[Mustafa Caliskan]

Hi Fernando,

I’m encountering an issue with the long2 command—it seems to be dropping some years from my dataset, and I haven’t been able to determine why. Could you help me figure out the cause?

I’m working with SIPP data, which is a rotating panel. Could that be a factor in this issue?

Thanks in advance for your help!


[CODE]

csdid creditcarddebt_log [weight=hweight] , ivar(hid) time(year) gvar(o_year) notyet
(importance weights assumed)
(importance weights assumed)
Panel is not balanced
Will use observations with Pair balanced (observed at t0 and t1)
...xxxxxx...x...xx...x...xx...x........x........x.
....(4 missing values generated)

Difference-in-difference with Multiple Time Periods

Number of obs = 43,022
Outcome model : regression adjustment
Treatment model: none
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
g2018 |
t_2014_2015 | .1288567 .1257227 1.02 0.305 -.1175552 .3752686
t_2015_2016 | -.011361 .1339865 -0.08 0.932 -.2739697 .2512477
t_2016_2017 | .1221403 .1358202 0.90 0.369 -.1440623 .3883429
t_2017_2018 | 0 (omitted)
t_2017_2019 | 0 (omitted)
t_2017_2020 | 0 (omitted)
t_2017_2021 | 0 (omitted)
t_2017_2022 | 0 (omitted)
t_2017_2023 | 0 (omitted)
-------------+----------------------------------------------------------------
g2019 |
t_2014_2015 | -.0540998 .0685279 -0.79 0.430 -.188412 .0802123
t_2015_2016 | .0822121 .0730696 1.13 0.261 -.0610017 .2254259
t_2016_2017 | .0206096 .0727599 0.28 0.777 -.1219971 .1632163
t_2017_2018 | 0 (omitted)
t_2018_2019 | -.0477295 .0678176 -0.70 0.482 -.1806495 .0851906
t_2018_2020 | .022488 .0821306 0.27 0.784 -.138485 .1834609
t_2018_2021 | -.0504147 .0997901 -0.51 0.613 -.2459997 .1451702
t_2018_2022 | 0 (omitted)
t_2018_2023 | 0 (omitted)
-------------+----------------------------------------------------------------
g2020 |
t_2014_2015 | .1201289 .0676236 1.78 0.076 -.0124109 .2526686
t_2015_2016 | -.1726418 .0733755 -2.35 0.019 -.3164551 -.0288285
t_2016_2017 | .025222 .0771589 0.33 0.744 -.1260067 .1764508
t_2017_2018 | 0 (omitted)
t_2018_2019 | .016374 .1043152 0.16 0.875 -.1880799 .220828
t_2019_2020 | -.0252896 .0932367 -0.27 0.786 -.2080301 .1574508
t_2019_2021 | .2404562 .1137706 2.11 0.035 .01747 .4634425
t_2019_2022 | 0 (omitted)
t_2019_2023 | 0 (omitted)
-------------+----------------------------------------------------------------
g2021 |
t_2014_2015 | .0188281 .0717065 0.26 0.793 -.1217141 .1593703
t_2015_2016 | -.0830524 .0781733 -1.06 0.288 -.2362692 .0701645
t_2016_2017 | .0437133 .0781423 0.56 0.576 -.1094427 .1968694
t_2017_2018 | 0 (omitted)
t_2018_2019 | -.0929588 .0826294 -1.13 0.261 -.2549095 .0689919
t_2019_2020 | .05824 .0907386 0.64 0.521 -.1196044 .2360843
t_2020_2021 | -.089361 .0767514 -1.16 0.244 -.239791 .061069
t_2020_2022 | .3382623 .1501639 2.25 0.024 .0439465 .6325781
t_2020_2023 | .274011 .205645 1.33 0.183 -.1290458 .6770679
-------------+----------------------------------------------------------------
g2022 |
t_2014_2015 | -.008298 .0802026 -0.10 0.918 -.1654923 .1488963
t_2015_2016 | .0440349 .0866937 0.51 0.611 -.1258817 .2139514
t_2016_2017 | .0201032 .081277 0.25 0.805 -.1391967 .1794032
t_2017_2018 | 0 (omitted)
t_2018_2019 | .0484307 .0830982 0.58 0.560 -.1144388 .2113002
t_2019_2020 | -.0461753 .0932762 -0.50 0.621 -.2289932 .1366427
t_2020_2021 | -.1121791 .0849456 -1.32 0.187 -.2786694 .0543112
t_2021_2022 | -.0251153 .1205218 -0.21 0.835 -.2613338 .2111031
t_2021_2023 | .0934505 .1449315 0.64 0.519 -.1906101 .3775112
-------------+----------------------------------------------------------------
g2023 |
t_2014_2015 | .1603671 .1059174 1.51 0.130 -.0472272 .3679614
t_2015_2016 | .169297 .1126223 1.50 0.133 -.0514387 .3900326
t_2016_2017 | -.0347464 .1174833 -0.30 0.767 -.2650095 .1955167
t_2017_2018 | 0 (omitted)
t_2018_2019 | .0845053 .1107733 0.76 0.446 -.1326064 .3016171
t_2019_2020 | .0161582 .1263434 0.13 0.898 -.2314704 .2637868
t_2020_2021 | -.1116175 .1313876 -0.85 0.396 -.3691324 .1458974
t_2021_2022 | -.0370114 .1881763 -0.20 0.844 -.4058301 .3318074
t_2022_2023 | -.1439625 .1314627 -1.10 0.273 -.4016247 .1136997
------------------------------------------------------------------------------
Control: Not yet Treated

