*5.2. Data Validity Analysis and Robustness Check*

First, we tested whether the mean values of the dependent variable industrial energy intensity between the treatment group and the control group were equal after BTH coordinated development strategy. The null hypothesis of the mean test is that there is no significant difference. The results show that the *p*-value equals zero from the base period, which indicates that there is a significant difference in industrial energy intensity between the treatment group and the control group.

Second, this paper conducted a further parallel trend check before the BTH coordinated development strategy. The DID model does not require the mean values to be the same but hypothesizes that the trends between the control group and the treatment group must be the same before policy implementation. To support this assumption, we set up a year dummy variable representing different years, and the cross term *BTH*·*Year* represents the possibly different variation trend of energy intensity in the treatment group compared with the control group. Figure 2 indicates that the trends of these two groups do not have significant differences in all three years before implementation of the BTH coordinated development strategy. The coefficients of the dummy variables are not statistically significant in the three years before implementation of the BTH coordinated development strategy. Furthermore, from the year of implementation of the BTH coordinated development strategy, the impact of the policy on industrial energy intensity was significantly negative and gradually decreasing with the passage of time. In the three years after implementation of the BTH coordinated development strategy, the industrial energy intensity decreased by 4.25% in the year of implementation and decreased by 6.37% three years after implementation.

**Figure 2.** Parallel trend test.

Third, even if the variation trends are the same as before the policy, we still need to test other policies that may lead to different trends between the two groups. Therefore, this paper performs the placebo test by setting the policy event in a period prior to 2014 to see whether there is still a significantly negative effect. As mentioned in the previous analysis, the premise of the DID method is that there is no significantly different trend in industrial energy intensity before implementation of the policy. If the policy event is set in a period before 2014, the estimated coefficient of the core variable may be not significant. If the results are contrary to our expectations, such as significantly negative, it means that there are some potential unobserved policy factors other than the BTH coordinated development strategy that could affect industrial energy intensity in this region. To ensure the robustness of empirical results, the policy impacting years are set as 2008, 2009, 2010, 2011, 2012, and 2013 and the years after them. The corresponding estimated results are reported in Table 3. It shows that the estimated coefficients of variable *BTH*·*Post* are all insignificant, which proves that the DID results are robust again.



Source: Authors' estimation. Notes: *p*-values are in brackets.
