The well-publicized Christie-Schultz collusion hypothesis provides an experiment for studying the determinants of market structure in Nasdaq markets. Some markets experienced substantial com- pression in the pro.t margins for market makers due to the change of quoting convention from odd-eighth avoidance to the use of the full spectrum of eighths. Contrary to what competitive theory predicts, the empirical results suggest that this change led to net entry of market makers, after controlling for a time .xed e?ect, trading activity, information aspects of trading, market size, volatility, and unobserved individual market e?ects. Moreover, the robustness and signi.- cance of this .nding do not change as di?erent estimation methods are employed to correct for possible self-selection bias of the estimated average treatment e?ect. Surprisingly, dealer .rms entered these markets despite the compression of pro.t margins. An explanation is provided based on collusion and investment in entry deterrence related to the practice of .preferencing.