Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: xtabond2 - Post estimation interpretation


From   Davide Mare <[email protected]>
To   [email protected]
Subject   Re: st: xtabond2 - Post estimation interpretation
Date   Tue, 29 Jan 2013 15:12:38 +0000

Dear Murod,

I am trying to work out the problem I have with the Sargan-Hansen
test. A few things come up to my mind:
1) See if the problem is related with error cross-sectional dependence
(CSD). To this end, following Sarafidis and Wansbeek (2012), I could
control for fixed time effects using dummy variables
2) Since I am using the Granger causality technique with 3 lags in the
dependent variable, I can use l(2/4) in the instrumental variables.

Thank you for your help.

Best,

Davide

2013/1/29 Murod Aliyev <[email protected]>:
> Dear Davide
>
> 1) I think this depends on the context of your work, and on the specification of your model. As far as it concerns xtabond2, you can include any variable.
>
> 2) You are right with your AR test results. Usually you expect AR(1) in differences to be present, and this is fine for the estimation method. You should mainly look at AR(2) in differences which is important, there should be no AR(2). If there is, you should go deeper and deeper with lags when specifying gmm instrument set.
> I think Sargan/Hansen test is instrument validity test, which means they test exogeneity of instruments, not their relevance (weakness/strength).
>
> Hope this helps.
>
> Best wishes,
> Murod
>
>
> ________________________________
> From: Davide Mare <[email protected]>
> To: [email protected]
> Sent: Monday, 28 January 2013, 18:10
> Subject: st: xtabond2 - Post estimation interpretation
>
> Dear all,
>
> I am struggling to interpret the results I get after using the
> xtabond2 command. I have two main questions:
>
> 1) Does it make sense to use macroeconomics variables in a cross
> country regression explicitly in the specification before the comma
> (e.g. unemployment to control for the possible effect of labour
> market)?
> 2) I am not sure on the interpretation of the results I get from the
> model I am implementing, specifically on the validity of the
> instruments. Find below the main statistics:
>
> Arellano-Bond test for AR(1) in first differences: z =  -1.67  Pr > z =  0.095
> Arellano-Bond test for AR(2) in first differences: z =  0.11  Pr > z =  0.916
>
> Sargan test of overid. restrictions: chi2(98)  = 289.36  Prob > chi2 =  0.000
> Hansen test of overid. restrictions: chi2(98)  = 188.54  Prob > chi2 =  0.000
>
> Difference-in-Hansen tests of exogeneity of instrument subsets:
> Hansen test excluding group:    chi2(88)  = 173.65  Prob > chi2 =  0.000
> Difference (null H = exogenous): chi2(10)  =  14.90  Prob > chi2 =  0.136
>
> If my understanding is correct, I have no serial correlation in the
> first order errors, second-order GMM residual serial correlation and
> the estimation fails to satisfy the Sargan/Hansen test statistics of
> overidentifying restrictions. Hence, the instrumental variables I am
> using in the estimation are weak.
>
> Thank you in advance for your help.
>
> Davide
> *
> *  For searches and help try:
> *  http://www.stata.com/help.cgi?search
> *  http://www.stata.com/support/faqs/resources/statalist-faq/
> *  http://www.ats.ucla.edu/stat/stata/
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/

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


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index