| As far as is known, this dissertation is the first reported study of weak-form efficiency of futures markets for foreign currency. Using daily data for a total of 15 British pound and Canadian dollar futures contracts traded on the International Monetary Market of the Chicago Mercantile Exchange for the period of 1975-1980, this study seeks to test Paul Samuelson's influential model of a competitive, informationally efficient futures market. The principal testable implication of Samuelson's model is that the first difference of natural logarithm of prices for any given futures contract should be approximately serially uncorrelated with mean zero.;Several methodological features of this study are also novel. To avoid possible bias in serial correlations test, only contracts free of price limit moves are analyzed. A nonparametric test for constancy of variance is employed to reveal violation of a key assumption on which parametric tests for autocorrelation depend. Nonstationary series are analyzed in stationary segments. Several serial correlation tests are used for the first time in the study of futures prices, including the Fisher and Siegel tests for periodicity, Kendall's turning points test, and an F test based on autoregressions. Univariate forecasting models are used as a basis for simulated trading in currency futures.;Results of three parametric and three nonparametric time domain tests for serial correlation, as well as three frequency domain tests, suggest that the random walk special case of Samuelson's model is inconsistent with the observed properties of 5 of 15 currency futures contracts analyzed. The observed serial correlation, however, is modest and unsystematic across contracts. Simulated trading based on autoregressive forecasting models for four contracts suggests that the observed dependencies in these series could have been profitably exploited by traders affiliated with class A clearing members of the I.M.M. Thus, the observed serial correlation, although slight, appears to be economically as well as statistically significant. |
