An empirical analysis of the effect of key policy rate changes on the unbiasedness hypothesis
MetadataShow full item record
- Master Thesis 
Interest rate parity is one of the most important theory in international finance which determines the relationship between the exchange rate and interest rates of two countries. However, there are many empirical findings showing that interest rate parity condition does not hold. A phenomenon called the forward premium puzzle commonly exists in the currency markets. It refers to the situation that the high-interest country’s currency tends to appreciate with respect to the low-interest country’s currency which contradicts the interest rate parity theory. Intrigued by this fascinating puzzle, we want to explore the causes of this puzzle, especially how the central bank announcements might affect the foreign exchange market. This thesis focuses on uncovering whether a country's central bank announcements of changes in the key policy rate would affect how the unbiasedness hypothesis holds. Using daily observations of spot exchange rates and 1-month and 3-month forward exchange rates enables us to remove the days of change from the dataset. This thesis distinguishes itself from the literature by its research question and methodology. Empirical literature usually find evidence against the unbiasedness hypothesis. This thesis on the other hand, finds some evidence supporting that unbiasedness hypothesis holds for the six currency pairs: CAD/USD, EUR/USD, JPY/USD, NOK/USD, CHF/USD and GBP/USD in the period 01/01/2002 to 19/10/2018. In order to solve econometric issues like serial correlation incurred by using daily observations of regression variables, necessary adjustments are made to correct for them. By comparing the regression results of different models, we can see the effect of removing days of key policy rate changes on the unbiasedness hypothesis. The main results from the empirical analysis show that removing the days surrounding changes in the key policy rate does not affect the regression results notably, except in the case of removing seven days for the CHF/USD.