Identify first year of a firm in the panel

[Miral Ayan]

Hi everyone,
How can I create a dummy variable for the first year of a firm in the panel data?

Thanks in advance

[Carlo Lazzaro]

Miral:
you may want to consider the following toy-example:

use "https://www.stata-press.com/data/r16/nlswork.dta"
. bysort idcode (year): gen wanted=1 if idcode==_n
. replace wanted=0 if wanted==. & idcode!=. & year!=.
.

[William Lisowski]

I think that in Carlo's example the generate command should be stated as

bysort idcode (year): gen wanted=1 if _n==1

to obtain a 1 for the first year within each idcode.

[Miral Ayan]

Dear Carlo Lazzaro ,
I used the code. It generated a variable with all 0 values.

[Miral Ayan]

Dear William Lisowski, the code worked perfectly. Thank you so much.

[Carlo Lazzaro]

Miral:
I obviously agreed with William about the first line of the code (my typo indeed. Thanks William for pointing this out).
But you asked for a two-level categorical variable, too.

. bysort idcode (year): gen wanted=1 if _n==1

. replace wanted=0 if wanted==. & idcode!=. & year!=.


. list wanted if idcode==1

       +--------+
       | wanted |
       |--------|
    1. |      1 |
    2. |      0 |
    3. |      0 |
    4. |      0 |
    5. |      0 |
       |--------|
    6. |      0 |
    7. |      0 |
    8. |      0 |
    9. |      0 |
   10. |      0 |
       |--------|
   11. |      0 |
   12. |      0 |
       +--------+

.

[Nick Cox]

Dummies (better: indicators) are most useful when they have distinct values of 0 and 1. That being so, a direct alternative to answers so far is

bysort id (year) : gen first = _n == 1

and another that is a little more general is

bysort id (year) : gen first = year == year[1]

which allows for panels in which repeated values of the time variable are possible.

EDIT: Carlo Lazzaro's #6 was not visible as I first typed this. Note that his #6 is in spirit a two-line version of my first suggestion..

[Carlo Lazzaro]

...and, as such, less efficient!!
Great, Nick!

[Miral Ayan]

Dear Carlo Lazzaro ,
Thank you. Can you correct me if it requires a dummy variable for the following statement?

"To further minimize reverse causality, we perform an instrumental-variable analysis. We identify the earliest year when each firm appears in our sample. TW co-option in the earliest year could not have resulted from the firm’s risk-taking in any of the subsequent years, making reverse causality unlikely"

[Carlo Lazzaro]

Miral:
provided that your clarification seems to classify your query as one belonging to the https://www.statalist.org/forums/help#adviceextras #4 Chapter, your trick should reduce the risk of reverse causality-led endogeneity.

[William Lisowski]

Let me clarify, for the benefit of anyone who should read this topic at a later date, that my advice in post #3 is intended to apply to only the generate command in post #2, leaving the replace command in post #2 as it was given.

I wasn't trying to rewrite or improve on Carlo's solution, just point out the typo that Carlo had overlooked.

[Carlo Lazzaro]

William:
your appreciated advice reached me the right way.
I was really grateful for your pointing out my typo in such a kind way.
Molte grazie davvero/Many thanks indeed

[Miral Ayan]

Thanks to all wonderful STATA community