Implementation of a single autoregressive term in a fixed effects model

[Joan Stein]

Hey to everyone,

as I am facing a problem of a wrongly build autoregressive model I tried to change the model structure accordingly.

Thus, I am trying to integrate an single autoregressive term into my fixed effects model. While I can find multiple discussions about Panel vector autoregressive models etc. where all independent variables are lagged, I cannot find discussions about such single autoregressive terms.

xtreg Depvar Indepvar L1.c.Indepvar##i.Dummy Controlvar1 Controlvar2 Controlvar3 Controlvar4, fe cluster(id)

Depvar and Indepvar are both stationary. Can I just run this model? Is there something else I have to check for (besides the common Gauss Markow Assumptions) or is the model itself in this form invalid?

I am not choosing the Panel autoregressive model since I would like to implement multiple control variables and the model seem to be too large to use the PVAR structure (there are much more control variables than in the exemplary code line). Moreover, I will estimate a building and simplified PVAR after the regression. So this one here, should represent a more detailed base.

Greetings and thank you in advance!

P.S. sorry for the frequent questions, I am kind of desperate right now

[Joro Kolev]

Do you have multiple time series, where the cross sectional dimension is small, and your time series dimension is large?

Or do you have a dataset with large cross sectional dimension and short time series?

[Joan Stein]

Hello Mr. Kolev thank you for your comment! I have a panel data set of large N and small T.

[Joro Kolev]

Nothing wrong in just putting the independent variable that you want to lag, with a lag operator as you have done in your first post.

You do not need to worry about stationarity, in your data the asymptotics go in the cross sectional dimension, so your variables can be as misbehaved as they want in the time series dimension.

Hello Mr. Kolev thank you for your comment! I have a panel data set of large N and small T.

[Joan Stein]

@Mr. Kovlev thank you very much! You are saving my day and probably my whole week!

May I ask you one more question?

1) Can I perform the model above also by lagging multiple variables or is this changing your assessment about the model validity or the importance of stationarity?

xtreg Depvar Indepvar0 Indepvar1 Indepvar2 Indepvar3 L1.Indepvar0 L1.Indepvar1 L1.Indepvar2 L1.Indepvar3 Controlvar1 Controlvar2 Controlvar3 Controlvar4, fe cluster(id)

[Joro Kolev]

Hi Joan, yes you can lag as many independent variables as you want and nothing changes in the econometrics of how this is implement. You can do what you are showing.

@Mr. Kovlev thank you very much! You are saving my day and probably my whole week!

May I ask you one more question?

1) Can I perform the model above also by lagging multiple variables or is this changing your assessment about the model validity or the importance of stationarity?

xtreg Depvar Indepvar0 Indepvar1 Indepvar2 Indepvar3 L1.Indepvar0 L1.Indepvar1 L1.Indepvar2 L1.Indepvar3 Controlvar1 Controlvar2 Controlvar3 Controlvar4, fe cluster(id)

[Joan Stein]

Thank you very much for your time!