We extend the statistical model used to estimate abundance from commercial catch-at-age data for many of the major commercial fish species in the world. The model we consider combines commercial catch-at-age data and research survey estimates of fish abundance; we extend the model to allow correlated errors among ages within a year for the survey estimates of fish abundance. We also formulate a method for modeling the fishing mortality on the oldest ages of the fish caught. Estimates are obtained using maximum likelihood. We conclude that the level of correlation among ages is sufficiently large to produce large biases in the standard methods for some stocks. The statistical model that includes correlated errors greatly reduces bias and increases efficiency if the correlation in the estimation error is large.