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]
st: Results of overidentification and underidentification test missing
From
Sutirtha Bagchi <[email protected]>
To
[email protected]
Subject
st: Results of overidentification and underidentification test missing
Date
Sat, 21 Sep 2013 17:56:29 -0400
Hello,
I am using the user-written command –xtivreg2— in Stata11 (Stata/SE
11.2 for Windows (32-bit)).
(*! xtivreg2 1.0.13 28Aug2011 *! author mes)
The issue I am facing is that in the Stata output, I find the results
of the Under identification and Weak Identification test missing. In
particular, the Kleibergen-Paap rk LM statistic and associated p-value
and the Kleibergen-Paap rk Wald F statistic are missing. Other test
statistics such as the Hansen J statistic for overidentification and
the Shea partial R2 are present in the output. I can verify that I
have updated Stata and so that alone is unlikely to fix this issue for
me.
Here are details of my data set on municipal pension plans where this comes up.
I have one observation per pension plan per municipality per time
period (decade). For simplicity, let us say, I have 2 pension plans
per municipality for ~ 1,000 municipalities for 3 decades – a total of
2 X 1,000 X 3 or ~ 6,000 observations. I am looking at the effect of
political orientation of the municipality (more specifically, the
independent variable is average Democratic vote share in mayoral
elections held in the last decade) on a measure of funding for the
pension plans offered by that municipality. However, I am concerned
about the possible endogeneity of the independent variable and I
therefore use demographic characteristics (percent of the population
that is self-employed and percent of the population that has a
disability) as instruments for the independent variable. As it turns
out, Democratic vote share goes up when the percent of the population
that is self-employed goes down or when the percent of the population
that has a disability goes up.
The Stata command I use is:
xi: xtivreg2 wmeanactfundratio_emplgrp2 (average_share_dems_votes7 =
pctslfemplydownbiznotincp pctpop16to64wdisability) i.currentdecade, fe
gmm2s first cluster(county)
where wmeanactfundratio_emplgrp2 = Mean funding ratio of pension plan
offered by a municipality for a particular employee group (with the
mean being taken over a decade);
average_share_dems_votes7 = Average Democratic vote share for mayoral
races held in the last decade;
pctslfemployedownbiznotincp = Percent of the population that is self-employed;
pctpop16to64wdisability = Percent of the population between 16 to 64
that has a disability;
i.currentdecade is a set of dummy variables for the decade; and finally,
county – These 1,000 municipalities can belong to one of ~ 65
counties. Clustering standard errors at the county level is the most
conservative and so I go with that.
Here is the output:
Warning - singleton groups detected. 117 observation(s) not used.
FIXED EFFECTS ESTIMATION
------------------------
Number of groups = 1135 Obs per group: min = 2
avg = 4.6
max = 9
First-stage regressions
-----------------------
First-stage regression of average_share_dems_votes7:
FIXED EFFECTS ESTIMATION
------------------------
Number of groups = 1135 Obs per group: min = 2
avg = 4.6
max = 9
OLS estimation
--------------
Estimates efficient for homoskedasticity only
Statistics robust to heteroskedasticity and clustering on county
Number of clusters (county) = 65 Number of obs = 5253
F( 4, 64) = 8.95
Prob > F = 0.0000
Total (centered) SS = 9.605866911 Centered R2 = 0.2599
Total (uncentered) SS = 9.605866911 Uncentered R2 = 0.2599
Residual SS = 7.109186342 Root MSE = .04157
------------------------------------------------------------------------------
| Robust
average_s~s7 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_Icurre~1990 | .0355722 .0077302 4.60 0.000 .0201294 .051015
_Icurre~2000 | .0432069 .0122536 3.53 0.001 .0187274 .0676864
pctslfempl~p | -.0001385 .0007944 -0.17 0.862 -.0017255 .0014485
pctpop16to~y | .0040835 .0013014 3.14 0.003 .0014837 .0066832
------------------------------------------------------------------------------
Included instruments: _Icurrentde_1990 _Icurrentde_2000
pctslfemplydownbiznotincp pctpop16to64wdisability
------------------------------------------------------------------------------
Partial R-squared of excluded instruments: 0.0327
Test of excluded instruments:
F( 2, 64) = 5.50
Prob > F = 0.0062
Summary results for first-stage regressions
-------------------------------------------
Variable | Shea Partial R2 | Partial R2 | F( 2, 64) P-value
average_shar| 0.0327 | 0.0327 | 5.50 0.0062
NB: first-stage F-stat cluster-robust
Underidentification tests
Ho: matrix of reduced form coefficients has rank=K1-1 (underidentified)
Ha: matrix has rank=K1 (identified)
Kleibergen-Paap rk LM statistic Chi-sq(2)=. P-val= .
Kleibergen-Paap rk Wald statistic Chi-sq(2)=. P-val= .
Weak identification test
Ho: equation is weakly identified
Kleibergen-Paap Wald rk F statistic .
See main output for Cragg-Donald weak id test critical values
Weak-instrument-robust inference
Tests of joint significance of endogenous regressors B1 in main equation
Ho: B1=0 and overidentifying restrictions are valid
Anderson-Rubin Wald test F(2,64)= 0.92 P-val=0.4038
Anderson-Rubin Wald test Chi-sq(2)=1.87 P-val=0.3927
Stock-Wright LM S statistic Chi-sq(2)=1.87 P-val=0.3927
NB: Underidentification, weak identification and
weak-identification-robust test statistics cluster-robust
Number of clusters N_clust = 65
Number of observations N = 5253
Number of regressors K = 3
Number of instruments L = 4
Number of excluded instruments L1 = 2
2-Step GMM estimation
---------------------
Estimates efficient for arbitrary heteroskedasticity and clustering on county
Statistics robust to heteroskedasticity and clustering on county
Number of clusters (county) = 65 Number of obs = 5253
F( 3, 64) = 14.86
Prob > F = 0.0000
Total (centered) SS = 29430177.08 Centered R2 = 0.0334
Total (uncentered) SS = 29430177.08 Uncentered R2 = 0.0334
Residual SS = 28447688.07 Root MSE = 83.12
------------------------------------------------------------------------------
| Robust
wmeanactfu~2 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
average_s~s7 | 209.2071 159.7796 1.31 0.190 -103.9552 522.3693
_Icurre~1990 | -30.10698 9.738608 -3.09 0.002 -49.19431 -11.01966
_Icurre~2000 | -47.3599 11.90716 -3.98 0.000 -70.6975 -24.0223
------------------------------------------------------------------------------
Underidentification test (Kleibergen-Paap rk LM statistic): .
Chi-sq(2) P-val = .
------------------------------------------------------------------------------
Weak identification test (Kleibergen-Paap rk Wald F statistic): .
Stock-Yogo weak ID test critical values: 10% maximal IV size 19.93
15% maximal IV size 11.59
20% maximal IV size 8.75
25% maximal IV size 7.25
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): 0.017
Chi-sq(1) P-val = 0.8959
------------------------------------------------------------------------------
Instrumented: average_share_dems_votes7
Included instruments: _Icurrentde_1990 _Icurrentde_2000
Excluded instruments: pctslfemplydownbiznotincp pctpop16to64wdisability
------------------------------------------------------------------------------
Please let me know if you need any further details. Thanks for any and
all suggestions,
Sutirtha Bagchi
*
* 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/