Empirical tests of canonical nonparametric American option-pricing methods

Alcock and Carmichael (2008, The Journal of Futures Markets, 28, 717-748) introduce a nonparametric method for pricing American-style options, that is derived from the canonical valuation developed by Stutzer (1996, The Journal of Finance, 51, 1633-1652). Although the statistical properties of this nonparametric pricing methodology have been studied in a controlled simulation environment, no study has yet examined the empirical validity of this method. We introduce an extension to this method that incorporates information contained in a small number of observed option prices. We explore the applicability of both the original method and our extension using a large sample of OEX American index options traded on the S&P100 index. Although the Alcock and Carmichael method fails to outperform a traditional implied-volatility-based Black-Scholes valuation or a binomial tree approach, our extension generates significantly lower pricing errors and performs comparably well to the implied-volatility Black-Scholes pricing, in particular for out-of-the-money American put options.