Robust Standard Errors to Spatial and Time Dependence in Panel Models When Neither N nor T is Very Large

This paper studies alternative approaches to consider time and spatial error dependence in aggregate panel models where neither N nor T is very large. I show that the variance of the two-way cluster standard errors (2CCE) is... [ view full abstract ]

This paper studies alternative approaches to consider time and spatial error dependence in aggregate panel models where neither N nor T is very large. I show that the variance of the two-way cluster standard errors (2CCE) is affected by both types of dependence. Therefore, these standard errors could be poorly estimated even in panels of moderate sample size. I show that the cluster can be expressed as a flexible panel version of the spatial autoregressive model (SAR). In a calibrated Monte Carlo exercise using state minimum wage data, I show that a parsimonious SAR panel model yields substantially better results than the 2CCE when N and T are as small as 50 and 30.