ABSTRACT
Recently, many time series at high frequency have become available for many markets. This paper tests one such high-frequency series from the foreign exchange market. Such series have the potential to reveal the detailed dynamics going on at the actual level of trading, using information which is lost at the daily, weekly, and monthly horizons. M
This paper performs several simple diagnostics on these new series. In Section 2 it describes the series used along with some of the problems involved in using them. These are not the ideal series one would want to study. Unfortunately, such series do not exist in the foreign exchange market. Finally, it provides some basic summary statistics and tests for linear dependence on both the price series and the times between recorded bids. Section 3 continues with some nonlinear diagnostic tests. Section 4 concentrates on the relation of the time between quotes to the magnitude of returns. Section 5 deals with the issue of prediction versus actual trading profits. A very simple trading rule is simulated and the results are discussed both in terms of their economic and statistical significance. Some important points are made concerning forecasting and the economic objective of trading profits. Finally, Section 6 concludes and suggests some work for the future.
