آیا نرخ ارز بصورت سریالی همبسته است؟ شواهد جدید از بازار FX یورو

This paper examines the serial uncorrelatedness hypothesis in the Euro FX markets by testing for autocorrelation in daily FX returns of 82 countries over the period of 1999–2010. We use three newly developed tests that are robust to conditional heteroskedasticity of unknown forms and that do not choose a lag parameter arbitrarily. They are Escanciano & Lobato (2009)'s automatic Box–Pierce Qp test, Nankervis & Savin (2010)'s generalized Andrews–Ploberger test and Deo (2000)'s robust Durlauf test. We find no significant autocorrelation in the FX returns of around 58 to 62 countries, suggesting that majority of the Euro FX markets are weak-form efficient.

The efficiency of financial markets has been a long standing debateable issue in the finance literature. This idea implies that (changes in) asset prices are serially uncorrelated, i.e., the serial uncorrelatedness hypothesis. Testing for autocorrelation therefore has held a central role in the empirical finance literature on the issue. However, Robinson (1991) notes that traditional tests for autocorrelation are limited in the sense that they ignore the possibility that a time series can be statistically uncorrelated but still not independent over time. In particular, they may not be robust to nonlinear dependency (such as conditional heteroskedasticity). Since then, there has been a growing interest in developing new tests for autocorrelation that allow for nonlinear dependence of unknown forms. See Francq, Roy, and Zakoian (2005) for a review of this literature. The presence of nonlinear dependence is well known in the foreign exchange markets. Famous examples include Hsieh (1989) and Panayotis and Apergis (1996). Hsieh (1989) first reported evidence of nonlinear dependence in daily foreign exchange rates. Panayotis and Apergis (1996) show that traditional tests provide mixed results on the efficient market hypothesis in the foreign exchange market. However, once the ARCH effect (a form of conditional heteroskedasticity) is properly accounted for, there is conclusive evidence supporting the efficient market hypothesis. The advent of the Euro is one of the important developments in the recent history of international finance. Since its introduction in 1999, the Euro has quickly established itself as the second most widely held international currency after the US dollar. Its importance suggests that a better understanding of the behaviour of Euro-based exchange rates is needed because the majority of previous studies examined FX rates against the US dollars only. This paper aims to re-examine the serial uncorrelatedness hypothesis in the Euro foreign currency markets with three newly developed tests that allow for conditional heteroskedasticity of unknown form. These tests have the additional advantage that we do not need to determine the lag length arbitrarily as it is data-driven. A comprehensive dataset that covers Euro-based FX rates of 82 countries over the period 1999 to 2010 is used. The results show that out of the 82 foreign currencies examined, we can reject the serial uncorrelatedness hypothesis in only 20 to 24 of them. These foreign currencies are typically currencies of small countries with undeveloped financial markets. Further analysis on sub-period rolling results confirms the main result. The rest of this paper is organized as follows. We first provide a selected literature review in Section 2 where the motivation for our study is given. Section 3 outlines the methodology and dataset, followed by Section 4 where the empirical results and sensitivity analyses are reported and discussed. Section 5 concludes this paper.

We examined the serial uncorrelatedness behaviour of 82 foreign currencies against the Euro using three newly developed tests that allow for conditional heteroskedasticity of unknown form and that do not require arbitrary specification of maximum lag length. These tests show that most Euro-based exchange rates do not exhibit any serial correlation problems, suggesting that these foreign currency markets are weak-form efficient. The result is robust to different subsample periods and some different grouping methods. This paper provides empirical evidence that sheds light on the market efficiency of the Euro-FX markets. It may be interesting to look at the same problem in a decision-making framework where transaction costs and out of sample performance are relevant. Further research is needed to deal with the above issues and we leave these problems to other researchers who pursue research along this line in the future.