Abstract:This paper discusses a practical approach that combines synthetic control with triple difference to address violations of the parallel trends assumption. By transforming triple difference into a DID structure, we can apply synthetic control to a triple-difference framework, enabling more robust estimates when parallel trends are violated across multiple dimensions. The proposed procedure is applied to a real-world dataset to illustrate when and how we should apply this practice, while cautions are presented afterwards. This method contributes to improving causal inference in policy evaluations and offers a valuable tool for researchers dealing with heterogeneous treatment effects across subgroups.
What problem does this paper attempt to address?
This paper attempts to solve the problem of the violation of the parallel trends assumption in the difference - in - differences - in - differences (DDD) framework. Specifically, the author proposes a method that combines synthetic control (SC) with triple difference to deal with the failure of the parallel trends assumption in multiple dimensions (such as treatment status, time, and subgroup characteristics).
### Problem Background
In causal inference, difference - in - differences (DID) and difference - in - differences - in - differences (DDD) are commonly used methods for estimating policy intervention or treatment effects. However, these methods rely on the parallel trends assumption, that is, in the absence of treatment, the expected change trends of the treatment group and the control group are the same. When this assumption does not hold, traditional DID and DDD estimates may be biased.
### Core Problems of the Paper
1. **Failure of the Parallel Trends Assumption**: In complex multi - dimensional situations, the parallel trends assumption may not be satisfied. Especially in the DDD framework, the treatment effects among different subgroups may be heterogeneous, making it more difficult to satisfy the parallel trends assumption.
2. **Limitations of Existing Methods**: Although existing methods (such as propensity score matching, entropy balancing, etc.) can improve the parallel trends assumption in the DID framework, their applications in the DDD framework are relatively few and their effects are limited.
### Solutions
To solve the above problems, the author proposes a new method that combines the synthetic control method with DDD. The specific steps are as follows:
1. **Transform DDD into DID Structure**: By adjusting the outcome variable, the DDD model is simplified into a DID structure, thus making it easier to apply the synthetic control method.
2. **Apply Synthetic Control**: Use the synthetic control method to construct more accurate counterfactual results, especially for those cases where the parallel trends assumption fails. This method can construct a synthetic control group closer to the characteristics of the treatment unit by weighted combination of control units.
3. **Empirical Analysis**: The author uses the data of China's "4 + 7 Cities" Centralized Drug Procurement Pilot Project for empirical analysis, showing how to apply this new method in practical research, and comparing the performance of standard DDD, DID, and synthetic DDD methods.
### Main Contributions
- **Theoretical Innovation**: Expand the application scope of the synthetic control method, enabling it to be applied to more complex DDD frameworks.
- **Empirical Application**: Provide specific empirical cases, demonstrating the application value of the new method in actual policy evaluation.
- **Improve Estimation Robustness**: By improving the counterfactual construction, the robustness of causal inference is improved, especially in cases where treatment effects are heterogeneous.
In short, this paper aims to provide a new tool to improve causal inference by combining synthetic control and triple difference, especially in cases where the parallel trends assumption fails, providing researchers with a more flexible and reliable method.