Can nominal exchange rates be characterized by deterministic chaos? In order to answer this question, the statistical framework proposed in Paper [ii] was used. The Swedish Krona versus Deutsche Mark, ECU, U.S. Dollar and Yen exchange rate series were examined using daily data from May 17, 1991 to August 31, 1995.
Because the Swedish Krona was pegged against the ECU between May 17, 1991 and November 19, 1992, the exchange rate series were divided into two parts where the first part includes exchange rates from May 17, 1991 to November 19, 1992 and the second part includes exchange rates from November 20, 1992 to August 31, 1995.
This was done to separate the dynamics from the fixed and flexible exchange rate periods. In most cases, the null hypothesis that the exchange rate series cannot be characterized by deterministic chaos was rejected. Therefore, one answer to the question posed may be yes. This conclusion must, however, be treated with caution for several reasons.
First, the results in this paper are not in accordance with the results in Jonsson (1997), who used daily data for the Swedish Krona versus the U.S. Dollar from November 20, 1992 to December 30, 1994. Since Jonsson (1997) used a method proposed by Nychka et al. (1992) that yields consistent estimates of the largest Lyapunov exponent, our results suggest that the method of estimation proposed by Rosenstein et al. (1993) may be upwardly biased.
The rate of convergence for Nychka's et al. (1992) method is not, however, known and the sample size used in Jonsson (1997) is small. Further investigation of these potential sources of difference
is clearly warranted.
Prof. Mikael Bask
Next: Technical analysis in the foreign exchange market
Summary: Index