Tewodaj Mogues <tmogues@students.wisc.edu> asked about a problem with
-xtabond- that occurs with lags of predermined variables. We have
researched the issue and concluded that he is correct. -xtabond- is using
levels lagged one or more periods as instruments for predetermined
variables, regardless of the number of lags of the predermined variables
included in the model. -xtabond- should only be using levels lagged p+1 or
more times, where p is the number of lags of the predetermined variables in
the model.
This problem implies that the current -xtabond- produces inconsistent
estimates of the parameters in models with lags of the predetermined
variables. For instance, estimates from the model
. xtabond y , pre(x , lag(1, .))
are not consistent.
This problem does not arise when there are no lags of the predetermined
variables. For instance, the current -xtabond- produces consistent
estimates from the model
. xtabond y , pre(x)
-xtabond- also produces consistent estimates for the models
. xtabond y , lags(2)
. xtabond y l(0/1).x1 , pre(x2)
. xtabond y l(0/1).x1 , pre(x2, lag(0, 1))
The problem will be fixed in an upcoming ado update that we expect to have
in distribution early next week.
--David
drukker@stata.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/