xtabond2 lag sub-option

[Joshua Evanss]

Dear Statalist,

I am examining quarterly data for an unbalanced panel of 11,000+ banks from 1996:Q1 to 2016:Q4 using Stata/IC. I am using a system gmm model through xtabond2. My regression code is as follows:

xtset id dateq

xtabond2 LCCF l.LCCF T1RAT SIZE RISK ROE MS MNA COMP POP INCG GDPG UNEM FFR q2_1997- q4_2016, ///
gmm(l.LCCF T1RAT SIZE RISK ROE MS MNA, lag(1 .) collapse) iv(COMP POP INCG GDPG UNEM FFR q2_1997- q4_2016) ///
twostep robust small nodiffsargan

For completeness, variables q2_1997- q4_2016 are dummies for each quarter to introduce time fixed effects. I have included all bank-level variables as endogenous in the -gmm- option, and all other variables as instruments in the -iv- option.

I want to try the model using the following lag specifications:

• lag(1 .)
• lag(2 .)
• lag(3 .)

But when doing so, as in the above code, I get the following error:

                     J():  3900  unable to allocate real <tmp>[1299463,588]
         xtabond2_mata():     -  function returned error
                 <istmt>:     -  function returned error

If I remove the -collapse- suboption in the -gmm- option, with the following code:

xtset id dateq

xtabond2 LCCF l.LCCF T1RAT SIZE RISK ROE MS MNA COMP POP INCG GDPG UNEM FFR q2_1997- q4_2016, ///
gmm(l.LCCF T1RAT SIZE RISK ROE MS MNA, lag(1 .)) iv(COMP POP INCG GDPG UNEM FFR q2_1997- q4_2016) ///
twostep robust small nodiffsargan

I get the following error:

                     J():  3900  unable to allocate real <tmp>[1299463,588]
         xtabond2_mata():     -  function returned error
                 <istmt>:     -  function returned error

Why do these errors occur? xtabond2 seems very limited the detail of its error reporting.

[Sebastian Kripfganz]

You might want to try the xtdpdgmm command as an alternative to xtabond2:

• XTDPDGMM: new Stata command for GMM estimation of linear (dynamic) panel data models
• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.