Seeking help with Dynamic Panel Regression using GMM

[Anthony CCH]

Hello everyone,

I am working on my Master's thesis which discusses the relationship between Geopolitical Risks (measured by Geopolitical Risk Index) and Bank Stability (measured by log-transformed Z-score).

Clearly, log-transformed Z scores are persistent and a dynamic panel regression is needed.

I watched some online videos and constructed my regression command this way:

xtabond2 log_z gpr log_total_assets div_ratio inflation l.log_z, gmm(log_z gpr log_total_assets div_ratio inflation, lag(2 .) collapse) robust h(3) two

gpr = Geopolitical Risk Index
bank controls = log_total_assets, diversification ratio
country controls = inflation

The result I get, unfortunately, fails the Sargan and Hansen tests...I have tried multiple lag combinations and have not found a set of valid test specifications.

Wondering if anyone could help?

[Sebastian Kripfganz]

There can be many reasons why these tests flag potential model misspecification. Instrument invalidity can result from relevant omitted variables, serially correlated error terms, a misclassification of variables as exogenous instead of predetermined or endogenous, etc. Another common reason could be that the additional assumptions of the system GMM estimator compared to the difference GMM estimator are violated.

The following presentation slides might be helpful:

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