The genetic gains were estimated for milk production and persistency, derived from random regression models, using eigenvector indices, and they were compared with the traditional selection index. The data set contained 4971 test day milk yield recorded for 691 buffalo cows, daughters of 120 sires and 532 dams. The model included the random effects of direct additive genetic, permanent environment and error, whereas the fixed effects were herd test day, year and season of calving and parity, and as a covariate, it was milk days. The first and the 2nd eigenvalues explained 73.1 and 22.9% of the variation of the random regression coefficients, respectively, suggesting that the use of the first two eigenvectors is sufficient. Genetic responses in total milk yield (TMY) based on the first eigenvector index (Ie1) and that based on the conventional selection (IMY) have close gain of about 171 kg in each index. The second eigenvector index (Ie2) showed an increase in TMY (9.91 kg), and thus an increase in the persistency (0.86 kg). The TMY and persistency are the two economically important traits in dairy production, additional genetic gains in persistency and high genetic gain for TMY could be obtained using the 2nd eigenvector index (I*2).