A refutation of "common ownership does not have anti-competitive effects in the airline industry"

We show that the main claim in Dennis, Gerardi, and Schenone (JF forthcoming) (DGS), namely "that the documented positive correlation between common ownership and ticket prices stems from the market share component of the common ownership measure, and not the ownership and control components," is factually incorrect. In particular, we show empirically that the placebo that according to DGS "keeps market shares fixed" is in fact highly negatively correlated with market shares. This correlation is mechanical and arises because the data set is an unbalanced panel, as we show analytically. We make a methodological contribution to the literature by showing how one can actually separate variation from market shares from variation in ownership. Contrary to DGS' claims, ownership changes do predict price changes once one constructs a valid placebo that actually separates the variation from market shares from the variation in ownership. AST's panel regressions in fact underestimated the price effect of common ownership, due to the endogeneity of market shares.