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: RE: xtivreg
From
"Schaffer, Mark E" <[email protected]>
To
<[email protected]>
Subject
RE: st: RE: xtivreg
Date
Fri, 11 Feb 2011 13:38:16 -0000
Conde,
It looks like you are using official xtivreg to estimate your equation using GLS, i.e., random effects.
There is a typo in your command line for xtivreg2, but I am sure you are not using GLS because the public version of xtivreg2 doesn't support random effects. Only FE and FD are supported in the public version.
If you try estimating using official xtivreg and the fixed effects or first differences estimator, you will probably get the same results you were getting with xtivreg2.
Moreover, the output of xtreg applied to your instrument h reports that sigma_u>0 but sigma_e=0. This says that there is no variation over time in h. The countries have different "fixed effects" (u_i varies over countries) but there is no time series variation (e_it is always zero).
My guess is that in the GLS (random effects) estimation, all the identifying variation in your instrument is cross-section (between) variation. This is why the GLS estimator, but not the FE estimator, can give you an IV estimate of the coefficient of interest.
--Mark
> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]] On Behalf Of
> Lanciné Condé
> Sent: 11 February 2011 12:34
> To: [email protected]
> Subject: Re: st: RE: xtivreg
>
> Dear Pr. Schaffer
> Thank you for your first advises. But, I still have the same
> difficulties. As I already said, I want to appreciate the effects of
> the monetary union participation on the economic performances, mesured
> through the growth rate of the real GDP by head (y). the right hand
> variables are the following:
>
>
> - The monetary union dummy (cu) which takes 1 if the country under
> consideration is a monetary union participant (Franc Zone countries,
> plus Rand area countries), 0 otherwise. Taking in account the
> historique relationship between monetary union countries, the umoney
> variable is supposed to be endogenous. So it instrumented by another
> dummy called antcol, which takes 1 when the country under
> consideration have been colonised by the major colonizer country of
> the sub-region, 0 otherwise.
> - muet1, a dummy which takes if the country does not have sea coast;
> - muet2, a dummy which takes 1 if an active monetary union exists in
> the region of the country under consideration;
> - muet3, a dummy which takes 1 if the country has been hit by
> political instabilities or civil wars;
> - x1, a proxy of Balassa-Samuelson effect;
> - x2, the total population growth rate
> - x3, a proxy of the total investissement
> - x4, the inflation rate
> - x5, the central gouvernment spenditure
> - x6, a proxy of the openness
>
> My equation of interest is the following:
> xtivreg d.y (cu=h) muet1 muet2 muet3 d.x1 d.x2 d.x3 d.x4 d.x5
> d.x6, first.
>
> The correlation between variables is not equal to 1 and the instrument
> (h) is varing over time. The implementation of following command -
> xtreg h,fe - return as answere:
>
> xtreg h, fe
> Fixed-effects (within) regression Number of obs
> = 950
> Group variable: country Number of
> groups = 38
> R-sq: within = . Obs per group:
> min = 25
> between = .
> avg = 25.0
> overall = .
> max = 25
>
> F(0,912) = 0.00
> corr(u_i, Xb) = . Prob > F
> = .
> --------------------------------------------------------------
> ----------------
> h | Coef. Std. Err. t P>|t| [95%
> Conf. Interval]
> -------------+------------------------------------------------
> ----------------
> _cons | .6578947 . . .
> . .
> -------------+------------------------------------------------
> ----------------
> sigma_u | .48078291
> sigma_e | 0
> rho | 1 (fraction of variance due to u_i)
> --------------------------------------------------------------
> ----------------
> F test that all u_i=0: F(37, 912) = .
> Prob > F = .
