Arnold Harberger, in a seminal 1954 paper, was the first to estimate empirically the deadweight social losses due to imperfect competition. By imposing five key assumptions (linear demand, unit price-elasticity, effective cartel pricing, constant marginal costs, and competitive profits = average profits), he concluded that US welfare losses were insignificant. In the past 15 years, advances in oligopoly theory have produced new algorithms that permit the calculation of economic losses and that relax some Harberger' s restrictive assumptions. Having accurate estimates of interindustry losses is important for the enforcement of competition laws. This paper examines the sensivity of economic losses due to market power across different oligopoly models with a variety of parametric values, data sources, and time periods. The alternative structural models include monopoly pricing, Cournot pricing with homogeneous or heterogeneous products, various forms of price leaderships, and general industry-wide oligopoly pricing. Own-price demand or supply elasticities may be fixed or may vary accross industries. All the estimates examined in this paper utilise US Census Bureau concentration indexes for the food manufacturing for various years between 1967 and 1987. Most estimates rely on Census price-cost margins , while others calculate price ratios from commercial micro-data sets. Finally, most estimates adopt perfect competition as the measurement criterion, whereas others prefer «workable» In view of these many differences in estimation methods, it is not surprising that average economic losses due to market power range widely. As expected from theory, the deadweight welfare losses are higher under monopoly pricing and under Cournot pricing ; average losses are greater when de¬ mand is convex to origin that when it is linear ; and consumer income transfers are about 40 times higher than the deadweight welfare losses. Somewhat surprising are the implausably high welfare losses when price leadership among the top four firms is assumed (equilibrium prices are frequently infinite), but we suspect the losses would be lower if the demand or supply elasticities were higher. We further analyzed the cross-industry correlations of the alternative loss estimates. We find that cardinal estimates vary substantially, but ordinal results very little. Thus, from the point of view of the enforcement of competition laws, the theoretical and empirical progress of the past 15 years has done little to alter the rank of the principal target industries. While the positive correlations among economic-loss estimates is comforting, much work remains to be done to improve estimation techniques and to decide rationally which behavorial models are correct.