Correct GMM-style instrument specification with deeper lags of dependent variable using Xtabond2

[Frederik Federspiel]

Hi there, I'm trying to understand how to correctly specify the gmm-style instrument generation in Xtabond2, when I have to use deeper lags of the dependent variable to ensure the null for the Arellano-Bond test for AR(2) in first differences can safely NOT be rejected - i.e. ensuring no first-order serial correlation in levels.

I basically just want to lag by one year and upwards for each independent variable, but now my lagged dependent variable is both L1 and L2, so I guess I want one set of instruments that goes L2, L3, L4, L5 and another that goes L3, L4, L5 - If I understand correctly?

Here's a simplified version of my code:

xtabond2 y L(1/2).y L.x1 x2 yr*, gmm(L(1/2).y L.x1, laglimits(1 5) collapse) iv(x2 yr*) twostep robust small

Is my lag-specification correct in the GMM-style instrument specification?

[Sebastian Kripfganz]

These sets of instruments are overlapping: L3.L.y is the same as L2.L2.y, and so on. This is not necessary a problem because redundant instruments are automatically dropped.

More on dynamic panel data GMM estimation:

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.