Evaluation of Accuracy of Exchange Rate Expectation Models for Understanding Observed Expectations

Exchange rate expectations are a crucial element in the main monetary models. Therefore, this paper analyses the mechanism behind their formation. To achieve this, we analyse traditional expectation models using data from the Survey of Professional Forecasters (SPF) for the CZK/EUR currency pair. The data used cover one-year expectations in the period from January 2001 to December 2022, which are provided monthly by the Czech National Bank (CNB). The paper demonstrates the poor performance of the perfect expectation model. Furthermore, it demonstrates that traditional models, such as static, extrapolative, regressive and adaptive expectations, exhibit some explanatory power but lack robustness. The only traditional model that exhibits robustness is the model based on the UIP puzzle, which also outperforms all other traditional models when evaluated using error metrics. Based on these observations, the paper introduces a non-traditional model in which agents simply shift the current spot value by a constant into the future. This model displays robustness and outperforms the others.