Using Fixed Effects Panel Model Estimates to Forecast Dependent Variable

[Marta Anacka]

Hello!
I was looking for this information at this forum as well as throughout the web and couldn't find proper answer.
I have estimated panel model with fixed effects (40 countries, 12 time periods) and I would like to forecast dependent variable 1 period ahead for each (or at least one) country/id. I haven't found explicit stata command to do so. Are there any tricks that I could introduce to get this forecast?

I am using STATA 12.

Thanks in advance,
MA

[Phil Bromiley]

If you look in the manual for xtreg postestimation, you'll find it predicts including the fixed effect automatically. It looks like: predict yhat,xbu . You'll need the values of the X variables to make this work.

[Marta Anacka]

Dear Phil,
thank you for your quick and clear reply.

I have tried the predict command but it looks that it works only for in-sample records. And I would like to forecast my dependent variable one period ahead.
I have used also:

predict m1_u, u

to compute my forecast manually, but I still have some doubts, as I am not sure if I can just add xb and u to get the forecast. I have read here that the unbiased estimator (BLUP) is somehow different (p. 6, eq. 13 in the referenced paper).

[Alfonso Sánchez-Peñalver]

I shall say that I have never had to do this, but my first thought is... wouldn't you have to forecast the independent variables to the next period (maybe doing ARIMA) and then use the model you just estimated to estimate the forecasted mean? The question then is whether this estimation is better than directly forecasting the dependent variable using ARIMA. Of course, if the FE are significant you would have to forecast by group. Again, this is not my area of expertise so I'm not sure what you would have to do, but it's what I thought of when reading your question.

After writing that, I checked help xtreg postestimation and saw there's a forecast, so I went to help forecast and saw there's an example using xtreg, fe. I hope that helps.