*4.2. Robustness Test*

4.2.1. Parallel Trend Test and Dynamic Effect

To ensure the effectiveness of the DID model, it is essential to prove that the common trend assumption is satisfied between the treatment and control groups. To test whether the parallel hypothesis was satisfied, we replace the dummy variable in Equation (1) with one corresponding to several years before and after implementing the CET policy [12,61,62]. The regression model is estimated as follows:

$$Ln(MV)\_{it} = \beta\_0 + \sum\_{k=-13}^{6} \beta\_k (treat\_i \times time\_l)^k + \beta\_2 \times X\_{it} + \mu\_i + \delta\_r + \gamma\_t + \varepsilon\_{it} \tag{2}$$

where (*treati* × *timet*) *<sup>k</sup>* is 1 when in the *k*-th year before the implementation of the pilot CET policy (*k* < 0) or in the *k*-th year after the implementation of the pilot CET policy (*k* > 0) for treatment groups, and 0 otherwise. Since the number of periods before 2013 is very large in our sample, the effects of the 9 to 13 years before 2013 are combined in a single group. If the coefficients of *did<sup>k</sup> irt* are insignificant before 2013 and significant after 2013, then the parallel trend hypothesis is satisfied. Figure 2 shows the test results of the parallel trend hypothesis and the dynamic trend of the pilot CET policy and companies' market value. The results indicate no significant difference between the treatment and control groups before the policy is implemented, while it implies a significant increase in companies' market value after implementing the pilot CET policy. The results confirm the parallel trend hypothesis.

**Figure 2.** The test of parallel trend hypothesis and the dynamic trend of the pilot CET policy and companies' market value.
