Estimating the effects of regulation when treated and control firms compete: a new method with application to the EU ETS

Governments often regulate particular firms differently from other firms in the same industry. For example, they might impose stricter environmental standards in regions with higher pollution, incentivise employment for firms in low-income areas, or offer subsidies to smaller firms. When evaluating the effects of these policies, researchers frequently compare the evolution of outcomes for regulated firms to those of unregulated firms within the same industry, but if firms compete in the output market, then virtually any model of imperfect competition would predict that the effects of industrial regulation spill over from regulated to unregulated firms through output prices. This might serve to obscure the effects of a regulation.

This paper presents a new method for estimating the effects of regulations when treated and control firms compete on the output market. The authors develop a Generalised Method of Moments (GMM) estimator and apply it to the European Emissions Trading System. They find reduced emissions at regulated plants without undermining revenues of regulated firms, relative to a counterfactual scenario in which there were no such regulation.

Key points for decision-makers

• Difference-in-differences estimators are commonly used to evaluate the effects of new environmental regulations, comparing the outcomes of newly regulated firms to others that are exempted. However, when firms compete in output markets, conventional difference-in-differences estimators will fail to identify the effect of a new policy on equilibrium outcomes, such as revenues and emissions.
• The authors show that difference-in-differences estimators cannot in general identify either the sign or the magnitude of treatment effects in this setting.
• They develop a new Generalised Method of Moments (GMM) estimator, which recovers unbiased estimates of treatment effects in simulations, while difference-in-differences estimators and other popular methods do not.
• The authors apply this method to estimate the effects of the EU Emissions Trading System on French manufacturing firms. A key feature of the EU ETS is that it only regulates large emitting installations, which means that regulation typically varies within industries, and even across plants within a firm.
• The estimator shows that the EU ETS led to a 6–9% increase in annual sales by regulated firms but reduced their CO₂ emissions by 5–25%, depending on the year. This is consistent with the Porter hypothesis: by investing in emissions-reducing technologies, regulated firms lowered their emissions and their costs at the same time.
• In aggregate, the paper finds that the EU ETS reduced CO₂ emissions from French manufacturers in the production of goods for the domestic market by 3–16% of observed domestic emissions, or 0.9–4.6 million tonnes annually.
• Critically, the authors find evidence of inter-firm spillover effects of the regulation, which renders conventional estimators unreliable. They find that the difference-in-differences estimator overstates the effect on revenues and understates the effect on emissions, while another popular estimator understates the effect on revenues and overstates the effect on emissions.