I am estimating a system GMM model with N=5000 and T=4. One time period is lost due to the dependent lagged variable, and so, in practice, T=3. The Arellano-Bond test for serial correlation needs 4 periods to test for second-order serial correlation. Hence, I cannot check for AR(2). What is the best way moving forward? Does this invalidate my results? Is there a way to work around this issue?

More Hernan Borrero's questions See All
Similar questions and discussions