Estimate dynamic panel data model

[Simon Petersen]

Hi Stata,

I'm trying to estimate the dynamic panel data model (shown in picture), but i'm really struggling since the model contains several dependent variables (the dimension if 25x9)

Do you guys have any idea how to estimate in Stata?

Best regards,

[Øyvind Snilsberg]

It's actually only 1 dependent variable, but it's measured over 25 time periods for each of the 9 units.
To estimate the model in Stata, the data should be in long format. E.g., for i = 1,2,

* Example generated by -dataex-. For more info, type help dataex
clear
input float(id time outcome)
1  1   .3488717
1  2   .2668857
1  3   .1366463
1  4 .028556867
1  5   .8689333
1  6   .3508549
1  7  .07110509
1  8  .32336795
1  9   .5551032
1 10    .875991
1 11  .20470947
1 12   .8927587
1 13   .5844658
1 14   .3697791
1 15   .8506309
1 16   .3913819
1 17  .11966132
1 18   .7542434
1 19   .6950234
1 20   .6866152
1 21   .9319346
1 22   .4548882
1 23   .0674011
1 24   .3379889
1 25   .9748848
2  1   .7264384
2  2  .04541512
2  3   .7459667
2  4   .4961259
2  5   .7167162
2  6    .859742
2  7  .13407555
2  8  .48844185
2  9   .8712187
2 10   .7664683
2 11  .25125554
2 12  .16636477
2 13   .7437958
2 14   .9805113
2 15   .7295772
2 16   .9011049
2 17  .26436493
2 18   .8856509
2 19    .882112
2 20    .748933
2 21   .9196262
2 22   .6934533
2 23   .2154026
2 24   .8285888
2 25  .04421536
end

[Simon Petersen]

Hi again,

thanks for your response, appreciated! But if measured over 25 time periods for each of the 9 units, then I'll have to do the calculation for each unit, which gives me different slopes for M (in each estimation). I have attached at image from the main-file, where a description is included (It looks like they estimate an 25 x9 (depedent variables) dynamic panel data model (i think).

[Simon Petersen]

(In the main-file, they get the same slopes for M)

[Simon Petersen]

Anyone? :-)

[Sebastian Kripfganz]

To me, it looks like the authors are just estimating a conventional fixed-effects regression; see xtreg, fe.

[Simon Petersen]

But if they are using fixed-effects regression, they must be estimating for one unit (sector) at a time (i dont think it's possible to have a 25x9 matrix as dependent variabel). But if they estimate one unit at a time, it's kinda weird the authors get the same estimates for beta (pleas correct me, if i'm wrong).

[Simon Petersen]

(Unless the authors use all units as dependent variable, but only one period (which I don't think they do))

[Simon Petersen]

Anyone? (sorry I'm rushing, but it's for an important policy analysis)