Asymmetries in the conditional mean and conditional variance in the exchange rate: evidence from within and across economic blocks

The paper tests the hypothesis that both the conditional mean and the conditional variance of exchange rates are asymmetric functions of past information. This hypothesis is tested by estimating an Asymmetric Threshold GARCH model for fifteen currencies. The empirical evidence suggests that both the conditional mean and the conditional variance respond asymmetrically to past information, with an AR(1) structure within blocks and an ARMA(1,1) structure for the EU currencies against the dollar. It is found that the conditional mean is an asymmetric function of past innovations, rising proportionately more during appreciation periods within and across blocks. This implies that, on average, the market incorporates positive news (depreciations) more quickly than negative news (appreciations). The conditional variance is an asymmetric function of past innovations as well, rising proportionately more during depreciations within blocks and appreciation periods across blocks. Furthermore, asymmetries in the conditional mean are linked to asymmetries in the conditional variance because the more rapid adjustment of the market to depreciations causes greater volatility during these periods. This, in turn, causes within blocks a slower speed of adjustment in the variance to devaluations than to appreciations. Finally, greater efficiency in currency markets is found within blocks than across blocks.