Although numerous efforts have been made to resolve the “Meese and Rogoff Puzzle”, the prevailing view still holds that no model consistently outperforms the random walk across all countries and time periods. However, a growing body of literature ...
Although numerous efforts have been made to resolve the “Meese and Rogoff Puzzle”, the prevailing view still holds that no model consistently outperforms the random walk across all countries and time periods. However, a growing body of literature suggests that under certain constraints, such as specific forecast horizons, and particular countries, certain models can outperform the random walk (RW) benchmark. Contributing to this optimistic strand, this study demonstrates that long-horizon nominal exchange rate forecasts based on the tradables-based real exchange rate (TRER), as well as very short-horizon forecasts in commodity-exporting countries based on a commodity price can outperform the RW model.
The first part of this study demonstrates that long-horizon nominal exchange rate forecasts based on the TRER can outperform the RW benchmark in countries with floating exchange rate regimes and stable prices. The TRER is constructed using producer price indices (PPI), which include a higher share of tradable goods compared to the conventional real exchange rate (RER) based on consumer price indices (CPI). This study shows that past values of the TRER are strongly negatively correlated with long-horizon changes in the nominal exchange rate, but not with the relative price index. Building on these in-sample regularities, this study uses the TRER to forecast the NER and find that it outperforms the random walk model in long-horizon forecasts. In addition, this study finds that nominal exchange rate forecasts based on TRER mostly performed better than those based on the conventional RER in long-horizon.
The second part of the study shows that an exchange rate forecasting model incorporating palm-oil prices can effectively forecast short-term exchange rates for the local currencies of major palm oil-exporting countries, particularly those of Indonesia and Malaysia. This model demonstrates superior performance compared to the random walk model in ex-ante out-of-sample forecasts for those countries at both daily and monthly frequencies. This study also finds that natural gas prices, another major export commodity for both countries, are not as effective as palm oil prices in predicting exchange rates.