Author:

Keywords:

Effects, Currency, Data, Risk, Time series

Abstract:

Forward rates of European currencies against the private and official ECU exhibit a bias similar to the one found in other data: the Cumby-Obstfeld-Fama (COF) regression coefficients are systematically below unity, and two thirds of them are negative. We use the discount of the private ECU relative to the official ECU as a measure of 'diffidence', a term that may cover both sharply fluctuating risk premia as well as Peso risk. Peso risk, in this context, covers not only fears of realignments but also the risk of a meltdown of the private ECU relative to the official one (a notion that receives some support from a time-series analysis of these data). Dichotomizing the data on the basis of the size of the discount in the private ECU), we find that the COF beta strongly depends on the degree of diffidence and that the negative COF coefficients are generated by typically less than 20 percent of the data. If the diffidence factor contains a risk premium, then this risk premium is definitely not the one predicted by Bansal (1997). Nor is the diffidence factor proxying for Huisman et al. (1997)'s transaction-cost effects. Thus, Peso risk remains as a strong candidate explanation for the forward bias in this sample.