Problem with csdid command

[Mateus Maciel]

I have the following dataset:

id_municipio    ano    uf    munic    imposto_renda    taxpayers    ano_3g    ano_4g    treatment_3g
1200013    2013    AC    Acrelândia    430014    128    2011    2018    2011
1200054    2013    AC    Assis Brasil    205203    97    2015    2021    0
1200104    2013    AC    Brasiléia    1708153    361    2015    2017    0
1200138    2013    AC    Bujari    243841    89    2010    2017    2010
1200179    2013    AC    Capixaba    236769    102    2015    2019    0
1200203    2013    AC    Cruzeiro do Sul    8797109    1031    2008    2016    2008
1200252    2013    AC    Epitaciolândia    2199140    252    2013    2017    2013
1200302    2013    AC    Feijó    447522    282    2015    2017    0
1200328    2013    AC    Jordão    36289    35    2016    2020    0
1200336    2013    AC    Mâncio Lima    597157    91    2011    2017    2011
1200344    2013    AC    Manoel Urbano    105124    66    2016    2021    0
1200351    2013    AC    Marechal Thaumaturgo    47086    20    2016    2020    0
1200385    2013    AC    Plácido de Castro    262876    146    2015    2019    0
1200807    2013    AC    Porto Acre    145717    87    2014    2018    0
1200393    2013    AC    Porto Walter    23606    15    2016    2021    0
1200401    2013    AC    Rio Branco    1.594e+08    10252    2008    2014    2008
1200427    2013    AC    Rodrigues Alves    54578    63    2015    2018    0
1200435    2013    AC    Santa Rosa do Purus    51197    34    2016    2020    0
1200500    2013    AC    Sena Madureira    1006259    347    2015    2017    0
1200450    2013    AC    Senador Guiomard    1706722    327    2011    2017    2011

I am trying to obtain a did estimator a la Callaway and Sant'anna running the following command:

clear

import delimited "C:\Users\mateu\OneDrive\Documentos\base_tax.csv"

keep if ano >= 2004 & ano <= 2013

generate treatment_3g = 0
replace treatment_3g = ano_3g if ano >= ano_3g

xtset id_municipio ano

csdid imposto_renda , ivar(id_municipio) time(ano) gvar(treatment_3g) method(dripw)

However, I obtain the following result:

 csdid imposto_renda , ivar(id_municipio) time(ano) gvar(treatment_3g) method(dripw)
Panel is not balanced
Will use observations with Pair balanced (observed at t0 and t1)
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxx
Difference-in-difference with Multiple Time Periods

                                                             Number of obs = 0
Outcome model  : least squares
Treatment model: inverse probability
------------------------------------------------------------------------------
             | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
g2008        |
 t_2004_2005 |          0  (omitted)
 t_2005_2006 |          0  (omitted)
 t_2006_2007 |          0  (omitted)
 t_2007_2008 |          0  (omitted)
 t_2007_2009 |          0  (omitted)
 t_2007_2010 |          0  (omitted)
 t_2007_2011 |          0  (omitted)
 t_2007_2012 |          0  (omitted)
 t_2007_2013 |          0  (omitted)
-------------+----------------------------------------------------------------
g2009        |
 t_2004_2005 |          0  (omitted)
 t_2005_2006 |          0  (omitted)
 t_2006_2007 |          0  (omitted)
 t_2007_2008 |          0  (omitted)
 t_2008_2009 |          0  (omitted)
 t_2008_2010 |          0  (omitted)
 t_2008_2011 |          0  (omitted)
 t_2008_2012 |          0  (omitted)
 t_2008_2013 |          0  (omitted)
-------------+----------------------------------------------------------------
g2010        |
 t_2004_2005 |          0  (omitted)
 t_2005_2006 |          0  (omitted)
 t_2006_2007 |          0  (omitted)
 t_2007_2008 |          0  (omitted)
 t_2008_2009 |          0  (omitted)
 t_2009_2010 |          0  (omitted)
 t_2009_2011 |          0  (omitted)
 t_2009_2012 |          0  (omitted)
 t_2009_2013 |          0  (omitted)
-------------+----------------------------------------------------------------
g2011        |
 t_2004_2005 |          0  (omitted)
 t_2005_2006 |          0  (omitted)
 t_2006_2007 |          0  (omitted)
 t_2007_2008 |          0  (omitted)
 t_2008_2009 |          0  (omitted)
 t_2009_2010 |          0  (omitted)
 t_2010_2011 |          0  (omitted)
 t_2010_2012 |          0  (omitted)
 t_2010_2013 |          0  (omitted)
-------------+----------------------------------------------------------------
g2012        |
 t_2004_2005 |          0  (omitted)
 t_2005_2006 |          0  (omitted)
 t_2006_2007 |          0  (omitted)
 t_2007_2008 |          0  (omitted)
 t_2008_2009 |          0  (omitted)
 t_2009_2010 |          0  (omitted)
 t_2010_2011 |          0  (omitted)
 t_2011_2012 |          0  (omitted)
 t_2011_2013 |          0  (omitted)
-------------+----------------------------------------------------------------
g2013        |
 t_2004_2005 |          0  (omitted)
 t_2005_2006 |          0  (omitted)
 t_2006_2007 |          0  (omitted)
 t_2007_2008 |          0  (omitted)
 t_2008_2009 |          0  (omitted)
 t_2009_2010 |          0  (omitted)
 t_2010_2011 |          0  (omitted)
 t_2011_2012 |          0  (omitted)
 t_2012_2013 |          0  (omitted)
------------------------------------------------------------------------------
Control: Never Treated

See Callaway and Sant'Anna (2021) for details

. 
end of do-file

I would appreciate if FernandoRios could help me.

[FernandoRios]

First things first
Show a cross tab between year and gvar

[Mateus Maciel]

Thanks for helping, Fernando. Here is the tab:

 tab ano treatment_3g

           |                           treatment_3g
       ano |         0       2008       2009       2010       2011       2012 |     Total
-----------+------------------------------------------------------------------+----------
      2004 |     4,770          0          0          0          0          0 |     4,770 
      2005 |     4,906          0          0          0          0          0 |     4,906 
      2006 |     4,960          0          0          0          0          0 |     4,960 
      2007 |     5,085          0          0          0          0          0 |     5,085 
      2008 |     4,750        440          0          0          0          0 |     5,190 
      2009 |     4,523        440        262          0          0          0 |     5,225 
      2010 |     4,060        440        262        566          0          0 |     5,328 
      2011 |     2,768        440        262        566      1,322          0 |     5,358 
      2012 |     2,169        440        262        566      1,327        643 |     5,407 
      2013 |     1,893        440        262        566      1,326        642 |     5,410 
-----------+------------------------------------------------------------------+----------
     Total |    39,884      2,640      1,310      2,264      3,975      1,285 |    51,639

[FernandoRios]

That’s the problem
you don’t see your treated cohorts before they are treated
min other words your construction of gvar is incorrect

[Mateus Maciel]

Dear Fernando, I think I understand your point. Nevertheless, would you mind giving me a hint regarding the "gvar" construction?

I do have information about the year when wich one of my units were treated. To build the "gvar", I used the following command:

 
 generate treatment_3g = 0 replace treatment_3g = ano_3g if ano >= ano_3g

Meaning that the unit received 1 if the year in the panel is equal to or bigger than the year when the unit was treated, and 0 otherwise.

[Mateus Maciel]

Never mind, Fernando. I found a way to solve it.

Thank you so much for helping me.

[FernandoRios]

Glad you saw how to fix it