Questions about the Arellano Bond estimator.

[Julia Hou]

Hi,

I am running a dynamic panel model using two-step Arrelano and Bond (1991). I have two questions.

(1) One of my regressions returns 0.972 as the Hansen test P value. Does it a concern? Because I've seen the optimal Hansen test value should between 0.2 and 0.3.
(2) Since my estimator is the two-step difference GMM, will people criticize me by not using the two-step system GMM? I actually tried the two-step system GMM, however, all the significance go away.

Thank you in advance for your help.

[Julia Hou]

Anybody? Any ideas? Thank you

[Jesse Wursten]

Hi Julia, in my experience, bumping a post is more likely to reduce the odds of getting an interesting answer than to increase it, especially if it's only been 2 hours since you posted.

[Julia Hou]

Hi Jesse,
Thank you for your advice. I am new here and do not know those rules.

[Richard Williams]

I would wait to bump until you are off the front page, and then try to provide some new information. Getting a nice friend to post something like "Yes, I was interested in this too, did you get an answer?" is a nice way to keep it alive without seeming obnoxious! Sometimes the right people aren't looking when you first post so I don't think a 2nd try after a reasonable amount of time is bad. But nobody should expect an answer in only 2 hours.

I once posted something at a terrible time -- a holiday weekend, or something like that. So, a few days later, I posted "Now that all the experts in this field have enjoyed a richly deserved weekend off, I was wondering..." I finally got a very good answer.

[Eric de Souza]

You should have a look at the paper (later published in the Oxford Bulletin for Economics and Statistics 2009) by David Roodman. In this paper, entitled "A Note on the Theme of Too Many Instruments", he makes a point that too many instruments vitiates the test. "It is easy in such simulations to produce J statistics with implausibly perfect p-values of 1.000." Since you provide no details about your estimation results, it is impossible to say much more.