Our initial sample consists of China’s A-share listed firms in the Shanghai and Shenzhen stock exchanges from 2010 to 2017. We selected observations in the light of the following criteria: 1) excluding observations in the banking, insurance, and other financial industries; 2) excluding observations with transaction statuses of special treatment (ST), suspension from trading (*ST) or particular transfer (PT); 3) excluding observations with only one-year data; and 4) excluding observations with missing and unavailable data. Finally, we obtained the unbalanced panel data of 13,119 observations as the final sample.

The financial data related to the listed firms come from the China Stock Market and Accounting Research (CSMAR) database and Wind database. City-level data came from the China City Statistical Yearbook. Corporate social responsibility data is an evaluation index sourced from the Hexun Database ¹ . This index started in 2010, and the civilized city pilot was implemented for two sessions before 2010. To eliminate the interference of the two civilized city selection activities on the empirical results, we excluded the sample of cities that were covered by the two selections mentioned above.

Considering that the national civilized cities are selected in batches, we set the following progressive difference-in-difference (DID) model: LnES G ict = a 0 + a 1 Civ _ Cit y ct + ∑ a j Control s cit + δ t + θ i + ε cit

Among them, L n E S G i c t represents the ESG performance of enterprise i in year t . C i v _ C i t y c t means whether city c was rated as a national civilized city in the year t . C o n t r o l s c i t is a set of control variables at the city and firm levels. The fixed effect of the firm ( θ i ) and year fixed effect ( δ t ) is added to the model, and ε c i t is a random disturbance term.

Corporate ESG performance is the dependent variable. We used the corporate social responsibility index evaluated by Hexun because this corporate social responsibility index includes the main content of ESG performance (Long and Zhang, 2021). In order to avoid the problem of heteroscedasticity caused by the absolute value, we used its logarithm value.

The national civilized city is the core explanatory variable, and we used a dummy variable to indicate whether the policy was implemented. The two civilized city selection times involved in the sample period were 21 December 2011, and 28 February 2015. Therefore, we set the corresponding city policy time to 2012 and 2015, respectively.

In addition, we also added the following control variables in the regression. At the regional level, we controlled GDP per capita (LnPGDP), measured by the logarithm value of GDP per capita, and population density (Population), measured by the ratio of urban population to the administrative area. At the firm level, we controlled firm size (Size), measured by the logarithm value of total assets, financial leverage (Lev), measured by the ratio of total liabilities to total assets, profitability (Roa), measured by the ratio of total profits to total assets, firm age (Age), measured by the logarithm value of the years since its establishment, and property right (Soe), measured by the dummy variable coded as 1 if the firm is a state-owned.

Table 1 shows the results of the impact of the selection as the national civilized cities on corporate ESG performance. In Model 1, the coefficient of Civ_City is significantly positive at the level of 5%, indicating that national civilized cities enhance corporate ESG performance. This is consistent with the conclusion of Chai et al. (2022) that the construction of a national civilized city led by the Chinese government has effectively strengthened the sense of corporate social responsibility. Under the government pressure and soft restraint mechanism brought about by the national civilized city selection policy, firms often help create a national civilized city by assuming more social responsibilities. Therefore, local governments constructing national civilized cities make a top-down pressure transmission mechanism for local firms, forcing them to invest in ESG actively.

The basic premise for applying the DID model is that there is no systematic difference in the corporate ESG performance between the experimental group and the control group before implementing the policy. To test whether this premise is confirmed, we refer to the practice of Li et al. (2016) and try whether the parallel trend condition is satisfied based on the event study analysis. Specifically, we proposed the following estimation model: LnES G ict = a 0 + ∑ k = − 3 3 + α k Civ _ Cit y k + ∑ a j Control s cit + δ t + θ i + ε cit

Among them, C i v _ C i t y k represents the dummy variable of the kth year of policy implementation, valuing 1 when the experimental group is in the kth year of policy implementation and 0 otherwise. Model 2 in Table 1 shows the results estimated by the above estimation model. In each year before the policy was implemented, the coefficients of C i v _ C i t y are not significant, which satisfies the assumption of parallel trends, and the DID model is suitable for the evaluation of policy effects in this study.

