Asymmetric Price Impacts of Order Flow on Exchange Rate Dynamics

We generalize the portfolio shifts model advanced by Evans and Lyons (2002a; b), and develop the dynamic asymmetric portfolio shifts (DAPS) model by explicitly allowing for possible market under- and overreactions and for asymmetric pricing impacts of order flows. Using the Reuters D2000-1 daily trading data for eight currency markets over a four-month period from 1 May to 31 August 1996, we find strong evidence of a nonlinear cointegrating relationship between exchange rates and (cumulative) order flows: The price impact of negative order flows (selling pressure) is overwhelmingly stronger than that of the positive ones (buying pressure). Through the dynamic multiplier analysis, we find two typical patterns of the price discovery process. The markets following overreactions tend to display a delayed overshooting and a volatile but faster adjustment towards equilibrium whereas the markets following underreactions are generally characterized by a gradual but persistent adjustment. In our model, these heterogeneous adjustment patterns reflect different liquidity provisions associated with different market conditions following under- and overreactions. In addition, the larger is the mispricing, the faster is the overall adjustment speed, a finding consistent with Abreu and Brunnermeier (2002) and Cai et al. (2011). We also find that underreactions are followed mostly by positive feedback trading while overreactions are characterized by delayed overshooting in the short run but corrected by negative feedback trading at longer horizons, the finding is consistent with Barberis et al. (1998) who show that positive short-run autocorrelations (momentum) signal underreaction while negative long-run autocorrelations (reversal) signal overreaction.