We document that cross-sectional FX correlation disparity is countercyclical, as exchange rate pairs with high average correlation become more correlated in bad times whereas pairs with low average correlation become less correlated. We show that currencies that perform badly (well) during periods of high cross-sectional disparity in conditional FX correlation yield high (low) average excess returns, suggesting that correlation risk is priced in currency markets. Furthermore, we find a negative cross-sectional relationship between average FX correlations and average FX correlation risk premia. Finally, we propose a no-arbitrage model that can match salient properties of FX correlations and correlation risk premia.
Systemic Risk Centre Discussion Papers DP 26