Hello all,

I was curious if someone code help me with Stata code in instrumenting a lagged dependent variable with prior values of itself. I have seen papers do this using Anderson and Hsiao (1981) as justification. My data is of a structure that it makes using any of the canned packages for dynamic panel models not feasible, so I am wondering how to do this manually.

Would it be appropriate to do something like so?

regress Δy(t-1) y(t-2) x1 x2 x3
predict p

regress y p x1 x2 x3 i.firm i.year


I do not think my mock example is correct, but I would greatly appreciate it if someone could help me out. Thanks!