Hi,
I hope that I can find some help on my issue. I have specified a dynamic panel GMM model using Stata's xtabond command.
I want to explain product sales and I have weekly(T=6) information the rating, review count and price of the products. My model looks like this:
log_sales(it)= α + β0*sales(t-1) + β1*rating(it) + β2*log_review_count(it) + β3*price(it) + μ(i) + ε(it).
In my dynamic panel GMM model rating is exogenous, price is endogenous and I suspect log_review_count to be endogenous as well.
Hence in Stata I specify:
xtabond log_sales rating, lags(1) maxldep(1) endogenous(price, lag(0,1)) endogenous(log_review_count, lag (0,1))
After the estimation I carry out a Sargan test, which is significant and indicates overidentification.
Yet what is puzzling me is that the Sargan test is insignificant if I treat the variable log_review_count as exogenous.
I should also mention that log_review_count is non-stationary, i.e. it increases over time for every product.
Does anybody have any idea why this happens. Does this indicate that log_review_count is actually exogneous and should be treates as such despite my theoretical considerations that it is endogenous?
Any help would be greatly appreciated!
I hope that I can find some help on my issue. I have specified a dynamic panel GMM model using Stata's xtabond command.
I want to explain product sales and I have weekly(T=6) information the rating, review count and price of the products. My model looks like this:
log_sales(it)= α + β0*sales(t-1) + β1*rating(it) + β2*log_review_count(it) + β3*price(it) + μ(i) + ε(it).
In my dynamic panel GMM model rating is exogenous, price is endogenous and I suspect log_review_count to be endogenous as well.
Hence in Stata I specify:
xtabond log_sales rating, lags(1) maxldep(1) endogenous(price, lag(0,1)) endogenous(log_review_count, lag (0,1))
After the estimation I carry out a Sargan test, which is significant and indicates overidentification.
Yet what is puzzling me is that the Sargan test is insignificant if I treat the variable log_review_count as exogenous.
I should also mention that log_review_count is non-stationary, i.e. it increases over time for every product.
Does anybody have any idea why this happens. Does this indicate that log_review_count is actually exogneous and should be treates as such despite my theoretical considerations that it is endogenous?
Any help would be greatly appreciated!