Hi together,
I have an interesting result, where I am not fully sure how to correctly deal with it:
I have a model in which I am explaining the use of fossil fuels and in another case the use of metals.
I have a model which is identical with respect to variables and instrumentation in xtabond2 for explaining the two variables.
However, I have an Innovation variable, lagged by one-time period. Now, from a theoretical side Innovation can both be considered predetermined, but also potentially endogenous. If it is predetermined we could include it as IV, due to using the lag. If it is endogenous, we would have to instrument as gmm.
Now I have the following results
Fossil Fuels: When I change the instrumentation from endogenous, to including Innovation as IV, the coefficient becomes similar to regular Fixed Effects estimation and significant. All relevant tests hold, and the other variables results do not change.
Metals: When changing the instrumentation here the Sargan Test signals that we have an instrumenting problem.
Now I am struggling with the results as I am asking myself:
1) Does it make sense to assume that for Metals Innovation is endogenous and that`s why instrumenting it as IV leads to such problems -> in case of Metals has to be included as endogenous ?
2) How to decide on the right choice in case of fossil fuels ? All other tests and results are ok in both cases, so now I am unsure whether we should argue that including it as IV is better (reduces the instrument count) or not (would no longer account for potential endogeneity). In this case it is especially relevant as the significance changes in this case.
If somebody has an idea how to soundly deal with these observations I would be very thankful for any support.
Thanks and best
Tobi
I have an interesting result, where I am not fully sure how to correctly deal with it:
I have a model in which I am explaining the use of fossil fuels and in another case the use of metals.
I have a model which is identical with respect to variables and instrumentation in xtabond2 for explaining the two variables.
However, I have an Innovation variable, lagged by one-time period. Now, from a theoretical side Innovation can both be considered predetermined, but also potentially endogenous. If it is predetermined we could include it as IV, due to using the lag. If it is endogenous, we would have to instrument as gmm.
Now I have the following results
Fossil Fuels: When I change the instrumentation from endogenous, to including Innovation as IV, the coefficient becomes similar to regular Fixed Effects estimation and significant. All relevant tests hold, and the other variables results do not change.
Metals: When changing the instrumentation here the Sargan Test signals that we have an instrumenting problem.
Now I am struggling with the results as I am asking myself:
1) Does it make sense to assume that for Metals Innovation is endogenous and that`s why instrumenting it as IV leads to such problems -> in case of Metals has to be included as endogenous ?
2) How to decide on the right choice in case of fossil fuels ? All other tests and results are ok in both cases, so now I am unsure whether we should argue that including it as IV is better (reduces the instrument count) or not (would no longer account for potential endogeneity). In this case it is especially relevant as the significance changes in this case.
If somebody has an idea how to soundly deal with these observations I would be very thankful for any support.
Thanks and best
Tobi
