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: XTIVREG2: Are my instruments proper?


From   Cameron McIntosh <[email protected]>
To   STATA LIST <[email protected]>
Subject   RE: st: XTIVREG2: Are my instruments proper?
Date   Mon, 26 Dec 2011 22:16:07 -0500

Ayesha,

If the year dummies and the endogenous predictor are correlated, this type of finding shouldn't really come as a surprise, and it doesn't speak at all for or against the validity of IVs.  

By the way, there is much more to confirming IVs as "proper" than just conducting statistical tests:

1. Graphical methods will tell you if a given variable (node) functions as an IV for a purported causal effect.2. Careful use of theory and reasoning will help you select appropriate instruments (surrogates for the experiment you wish you could have done). 3. Statistical tests will tell you about strength of instruments, differences in coefficients across models, and whether IVs are correlated with disturbances (in that case, no, not proper). 

Chan, H., & Kuroki, M. (2010). Using Descendants as Instrumental Variables for the Identification of Direct Causal Effects in Linear SEMs.In Proceedings of the 13th International Conference on Artificial Intelligence and Statistics (AISTATS) 2010, Chia Laguna Resort, Sardinia, Italy. Journal of Machine Learning Research, Workshop & Conference Proceedings, 9, 73-80. http://jmlr.csail.mit.edu/proceedings/papers/v9/chan10a/chan10a.pdf

Brito, C., & Pearl, J. (2002). Generalized instrumental variables. In A. Darwiche and N Friedman (Eds.), Uncertainty in Artificial Intelligence: Proceedings of the Eighteenth Conference (pp. 85-93). San Francisco, CA:  Morgan Kaufmann.http://ftp.cs.ucla.edu/pub/stat_ser/r370.pdf

Leamer, E.E. (2010). Tantalus on the road to asymptopia. Journal of Economic Perspectives, 24(2),31-46.http://www.anderson.ucla.edu/faculty/edward.leamer/selected_research/Tantalus%20by%20Leamer.pdf

Hahn, J., & Hausman, J. (2005). Estimation with Valid and Invalid Instruments. Annales d’économie et de statistique, 79/80, 25-57.http://economics.mit.edu/files/5611

Hahn, J., & Hausman, J. (2002). A New Specification Test for the Validity of Instrumental Variables. Econometrica, 70(1), 163-189. 

Abrevaya, J., Hausman, J.A., & Khan, S. (2010). Testing for Causal Effects in a Generalized Regression Model With Endogenous Regressors. Econometrica, 78(6), 2043–2061.

Hausman, J.A. (1983). Specification and estimation of simultaneous equation models. In Z. Griliches & Intriligator, M.D. (Ed.), Handbook  of  Econometrics (vol I, pp. 391-448). North-Holland Publishing Company.http://pria.uran.ru/ebooks/Unsorted/07%20Hausman%20-%20Specification%20and%20Estimation%20of%20Simultaneous%20Equation%20Models.pdf

Murray, M.P. (2006). Avoiding Invalid Instruments and Coping with Weak Instruments. Journal of Economic Perspectives, 20(4), 111-132.http://www.eui.eu/Personal/Guiso/Courses/Econometrics/Murray_IV_jep_06.pdf

