There is strong evidence that interest rates and bond yield movements exhibit both stochastic volatility and unanticipated jumps. The presence of frequent jumps makes it natural to ask whether there is a premium for jump risk embedded in observed bond yields. This paper identifies a class of jump-diffusion models that are successful in approximating the term structure of interest rates of emerging markets. The parameters of the term structure of interest rates are reconciled with the associated bond yields by estimating the volatility and jump risk premia in highly volatile markets. Using the simulated method of moments (SMM), results suggest that all variants of models which do not take into account stochastic volatility and unanticipated jumps cannot generate the non-normalities consistent with the observed interest rates. Jumps occur (8,10) times a year in Argentina and Brazil, respectively. The size and variance of these jumps is also of statistical significance.