Empirical literature and related legal practice using concentration as a proxy for competition measurement are prone to a fallacy of division, as concentration measures are appropriate for perfect competition and perfect collusion but not intermediate levels of competition. Extending the classic Cournot model used to derive the Hirschman–Herfindahl Index (HHI) of market concentration, we propose an adaptation of this model intended to remove the aggregation bias and omitted variable bias that exist in many empirical studies. Application of our model and new critical mass measures to data for US commercial banks in the period 1984–2013 confirms that concentration measures are indeed unreliable competition metrics. Our results lead us to conclude that critical mass builds upon HHI as a market power metric that incorporates competitive information not contained in HHI. Policy and future research implications are briefly discussed.