Paul E. Peterson

Using mechanical trading systems to evaluate the weak form efficiency of futures markets

An efficient market has been described by where irt+1 is the profit in the next period, and At Fama (1970) as one in which prices always fully is the information available in the current period. reflect all available information. Of the three Samuelson (p. 44) explained it this way: Let us tests of efficiency discussed, the weak form test observe numerous sequences of futures prices is concerned with the randomness of price generated by (a martingale) up until their termimovements and measures the ability to predict nal date. They will turn out, on the average, to future price changes from past and present have no upward or downward drift anywhere! changes. There are two general ways to evaluate emphasis his). Given this, . .. there is no way weak form efficiency: statistical tests and meof making an expected profit by extrapolating chanical trading rules. Statistical methods, inpast changes in the futures price, by chart or any cluding serial correlation, spectral analysis and esoteric devices of magic or mathematics. The nonparametric runs tests, permit hypothesis testmarket quotation already contains in itself all ing, but Fama and Blume (p. 227) point out that that can be known about the future, and in that they may be of limited value with complex or sense it has discounted future contingencies as irregular price structures. much as is humanly possible. Fama (1970, p. Mechanical trading systems, such as filter 385) added that assumptions of weak form effirules and moving averages, provide a more sensiciency rule out the possibility of trading systive test for nonrandomness, because they do not tems based only on information in (t that have depend on the pattern or cause of the price expected returns in excess of equilibrium exchanges (Bear and Stevenson, p. 980). However, pected profits or returns. We would then expect Cargill and Rausser (1975, pp. 1045-1046) noted a weak form efficient market to yield zero profits that while results from filter rule tests generally to any mechanical trading scheme, thus the null parallel those from serial correlation tests, the hypothesis for any statistical test would be zero lack of agreement on the level of expected profits, with any excess or nonzero returns profits and the inability to make probabilistic indicating some degree of weak form ineffistatements severely limit their use. Conseciency. quently, there have been only a few weak form The choice of a benchmark of zero for the studies (Houthakker; Leuthold, 1972; Smidt; equilibrium expected profits or returns reStevenson and Bear) that employ trading rules. quires elaboration. Praetz (1976, 1979) proposed This paper develops a general framework for that the returns to a buy and hold strategy be using mechanical trading systems as a test of used as a benchmark in futures market analysis, weak form efficiency in futures markets, and as it is in studies of the securities markets. Two creates a procedure for statistical analysis of the futures market studies (Houthakker; Stevenson results produced by these methods. An example and Bear) have in fact used the buy and hold is given using filter rules to test the weak form benchmark. However, we seriously question the efficiency of the hog futures market from 1973 to validity of the buy and hold strategy in connec1977. tion with futures market research. The efficient market hypothesis was originally used to evaluate securities markets, where THEORETICAL BACKGROUND stockholders buy shares and expect to benefit from share price increases in addition to diviIt has been shown by Samuelson and Mandeldends. Regular dividends represent the equilibbrot (1966) that speculative prices follow a marrium expected profits or returns, and share tingale process price increases becomes the excess returns discussed by Fama (1970).l Since futures conE(7rt+lI(t) = 0 tracts have no guaranteed return, there is nothing