Capturing asymmetry in real exchange rate with quantile autoregression

Quantile Autoregression (QAR) is used to explore asymmetries in the adjustment process of pairwise Real Exchange Rate (RER) between the Italian lire, French franc, Deutsch mark and the British pound. Based on the best specification for each quantile we construct predicted conditional density functions, which guided us to identify two sources of asymmetry: (1) dispersion depends on the conditioned value of the RER, i.e. 'conditional' heteroskedasticity; (2) the probability of increases and falls also changes according to the conditioned value, i.e. there is higher probability for the RER to appreciate (depreciate) given the currency is depreciated (appreciated). We only verified strong heterokedasticity in relations among the lire, franc and mark, which was resolved by estimating quadratic autoregressive model for some quantiles. Relations involving the pound presented stable but higher dispersion indicating larger probability of wider oscillation.