Persistence Characteristics of Latin American Financial Markets

Static time series models usually assume stationarity, normality, and independence for the increments of financial rates of return. This paper investigates the empirical characteristics of financial rates of return from Latin American stock and currency markets and documents that their empirical rates of return are non-normal, non- stationary and non-ergodic, and that they exhibit long-term dependence. This paper measures the degree of long-term dependence of these financial time series by calculating their global, or homogeneous, Hurst exponents from their wavelet multiresolution analyses (MRA), i.e. from the wavelet resonance coefficients. Visualizations of these resonance coefficients and their power spectra are provided by scalograms and scalegrams, respectively. These visualizations help to identify the long-term dependence characteristics, which cannot be identified by the classical time series analysis, which is based on the stationarity and independence assumptions. Our findings are consistent with some empirical findings from financial market data in the USA, in Europe and in Asia, but extend their domain of empirical investigation.