Spillover effects of central cities on sustainability efforts in a metropolitan area

Abstract

Metropolitan cities can serve as laboratories of sustainable development by experimenting with innovative sustainability programs while leveraging the advantages of metropolitan areas. With the importance of cities’ sustainability efforts, scholars have increasingly explored what factors motivate local governments to implement voluntary sustainability programs by focusing on internal government and community characteristics. However, what is missing in the previous discussion is whether city governments respond to sustainability efforts by other government entities, especially neighboring local governments. Drawing on institutional and policy diffusion theories, we analyzed the sustainability programs of 251 suburban cities in 66 metropolitan areas. We find that suburban cities are likely to consider sustainability efforts of central cities in the same metropolitan area when determining the extent to which they implement their sustainability programs. However, they are not necessarily responsive to sustainability efforts of nearby suburban cities. Our research sheds light on how local governments’ sustainability policy decisions are affected by their relationships with other local governments. It also provides policy implications for the important role of central cities in facilitating collective sustainability efforts in metropolitan areas.

Appendix: determining a spatial weight matrix

There are, in large, two ways of specifying the spatial weight matrix: (1) binary weight, where w_ij = 1 if the geographical distance between ${\text{location}}_{i}$ and ${\text{location}}_{j}$ is within a certain cutoff distance, otherwise 0 and (2) the inverse geographical distance between ${\text{location}}_{i}$ and ${\text{location}}_{j}$ if they are within a certain cutoff distance, otherwise 0 (Anselin and Le Gallo 2006). In either way, the weights matrix is usually row-standardized, such that $\sum\nolimits_{j} {w_{ij} = 1}$ (Drukker et al. 2013).

Given our focus on metropolitan areas, we relied on the land area size of MSAs included in our sample to determine the cutoff distance within which cities are considered as neighbors. For example, average land area size of MSAs in our sample is about 2318 km² (about 48 km radius) with minimum 38 km² (about 6 km radius) and maximum 15,612 km² (about 124 km radius). It implies that the maximum distance between two suburban cities in the same MSA is about 96 km on average. Taking this into account, we estimated our empirical models by using seven different spatial weight matrices: binary matrices with cutoff distance 50 km, 100 km, and 150 km, respectively, and the inverse distance matrices with cutoff distance 50 km, 100 km, 150 km, and no cutoff distance limit, respectively. Table 4 provides a summary of statistics for the empirical models using the seven different spatial weight matrices for our two dependent variables, respectively. Moran’s MI statistics test the hypothesis of no spatially autocorrelated disturbances, while LM-lag statistics test the hypothesis of no spatial lags (Zafra-Gómez and Chica-Olmo 2019). As shown in Table 4, the binary weight and the inverse distance matrices produce similar results for the number of program measure. However, for the breadth of program measure, Moran’s MI statistics indicate greater likelihood of disturbances to be spatially autocorrelated when using spatial matrix using no cutoff distance (p > 0.10), supporting our use of the SAR model that controls for spatial errors (Zafra-Gómez and Chica-Olmo 2019). The Akaike information criterion (AIC)s provide similar results across different cutoff distances, while the smaller AICs of the models using the inverse distance matrices suggest better model fits. Taken all these together, we determined to use inverse distance matrices using 50 km cutoff and no cutoff distance and then compared the results to ensure consistency of our findings across these two different scenarios.

Table 4 Statistics to determine a spatial weight matrix