Estimate panel data eq w. lag value and timedummy

[DesrAw]

Hey everyone

I've gotten some pre exam material, where I'm trying to reproduce an equation from an article about worker trainings effect on scrap rate.
The data set it about who get money from the goverment and when.
It is a panel data set, that has data from three years of a number of companies. Year: 1987-1989.
GG= (0; no money granted, 1; money granted in 1988, 2; money granted in 1989)
What I'm trying to estimate of "dlog(scrap rate) c dlog(training) dlog(training(-1) year=1989" filter: "1987 1989 if GG<2"
I get the error "near singular matrix". There has to be a C, and I'm pretty sure the filter is right.
If you understand what I'm trying to estimate, from the equation above, is there another way I can estimate it and it will work?

I know this is a brief, but the exam is in 10 hours ;).

I hope for a good and fast answer. I have the workfile if that helps :)

Regards
DesrAw