An Empirical Analysis of Ex Ante Profits from Forward Speculation in Foreign Exchange Markets
构建了六种外汇市场远期投机事前利润的时间序列带，用非参数检验发现事前利润不可预测，并将其波动归因于名义汇率偏离鞅过程及贸易条件变动，对研究外汇市场效率的学者有参考价值。
This paper constructs a time series band for ex ante profits from forward speculation and examines the permanent component of the median of the band for six different exchange markets. The unpredictability of ex ante profits is rejected using nonparametric tests. Deviations of ex ante profits from forward premia are attributed to deviations of nominal exchange rates from martingale processes. It is shown that movements in the terms of trade are responsible for most of the variability and serial correlation properties of ex ante profits. T is by now widely recognized that the forward rate is not an unbiased predictor of future spot rates and that realized nominal profits from spec- ulating in the forward market are nonnegligible and highly volatile (see, e.g., Hodrick (1987), Frankel and Meese (1987)). However, the nature of these profits is as of yet undetermined. This paper attempts to shed light on this issue by examining the time series properties of ex ante profits from forward speculations in six foreign exchange markets. I focus on three aspects of the problem. First, I provide a measure of the range and the variability of ex ante profits, two issues with contradictory evidence in the literature (see Hodrick (1987) and Frankel (1988) for refer- ences). Second, I relate deviations from uncov- ered interest parity to risk and to deviations of nominal spot rates from martingale processes. Third, I link the time series properties of ex ante profits to the behavior of the real exchange rate. Since ex ante profits are not observable, the empirical analysis requires the construction of a time series for the expected future spot rate. Frankel and Froot (1987), Ito (1988), Cumby (1988), and Diebold and Nason (1990) have all suggested ways of computing this series.' Here I provide an alternative method for constructing a time series for expected future spot rates and for statistically assessing the properties of ex ante profits. Expected future spot rates are measured as a 90% confidence band obtained from the simulated recursive distribution for the linear predictor of the series. I simulate the distribution of the linear predictor rather than use simple point forecasts in order to reduce the forecast error due to parameter uncertainty. An addi- tional measure for expected future spot rates considered is a point estimate of the permanent component of the median of the band. I use nonparametric tests to examine the rela- tionship between ex ante profits and other series. Nonparametric tests are employed here for two reasons. First, I am interested in the entire popu- lation properties of the data. Existing results are derived by examining only the first and second moments of the data. Second, efficiency tests are generally based on linear parametric functional forms. Pagan and Ullah (1988) have pointed out problems with these tests when the conditional mean of the data is nonlinear. Nonparametric tests overcome these problems. The results indicate that nominal (and real) rates do not follow martingale processes, that a risk premium rarely explains the behavior of the ex ante profits and that variations in expected terms of trade account for the time series proper- ties of ex ante profits.