Repeated time values within panel: time-series cross-sectional data with country & year

[Mattia Gatti]

Hi there,

I am having a problem with a dataset I am using for my thesis on Stata 18. I am using the Chapel Hill Expert Survey (CHES) dataset with parties nested in countries and two time points: 2017 and 2019.
It is my first time using panel data and, more generally, carrying out multi-level modelling. I am interested to run fixed effect models, yet I get stuck at the beginning because of the following error:

 xtset country year, yearly
repeated time values within panel
r(451);

The issue relates to the fact that each country variable is associated to multiple (different) observations (in this case, parties or better party_id). In short, country is not a unique identifier for my observations.

Nonetheless, country is my only second level predictor and I would like to take into account the clustering effect present in the data.

Here is a snapshot of my dataset:

* Example generated by -dataex-. For more info, type help dataex
clear
input double(country year party_id)
 3 2019  301
 3 2017  301
 3 2017  302
 3 2019  302
 3 2019  303
 3 2017  303
 3 2019  304
 3 2017  304
 3 2019  306
 3 2017  306
 3 2019  308
 3 2017  308
 3 2019  310
 3 2017  310
 3 2019  311
 3 2019  312
 3 2017  350
 3 2017  351
 4 2019  401
 4 2017  401
 4 2017  402
 4 2019  402
 4 2017  403
 4 2019  403
 4 2017  404
 4 2019  404
 4 2017  412
 4 2017  413
 4 2017  414
 4 2019  415
 4 2017  415
 4 2019  416
 4 2019  417
 4 2019  418
 4 2017  450
 5 2019  501
 5 2017  501
 5 2017  502
 5 2019  502
 5 2019  504
 5 2017  504
 5 2019  506
 5 2017  506
 5 2017  507
 5 2017  511
 5 2019  511
 5 2019  513
 5 2017  517
 5 2019  517
 5 2019  524
 5 2017  525
 5 2019  525
 5 2017  526
 5 2019  526
 5 2019  527
 5 2019  528
 5 2017  550
 5 2019  550
end

Suggestions would be highly appreciated.

Thanks in advance
Mattia

[Carlo Lazzaro]

Mattia:
this is not a relevant issue.
Provided that you do not plan to use time-series operators in your -xtreg- code, you can simply -xtset- your dataset via the -panelid- only.
This trick still makes including -i.timevar- as a predictor feasible.

[Mattia Gatti]

Hi Carlo,

thanks a lot for the quick answer.

If I have understood correctly, I could go on by just coding:

xtset country

While, to account for time, I should include it as a dichotomous var in the xtreg I will perform.

Thanks

Mattia