Hello!
I'm running an IV regression in Stata on a monthly panel dataset for the Brazilian Amazon. Specifically, at the municipality-level, I am regressing the % vote share for specific candidates in local municipal elections (with observations once every 4 years) on the number of environmental fines in the municipality, using monthly average cloud cover as the instrument. But because electoral observations are every 4 years, I computed this manually - by:
1) Fixed effects regression of monthly fines on monthly cloud in the first stage of the IV
2) Save the predicted values
3) Summing the predicted values for all 4 years before an election (so for the 2012 election, the value for 'predicted fines' would be the sum of all fines from 2008-12)
4) Fixed effects regression of the % Vote Share on the Sum of Predicted Fines, at the municipality-level
However, I am aware that using this manual method would result in incorrect standard errors - I've seen the issue raised in other posts, but unfortunately none I have seen have been able to solve this. Could anyone please advise how to correct the standard errors?
Please find attached sample data and some code below.
I'm running an IV regression in Stata on a monthly panel dataset for the Brazilian Amazon. Specifically, at the municipality-level, I am regressing the % vote share for specific candidates in local municipal elections (with observations once every 4 years) on the number of environmental fines in the municipality, using monthly average cloud cover as the instrument. But because electoral observations are every 4 years, I computed this manually - by:
1) Fixed effects regression of monthly fines on monthly cloud in the first stage of the IV
2) Save the predicted values
3) Summing the predicted values for all 4 years before an election (so for the 2012 election, the value for 'predicted fines' would be the sum of all fines from 2008-12)
4) Fixed effects regression of the % Vote Share on the Sum of Predicted Fines, at the municipality-level
However, I am aware that using this manual method would result in incorrect standard errors - I've seen the issue raised in other posts, but unfortunately none I have seen have been able to solve this. Could anyone please advise how to correct the standard errors?
Please find attached sample data and some code below.
Code:
clear xtset muni_code year_month /// First Stage: xtreg fines_count deter_cloud i.year_month , fe vce(cluster muni_code) predict xhat bys Concat_MuniCode_ElectPeriod: egen xhat_sum = total(xhat) /// Second Stage: xtreg vote_pct_farmer_council xhat_sum i.year_month , fe vce(cluster muni_code)
