issues using xtabond -artests() cannot exceed the number of time periods-

[marcos gomez]

Hello,

I have a cross-section of thousand of counties observed at time t=1991, 2000 and 2010, and I'm trying to estimate the following econometric growth model
Ln(county_it) = rho Ln(county_it-1) + beta X_it + theta_i + theta_t
where
- X: interest and control county variables (education, expenditure, unemployment, etc.)
- theta_i: is the city fixed effect
- theta_t: is the time fixed effect.

I want to regress the model via Arellano and Bond, so, in STATA, I have tried the following commands

 xtset ibge ano
xtabond lpop_total L.lpop_total $basico
xtabond lpop_total $basico
xtabond lpop_total lpop_total_1 $basico

where (lpop_total_1 = l.lpop_total) and $basico is a sequence of time- and county-independent control variables.
The variables ano is the year 1991, 2000 and 2010, and ibge is the city code.
In all the three situations, STATA gives the same error message:

artests() cannot exceed the number of time periods
r(498);

What's the problem. Is the panel too short, or am I missing something?

Thank you!

[marcos gomez]

Some extra information that might help

xtset ibge ano
       panel variable:  ibge (strongly balanced)
        time variable:  ano, 1 to 3
                delta:  1 unit

xtabond lpop_total L.lpop_total $basico
note: L.lpop_total dropped because of collinearity
artests() cannot exceed the number of time periods
r(498);