An empirical analysis of the unbiasedness hypothesis

Abstract

This thesis has two main aims, split into two parts. The aim of the first part is to see if the unbiasedness hypothesis holds for some of the world’s most liquid currency pairs. The objective is to gather new data on spot and forward rates with three different maturity lengths from 1996 up until 2018, and test these data using mean comparison t-tests and time series regression analyses. The results reveal that the unbiasedness hypothesis does not seem to hold for most currency pairs for the one month, one week or overnight maturities. There are also some evidence indicating the presence of the forward premium puzzle for some currency pairs, especially in the monthly and overnight maturities. The aim of the second part is to uncover potential statistical relationships between the deviation from the unbiasedness hypothesis and leading explanatory variables. In extension, these relationships could be used to predict future deviations from the unbiasedness hypothesis and thus increase excess return from carry trade. The objective for this part of the thesis is to collect data on various relevant economic variables and test the explanatory power of these variables, using regression analysis and a direction of change-model. Most notably, we find that the Baltic Dry Index has served as a positive, leading indicator of deviations from the unbiasedness hypothesis in the period of our analysis. We also find evidence of a relationship for the CBOE Volatility Index which varies with the length of maturity, and a negative relationship for the S&P 500 index. However, we do not find any relationship between the deviation from the unbiasedness hypothesis and the USD denominated LIBOR, the Brent crude oil price or the bid-ask spread for the foreign exchange spot.