Spatial Autoregression with an Unknown Transformation of the Weight Matrix

This paper extends the spatial autoregressive model with a known matrix analytic transformation of the weight matrix to the case where this transformation is unknown given that $T$ is large. As $n$ is fixed, this can be done efficiently without the use of full nonparametric methods. We analyze the quasi-maximum likelihood estimator and show its consistency and asymptotic normality. In simulation, our estimator can accurately construct the transformed spatial weights matrix, and we apply our procedure to the estimation of unemployment in New Hampshire with county based spatial dependence. JEL Classification: C1, C5