CMStatistics 2021: Start Registration
View Submission - CFE
A0598
Title: A network regression model with an estimated interaction matrix Authors:  Jose Olmo - Universidad de Zaragoza (Spain) [presenting]
Marcos Sanso-Navarro - Universidad de Zaragoza (Spain)
Abstract: A network regression model is proposed that incorporates exogenous neighboring effects into standard cross-sectional specifications. The interaction matrix is given by realizations of a functional coefficient that captures the network effects between the neighboring covariates and the outcome variable. This matrix is estimated using sieve regression methods and a Taylor expansion about a grid of reference points spanning the support of the distance variable that establishes the similarity between observations. The standardized estimator of the functional coefficient follows a zero-mean Gaussian process and the associated network parameter estimates are consistent and asymptotically normal. We also implement a uniform test to statistically assess for the presence of network effects. The empirical application studies environmental Engel curves, recently discussed in the literature, and finds strong evidence of neighboring effects in the relationship between households' income and the amount of pollution embodied in the goods and services they consume.