Hi everyone,
I am using Arrelano-Bond/ Anderson-Hsiao GMM/IV 2SLS models to estimate the impact on US anti-dumping measures on China's exports of product i.
Within my model I have lagged the dependent variable as an explanatory variable along with a vector of other controls Xit (macro-level and industry-level).
Yiht = a + Yiht-1 + Xit + eiht.
My problem is that the certain industry-level controls are inadequate and poorly disclosed, hence I am using industry-time fixed effects as an alternative approach.
Yiht = a + Yiht-1 + Xit + Sit + eiht - where Sit is the industry-time dummy.
However, since I am using a dynamic panel approach, I understand that first-differencing a dummy removes it. So my question is:
1 - Because my industry-time dummy is time-variant, does it remain?
2 - If so, I'd really appreciate the command code for Arrelano-Bond GMM and Andersan-Hsiao IV/2 SLS estimators in this setting.
I have taken a look through Roodman's Xtabond2, which is very insightful, however there was no information about first-differencing an time interaction dummy.
Thank you in advance!
Best
Ray
I am using Arrelano-Bond/ Anderson-Hsiao GMM/IV 2SLS models to estimate the impact on US anti-dumping measures on China's exports of product i.
Within my model I have lagged the dependent variable as an explanatory variable along with a vector of other controls Xit (macro-level and industry-level).
Yiht = a + Yiht-1 + Xit + eiht.
My problem is that the certain industry-level controls are inadequate and poorly disclosed, hence I am using industry-time fixed effects as an alternative approach.
Yiht = a + Yiht-1 + Xit + Sit + eiht - where Sit is the industry-time dummy.
However, since I am using a dynamic panel approach, I understand that first-differencing a dummy removes it. So my question is:
1 - Because my industry-time dummy is time-variant, does it remain?
2 - If so, I'd really appreciate the command code for Arrelano-Bond GMM and Andersan-Hsiao IV/2 SLS estimators in this setting.
I have taken a look through Roodman's Xtabond2, which is very insightful, however there was no information about first-differencing an time interaction dummy.
Thank you in advance!
Best
Ray
