1. Introduction
Sovereign credit ratings are a condensed assessment of a government’s ability and
willingness to repay its public debt both in principal and in interests on time. In this,
they are forward-looking qualitative measures of the probability of default put forward
by rating agencies. Naturally, one should try to understand the determinants of credit
ratings, given their relevance for international financial markets, economic agents and
governments. Indeed, sovereign credit ratings are important in three ways. First,
sovereign ratings are a key determinant of the interest rates a country faces in the
international financial market and therefore of its borrowing costs. Second, the
sovereign rating may have a constraining impact on the ratings assigned to domestic
banks or companies. Third, some institutional investors have lower bounds for the risk
they can assume in their investments and they will choose their bond portfolio
composition taking into account the credit risk perceived via the rating notations. For
instance, the European Central Bank when conducting open market operations can only
take as collateral bonds that have at least a single A attributed by at least one of the
major rating agencies.
In this paper we perform an empirical analysis of foreign currency sovereign debt
ratings, using rating data from the three main international rating agencies: Fitch
Ratings, Moody’s, and Standard & Poor’s. We have compiled a comprehensive data set
on sovereign debt ratings, macroeconomic data, and qualitative variables for a wide
range of countries starting in 1990. Regarding the empirical modelling strategy, we
follow the two main strands in the literature. We make use of linear regression methods
on a linear transformation of the ratings and we also estimate our specifications under
an ordered probit response framework.
Our main contribution to the existing literature is the innovation of the estimation
method used and the functional form specification, and the large dataset employed.
Under the linear framework, we argue that random effects estimation will be inadequate
due to the correlation between the country specific error and the regressors, but also that
its alternative, fixed effects estimation will not be very informative. We salvage the
random effects approach by means of modelling the country specific error, which in
practical terms implies adding time-averages of the explanatory variables as additional