridor CIVs and VXD in terms of the forecasting ability, as wide-corridor CIVs and VXD impound information from deep out-of-the-money options whose prices contain large volatility risk premiums and may not reflect a fair market expectation of volatility. In the economic sense, wide-corridor CIVs and VXD outperform narrow-corridor CIVs in turbulent periods, while narrow-corridor CIVs outperform wide-corridor CIVs and VXD in medium-and low-volatility periods. The profitability pattern is consistent both in-sample and out-of-sample, before and after transaction costs are considered, and is also robust to option strategy choices.