(Universidad Icesi, 2013-01-01) Alonso Cifuentes, Julio César; Chaves, Juan Manuel
This paper evaluates the performance of 20 different methods (parametric, and semi-parametric, and nonparametric),
as well as the historical simulation method, to estimate the next-trading-day value-at risk (VaR)
of a representative portfolio for 5 different Latin American countries (Argentina, Brasil, Colombia and Peru).
We found that the non-parametric (i.e. historic simulation), and the semi-parametric methods were the best
way to estimate the risk among the twenty different methods evaluated for all the countries in the sample.