See Callaway and Sant'Anna (2021) for details

. estat event
ATT by Periods Before and After treatment
Event Study:Dynamic effects
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
Pre_avg | .0263242 .0186142 1.41 0.157 -.0101589 .0628074
Post_avg | .0387251 .0463956 0.83 0.404 -.0522086 .1296588
Tm8 | .1603671 .1059174 1.51 0.130 -.0472272 .3679614
Tm7 | .0447273 .0654551 0.68 0.494 -.0835624 .173017
Tm6 | .0193928 .0508632 0.38 0.703 -.0802972 .1190828
Tm5 | .0178399 .0448449 0.40 0.691 -.0700545 .1057343
Tm4 | -.0342753 .0396317 -0.86 0.387 -.111952 .0434014
Tm3 | .0613735 .0407487 1.51 0.132 -.0184925 .1412395
Tm2 | -.0438213 .0424084 -1.03 0.301 -.1269403 .0392977
Tm1 | -.0150102 .0488613 -0.31 0.759 -.1107767 .0807563
Tp0 | -.0683828 .0408175 -1.68 0.094 -.1483837 .011618
Tp1 | .1377402 .0584609 2.36 0.018 .023159 .2523213
Tp2 | .046818 .0943233 0.50 0.620 -.1380524 .2316883
------------------------------------------------------------------------------






csdid creditcarddebt_log, ivar(hid) time(year) treatment(treatment_active) gvar(o_year) agg(event) long2
Panel is not balanced
Will use observations with Pair balanced (observed at t0 and t1)
xxxx...xxxxxx...xxxxxx.....xxxx.....xxxxxx...xxxxx
x...
Difference-in-difference with Multiple Time Periods

Number of obs = 21,171
Outcome model : regression adjustment
Treatment model: none
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
Pre_avg | .1072762 .0532004 2.02 0.044 .0030053 .2115471
Post_avg | .0114637 .0662737 0.17 0.863 -.1184304 .1413578
Tm4 | .2430794 .1101621 2.21 0.027 .0271656 .4589932
Tm3 | .0213244 .0673931 0.32 0.752 -.1107636 .1534125
Tm2 | .0574248 .0391181 1.47 0.142 -.0192452 .1340948
Tp0 | -.0012944 .0405202 -0.03 0.975 -.0807127 .0781238
Tp1 | .1258476 .0713153 1.76 0.078 -.0139277 .2656229
Tp2 | -.090162 .1567307 -0.58 0.565 -.3973485 .2170245
------------------------------------------------------------------------------
Control: Never Treated

See Callaway and Sant'Anna (2021) for details


[CODE]

[FernandoRios]

start with a simple task
tabulate tvar and gvar.

[Mustafa Caliskan]

tab year treatment

| treatment
year | 0 2018 2019 2020 2021 2022 | Total
-----------+------------------------------------------------------------------+----------
2013 | 11,116 445 2,107 1,885 1,772 2,505 | 19,830
2014 | 11,039 431 2,083 1,908 1,752 2,453 | 19,666
2015 | 9,085 344 1,783 1,541 1,503 2,000 | 16,256
2016 | 7,515 264 1,507 1,278 1,229 1,626 | 13,419
2017 | 9,690 410 1,544 1,249 1,483 1,694 | 16,070
2018 | 7,474 313 1,312 988 1,211 1,311 | 12,609
2019 | 9,723 414 1,583 1,337 1,553 1,695 | 16,305
2020 | 10,763 427 1,592 1,380 1,653 1,707 | 17,522
2021 | 7,014 301 1,041 843 970 1,235 | 11,404
2022 | 6,083 273 900 770 826 1,085 | 9,937
-----------+------------------------------------------------------------------+----------
Total | 89,502 3,622 15,452 13,179 13,952 17,311 | 153,018

[FernandoRios]

Ok it’s probably because you have a kind of rotating panel
so households are not there for all periods

[Anna Chesa]

Hello again Fernando! Is there any explanation to why do I always get very high standard errors in the pre-treatment phase compared to the treated one? Even when I try with a balanced panel, the standard errors are way bigger in the left of the eventstudy plot and I do not understand why, thank you!

[FernandoRios]

No general explanation
it is what it is
however Pretreatment test are known for low power

[Anna Chesa]

Use jwdid for that. It has an option for adding fixed effects other than individual
In any case when explaining and presenting results be aware that identification may be difficult that way, and the sssumptions you are imposing when doing that

Hello Fernando! I am not sure whether I understand this, can you link me to a paper or any reference that explains why I cannot use year fixed effects when using Callaway and Sant'Anna (2021)? Thanks a lot!

[FernandoRios]

It’s about how it is constructed
A) each attgt considers only 2x2 periods
b) for each period a regression is run
c) for that regression there is no time component because you only have one period and one group

[Rory Todd]

Hi Fernando,

Quick question about exporting csdid results to LaTeX using esttab. I'd like to have only the GAverage coefficient (and standard error/p-value) in my table. But if I do

eststo: csdid tempO ln_GNI_pc ln_wdi_pop, ivar(ccode) time(year) gvar(firstZyear) notyet agg(group)

then it gives me coefficients for all treatment timing groups:

. esttab

----------------------------
(1)

----------------------------
GAverage -7.456***
(-4.36)

G1997 -0.628
(-0.24)

G1998 0.328
(1.00)

G2000 -6.730*
(-2.38)

G2001 -7.825
(-1.54)

G2002 -18.72**
(-2.64)

G2005 -39.57***
(-177.71)

G2006 -15.36
(-1.38)

G2007 3.426***
(10.30)

G2008 2.007***
(6.04)

G2009 2.375***
(4.31)

G2010 -4.146***
(-3.34)
----------------------------
N 1980
----------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001


Do you know how I only show the GAverage one?

Thank you!