Fractional Cointegration and Price Discovery in FX Markets

Abstract: I employ bivariate fractionally cointegrated vector autoregressive models to analyze price discovery on the EUR/GBP market. Using daily spot rates between 2010 and 2022 along with corresponding one-month and three-month forward rates, I extract parameter estimates for pairwise long-run relationships, each pair containing a spot and a forward. These parameters tell a story both about the equilibrium relation between spots and forwards through the cointegrating vector and about the way that spots and forwards adjust towards equilibrium through the loading matrix, which here comes in the form of a vector. Using permanent- transitory decomposition, the orthogonal vector to the latter is found, which contains the weights with which each market contributes to price discovery. The core finding looking at spots and one-month forwards is that about 55.2% of price discovery occurs in the market for the former, with 44.8% in the market for the latter. Considering spots and three-month forwards, the weight on the forward market increases to 67.8%, leaving the remaining 32.2% to the spot market. I interpret these results as likely indicating that markets for longer-term forwards bring more unique information to the spot market than shorter-term forwards do. This could both be due to differences in what news are important for traders to consider, and due to market participants digesting information differently. I also find evidence of covered interest rate parity violations for three-month forwards.