We begin with a retrospective analysis of three major U.S. airline mergers and document the sensitivity of the findings, particularly questioning whether market conditions evolve similarly for treated and control markets. We then develop a structural model that clarifies this and other assumptions implicit in retrospective analyses and separates efficiency gains from increases in firms' conduct. Using only pre-merger data, we propose a reduced-form approach that leverages exogenous changes in market structure to forecast merger effects. Finally, we use structural prospective merger simulations with our other estimates for a comprehensive evaluation. This bridging of approaches uncovers a fundamental tension: either efficiency gains were limited or, if they were significant, they were accompanied and offset by coordinated effects.