| Summary: | 碩士 === 國立高雄應用科技大學 === 國際企業系 === 98 === Since the classic paper published by Meese and Rogoff (1983), exchange rate forecast becomes an important issue and particularly in model selection and econometric methods that might influence forecast results.
In this study, using the microstructure approach, we adopt order flow as the determinate variable, which is proposed by Evans and Lyons (2002); then investigate whether order flow can be a determination and explain exchange rate behavior. Our data comes from EBS (Electronic Brokerage Services) database, including U.S. dollar/ Japanese Yen that occurs from 2007/1/3 to2007/12/28. By using hourly and daily frequency data, we discuss the forecast ability of order flow to exchange rate, and test for forecast ability using RMSE and MAE. We expect order flow can provide better forecasting performance than a random walk model.
Our experiment results show that, in the two different frequencies, the coefficients of order flow are significant. For out-of-sample forecast: (i) The microstructure approach in daily frequency has a better forecast ability relative to a random walk model, which is consistent with Evans and Lyons (2005). (ii) During our sample, information which is conveyed by order flow, is not absorbed by the dealer as quick as we think. The delay effect can affect the dealer’s behavior, and help to increase the forecast ability.
|
|---|
