New statistic for financial return distributions: power-law or exponential?