An econometric analysis on the time-variant performance of currency forwards: Have forward exchange rates become less useful predictors of future spot rates?

Based on a priori assumption that the prevailing digital upheaval has boosted market efficiency, this paper investigates if the predicting performance of currency forwards has significantly changed. The advent of the internet has alleviated information retrieval and thus it left its marks on some attributes of currency forwards that are worth examining.

The pivotal beginning of forecasting accuracy in an efficient market can be traced back to Messe and Rogoff (1983). In their seminal paper they argue that no short or medium term forecast can outperform a random walk model. Hence, their claim sparked a debate about the predictability of spot rates and galvanized a great many researchers to examine the forecasting performance of various markets. After all, it is tempting to challenge the very idea inherent in the random walk model, that today's spot rate is superior to fundamental economic knowledge or even professional forecasting services employing a wide variety of technical analyses. Eugene Fama's (1991) path-breaking study on market efficiency embedded eventually the random walk model within a theoretical framework. His hypothesis of an efficient market, reflecting all relevant information available, has long stood the test of time.

Studies in the field of financial forecasting have therefore predominantly revolved around the theoretical question whether a market is efficient at all and which model to employ for this task ahead. However, research has only modestly addressed the time dependent performance of currency forecasts. Hence, this paper intends to contribute towards current research by applying one particular model and monitor the performance of forecasts over a rolling time span.

Chiang (1985) suggested that a time-variant analysis on the regression coefficients could yield rich information that cannot possibly be ignored; however, his enquiry was drowned too soon by an overwhelming wave of cointegration set off by Engle and Granger (1987). Nevertheless recent evidence in the stock exchange market provided by Degiannakis, Livada and Panas (2008) continues to point at the increased sensitivity of regression coefficients in rolling time samples.

By tapping into the Reuters database, this study intends to collect raw data of spot and three months forward rates of major currencies, ranging from as far back as the collapse of the Bretton Woods system to present day. Data will be further processed by removing heteroscedasticity and skimmed for structural breaks. With the aid of an appropriate autoregressive model, low and high volatility periods are to be defined and checked for unit roots using the Augmented Dickey-Fuller test. This is justified since most of time series are indeed difference stationary processes; an idea originally put forward by Mohanty and Song (1995). The processed data will then be segmented into an appropriate number of sub samples and systematically checked for cointegration between spot and forward rates by using Engle and Granger's aforementioned method.

It should be remarked that literature has been rather polarised on the question whether a cointegration relationship exists. For example Clarida and Taylor (1993) critically reject the hypothesis of a cointegrating movement between spot and forward rates, whereas more innovative studies vindicate clearly such a liaison only when structural breaks are taken into consideration (Villanueva, 2007). However, such ambiguities shall not curb this papers intention to shed further light onto this subject. Even if the majority of collected samples reject the hypothesis of cointegration, still valuable clues about the performance of spot rate predictability can be extracted from various statistics like the coefficient of determination, residuals sum of square or Durbin-Watson parameter. If cointegration is found, a set of appropriate error correction models may give further insights about spot rate predictability. An Engle-Granger causality test can be an additional asset to fathom the relationship between spot and forward rates.

This paper anticipates market efficiency to have considerably increased in the last decades in favour of the random walk model. The statistical package PCGive will be used as main frame for data mining, unit root testing and cointegration; however, to save time gathering data, Microsoft Excel spreadsheets will aid in performing bulk calculations on the rolling sample coefficients in consideration. Ultimately, this paper lives on a large number of inspiring journal articles, whose aggregated methodology will set a practical guide, but in terms of unit root testing and cointegration, it will refer to standard literature in applied econometrics and financial economics.

Source: Essay UK -

Not what you're looking for?


About this resource

This Economics essay was submitted to us by a student in order to help you with your studies.


No ratings yet!

  • Order a custom essay
  • Download this page
  • Print this page
  • Search again

Word count:

This page has approximately words.



If you use part of this page in your own work, you need to provide a citation, as follows:

Essay UK, econometric analysis. Available from: <> [19-03-18].

More information:

If you are the original author of this content and no longer wish to have it published on our website then please click on the link below to request removal:

Essay and dissertation help