How can climate adaptation policy enhance urban climate resilience? Empirical evidence from China

PLOS One | https://doi.org/10.1371/journal.pone.0335736 November 6, 2025 12 / 21 to 0, and except for 2016, the estimated policy effect was not statistically significant. The test results indicate that, prior to the implementation of the CACCP policy, there was no significant change in the gap between the experimental group and the synthetic control group in the outcome variable; that is, the parallel trends assumption was met, confirming the high validity of the Synthetic DID results in this study. 4.2.2. Instrumental variable test. We employed the instrumental variables method to address potential endogeneity issues in evaluating the CACCP policy. We selected the number of extreme rainfall days lagged by five years (L5_ERD ) as the instrumental variable for this study, primarily for the following reasons: first, policymakers may prioritize cities with historically higher disaster risks (such as extreme rainfall) as pilot sites for climate adaptation initiatives; Policy makers may also tend to select cities with relatively better infrastructure, moderate risks, and greater potential for policy demonstration effects as pilot projects for climate adaptation construction. Therefore, L5_ERD meets the relevance condition (the instrumental variable may positively or negatively correlate with the policy). Second, the direct impact of historical disaster risks on current urban climate resilience has diminished over time and is now negligible. Therefore, L5_ERD satisfies the exogeneity condition. This study utilized the data on extreme rainfall days in Chinese cities published

by Guo, Ji, and Zhang (2024) [68]. After standardizing the data from 2005 to 2017, an instrumental variable method was employed for estimation, with the results presented in Table 4. The results indicate that the instrumental variable L5_ERD passed the relevant tests, and the CACCP policy promoted climate resilience in the pilot cities. This result is consistent with the conclusions of this study, indicating that there are no serious endogeneity issues, and the results are robust. 4.2.3. Mixed placebo test. To enhance the credibility of the policy effect estimates in this study, we used a spatiotemporal mixed placebo method to test them. We simulated 500 random shocks and randomly generated pseudo- experimental groups and pseudo-policy treatment times, ultimately obtaining 500 estimates of the policy treatment. The kernel density curves of these estimates are shown in Fig 5. As illustrated in Fig 5, the estimates of policy effects derived from stochastic simulations exhibit a normal distribution. The coefficient estimates are predominantly distributed around 0, exhibiting a considerably smaller value than the estimates of the actual policy effect (the red vertical line indicates Table 3. The average treatment effect of CACCP policy. Resilience ATT Std. Err. t P > |t| [95% Conf. Interval] treated 0.0118** 0.0048 2.48 0.013 0.0025 0.0212 Note: ** indicates a 5% statistical significance level. https://doi.org/10.1371/journal.pone.0335736.t003 Fig 4. Parallel trend test. https://doi.org/10.1371/journal.pone.0335736.g004