SGMM and unit roots

[gicanzo]

Hi guys,

I am running SGMM (Blundell-Bond) regressions on panel data with lots of unit roots, so I was wondering if it is reasonable to take first differences of the variables and run SGMM on them, so that the level equation used by sgmm would be in reality a first-differenced equation and the difference equation would be second-differenced?
Would this solve my problems with unit roots? I hope my exposition is clear enough.
Thanks for your attention.

Best Regards,
Giuseppe