*4.1. Spatial Autocorrelation Test and Spatial Econometric Model Selection*

Urban carbon emission intensity and green economic efficiency are spatially correlated, which is a prerequisite for spatial econometric modeling [60]. Stata.23 is used to calculate the global Moran index to explore the spatial agglomeration characteristics. As shown in Table 2, the GEE Moran index is positive for all the years from 2007 to 2019, and the spatial agglomeration strengthens year by year except for 2013–2015, all of which pass the 1% significance level test. During the observation period, the Moran index of Carin is positive in all cases, and the degree of spatial agglomeration increases in fluctuation. Overall, both urban green economic efficiency and carbon emission intensity have significant spatial agglomeration characteristics.


**Table 2.** Global Moran Index test results for green economy efficiency and carbon emission intensity.

Note: \*\*\*\*, \*\*, and \* represent significance levels of 1%, 5%, and 10%, respectively.

Based on passing Moran's I test, the spatial effect econometric model can be selected. Additionally, to measure the nonlinear characteristics of tourism development on urban carbon emission intensity and green economic efficiency, models (3) and (4) containing quadratic terms of tourism development variables were introduced. First, the LM test, LR

test and Wald test were performed to identify the spatial econometric models. Both the LM spatial lag test and LM spatial error test showed high significance, so both the SAR model and SEM were suitable for this study, and we chose the SDM model that combined both. Then, the LR test and Wald test were applied to further determine whether the SDM could be degraded to the SAR model or SEM. The comprehensive test results showed that the SDM was better than the SAR and SEM, so the SDM was selected as the baseline regression model in this paper. Based on the SDM model [62], all matrices passed the Hausman test at a 1% significance level except for model (1), which had a negative value, so the fixed-effects model was selected. As shown in Table 3, the overall R2 of the individual fixed-effects model is significantly better than that of the time-point fixed-effects and double fixed-effects models. Finally, the individual fixed-effects spatial Durbin model is chosen to analyze the impact of tourism development on urban carbon emission intensity and green economic efficiency.

**Table 3.** Spatial econometric model selection.


Note: \*\*\*\*, \*\*, and \* represent significance levels of 1%, 5%, and 10%, respectively.
