Why Do Christie and Schultz Infer Collusion from Their Data?

Christie and Schultz (Camp;S) and Christie, Harris and Schultz (CHamp;S) suggest that Nasdaq dealers might tacitly collude to maintain wide bid-ask spreads. Each of the steps by which they reach their conclusions is wrong or misleading. (i) The inference of collusion that widens Nasdaq spreads, based on evidence of relative infrequency of odd-eighth quotes, is ill-founded on reverse logic. Relatively infrequent use of odd quotes is concomitant with wider quoted spreads, even on NYSE/AMEX. Competitive quoted spreads on Nasdaq are expected to be greater than on NYSE/AMEX for economic reasons including the nature of the stocks, quote depth, risk of market-making, and the structure of commissions. Econometrically, spreads are significantly related to stock characteristics, with a modest difference between NYSE/AMEX and Nasdaq. (ii) Contrary to Camp;S, competitive economic factors do explain differences among frequencies of odd-eighth quotes. The lagged dependent variable (i.e., the frequency with which odd-eighths were used in the past) is also statistically significant, implying significant unmeasured economic factors that persistently characterize different stocks. This implies nothing about collusion. We obtained equivalent results for Camp;S's NYSE/AMEX sample. Our standard antitrust/industrial organization analysis shows Camp;S are also wrong to claim that tacit collusion on Nasdaq is plausible. (iii) A decline in spreads with corresponding increased usage of odd-eighth quotes for a few Nasdaq stocks after publicity and regulatory pressure concerning spreads in May 1994 does not imply collusion. Stocks do experience changes in spreads with phases of multiple competitive equilibrium levels of spread and depth. It would be surprising if effective regulatory pressure pushed no stocks from one competitive level of spread to another