Does LKR/AUD exchange rate exhibit long memory? a fractional integration approach

Abstract

This study employs the new science of "econophysics" approach to explain the behavior of exchange rate between Sri Lankan rupees and Australian dollar (LKR/AUD). The study covers the period from January 1, 1990 to December 12, 2017, consisting of 7212 observations. The return of the LKR/AUD is defined as [rt = [ln(ERt) - In(ERt_1)]*100]. ARFIMA(p,d,q)) model and FIGARCH(p,d,q) model are used to examine the presence of fractional integration in the return series. The time domain exact maximum likelihood is used to estimate the ARFIMA model. Volatility of LKRAUD exchange rate return series are proxied by absolute return, squared return and conditional variance derived from FIGARCH model. The autocorrelation of all proxies decay hyperbolically for lags 1 through 200. The results show that return series does not exhibit long memory while conditional variance of return series, absolute return series and squared return series do. The estimate of Long memory parameter for the return series indicates that return series seems to have short memory. However, the squared return series, conditional variance series and absolute return series exhibit long memory as the statistic of memory parameter 'if are statistically significant at 1% significance level and lie within the interval 0 to 0.5. Visual inspection and inferential results reveal strong evidence of long memory property in the volatility of daily LKR/AUD exchange rate return. Shocks to the exchange rate persist over a long period of time. The findings indicate that these markets are not efficient. Hence, the results provide information to the investors, traders and government policy makers to add some risk in their strategies.