A counterfactual is the scenario that would have existed in the absence of an event happening. For example, if we are trying to understand the impact of government policy related to foreign direct investment (FDI), the counterfactual would be the level of FDI that would have existed if the policy was not implemented. The counterfactual is also referred to as the ‘but-for’ world, i.e., what the world would have looked like but-for a particular event.
In the context of competition law, counterfactuals are used to understand the impact of actions, i.e., anticompetitive conduct and combinations (mergers, amalgamations, and acquisitions), on parameters including competition, customer choice, prices, and supply of products and services. For instance, in the case of anticompetitive conduct (such as exclusionary conduct) by dominant undertakings, the European Communities Commission’s Guidance Paper states, “This assessment will usually be made by comparing the actual or likely future situation in the relevant market (with the dominant undertaking’s conduct in place) with an appropriate counterfactual, such as the simple absence of the conduct in question or with another realistic alternative scenario, having regard to established business practices.”1 In terms of assessing the competition impact of mergers, according to the U.K. CAT, “the CMA will compare the prospects for competition with the merger against the competitive situation without the merger – i.e., in the counterfactual,”2
To estimate the effects or likely effects of such actions, one must compare observed actual outcomes to estimates of counterfactual outcomes, i.e., what would have happened but-for the action. Properly estimated counterfactuals allow the researcher to assess whether antitrust harm occurred or is likely to occur, and the extent of loss to consumers, suppliers, or competitors, if any. According to a U.S. court ruling, “an antitrust plaintiff's damages should reflect the difference between its performance in a hypothetical market free of all antitrust violations and its actual performance in the market infected by the anticompetitive conduct.”3
A variety of approaches exist for reconstructing the counterfactual but-for world. They vary in the level of sophistication and the plausibility of their assumptions. The methods available can be divided broadly into two categories, “yardstick” and “before and after” techniques.
Yardstick approaches compare at-issue outcomes to other outcomes during the same time period that are assumed to be unimpacted by the anticompetitive conduct or combination. Such untainted outcomes typically come from related industries in the same geographic market or the same industry in different geographic markets.4 A yardstick approach can be simple or advanced. In the case of a simple yardstick, we would assume that the counterfactual or but-for outcome - price, rate of return or other parameters of interest - in the market affected by the action would be same as those in the unaffected comparator market. A more advanced yardstick approach is the regression-based difference-in-differences (DiD) technique. DiD measures the effect of the action by comparing the change over time of an outcome in the market affected by the action to the change over time in the unaffected comparator market. The validity of DiD rests on the “parallel trends” assumption, which states that the outcomes in both groups would change by the same amount over the same time period in the but-for world. Under this assumption, any difference in the changes over time between the two groups is attributable to the action.5
Before and after approaches compare outcomes in the at-issue market during the period that is affected by the action to outcomes in the same market during a competitive benchmark period that is unaffected by the action. The “dummy-variable” model and forecast model are two commonly used techniques to undertake a before and after analysis. The statistical technique of regression analysis is used to estimate both of these models.
Under the dummy-variable model, the relationship between the outcome variable, say price of a product or service, and explanatory factors that affect price, such as input costs, price of substitutes, price of complements, income of consumers, etc. is specified along with a variable (dummy variable) that takes the value 0 during the competitive benchmark period and the value 1 during period when the action was taken. The relationships between the outcome variable and the explanatory factors are estimated using real-world data across the competitive benchmark and the period affected by the action using regression techniques. The estimated relationship between the outcome variable and the dummy variable measures the effect of the action on the outcome variable, after controlling for the effects of the other explanatory factors on the outcome variable. In other words, the estimated effect of the dummy variable can be interpreted as the difference between actual and counterfactual outcomes.
An alternative to the dummy variable regression is the forecasting model. Whereas the dummy-variable model is estimated using data from both the competitive benchmark and action periods, a forecasting model is estimated using only data from the competitive benchmark period. Based on the estimated relationships between the outcome variable and the explanatory variables, the values of the outcome variable during the anticompetitive action period are forecast. These forecast values of the outcome variable represent estimates of the counterfactual outcomes. The difference between the forecasted counterfactual outcome and the actual value of the outcome during the anticompetitive period represents the effect of the anticompetitive action.6
Guidance on the Commission’s Enforcement Priorities in Applying Article 82 of the EC Treaty to Abusive Exclusionary Conduct by Dominant Undertakings, Guidance Paper. Commission of the European Communities, [2008], Paragraph 20, page 10.↩︎
Spreadex Ltd v Competition and Markets Authority [2026] CAT 24 [25], 12.↩︎
National Farmers’ Organization Inc v Associated Milk Producers Inc, 850 F 2d 1286 (8th Cir 1988) 1306.↩︎
Daniel L Rubinfeld, ‘Antitrust Damages’ in Einer R Elhauge (ed), Research Handbook on the Economics of Antitrust Law (Edward Elgar Publishing 2012) 377, 381.↩︎
National Academy of Sciences, National Academy of Engineering and Institute of Medicine, Reference Manual on Scientific Evidence (4th edn, National Academies Press 2011) 624–626, 656–658.↩︎
Justin McCrary and Daniel L Rubinfeld, ‘Measuring Benchmark Damages in Antitrust Litigation’ (2014) 3 Journal of Econometric Methods 63.↩︎
