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: RE: Blown up IV coefficient


From   Shikha Sinha <[email protected]>
To   [email protected]
Subject   Re: st: RE: Blown up IV coefficient
Date   Wed, 9 Mar 2011 13:52:47 -0500

Now, I am running the model without the fixed effect. First stage
looks good, rank is significant and has the expected sign. OLS
coefficient is (-1.868918), while IV coeff is still blown up
(-19.42056), in spite of a strong IV. What could be the reason for
this? Also, sometimes, sign changes in IV estimation, whats the
economteric argument for the sign change?

Thanks,
Shikha

reg roaddum rank  ims_hab_pop, robust

Linear regression                                      Number of obs =     266
                                                       F(  2,   263) =   17.60
                                                       Prob > F      =  0.0000
                                                       R-squared     =  0.0705
                                                       Root MSE      =  .48247

------------------------------------------------------------------------------
             |               Robust
     roaddum |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        rank |   .0055897   .0009531     5.86   0.000     .0037131    .0074663
 ims_hab_pop |  -.0001736   .0005512    -0.31   0.753     -.001259    .0009117
       _cons |   .4812974   .5810408     0.83   0.408    -.6627865    1.625381
------------------------------------------------------------------------------

. reg avgdalprice roaddum ims_hab_pop, robust

Linear regression                                      Number of obs =     251
                                                       F(  2,   248) =    1.02
                                                       Prob > F      =  0.3609
                                                       R-squared     =  0.0085
                                                       Root MSE      =  12.093

------------------------------------------------------------------------------
             |               Robust
 avgdalprice |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
     roaddum |  -1.868918    1.55255    -1.20   0.230    -4.926783    1.188947
 ims_hab_pop |   .0117034   .0144311     0.81   0.418    -.0167197    .0401265
       _cons |   116.8683   15.27632     7.65   0.000     86.78041    146.9561
------------------------------------------------------------------------------

. ivreg2 avgdalprice ims_hab_pop (roaddum=rank), robust

IV (2SLS) estimation
--------------------

Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity

                                                      Number of obs =      251
                                                      F(  2,   248) =     4.33
                                                      Prob > F      =   0.0142
Total (centered) SS     =  36580.18327                Centered R2   =  -0.5124
Total (uncentered) SS   =      4167930                Uncentered R2 =   0.9867
Residual SS             =  55324.45189                Root MSE      =    14.85

------------------------------------------------------------------------------
             |               Robust
 avgdalprice |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
     roaddum |  -19.42056   6.622308    -2.93   0.003    -32.40005   -6.441074
 ims_hab_pop |    .016525   .0172844     0.96   0.339    -.0173519    .0504019
       _cons |    121.521   18.02514     6.74   0.000     86.19241    156.8496
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic):             18.413
                                                   Chi-sq(1) P-val =    0.0000
------------------------------------------------------------------------------
Weak identification test (Cragg-Donald Wald F statistic):               17.930
                         (Kleibergen-Paap rk Wald F statistic):         32.974
Stock-Yogo weak ID test critical values: 10% maximal IV size             16.38
                                         15% maximal IV size              8.96
                                         20% maximal IV size              6.66
                                         25% maximal IV size              5.53
Source: Stock-Yogo (2005).  Reproduced by permission.
NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors.
------------------------------------------------------------------------------
Hansen J statistic (overidentification test of all instruments):         0.000
                                                 (equation exactly identified)
------------------------------------------------------------------------------
Instrumented:         roaddum
Included instruments: ims_hab_pop
Excluded instruments: rank
------------------------------------------------------------------------------



