AbstractStochastic simulation was used to assess the suitability of alternative dependent variables in international sire evaluation. Three such variables were considered: daughter yield deviations (DYD), national proofs (ETA), and de-regressed national proofs (DPRF). Data from two dairy cattle populations considering 10 generations were simulated. Genetic ties necessary for a joint evaluation of the two populations were established with bull exchange in each generation. De-regressed national proofs were the best and ETA were the worst in estimating genetic difference between the two populations and ranking bulls across country. Further, de-regressed national proofs can be easily computed and, contrary to DYD, are always available. De-regressed national proofs are recommended as dependent variables in across country sire evaluation.