egen (sum) is not summing up absolute figures but counts a number of observations

[Julia Korosteleva]

Greetings All,
I seem to have a problem in Stata that should be a pretty much straightforward to resolve, but nothing I have tried so far (including egen sum or collapse command), seems to work. I have a firm-level panel data for one country only. I am trying to calculate total employment over 5-year time periods. First, I use an egen sum command to sum up employment annually: egen yr_employment=sum(employment), by(year). However, instead of getting the figures on total annual employment, I get firm observations on employment counted. I have tried all possible solutions, and yet nothing worked yet. 'Employment' has a data storage format 'long'. I have read somewhere on Stata forum discussion that sometimes a 'long' format could cause issues as a variable seems to be wrongly treated as string, though it is a numeric. I have generated a new var with a float format, but this hasn't resolved the problem. And I am not sure if it's a 'long' data storage type that causes a problem in the first place. Any suggestions of how I could resolve this problem would be highly appreciated.

[Nick Cox]

Please show a data example that backs up the report here. The option by(year) would add over firms, not over years.

long is a variable or storage type, not a display format.

This works for me:

. clear

. set obs 10

. gen year = cond(_n <= 5, 2010 + _n, 2010 + _n - 5)

. gen firm = cond(_n <= 5, 1, 2)

. gen long employment = 1e6 + _n

. egen check = sum(employment), by(firm)

. l

     +----------------------------------+
     | year   firm   employ~t     check |
     |----------------------------------|
  1. | 2011      1    1000001   5000015 |
  2. | 2012      1    1000002   5000015 |
  3. | 2013      1    1000003   5000015 |
  4. | 2014      1    1000004   5000015 |
  5. | 2015      1    1000005   5000015 |
     |----------------------------------|
  6. | 2011      2    1000006   5000040 |
  7. | 2012      2    1000007   5000040 |
  8. | 2013      2    1000008   5000040 |
  9. | 2014      2    1000009   5000040 |
 10. | 2015      2    1000010   5000040 |
     +----------------------------------+

Note that the egen function sum() went undocumented in Stata 9. It still works, but total() has been the documented name since 2005.

[Julia Korosteleva]

Dear Nick, many thanks for your prompt response which is very useful - all works fine. I am not sure what was an issue though (could be, perhaps, a glitch) as all has worked perfectly fine this morning with the use of both 'sum ' and 'total' and without generating a new 'storage type' variable to replace the one with a 'long' format. Thank you once again.