To further test whether the above results are robust, we conducted the following tests: 1) to reduce the effect of selection bias on the empirical results, we provided the results of the estimation of samples based on the propensity score matching method after 1:1 proximity no-put back matching in Model 3; 2) to eliminate the interference of the development trend of different industries on the empirical results, we controlled the fixed effect of industry-year multiplication in Model 4; 3) to alleviate the problems caused by missing values of data, we further increased the variables at the city level in Model 5, including the proportion of the added value of the tertiary industry to GDP, the size of the government (the ratio of fiscal expenditure to GDP), and the level of opening up (the ratio of foreign direct investment to GDP). At the firm level, we added the proportion of intangible assets to total assets, the proportion of tangible assets to total assets, and the proportion of shares held by the top ten shareholders; and 4) to alleviate the serial autocorrelation problem of residuals among firms, we clustered the standard errors of coefficients at the firm level in Model 6. The above results show that the coefficients of   C i v _ C i t y are always positive at the 5% level of significance, indicating that national civilized cities will improve corporate ESG performance, which is consistent with our main findings.

Our dependent variable, corporate ESG performance, is measured by the corporate social responsibility index evaluated by Hexun, which mainly consists of the environmental responsibility index and shareholder responsibility index. We separated the two indexes to estimate the regressions. Models 1 and 2 in Table 2 show the results. It can be seen that after the city is selected as the national civilized city, corporate environmental responsibility has been significantly increased, but the impact on shareholder responsibilities is not significant. This finding is contrary to Chai et al. (2022) but is consistent with the logic of corporate behaviors under government regulation. Firms assuming social responsibility have the characteristics of acting on the spot. Specifically, when the city is selected as the national civilized city, the local government requires firms to assume more social responsibility, primarily environmental responsibility. Firms pursuing economic interests are more willing to take on environmental responsibilities that are highly quantifiable while ignoring shareholder responsibilities that the government is less concerned about and hard to supervise.

From the perspective of property rights, state-owned firms have more political connections and are more closely related to the government (Lou et al., 2021). Compared with non-state-owned ones, they are more likely to reach agreements with government goals (Kim and Sumner, 2021). The civilized city policy may be more able to increase the ESG performance of state-owned firms. We divided the sample into two sub-samples, one for state-owned firms and another for non-state-owned firms. We found that national civilized cities have improved the ESG performance of state-owned firms, but the impact on non-state-owned firms is not significant. Models 3 and 4 in Table 2 report this heterogeneous impact. This result is consistent with our logical conjecture. It also confirmed the findings of Cheng et al. (2021) that state-owned firms play an essential role in implementing government policies, while non-state-owned ones have relatively little feedback and implementation.

The media is an external supplement to the corporate governance mechanism (Xiong and Luo, 2021). The brand effect of civilized cities will attract more media attention. When the city is selected as the national civilized city, corporate behaviors, especially social responsibility behaviors, receive special attention from the media, and the firms are required to perform more social responsibility (Zhou et al., 2019). We used the number of times a firm was reported in the news as a proxy variable for its media attention. Based on the median value of the number of reporting times in one industry each year, we divided the sample into two sub-samples, one for high media attention and another for low media attention. We found that firms that have received more media attention will have a better ESG performance when their cities are selected as the national civilized cities. However, the ESG performance of firms with low media attention has not been affected by the national civilized cities, as shown in models 5 and 6 in Table 2. This is a unique discovery of this paper. It can be seen that the spotlight effect of the media is an essential guarantee for civilized city policy to enhance corporate ESG performance. This result is similar to the findings of Xiong and Luo (2021), that is, the media spotlight effect will increase the transparency of corporate behavior, inhibit corporate anomie behavior, and encourage firms to fulfill their social responsibilities.