Best,

Cam

> Date: Mon, 26 Dec 2011 17:23:01 -0700
> Subject: st: XTIVREG2: Are my instruments proper?
> From: [email protected]
> To: [email protected]
> CC: [email protected]
> 
> Dear Statalist,
> 
> Happy Holidays!
> 
> I have two sets of results below. In the first fe regression, the main
> endogenous variable "wlggluwmktsh" is a significant predictor and the
> instruments are acceptable based on the 3 tests reported by xtivreg2.
> 
> In the second regression, I add year dummies. Now, the endogenous variable
> is NOT significant and the 3 tests for valid/strong instruments are still
> met.
> 
> I'm not clear why the regressor becomes insignificant in the second
> regression. Isn't it the purpose of the first IV regression to account for
> omitted variable issues?
> 
> Thanks in advance for your guidance.
> 
> Best wishes,
> 
> Ayesha
> Ayesha Malhotra
> Assistant Professor, Strategy & Global Management
> Haskayne School of Business, University of Calgary
> 
> FIRST REGRESSION
> . xtivreg2 wuwmktsh wlgrivalussh wlgrivalusshsq lgintmarg lglogassetsmd
> lgsigmrg wlgassetgr lgassetdiversity wlgroa lgmnyctr (wlggluwmktsh
> > wlgciussh=wlgsigfordep l.wlggluwmktsh l.lgciussh), fe cluster(entity)
> Warning - singleton groups detected.  6 observation(s) not used.
> 
> FIXED EFFECTS ESTIMATION
> ------------------------
> Number of groups =        77                    Obs per group: miN=         2
>                                                                avg =     
> 11.2
>                                                                max =      
>  22
> 
> IV (2SLS) estimation
> --------------------
> 
> Estimates efficient for homoskedasticity only
> Statistics robust to heteroskedasticity and clustering on entity
> 
> Number of clusters (entity) = 77                      Number of obs = 859
>                                                       F( 11,    76) =  7.43
>                                                       Prob > F      =  0.0000
> Total (centered) SS     =  1417.794625                Centered R2   =  
> 0.6587
> Total (uncentered) SS   =  1417.794625                Uncentered R2 =  
> 0.6587
> Residual SS             =  483.9461722                Root MSE      =   
> .7867
> 
> ------------------------------------------------------------------------------
>              |               Robust
>     wuwmktsh |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
> Interval]
> -------------+-------------------------------------------------------------wlggluwmktsh
> |   1.090995    .086687    12.59   0.000     .9210911   260898
> Other results removed...
> 
> ------------------------------------------------------------------------------
> Underidentification test (Kleibergen-Paap rk LM statistic):  10.050
>                                                    Chi-sq(2) P-val = 0.0066
> ------------------------------------------------------------------------------
> Weak identification test (Kleibergen-Paap rk Wald F statistic):    14.625
> Stock-Yogo weak ID test critical values: 10% maximal IV size       13.43
>                                          15% maximal IV size        8.18
>                                          20% maximal IV size        6.40
>                                          25% maximal IV size         5.45
> 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):    0.980
>                                                    Chi-sq(1) P-val = 0.3221
> ------------------------------------------------------------------------------
> Instrumented:         wlggluwmktsh wlgciussh
> Included instruments: wlgrivalussh wlgrivalusshsq lgintmarg lglogassetsmd
>                       lgsigmrg wlgassetgr lgassetdiversity wlgroa lgmnyctr
> Excluded instruments: wlgsigfordep L.wlggluwmktsh L.lgciussh
> 
> SECOND REGRESSION - HAS TIME DUMMIES
> 
>  xtivreg2 wuwmktsh wlgrivalussh wlgrivalusshsq lgintmarg lglogassetsmd
> lgsigmrg wlgassetgr lgassetdiversity wlgroa lgmnyctr yrdum1-yrdum23
> >  (wlggluwmktsh wlgciussh=wlgsigfordep l.wlggluwmktsh l.lgciussh), fe
> cluster(entity)
> Warning - singleton groups detected.  6 observation(s) not used.
> Warning - collinearities detected
> Vars dropped:  yrdum1 yrdum23
> 
> FIXED EFFECTS ESTIMATION
> ------------------------
> Number of groups =        77                    Obs per group: min =      
>   2
>                                                                avg =     
> 11.2
