Indirect inference estimation of higher-order spatial autoregressive models
提出一种通过匹配普通最小二乘估计量与其解析近似期望来估计高阶空间自回归模型参数的方法，该估计量一致、渐近正态、无需模拟且对未知异方差稳健，蒙特卡洛模拟和Airbnb租金实证验证了其有效性。
This paper proposes estimating parameters in higher-order spatial autoregressive models, where the error term also follows a spatial autoregression and its innovations are heteroskedastic, by matching the simple ordinary least squares estimator with its analytical approximate expectation, following the principle of indirect inference. The resulting estimator is shown to be consistent, asymptotically normal, simulation-free, and robust to unknown heteroskedasticity. Monte Carlo simulations demonstrate its good finite-sample properties in comparison with existing estimators. An empirical study of Airbnb rental prices in the city of Asheville illustrates that the structure of spatial correlation and effects of various factors at the early stage of the COVID-19 pandemic are quite different from those during the second summer. Notably, during the pandemic, safety is valued more and on-line reviews are valued much less.