The Quality of Volatility Traded on the Over-the-Counter Currency Market : A Multiple Horizons Study

Previous studies of the quality of market forecasted future volatility in currency options use implied volatilities from exchange-traded currency options markets, and find that though implied volatility has substantial informational content, it is a biased predictor of future volatility. However, such empirical results are likely to be affected by two sources of well-documented errors: measurement errors in model inputs and errors in the option pricing model that is used for computation.This paper focuses on the former source of errors, and differs from previous studies in that it uses quoted implied volatility data from the OTC currency option market. The institutional features of the OTC market alleviate the measurement problems that are found in studies which use implied volatility that is derived from exchange-traded option prices. Unlike exchange traded currency option markets, in which market players quote option prices in terms of option premiums, in the OTC currency option market the price quotes are actually made in terms of volatility, which is expressed as a percentage per annum. Furthermore, the OTC currency options have daily quotes for standard maturities, allowing us to study the market's ability to forecast future volatility for different time horizons.The evidence shows that quoted implied volatility is an unbiased estimator of future volatility at the one-month horizon, but its predictive power decreases with longer horizons. The results are consistent with the Figlewski (1997) hypothesis that the informational content of quoted implied volatility is positively related to liquidity. The results also indicate that the quoted implied volatility has more predictive power than the historical standard deviation, RiskMetrics, and GARCH-based volatility forecasts across all of the currency pairs and forecasting horizons.These results are consistent with the argument that measurement errors have a substantial effect on the implied volatility estimator and the quality of the inferences that are based on it