Out of sample prediction in fixed effects panel model with AUTOREGRESSIVE term

[Nooh Ahmad]

Hi
I have a panel model that I estimated as follows:
areg N_ratio L1.N_ratio L1.M_ratio, abs(firm)

I have the explanatory variables for the next 12 quarters, to predict N_ratio for out of sample, here is what I did:

global end_date = 150
predict fe_x, d
bys firm: egen fe = max(fe_x)

mat beta=e(b)
svmat double beta

global b_N = beta1[1]
global b_M = beta2[1]
global b_con = beta3[1]

drop beta*

bys firm (date): N_ratio_predicted = $b_N*N_Ratio[_n-1] + $b_M*M_Ratio[_n-1] + $b_con + fe if date > $end_date

The problem is I get the forecast value of only ONE period ahead. That is apparently because of the autoregressive term. If I remove the AR term, I am able to generate the forecast....What am I doing wrong here and how to fix
Thanks

[Nooh Ahmad]

Bump

[George Ford]

the way you are doing it looks back for values of the lag and can only find the last value. so you get one forecast value.

you'll need to use forecast (help forecast).