We agree that there are possible biases in our sampling procedure. Any calculations based on imperfect historical data and limited samples are subject to error. The job of constructive research is not simply to cry 'bias', but to seek to minimize it and to estimate its likely effects. This we have done. Professor Prais offers no quantitative estimate of the significance of the issues to which he refers. He correctly points out that we acknowledge that our sampling procedure may have led to some upward bias. He fails to mention that there is also an opposite source of bias, in that concentra- tion increase resulting from mergers we have failed to detect is attributed to internal growth....