>
> When I implement the xtivreg command, I am using le default option GLS
> and the return is as follow:
>
> xtivreg d.y (cu=h) muet1 muet2 muet3 d.x1 d.x2 d.x3 d.x4 d.x5
> d.x6, first, first
>
> First-stage G2SLS regression
> Number of obs
> = 912
> Wald chi(10)
> = 539
> Prob > chi2
> = 0.0000
> --------------------------------------------------------------
> ----------------
> cu | Coef. Std. Err. z P>|z| [95%
> Conf. Interval]
> -------------+------------------------------------------------
> ----------------
> muet1 | -.0206403 .0276052 -0.75 0.455
> -.0747454 .0334648
> muet2 | .511889 .0442444 11.57 0.000
> .4251716 .5986065
> muet3 | -.0275062 .0350349 -0.79 0.432
> -.0961733 .041161
> |
> x1 |
> D1. | -10.24638 1.991216 -5.15 0.000
> -14.1491 -6.343671
> |
> x2 |
> D1. | -.0026825 .0238177 -0.11 0.910
> -.0493643 .0439993
> |
> x3 |
> D1. | .0000864 .0003503 0.25 0.805
> -.0006001 .000773
> |
> x4 |
> D1. | -1.45e-07 .0000107 -0.01 0.989
> -.0000211 .0000208
> |
> x5 |
> D1. | .0000201 .0049138 0.00 0.997
> -.0096108 .009651
> |
> x6 |
> D1. | -.0007695 .0012583 -0.61 0.541
> -.0032358 .0016967
> |
> h | .5528577 .0284752 19.42 0.000
> .4970473 .6086681
> _cons | -.3905693 .0497611 -7.85 0.000
> -.4880993 -.2930394
> --------------------------------------------------------------
> ----------------
> G2SLS random-effects IV regression Number of obs
> = 912
> Group variable: country Number of
> groups = 38
> R-sq: within = 0.2238 Obs per
> group: min = 24
> between = 0.0063
> avg = 24.0
> overall = 0.2216
> max = 24
> Wald chi2(10)
> = 256.72
> corr(u_i, X) = 0 (assumed) Prob > chi2
> = 0.0000
> --------------------------------------------------------------
> ----------------
> __000003 | Coef. Std. Err. z P>|z| [95%
> Conf. Interval]
> -------------+------------------------------------------------
> ----------------
> cu | -.3224284 2.902101 -0.11 0.912
> -6.010442 5.365585
> muet1 | -.5010761 1.542581 -0.32 0.745
> -3.52448 2.522328
> muet2 | -.1100118 2.807213 -0.04 0.969
> -5.612047 5.392024
> muet3 | .1415018 1.98299 0.07 0.943
> -3.745088 4.028092
> |
> x1|
> D1. | 305.6057 115.0715 2.66 0.008
> 80.06972 531.1418
> |
> x2 |
> D1. | .6515786 1.342214 0.49 0.627
> -1.979112 3.282269
> |
> x3 |
> D1. | .3058186 .0197351 15.50 0.000
> .2671385 .3444987
> |
> x4 |
> D1. | -.000037 .0006026 -0.06 0.951
> -.001218 .001144
> |
> x5 |
> D1. | -.0870872 .276872 -0.31 0.753
> -.6297464 .4555721
> |
> x6 |
> D1. | -.1354545 .0710059 -1.91 0.056
> -.2746234 .0037144
> |
> _cons | .6136019 2.57388 0.24 0.812
> -4.431111 5.658315
> -------------+------------------------------------------------
> ----------------
> sigma_u | 0
> sigma_e | 22.526807
> rho | 0 (fraction of variance due to u_i)
> --------------------------------------------------------------
> ----------------
> Instrumented: cu
> Instruments: muet1 muet2 muet3 D.x1 D.x2 D.x3 D.x4 D.x5 D.x6 h
> --------------------------------------------------------------
> ----------------
>
>
> To got some necessary statistics, the implementation of the command
> below are not work:
>
> xtivreg2 d.y (cu=h) muet1 muet2 muet3 d.x1 d.x2 d.x3 d.x4 d.x5 d.x6,
> first, first
>
> Sincerely
> CONDE
>
> *
> * 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/
>
--
Heriot-Watt University is a Scottish charity
registered under charity number SC000278.
*
* 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/