Dynamic Panel Models

[ancia dettox]

Hi,

I try to build model in Stata with xtabond. When I should use option vce(robust)? Or when should I use onestep or twostep?

Firstly, I made xtabond..., vce(robust). Sargan test has good results (p-value greater than 0,2, of course it works without vce(robust) option). Arellano-Bond test also:

|Order | z Prob > z|
| 1 |-4.0583 0.0000 |
| 2 | .71112 0.4770

But when I made xtabobd...., vce(robust) twostep all my independent variables are not significant... Sargan test doesnt work because p-value equals to 1.000 (to many instruments). Arellano-Bond looks like:

Order | z Prob > z|
| 1 |-1.8794 0.0602 |
| 2 | .27934 0.7800

Should I use twostep option if Sargan test and Arellano-Bond are good when I have onestep xtabond?

[Phil Bromiley]

David Roodman did a nice intro to xtabond2 see: https://www.stata-journal.com/articl...article=st0159
He may also have a longer version on-line - http://citeseerx.ist.psu.edu/viewdoc...=rep1&type=pdf