A robust test for network generated dependence

The paper introduces a robust testing procedure for network generated cross sectional dependence in the endogenous variables, exogenous variables and/or disturbances. Empirical researchers often face situations where they are unsure about how to model the proximity between cross sectional units in a network. The tests considered provide the empirical researcher an important degree of robustness in such situations. They generalize the Moran (1950) I test for dependence in spatial networks. The asymptotic properties of the tests are established under general conditions. The paper also discusses the use of the test statistics in situations where the network topology is endogenous.