Principal component measures of exchange market pressure: comparisons with variance-weighted measures

Abstract

In studies of currency crises, Exchange Market Pressure (EMP) captures depreciations and central-bank interventions in a single index. However, while the measure's three components are commonly given variance-smoothing weights, this approach has been criticized as problematic. One proposed alternative is to use Principal Components Analysis (PCA) to derive the proper weights. This article examines EMP for 21 countries over the period from 2001 to 2012. While the first principal component never produces weights of the correct sign, some countries’ second (and sometimes third) principal components can be used. We then compare the two measures, finding that the PCA-based measure is highly correlated with the variance-smoothing EMP measure, but that its values are often less extreme. This allows for different definitions of ‘crisis’ periods between the two, and different results in econometric estimations of EMP determinants. As a result, we find that the method of calculating EMP does indeed affect empirical analyses.

Keywords:

• exchange market pressure
• principal components
• time series
• bounds testing

JEL Classification:

• F31
• C10

Notes

1 For more on this commonly used approach, see Tsay (2005), Chapter 9 (p. 421).

2 Angola, Bolivia, Uganda and Ukraine have larger variance in the interest-rate component than in the reserve component.

3 Since EMP is a weighted sum of percentages, we feel that no other transformation is necessary.