Spatial Archimedean-Copulas Maximum Likelihood Estimation With Endogenous Spatial Weights Matrices

Lee, Jieun

Permalink

https://hdl.handle.net/2142/114499 Copy

Description

Title

Spatial Archimedean-Copulas Maximum Likelihood Estimation With Endogenous Spatial Weights Matrices

Author(s)

Lee, Jieun

Issue Date

2022-09-16

Keyword(s)

Archimedean copula, maximum likelihood estimation, dependence in the error terms, dependence at the extreme values (tail dependence), endogenous spatial weights matrices.

Abstract

Spatial ecoometrics involves spatial weights matrices (W) to capture the effect of interactions among spatial units on the economic outcomes. Commonly, the intuitive use in W has been found in the geographical features. If W is constructed by non-predetermined distances, however, it could be endogenous and thus the conventional spatial autoregressive estimator (SAR) may not be consistent. The Quasi Maximum Likelihood Estimation (QMLE) specified under a normal distribution is known to be consistent and asymptotically normal (Qu and Lee, 2015). However, there do exist some situations where such specification is not enough. Particularly, in case where the variables are dependent at the extreme values or have heavy tail dependence, estimators specified under a normal distribution do not work well as the tail dependence is not inferred. In this paper, I propose the Spatial Archimedean-Copulas Maximum Likelihood (ML) estimator to control the endogeneity of W when the disturbance terms are dependent at the extreme values. A Monte Carlo simulation supports the analytics and shows its nice finite sample properties.

Type of Resource

Work in progress

Language

en

Handle URL

https://hdl.handle.net/2142/114499

Owning Collections