Abstract
The financial rates of return from Middle East and North African markets are found to be nonnormal, nonstationary and long-range dependent, i.e. they have long memory. The degree of long-term dependence is measured by Hurst exponents using local Whittle method which is a semi-parametric method that presents robustness to data seasonality and short-range dependence. Our long-term results are consistent with similar empirical findings from American, European and Asian financial markets. Therefore, the article extends the domain of the empirical investigation of the dynamics characteristics of the global financial markets and disproves the hypothesis of perfectly efficient financial markets.
Notes
1 Mandelbort (Citation1969, Citation1972) was the first author who introduced the concept of such long-term persistence.
2 Long memory have been also been also examined in returns of other assets. For example, Cheung (Citation1993), Pan et al. (Citation1996), Barkoulas et al. (Citation2004), Karuppiah and Los (Citation2005) and Jin et al. (Citation2006) found evidence of long memory in currency return series.
3 See Baillie (Citation1996) and Robinson (Citation2003) for some recent, extensive, literature surveys.
4 For recent evidence about long memory in emerging markets see for example, Henry (Citation2002), Kilic (Citation2004), Cajueiro and Tabak (Citation2005) and Kyaw et al. (Citation2006).
5 Recently, Hurvich and Ray (Citation2005) and Davidson and Sibbertsen (Citation2005) argue that Local Whittle estimator is much less biased and the finite-sample SEs and yields more accurate confidence intervals than the widely-used GPH estimator.
6 There is little theoretical justification to prefer one volatility measure over any of the others. In this article we follow Bollerslev and Wiright (Citation2000) in using the three volatility measures.
7 Excess Kurtosis in equity returns has been well-documented in emerging markets by a number of other studies including Bekaret and Harvey (Citation1997
8 This test is a nonparametric statistical test proposed to deal with nuisance parameters for a class of short memory linear processes which may be observed or be the disturbances of a deterministic linear regression model. According to Harris et al. (Citation2006), this statistic has superior empirical size and power properties to the KPSS test for a short memory null hypothesis against an I(1) alternative.