All for One and One for All?: A Principal Component Analysis of
Latin American Brady Bond Debt from 1994 to 2000
Kevin Paul Scherer and Marco Avellaneda
IJTAF 2001
We use Principal Component Analysis
(PCA) to study the Brady Bond Debt of the four primary Latin American sovereign
issuers: Argentina, Brazil, Mexico and Venezuela. Our dataset covers a period
of 5 1/2 years starting in JUly 1994 and consists of daily
sovereign (``stripped'') yield levels for the Par and Discount debt
securities of each country. We examine the behavior of the characteristic
roots and eigenvectors of the empirical covariance matrices computed sequentially
over different periods. We show that, by and large, there exist two
statistically significant components, or factors, which explain
up to 90% of the realized variance. The eigenvector with the largest eigenvalue
corresponds to the variance attributable to ``regional'' (``Latin'')
risk. The second component strongly suggests the existence of a volatility
risk factor associated to Venezuelan debt in relation to the rest of the
region. A time-dependent factor analysis reveals that the
importance of the variance explained by the factor changes over time
and that this variation can be interpreted to some extent in terms
of market events. In particular, we investigate the relation between
the evolution of the PCA factors with the market dislocations that occurred
during the observation period, including the so-called Tequila effect, Asian
flu, Ruble devaluation and Real devaluation.