On Wed, Mar 9, 2011 at 1:32 PM, Wooldridge, Jeffrey <[email protected]> wrote:
> No, sorry. You of course had it right. That's what I get for copying and pasting and not editing very well.
>
> I suspected rank would not be statistically significant. Seems hard to justify IV in that case.
>
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of Shikha Sinha
> Sent: Wednesday, March 09, 2011 1:25 PM
> To: [email protected]
> Subject: Re: st: RE: Blown up IV coefficient
>
> Jeff, thanks.
>
> Should the reduced-form also include the dependent variable
> (agr_share) as mentioned by you?
>  areg roaddum rank  ims_hab_pop, absorb(dcode) robust
>
> Sign on rank is expected, but it is statistically insignificant...
>
> Thanks,
> Shikha
>
> On Wed, Mar 9, 2011 at 12:24 PM, Wooldridge, Jeffrey <[email protected]> wrote:
>> For the IV estimates to be at all convincing, first run the reduced
>> form,
>>
>> areg roaddum rank agr_share  ims_hab_pop, absorb(dcode) robust
>>
>> and look at the sign of the coefficient on rank and its t statistic.
>> Hopefully you have a predicted sign for the effect of rank on roaddum.
>> If the sign is what you expect, is rank strongly statistically
>> significant? The statistics for weak identification provide this
>> information, too, but you should estimate the reduced form, anyway. It
>> is easier to see what is happening.
>>
>> My guess is the instrument is weak. Or, the IV could actually be
>> endogenous (which cannot be tested, unfortunately).
>>
>> Jeff
>>
>> -----Original Message-----
>> From: [email protected]
>> [mailto:[email protected]] On Behalf Of Shikha Sinha
>> Sent: Wednesday, March 09, 2011 12:17 PM
>> To: [email protected]
>> Subject: st: Blown up IV coefficient
>>
>> Dear all,
>>
>> I am estimating an instrumental variable model by using -ivreg2 in
>> Stata. The IV estimates are way more than OLS estimate (4.006567).
>> Could you advise what could be possible reasons for such a blown up IV
>> coeff (49.97424 )? or my interpretation of coeff is incorrect.
>>
>>  areg agr_share roaddum ims_hab_pop, robust absorb(dcode)
>>
>> Linear regression, absorbing indicators                Number of obs =
>> 266
>>                                                       F(  2,   250) =
>> 0.66
>>                                                       Prob > F      =
>> 0.5182
>>                                                       R-squared     =
>> 0.0680
>>                                                       Adj R-squared =
>> 0.0121
>>                                                       Root MSE      =
>> 26.294
>>
>> ------------------------------------------------------------------------
>> ------
>>             |               Robust
>>   agr_share |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
>> Interval]
>> -------------+----------------------------------------------------------
>> ------
>>     roaddum |   4.006567   3.572553     1.12   0.263     -3.02957
>> 11.0427
>>  ims_hab_pop |    .001926   .0344155     0.06   0.955    -.0658552
>> .0697071
>>       _cons |   34.78717   36.38659     0.96   0.340    -36.87615
>> 106.4505
>> -------------+----------------------------------------------------------
>> ------
>>       dcode |   absorbed                                      (14
>> categories)
>>
>> . xi: ivreg2 agr_share  ims_hab_pop i.dcode (roaddum=rank), robust
>> i.dcode           _Idcode_1-14        (_Idcode_1 for dcode==UP09
>> omitted)
>>
>> IV (2SLS) estimation
>> --------------------
>>
>> Estimates efficient for homoskedasticity only
>> Statistics robust to heteroskedasticity
>>
>>                                                      Number of obs =
>> 266
>>                                                      F( 15,   250) =
>> 0.88
>>                                                      Prob > F      =
>> 0.5891
>> Total (centered) SS     =  185457.6241                Centered R2   =
>> -0.5185
>> Total (uncentered) SS   =       590824                Uncentered R2 =
>> 0.5233
>> Residual SS             =   281625.775                Root MSE      =
>> 32.54
>>
>> ------------------------------------------------------------------------
>> ------
>>             |               Robust
>>   agr_share |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
>> Interval]
>> -------------+----------------------------------------------------------
>> ------
>>     roaddum |   49.97424   35.09856     1.42   0.154    -18.81768
>> 118.7662
>>  ims_hab_pop |  -.0354515   .0511597    -0.69   0.488    -.1357226
>> .0648196
>>   _Idcode_2 |   -39.0505   25.58464    -1.53   0.127    -89.19546
>> 11.09446
>>   _Idcode_3 |  -41.62481   26.03165    -1.60   0.110    -92.64591
>> 9.396298
>>   _Idcode_4 |  -16.54727   20.19038    -0.82   0.412    -56.11968
>> 23.02513
>>   _Idcode_5 |  -41.45807   29.51111    -1.40   0.160    -99.29878
>> 16.38264
>>   _Idcode_6 |  -32.93124   16.32568    -2.02   0.044      -64.929
>> -.9334909
>>   _Idcode_7 |  -8.880318   13.20529    -0.67   0.501    -34.76221
>> 17.00157
>>   _Idcode_8 |  -44.38007   24.23395    -1.83   0.067    -91.87774
>> 3.117595
>>   _Idcode_9 |  -44.81575   24.87717    -1.80   0.072    -93.57409
>> 3.942605
>>  _Idcode_10 |  -21.14303   12.84912    -1.65   0.100    -46.32684
>> 4.040777
>>  _Idcode_11 |  -22.56438   9.922245    -2.27   0.023    -42.01162
>> -3.117138
>>  _Idcode_12 |  -12.55903   12.52187    -1.00   0.316    -37.10144
>> 11.98338
>>  _Idcode_13 |  -14.25003   9.233117    -1.54   0.123    -32.34661
>> 3.846547
>>  _Idcode_14 |  -26.27491   11.42556    -2.30   0.021    -48.66859
>> -3.881229
>>       _cons |   74.88709   53.28162     1.41   0.160    -29.54296
>> 179.3171
>> ------------------------------------------------------------------------
>> ------
>> Underidentification test (Kleibergen-Paap rk LM statistic):
>> 1.825
>>                                                   Chi-sq(1) P-val =
>> 0.1767
>> ------------------------------------------------------------------------
>> ------
>> Weak identification test (Cragg-Donald Wald F statistic):
>> 2.977
>>                         (Kleibergen-Paap rk Wald F statistic):
>> 2.488
>> Stock-Yogo weak ID test critical values: 10% maximal IV size
>> 16.38
>>                                         15% maximal IV size
>> 8.96
>>                                         20% maximal IV size
>> 6.66
>>                                         25% maximal IV size
>> 5.53
>> Source: Stock-Yogo (2005).  Reproduced by permission.
>> NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors.
>> ------------------------------------------------------------------------
>> ------
>> Hansen J statistic (overidentification test of all instruments):
>> 0.000
>>                                                 (equation exactly
>> identified)
>> ------------------------------------------------------------------------
>> ------
>> Instrumented:         roaddum
>> Included instruments: ims_hab_pop _Idcode_2 _Idcode_3 _Idcode_4
>> _Idcode_5
>>                      _Idcode_6 _Idcode_7 _Idcode_8 _Idcode_9 _Idcode_10
>>                      _Idcode_11 _Idcode_12 _Idcode_13 _Idcode_14
>> Excluded instruments: rank
>> ------------------------------------------------------------------------
>> ------
>>
>> Thanks,
>> Shikha
>> *
>> *   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/
>>
>> *
>> *   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/
>>
>
> *
> *   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/
>
> *
> *   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/
>

*
*   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/


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