How frequent a BEER? Assessing the impact of data frequency on real exchange rate misalignment estimation

This paper explores the robustness of Behavioural Equilibrium Exchange Rate (BEER) models – employed to estimate real effective exchange rate (REER) deviations from “equilibrium” values consistent with macroeconomic fundamentals – to the frequency (annual vs. quarterly) of the underlying data. Indeed, data frequency influences both the length of the sample period (which is typically shorter in a quarterly model) and the set of relevant fundamentals to be included in the specification, and can affect the plausibility of some of the BEER modelling assumptions, which are especially restrictive at the quarterly frequency. The paper compares REER misalignment estimates stemming from a carefully specified annual model, estimated since 1980 for 55 countries, and a comparable quarterly model, estimated since 1999, which is a variant of that currently in use at the Bank of Italy (Giordano, 2018). In the overlapping period the annualised quarterly-model misalignments are quite similar to those based on the annual model. Moreover, the in-sample power of quarterly REER misalignments in explaining subsequent, actual REER developments is found to be higher than that of the annual estimates, signalling their greater usefulness in assessing a country’s external economic outlook. This paper therefore confirms the robustness of the quarterly BEER model currently employed at the Bank of Italy; moreover, it suggests that the “optimal” frequency of a BEER model depends on the use (research vs. monitoring and policy-making) one makes of the resulting measures.