[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"V. Sarafidis" <vs242@cam.ac.uk> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
st: nocons and Arellano-Bond test in xtabond2 command |

Date |
04 Oct 2005 19:20:55 +0100 |

Hello,

I would greatly appreciate any help on this issue: I am using a large balanced panel with no missing data, T=12. I am running the same model twice using the command xtabond2 with the only difference being the inclusion of a constant in the levels equations or not. This is implemented by typing "noconstant" in the options.

The constant is statistically significant and in fact the value of the autoregressive coefficient is very sensitive to this. In particular, when I include a constant, the estimated lagged dependent variable coefficient is equal to .5870835 and when I don't include a constant the value equals .7270153. Notice that in the fixed effects model the LDV coefficient equals .5742592.

So running system GMM with a constant seems to provide only minor "improvement" on the LDV coefficient. Normally, the constant is included in the levels equations if there are individual-specific fixed effects, but in this case, and although it appears to be statistically significant, it seems that it does not "work beneficially" in the loose meaning of the words. What would you think on this?

In addition, I get the following results on the Arellano-Bond AR tests for the first-differenced disturbance.

Model with a constant:

Arellano-Bond test for AR(1) in first differences: z = . Pr > z = . Arellano-Bond test for AR(2) in first differences: z = -4.92 Pr > z = 0.000 Arellano-Bond test for AR(3) in first differences: z = -0.13 Pr > z = 0.895

Model without a constant:

Arellano-Bond test for AR(1) in first differences: z = -17.45 Pr > z = 0.000 Arellano-Bond test for AR(2) in first differences: z = -4.00 Pr > z = 0.000 Arellano-Bond test for AR(3) in first differences: z = 0.02 Pr > z = 0.981

So in the model that includes a constant, I get a "." value for the AR(1) test. What does this mean?

Many thanks in advance

Vasilis

*

* 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/

**References**:**st: Re: AR1***From:*"Michael Blasnik" <michael.blasnik@verizon.net>

- Prev by Date:
**st: combining pweights and aweights** - Next by Date:
**st: RE: Neural Networks** - Previous by thread:
**st: Re: AR1** - Next by thread:
**Re: st: AR1** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |