Hmm, that is bad news about there being so much missing data for the local government budget instrument. I'm not sure what to do.

 Maybe we should try using that instrument and see if it makes a difference. If it doesn't, we might conclude that we don't need an instrument, and use the full dataset.


Eric




[email protected] wrote:

Dear Sir,

      I extracted local govt budget and number of full time employees data. This is county level data. I am attaching the file. The bad news is that data for 733 counties are missing. That is about 1/3 of all observations. I think we shouldn't use this instrument because of missing data.
      This data was in two parts, one file in fixed format containing the actual data, the other file also in fixed format containing names of county etc.
      The first file did not have around 733 observations which were in second dataset. After I merged the two, there are missing values for 733 observations which were originally in first dataset. The _merge variable will say "only in master dataset" for these observations. You can use tab _merge to see this.
      The URL for these dataset is
 http://www.census.gov/govs/www/apesstl02.html

Sincerely,
Manu Raghav