Currency Composition of Global Foreign Exchange Reserves: Stylized Facts from Multivariate Structural Time Series Modeling

Abstract

The study is an empirical investigation dealing with the stylized fact about the currency composition of foreign exchange rezerves. IMF’s COFER database statistics featuring aggregated data for two groupings of countries, namely advanced economies (33 countries), and Emerging and developing economies (107 countries) are used in the computations.The methodology is based on multivariate structural time series (MSTS) applied to quarterly foreign exchange reserve series involving four main currencies namely, U.S. dollar, Euro, Pound Sterling and Japanese Yen for the 1999:1-2009:4 period. Rather than using the level of these series, their shares in total allocated reserves are calculated and hence used here. Empirical findings for advanced economies group indicate that a strong but negative correlation can be detected between the trend disturbances of UsD and Euro-Pound rezerve shares while a high positive correlation exists between the latter two. Consequently, this imposes a dependence structure in the form of a common trend between rezerve shares in these currencies. The common factor dependence structure changes noticeably for emerging economies group.