I constructed a panel model on the impact of immigration on crime, with 12 regions over 8 years.
While using Fixed effects, my results seemed reasonable.
However, after conducting IV estimation and using the Arellano-Bond model, my coefficient changed signs and the impact on immigration remained insignificant. After testing for instrument relevance, all my instruments came out to be irrelevant.
Is there any way around this? I used lagged immigration and current disability as instruments. Additionally, I tried using log of GVA per capita, but removed that as it essentially is an income variable which should feature in the RHS of my regression.
Is there any way around this? Why are my instruments impacting my main variable that way and why are all the Sargan statistics significant?
Am I missing out on something?
Thank you for your time!
While using Fixed effects, my results seemed reasonable.
However, after conducting IV estimation and using the Arellano-Bond model, my coefficient changed signs and the impact on immigration remained insignificant. After testing for instrument relevance, all my instruments came out to be irrelevant.
Is there any way around this? I used lagged immigration and current disability as instruments. Additionally, I tried using log of GVA per capita, but removed that as it essentially is an income variable which should feature in the RHS of my regression.
Is there any way around this? Why are my instruments impacting my main variable that way and why are all the Sargan statistics significant?
Am I missing out on something?
Thank you for your time!