based on third generation of currency crisis models has motivated various reports from the
IMF on the “architecture” of the international financial system, where the emphasis is on the
importance of sound debt and liquidity management in helping to prevent external crises.
For instance, the IMF report on “Debt- and Reserve-Related Indicators of External Vulner-
ability”, published in 2000 stresses the importance, for Central Banks, of holding foreign
reserves in order to maintaining liquidity and allowing time to absorb shocks in situations
where access to borrowing is curtailed or very costly. It is, therefore, important to monitor
a number of vulnerability indicators (such as the ratio of either the total stock of external
debt to the stock of international reserve or the ratio of the short term external debt to
the stock of foreign reserves) to examine whether they can be considered as accurate lead-
ing indicator of currency crisis, as suggested by the Early Warning System literature (EWS).
Currency turbulence in this paper is proxied by using the Exchange Market Pressure
Index (EMP). This index was first used by Girton and Roper (1977), and subsequently by
a number of authors in the context of exchange rate crises (see Tanner, 2002, for a recent
use). Girton and Roper use a simple monetary model to derive a definition of EMP as
the sum of exchange rate depreciation and reserve outflows, scaled by base money. This
index summarizes the flow of excess supply of money (i.e., the difference between the growth
rates of the domestic component of the monetary base and money demand) in a managed
exchange rate regime, reflected in both exchange rate and reserve movements. Hence an
increase in the value of a country’s EMP indicates that the net demand for that country’s
currency is weakening and hence that the currency may be liable to a speculative attack or
that such an attack is already under way.
There are two main methods used in the EWS literature to predict currency market
turbulence. One method relies on the signal approach proposed by Kaminsky et al. (1998a),
and the other one relies on “parametric” modelling, given that it is based upon limited
dependent variable regression modelling (see the study of Frankel and Rose, 1996, among
the others). The EWS modelling approach we follow in this paper is the “parametric”
one, and it is based upon the investigation of the out-of-sample predictive performance of
a composite leading indicator via regression analysis. In particular, the composite indicator
is constructed by extracting common factors from the Bank for International Settlements,
BIS, dataset which gives detailed information on the composition of external debt. To our
knowledge this dataset has not been previously exploited to construct composite leading in-
dicator of a currency crisis event, defined in terms of the EMP index1 . The choice of using
1The use of annual data by Frenkel and Rose (1996) allows to exploit information on the composition of