A new approach to exchange rate forecast: The role of global financial cycle and time‐varying parameters

Abstract

The exchange rate disconnect puzzle argues that macroeconomic fundamentals are not able to accurately predict exchange rate. Recent studies have shown that the puzzle could be upturned if: (i) the dataset is structured in a panel form; (ii) the model is based on the portfolio balance theory (PBT); (iii) factor models are employed and (iv) time-varying parameter (TVP) regression is used. This study combines these strands of the literature. Essentially, the study conjectures that Global Financial Cycle (GFCy), drawing inspiration from PBT, has some predictive information content on exchange rate. Using dataset for 25 countries, we produced some mixed results. On the whole, the GFCy is able to produce lower forecast error, as compared to the benchmark model. However, its effectiveness is dependent upon the regression type (TVP vs Panel Fixed Effect); forecast horizons (short vs long); the sample period (early vs. late) and measures of GFCy. The results are robust to a number of checks.