RRR 9-2003, March, 2003:
A NON-RECURSIVE REGRESSION MODEL FOR COUNTRY RISK RATING

S. Alexe	
RUTCOR - Rutgers University Center for Operations Research, Piscataway, NJ, USA, 
Email: salexe@rutcor.rutgres.edu 

P L. Hammer
RUTCOR - Rutgers University Center for Operations Research, Piscataway, NJ, USA, 
Email: hammer@rutcor.rutgers.edu

A.Kogan
RUTCOR- Rutgers University Center for Operations Research, Piscataway, NJ, USA, & 
Rutgers Business School, Rutgers University, 180 University Avenue, Newark, NJ, USA, 
Email: kogan@rutcor.rutgers.edu

M.A. Lejeune
Rutgers Business School, Rutgers University, 180 University Avenue, Newark, NJ, USA, 
Email: mlejeune@andromeda.rutgers.edu

Abstract. The central objective of this paper is to develop a rating system equivalent to and explanatory of the results provided by Standard & Poor's country risk ratings. An important requirement imposed on the rating model constructed here is that of non-reliance on lagged ratings, i.e. the exclusion of the lagged country risk ratings from the set of independent variables. We use the 1998 Standard & Poor country risk ratings to develop a non-recursive multiple regression model for the ratings considered as the dependent variable, regressed on a set of economic and political variables. We use the k-folding cross-validation technique to evaluate the accuracy of linear regression predictions. The stability of the constructed non-recursive regression model is evaluated in three ways. First, we show that it correlates well not only with the ratings of Standard & Poor, but also with those of other agencies (Moody's and The Institutional Investor). Second, we show its temporal stability by applying the non-recursive multiple regression model derived from the 1998 dataset to the 1999 data. Third, we show that the proposed model can successfully predict the ratings of several previously non-rated countries.