*3.3. Data Description*

Two hundred and eighty prefecture-level cities in China are the subject of the data in this paper, which involve a total of 21 variables in the fields of energy, tourism, environment, and economy; the data collection was challenging; most city statistics were not updated in a timely manner, leading to a significant amount of missing data in 2020; and artificially completing the data would interfere with the validity of the research findings. Likewise, the pace of China's tourism development has slowed down dramatically as a result of the new crown pandemic, taking into account that force majeure circumstances will compromise the validity of the findings and have a negative influence on tourism's contribution to the green economy. Given the aforementioned justifications, based on the availability and consistency of data, this paper selects the relevant data of 280 cities in 30 provinces from 2007 to 2019, excluding the data of Tibet, Hong Kong, Macao and Taiwan. Due to the long period and the change in some city data, this paper takes 2019 prefecture-level cities as the benchmark, excludes the data of merged cities, and retains the data related to the removal of counties and promotion of cities. The relevant data are obtained from *The China City Statistical Yearbook, China City Construction Statistical Yearbook, China Energy Statistical Yearbook* and *Tourism Statistical Yearbook* of each city. Carbon emission data were calculated based on county-level data from the literature. The list of pilot low-carbon cities is based on the list announced by the National Development and Reform Commission. To avoid the interference of multicollinearity and pseudo-regression, the VIF test and unit variance test were conducted on the panel data, and the results showed that the variance inflation factor of the panel data was less than three, and they all passed the LLC test and Fisher-ADF test at the 1% significance level, so there was no multicollinearity problem, and the data were smooth. In addition, the effect of heteroskedasticity was eliminated by taking the logarithms of all variables.
