Comparing Conditional Hedging Strategies

Abstract

The traditional approach to discriminate amongst two competing hedging strategies is to compare the sample portfolio return variance implied by each strategy. This simple approach suffers from two drawbacks. First, it is an unconditional performance measure which is theoretically not coherent with a dynamic hedging strategy that minimizes the conditional portfolio return variance. Second, estimating unconditional performance over the entire period may not be sufficcient since a strategy with a good unconditional hedging performance may not perform well at a particular point in time. In this paper, I use the Giacomini and White (2006), the Wald, and the Diebold and Mariano (1995) statistical tests in order to conditionally (and as a special case, unconditionally) compare the portfolio return variances implied by two competing hedging strategies. The attractive feature of the conditional perspective is that, in case of rejection of equal conditional hedging effectiveness among two initial strategies, it provides us with a new hedging strategy that selects at each date the initial strategy that will perform the best next period, conditional on current information. An application to several agricultural commodities illustrates the technique. For daily hedging horizons, it is found that most of the time Ederington's (1979) static strategy is superior to more elaborate dynamic strategies. This calls into question earlier results reported in the literature that were based on a much smaller database.

Keywords: GARCH, Hedging, Strategy, Portfolio, Variance, IT, Performance, Time, Tests, Order, Effectiveness, Information, Database