An interesting view on the policy intervention and the treatment effects
For many of the DiD studies, authors generally argue that the implementation of the policy is largely exogeneous. The argument for this is that normally the policy is decided at the aggregate level and the policy detail is not targeting individual companies. However, different political groups, in both the east and west, and experts affiliated with different business entities influence the actual details in the policy. To this extent, you actually need all treated units have no influence on the policy. In reality, political connections and market structure will absolutely have an impact on this.
Ok, even if we agree that the policy is exogeneous, people should already anticipate such treatment and thus may alter the behavior of treatment units. For instance, there is a regulation about firms ESG policy, which is a very hot topic in the current literature. This topic will lead to heavy fines if the ESG requirement is not met. In the most idea world, we should compare firms’ R&D on ESG technology between the high and low ESG score firms, before and after the enforcement of the policy. However, in reality, banks and the market, bond and equity, may react to that information before actual enforcement time. They may cut back loans to the low ESG companies and equity invests may short their stocks and reduce the stock prices. So, even before the implementation of the policy, firms may already react to capital market. For low ESG firms with sufficient capital buffer, they may invest in R&D before the actual implementation. For low ESG firms with financial constraints, they do not have such resource for green R&D and perform even worse.
In a standard DiD, the estimator identifies the ATT, comparing treated units to untreated ones. We use the expectation of the control group after the treatment as the estimate of the true expectation of the treatment group as if they are not being treated. This equality holds under the parallel trend assumption that the treated and control groups are indifferent in ALL other aspects.