st: dgmmiv use when two period lags in xtdpd

[Hewan Belay <hewan_belay@yahoo.com>]

Dear Statalist,

I am wondering whether there is an error in the Stata11 manual’s specification of some xtdpd examples—or else I may not be properly understanding the command. My question has to do with the proper instrumentation of the lagged dependent variable, using -dgmmiv-. 

Please compare the regression command in Example 1, which reads

xtdpd L(0/2).n L(0/1).w L(0/2).(k ys) yr1980-yr1984 year, noconstant div(L(0/1).w L(0/2).(k ys) yr1980-yr1984 year) dgmmiv(n)

with the regression command for Example 4:

xtdpd L(0/1).n L(0/2).(w k) yr1980-yr1984 year, div(yr1980-yr1984 year) dgmmiv(n) dgmmiv(L2.(w k), lag(1 .)) lgmmiv(n L1.(w k)) vce(robust) hascons

The first regression uses both n_t-1 and n_t-2 as regressors, whereas the second only uses n_t-1 as regressor. Yet, for both the instrument set based on further lags of the dependent variable is the same, specificied as -dgmmiv(n)-. If the second regression (the one in Example 4) uses all permitted further lags of Y as instruments through the specification  -dgmmiv(n)-, clearly the Example 1 regression is using some non-permissible lags as instruments. If this is thus not an error in the manual, please let me know what I got wrong in this assessment.

A footnote: the default lagrange is such that writing -dgmmiv(n)- is the same as writing -dgmmiv(n, lag(2 .)). But whatever the default, I would think that the two examples above should not be using the same instrument range, as discussed above.

Thanks in advance for your response!
Hewan


      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/