How can I keep firms have at least 2 years of pre and post-event financial data

[Celine Tran]

Hi,

I have an unbalanced panel data, which firms have a different number of year-observations. However, for my research, I want to keep firms have at least 2 years of pre and post-event financial data. The year that event occurs is 0. One-year pre, two-year pre are -1 and -2, etc. One-year post, two-year post are 1 and 2, etc...

Thank you very much in advance.

Regards,
Anh

[David Benson]

There is probably a few ways to do this (I'd have to see a copy of what your data looks like to get this to work:

1. Do all firms that have at least 5 yrs of data have -2, -1, 0, 1, and 2? (Or could a firm have 5 yrs of data, but enter at event_year==+2?). If it's the first, then you could do something like this

bysort firm_id (year): gen years_in_data = _N  
drop if years_in_data < 5
* NOTE: Above I assume year is an event_year (-2, -1, 0, +1, +2, etc), not 2005, 2006, 2007, etc

2. Alternatively, if a firm is in your sample at event_year==-2 and event_year==+2 does that mean it is also in your data for all years in between (i.e. it somehow isn't missing -1 or +1 so something like that?

bysort firm_id: egen first_year = min(year)
bysort firm_id: egen last_year = max(year)
keep if first_year<=-2 and last_year>=2 & last_year!=.

Hope that helps!

[Celine Tran]

Dear David,

Thank you very much for your help. However, it is not the one I need. I would like to investigate, for example, firm performance before and after an event (such as, global crisis - year 0). My sample is unbalanced, however I need that each firm must have at least two-year financial data before and after that event. Therefore, firms which have only one-year before event and more than one-year financial data; or more than one-year before event and only one-year after event, must be remove. Please ignore missing value.


For example: My sample is something like that:

id year
1 -2
1 -1
1 0
1 1
1 2
1 3
1 4
2 -3
2 -2
2 -1
2 0
2 1
2 2
3 -1
3 0
3 1
3 2
3 3
4 -3
4 -2
4 -1
4 0
4 1

I want to keep only firm id1 and firm id 2 because they have at least two-year financial data around year 0. Firm id 3 and 4 should be dropped from my sample because it has only one-year financial data before year 0 and only one-year financial data after event, respectively.

Thank you very much.

[David Benson]

Hi Celine,

Actually, I think that second set of code should work (with one modification):

bysort id: egen first_year = min(year)
bysort id: egen last_year = max(year)
gen to_keep=1 if first_year<=-2 & last_year>=2  // in your example above, to_keep==1 for firms 1 & 2, and missing for 3 & 4

* I don't like to drop things until I know it is what I want, once you've verified everything:
keep if to_keep==1

I've also posted the dataex stuff so you can copy and paste this to the General Forum page.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte(id year)
1 -2
1 -1
1  0
1  1
1  2
1  3
1  4
2 -3
2 -2
2 -1
2  0
2  1
2  2
3 -1
3  0
3  1
3  2
3  3
4 -3
4 -2
4 -1
4  0
4  1
end

[Celine Tran]

Dear David,

Thank you very much. I have saved me . The code works through.

Regards,
Anh.

[Celine Tran]

David Benson

Hi David,

Today, I am writing to ask you some questions related to the one above. If I want to keep firms have at least two firm-year observations before and after an event with ignoring the consecutiveness of year, what should I do?

For example:

       id      year
  1. 1 -2
  2. 1 -1
  3. 1 0
  4. 1 1
  5. 1 2
  6. 1 3
  7. 1 4
  8. 2 -4
  9. 2 -2
 10. 2 0
 11. 2 1
 12. 2 2
 13. 3 -1
 14. 3 0
 15. 3 1
 16. 3 2
 17. 3 3
 18. 4 -4
 19. 4 -1
 20. 4 0
 21. 4 1

I want to keep firm 1 and firm 2 because they have at least two firm-year observations around year =0.

Thank you very much in advance !

[Carlo Lazzaro]

Celine:
please re-post your query to the General forum, as this one is intended for practising purposes only. Thanks.

[Amelia Azjahra]

Carlo Lazzaro

Hi Carlo, i wanna ask you about panel data but not related to the one above.

I have micro panel data with 15.689 observation in 7 years. when i run hausman test, the results is to reject the null-hypothesis.
but, when i run Breusch-Pagan LM test, the results is also reject the null-hypothesis.

I am confused, what is the best model for my data? because i have one time-invariant variable so it's impossible to use FEM.
but i read from a book, if the true model is fixed effects, the random effects estimator is inconsistent.

Thanks before

[Carlo Lazzaro]

Amelia:
welcome to this forum.
Please re-post your query on the General forum, as this one is intended for practising purposes only. Thanks.