>                                                                max =      
>  22
> 
> IV (2SLS) estimation
> --------------------
> 
> Estimates efficient for homoskedasticity only
> Statistics robust to heteroskedasticity and clustering on entity
> 
> Number of clusters (entity) = 77                      Number of obs =     
> 859
>                                                       F( 32,    76) = 
> 6476.04
>                                                       Prob > F      =  
> 0.0000
> Total (centered) SS     =  1417.794625                Centered R2   =  
> 0.8120
> Total (uncentered) SS   =  1417.794625                Uncentered R2 =  
> 0.8120
> Residual SS             =  266.5785924                Root MSE      =   
> .5839
> 
> ---------------------------------------------------------------------------
>              |               Robust
>     wuwmktsh |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
> Interval]
> -------------+----------------------------------------------------------------
> wlggluwmktsh |   .0616593   .1000854     0.62   0.538  .1345045    .2578231
> 
> Some other results removed....
>       yrdum2 |  -30.33265   2.549937   -11.90   0.000    -35.33044  
> -25.33487
>       yrdum3 |  -30.94688   2.592648   -11.94   0.000    -36.02838  
> -25.86539
>       yrdum4 |  -29.49283   2.483674   -11.87   0.000    -34.36075  
> -24.62492
>       yrdum5 |  -29.74965   2.509561   -11.85   0.000     -34.6683  
> -24.83101
>       yrdum6 |  -29.20278   2.466175   -11.84   0.000    -34.03639  
> -24.36916
>       yrdum7 |   -28.2574   2.399218   -11.78   0.000    -32.95978  
> -23.55502
>       yrdum8 |  -25.69739   2.217672   -11.59   0.000    -30.04395  
> -21.35083
>       yrdum9 |  -21.18458   1.941362   -10.91   0.000    -24.98958  
> -17.37958
>      yrdum10 |  -21.14397   1.959225   -10.79   0.000    -24.98398  
> -17.30396
>      yrdum11 |  -19.71581   2.089331    -9.44   0.000    -23.81082   
> -15.6208
>      yrdum12 |  -12.39363   1.355772    -9.14   0.000    -15.05089  
> -9.736362
>      yrdum13 |  -9.591614   1.271486    -7.54   0.000    -12.08368  
> -7.099548
>      yrdum14 |  -7.524386   .8318182    -9.05   0.000     -9.15472  
> -5.894053
>      yrdum15 |  -7.845728   .8834701    -8.88   0.000    -9.577297  
> -6.114158
>      yrdum16 |  -8.041389   .9002762    -8.93   0.000    -9.805898   
> -6.27688
>      yrdum17 |  -8.088559   .9122472    -8.87   0.000     -9.87653  
> -6.300587
>      yrdum18 |  -10.25239   1.067351    -9.61   0.000    -12.34436  
> -8.160416
>      yrdum19 |  -11.51903   1.186703    -9.71   0.000    -13.84492  
> -9.193134
>      yrdum20 |  -9.931895   1.189907    -8.35   0.000    -12.26407  
> -7.599721
>      yrdum21 |  -4.975506   .7196383    -6.91   0.000    -6.385971  
> -3.565041
>      yrdum22 |  -4.745704   .6567907    -7.23   0.000    -6.032991  
> -3.458418
> ------------------------------------------------------------------------------
> Underidentification test (Kleibergen-Paap rk LM statistic):   11.455
>                                                    Chi-sq(2) P-val = 0.0033
> ------------------------------------------------------------------------------
> Weak identification test (Kleibergen-Paap rk Wald F statistic):  36.383
> Stock-Yogo weak ID test critical values: 10% maximal IV size      13.43
>                                          15% maximal IV size       8.18
>                                          20% maximal IV size        6.40
>                                          25% maximal IV size        5.45
> 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): 2.155
>                                                    Chi-sq(1) P-val = 0.1421
> ------------------------------------------------------------------------------
> Instrumented:         wlggluwmktsh wlgciussh
> Included instruments: wlgrivalussh wlgrivalusshsq lgintmarg lglogassetsmd
>                       lgsigmrg wlgassetgr lgassetdiversity wlgroa lgmnyctr
>                       yrdum2 yrdum3 yrdum4 yrdum5 yrdum6 yrdum7 yrdum8 yrdum9
>                       yrdum10 yrdum11 yrdum12 yrdum13 yrdum14 yrdum15 yrdum16
>                       yrdum17 yrdum18 yrdum19 yrdum20 yrdum21 yrdum22
> Excluded instruments: wlgsigfordep L.wlggluwmktsh L.lgciussh
> Dropped collinear:    yrdum1 yrdum23
> ------------------------------------------------------------------------------
> 
> .
> 
> 
> *
> *   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