Evaluating benchmarks for Norwegian exchange rate forecasting

Abstract

In this thesis, we compare the out-of-sample forecasting abilities of three fundamental exchange rate models (EqCM) against the random walk (without drift), RW. The objective of the thesis is to see how well the RW model preforms against fundamental exchange rate models that in the literature have proven to be better at forecasting the Norwegian exchange rate. These models were tested on an out-of-sample period (2009:1-2015:4) that include two characteristic exchange rate regimes. The models estimated are well specified, but as we have found, this does not enhance the forecasting abilities. We find that the RW model is the benchmark of choice for Norwegian exchange rate forecasting. That is, the overall performance of the RW model is significantly better than two out of the three models tested over short-run horizons (1-4 quarters), and better than all the models over long-run horizons (16 and 28 quarters). These results are present even as we use a naive static 1-step ahead forecasting procedure. In the short run, oil prices are found to be a significant determinant for the exchange rate. In the long-run, exchange rates reflects the ratio between domestic and foreign prices. Which is in accordance with the PPP hypothesis.