Characterization and estimation of heterogeneous spatial autocorrelation in spatial autoregressive models

Area of Science:

• Spatial econometrics
• Regional science
• Environmental economics

Background:

• Traditional Spatial Autoregressive (SAR) models assume constant spatial autocorrelation.
• This assumption limits their ability to capture spatial heterogeneity in interactions.
• Existing models struggle with dynamic spatial dependence structures.

Purpose of the Study:

• To propose a novel Spatial Single-Index Varying Coefficient Autoregressive (SSIVCAR) model.
• To address the limitations of traditional SAR models in capturing spatial heterogeneity.
• To provide a more accurate framework for analyzing spatial dependence.

Main Methods:

• Introduction of a single-index varying coefficient function.
• Estimation using a combination of spline methods and two-stage least squares.
• Performance assessment via Monte Carlo simulations under finite sample conditions.

Main Results:

• The proposed SSIVCAR model significantly improves the capture of spatial heterogeneity.
• Estimation accuracy is enhanced compared to traditional models.
• Simulations confirm the model's effectiveness in finite sample conditions.

Conclusions:

• The SSIVCAR model offers a robust framework for analyzing complex spatial dependence.
• The model reveals heterogeneous impacts of digital economy on environmental quality across regions.
• Findings provide valuable insights for regional governance and policy-making.