We compare the empirical performance of a structural and reduced form default risky bond pricing model using Brady bonds from different countries. Using their collateral to estimate recovery rates, enables us to estimate pricing models in this environment with greater precision. Goodness of fit statistics indicate comparable model performance whilst our out of sample tests favour the reduced form model. We also find that allowing credit spreads to depend on riskless term structure factors enhances explanatory power. We test for a common factor driving default probabilities across countries using our reduced form model. We find that this factor is statistically significant.