A novel approach to the estimation of an actively managed component of foreign exchange reserves

This paper offers a novel approach to the estimation of an active component of reserves making use of a time-varying coefficient model estimated with Bayesian techniques. The approach substantially extends the time and country coverage of estimates beyond that available under the existing approach. We find that the estimates of an active component for 20 countries over 1995–2017 period are highly correlated with those obtained from the existing approach. The new estimates are cross-checked against the available data on FX market interventions in Argentina, Chile, the Czech Republic, Mexico, Russia and Turkey. We demonstrate that these estimates are a better proxy of FX interventions than simple changes in reserves and are at least as good as the estimates from the existing approach. Our novel approach contributes to a better understanding of changes in FX reserves in many countries, including large reserve holders for which the relevant data are hardly available.