J Test and Instrument Rank

[andre_6043]

Dear all,

i am currently doing a GMM Regression with 4 potentially endogenous independant variables and 3 exogenous independant variables. The 3 endogenous variables are instrumented with the same three lagged (1 period) variables and one additional instrument.
The question I have concerns the test whether my instruments are weak or not. I have read that I should check the J-statistic and the instrument rank. Unfortunately I do not know how to assess whether the values are good or not. What values of the J-Statistic and the instrument rank suggests good instruments?

Thanks in advance for your advice!

[startz]

Neither of these is really a test for weak instruments. If you have four endogenous variables and four instruments, the rank had better be 4 or GMM doesn't make any sense. The J-statistic can be used to test overidentifying restrictions.

[startz]

More helpfully, you may want to take a look at the EViews 7 beta. It has some helpful diagnostics.

[andre_6043]

Dear all,

i am currently doing a dynamic regression using as instruments dynamic lagged values (@dyn(-2,-5)) of the variable.
Does it makes sense to check Hansen's J-Stat to see whether I use relevant and valid instruments? In case it does make sense,
how do I know how many degrees of freedom my dynamic regression has? Besides, does someone know good literatur about Hansens J-Statistic
(alternatively Sargan Test) for dynamic regressions? Or simply literature about testing instrument validity and relevance in dynamic
regressions?

Thanks in advance for your efforts!

[theologos]

Hello,


There is a nice chapter for Instrumental Variables in Stock and Watson Introduction to Econometrics, 2ed.

Regards