No abstract

This study investigates the nature of price changes in a variety of major and minor foreign exchange markets. The results suggest that the log of price changes over one (trading) day intervals seems to follow a non-normal stable distribution function. Different measures of location (and to lesser extent scale) are present for different days of the week. Dollar denominated price changes are high on Mondays and Wednesdays and low on Thursdays and Fridays for all currencies. The Wednesday-Thursday result is consistent with the settlement procedures used in foreign exchange transactions in the dollar. The Friday-Monday result is consistent with an increase in demand for the dollar prior to the weekend.THIS STUDY OFFERS SOME previously unavailable results on the nature of price changes in a certain set of speculative markets. In particular, this is a study of the distributional characteristics of daily and weekly price changes for a number of major and minor foreign exchange markets. The results document the existence of non-normal stable distributions for price changes in foreign exchange markets. These price change distributions differ by day of the week in ways that are only partly attributable to the institutional structure of the market. The results of the study are interpreted within the context of understanding how foreign exchange markets operate, and within the context of prior studies of stock market data concerned with summarizing the probability distribution of price changes for different intervals, including days of the week.Since Bachelier's [1] original treatise that contained the proposition that speculative market prices follow a process that could be described by a normal distribution, there have been numerous alternative attempts to describe the stochastic processes which characterize speculative market prices. These studies have not resulted in any generalized description of the probability distribution of price changes in such markets. Indeed, the evidence suggests substantial differences in the process among various types of financial instruments. The failure to achieve consensus can be traced to the lack of robustness in the statistical procedures employed to estimate processes and to differences in the underlying institutional or economic arrangements which are unique to that market. The possibility of differences in the information set used to assess value or differences in technology for transmitting information makes it important to secure additional for their comments. 694The Journal of Finance knowledge of the pricing process in different speculative markets. Such knowledge is essential to the formation of estimates of risk and return for the construction of optimal investor portfolios, as well as for understanding how speculative markets operate.The implications of the results of this investigation for the nature of speculative markets are both general and specific. General implications are reached concerning the process by which foreign exchange markets adjust to either information flows...

No abstract

THE COMMENT BY SO [12] on our 1982 paper (McFarland, Pettit, and Sung [8])provides an interesting extension to our work on foreign exchange markets and yields additional insights into the distributional properties of foreign exchange price changes. The contribution of this comment lies in the application of recently developed statistical methods, not available at the time of our study, for estimating the parameters of asymmetric stable distributions.' In estimating the skewness parameter in addition to the estimates for the characteristic exponent, scale, and location, the author finds that, in some cases, foreign exchange price changes are better represented by asymmetric stable distributions than by symmetric stable distributions. As in our study, only a small number of the skewness parameters are significantly different from zero. Furthermore, the author concludes, "The location estimates show negative Monday returns for all currencies except for the yen," and "These findings are not necessarily in conflict with MPS." None of these location estimates, however, is significantly different from zero ([12], Tables III and IV). These new estimates indicate that day-of-theweek effects are reduced. If true, the results imply that our economic rationale for the Monday effect may no longer be appropriate (based on risk premia or settlement practices in these markets), and an alternative may be needed. needed.In a recent study, Akgiray, Booth, and Seifert [2] test the assumption of asymmetric stable distributions for monthly exchange rates for twelve Latin American countries and compare them with five major trading nations. In summarizing their results, they state:The infinite variance (asymmetric) stable law hypothesis is tested using a regression-type technique to estimate the parameters of the stable Paretian distribution. Strong evidence is found supporting this hypothesis for Latin American exchange rates. In contrast, it is rejected for most of the major trading nations. A separate analysis of extreme observations using a robust statistical procedure further supports these findings. 1 There appear to be some rather dramatic differences, at least based on casual observation, between our data base and the one used in this comment. The lack of comparability of data bases in this area in terms of the sources, the selection of particular series, and the construction of variables plagues attempts to compare empirical results. 189