Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | sdemiralp <sdemiralp@ku.edu.tr> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: Re: ivreg2 - weak instruments? |
Date | Wed, 11 Apr 2012 07:29:44 -0700 (PDT) |
I am experiencing a very similar problem. I use ivreg2 to estimate a FE model with three endogenous variables. I use clustered errors. While each first stage regression seems to pass the Angrist-Pischke multivariate F test of excluded instruments with p=0.0000, when it comes to the overall test of underidentification (Kleibergen-Paap rk LM statistic), the model fails with p=0.19. Does that make sense? Thank you very much, Selva Here is my output file. IXED EFFECTS ESTIMATION ------------------------ Number of groups = 210 Obs per group: min = 2 avg = 4.7 max = 5 First-stage regressions ----------------------- First-stage regression of managed_ratio_ch: FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 210 Obs per group: min = 2 avg = 4.7 max = 5 OLS estimation -------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on rssd9348 Number of clusters (rssd9348) = 210 Number of obs = 987 F( 16, 209) = 10.55 Prob > F = 0.0000 Total (centered) SS = .4381141535 Centered R2 = 0.3002 Total (uncentered) SS = .4381141535 Uncentered R2 = 0.3002 Residual SS = .3065949304 Root MSE = .02007 ------------------------------------------------------------------------------ | Robust managed~o_ch | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- core_ratio_1 | -.1141758 .0491075 -2.33 0.021 -.2109853 -.0173663 core_ratio~h | .0621601 .0589716 1.05 0.293 -.0540954 .1784155 leverage_1 | -.0014088 .0012026 -1.17 0.243 -.0037796 .000962 ln_asset_1 | -.0819344 .024942 -3.28 0.001 -.1311046 -.0327641 commit_ra~_1 | .0108455 .0871669 0.12 0.901 -.1609935 .1826845 illiquid_r~1 | .0004838 .0003886 1.24 0.215 -.0002824 .00125 house_1 | -.0006154 .0003155 -1.95 0.052 -.0012373 6.52e-06 delin_1 | .0000671 .0006693 0.10 0.920 -.0012523 .0013866 qr6 | -.0098983 .0059937 -1.65 0.100 -.0217141 .0019175 qr7 | -.0099937 .0046444 -2.15 0.033 -.0191495 -.000838 qr8 | -.0073085 .0040408 -1.81 0.072 -.0152745 .0006574 qr9 | -.0076849 .0023547 -3.26 0.001 -.0123268 -.003043 managed_~o_1 | -.61656 .0800671 -7.70 0.000 -.7744026 -.4587175 tarp_ratio_1 | -.1237328 .1287304 -0.96 0.338 -.3775093 .1300437 risky_1 | .0014801 .0008822 1.68 0.095 -.0002591 .0032193 secur_1 | 2.060793 1.502494 1.37 0.172 -.9011926 5.022778 ------------------------------------------------------------------------------ Included instruments: core_ratio_1 core_ratio_ch leverage_1 ln_asset_1 commit_ratio_1 illiquid_ratio_1 house_1 delin_1 qr6 qr7 qr8 qr9 managed_ratio_1 tarp_ratio_1 risky_1 secur_1 ------------------------------------------------------------------------------ F test of excluded instruments: F( 4, 209) = 17.45 Prob > F = 0.0000 Angrist-Pischke multivariate F test of excluded instruments: F( 2, 209) = 34.35 Prob > F = 0.0000 First-stage regression of taf_ratio_ch_av: FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 210 Obs per group: min = 2 avg = 4.7 max = 5 OLS estimation -------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on rssd9348 Number of clusters (rssd9348) = 210 Number of obs = 987 F( 16, 209) = 3.49 Prob > F = 0.0000 Total (centered) SS = .0823897245 Centered R2 = 0.1853 Total (uncentered) SS = .0823897245 Uncentered R2 = 0.1853 Residual SS = .0671216217 Root MSE = .009392 ------------------------------------------------------------------------------ | Robust taf_rat~h_av | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- core_ratio_1 | -.0387296 .0302408 -1.28 0.202 -.0983458 .0208865 core_ratio~h | -.0043928 .0101968 -0.43 0.667 -.0244946 .0157089 leverage_1 | -.000837 .0005165 -1.62 0.107 -.0018553 .0001813 ln_asset_1 | .0013435 .0089253 0.15 0.880 -.0162517 .0189386 commit_ra~_1 | .0947031 .0550451 1.72 0.087 -.0138116 .2032179 illiquid_r~1 | -.0000196 .0002181 -0.09 0.929 -.0004495 .0004104 house_1 | .0001778 .0000938 1.90 0.059 -7.16e-06 .0003627 delin_1 | -.0001103 .0001618 -0.68 0.496 -.0004292 .0002086 qr6 | .0004993 .0024166 0.21 0.837 -.0042648 .0052633 qr7 | .0002592 .001563 0.17 0.868 -.002822 .0033403 qr8 | -.000108 .0011579 -0.09 0.926 -.0023906 .0021746 qr9 | -.0017711 .000892 -1.99 0.048 -.0035295 -.0000127 managed_~o_1 | -.0841551 .0470226 -1.79 0.075 -.1768544 .0085443 tarp_ratio_1 | .0243501 .045094 0.54 0.590 -.0645473 .1132475 risky_1 | -.0000277 .0002501 -0.11 0.912 -.0005208 .0004655 secur_1 | -6.216941 2.092633 -2.97 0.003 -10.34231 -2.091567 ------------------------------------------------------------------------------ Included instruments: core_ratio_1 core_ratio_ch leverage_1 ln_asset_1 commit_ratio_1 illiquid_ratio_1 house_1 delin_1 qr6 qr7 qr8 qr9 managed_ratio_1 tarp_ratio_1 risky_1 secur_1 ------------------------------------------------------------------------------ F test of excluded instruments: F( 4, 209) = 3.75 Prob > F = 0.0057 Angrist-Pischke multivariate F test of excluded instruments: F( 2, 209) = 6.59 Prob > F = 0.0017 First-stage regression of tarp_ratio_ch: FIXED EFFECTS ESTIMATION ------------------------ Number of groups = 210 Obs per group: min = 2 avg = 4.7 max = 5 OLS estimation -------------- Estimates efficient for homoskedasticity only Statistics robust to heteroskedasticity and clustering on rssd9348 Number of clusters (rssd9348) = 210 Number of obs = 987 F( 16, 209) = 86.95 Prob > F = 0.0000 Total (centered) SS = .0303474345 Centered R2 = 0.6754 Total (uncentered) SS = .0303474345 Uncentered R2 = 0.6754 Residual SS = .009851877 Root MSE = .003598 ------------------------------------------------------------------------------ | Robust tarp_ratio~h | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- core_ratio_1 | -.0094204 .0083818 -1.12 0.262 -.0259441 .0071034 core_ratio~h | .0018119 .0045222 0.40 0.689 -.007103 .0107269 leverage_1 | -.0003055 .000225 -1.36 0.176 -.0007491 .000138 ln_asset_1 | -.0170686 .0037394 -4.56 0.000 -.0244405 -.0096968 commit_ra~_1 | -.0041764 .0153924 -0.27 0.786 -.0345207 .0261679 illiquid_r~1 | -.0000448 .0000612 -0.73 0.465 -.0001654 .0000758 house_1 | -.0001395 .0000633 -2.20 0.029 -.0002642 -.0000147 delin_1 | .0000859 .0001097 0.78 0.434 -.0001304 .0003022 qr6 | .0012991 .0010841 1.20 0.232 -.0008381 .0034363 qr7 | -.0003277 .000765 -0.43 0.669 -.0018358 .0011804 qr8 | -.0002455 .0005772 -0.43 0.671 -.0013834 .0008923 qr9 | -.0001685 .0003369 -0.50 0.617 -.0008328 .0004957 managed_~o_1 | .0096373 .0071035 1.36 0.176 -.0043663 .0236409 tarp_ratio_1 | -.8548846 .038516 -22.20 0.000 -.9308142 -.778955 risky_1 | .0003544 .0001179 3.01 0.003 .000122 .0005869 secur_1 | .4395913 .2525309 1.74 0.083 -.0582429 .9374254 ------------------------------------------------------------------------------ Included instruments: core_ratio_1 core_ratio_ch leverage_1 ln_asset_1 commit_ratio_1 illiquid_ratio_1 house_1 delin_1 qr6 qr7 qr8 qr9 managed_ratio_1 tarp_ratio_1 risky_1 secur_1 ------------------------------------------------------------------------------ F test of excluded instruments: F( 4, 209) = 123.81 Prob > F = 0.0000 Angrist-Pischke multivariate F test of excluded instruments: F( 2, 209) = 247.39 Prob > F = 0.0000 Summary results for first-stage regressions ------------------------------------------- (Underid) (Weak id) Variable | F( 4, 209) P-val | AP Chi-sq( 2) P-val | AP F( 2, 209) managed_rati | 17.45 0.0000 | 70.09 0.0000 | 34.35 taf_ratio_ch | 3.75 0.0057 | 13.46 0.0012 | 6.59 tarp_ratio_c | 123.81 0.0000 | 504.83 0.0000 | 247.39 NB: first-stage test statistics cluster-robust Stock-Yogo weak ID test critical values for single endogenous regressor: 10% maximal LIML size 8.68 15% maximal LIML size 5.33 20% maximal LIML size 4.42 25% maximal LIML size 3.92 Source: Stock-Yogo (2005). Reproduced by permission. NB: Critical values are for Cragg-Donald F statistic and i.i.d. errors. Underidentification test 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)=3.26 P-val=0.1964 Weak identification test Ho: equation is weakly identified Cragg-Donald Wald F statistic 24.31 Kleibergen-Paap Wald rk F statistic 3.08 Stock-Yogo weak ID test critical values for K1=3 and L1=4: <not available> Weak-instrument-robust inference Tests of joint significance of endogenous regressors B1 in main equation Ho: B1=0 and orthogonality conditions are valid Anderson-Rubin Wald test F(4,209)= 15.58 P-val=0.0000 Anderson-Rubin Wald test Chi-sq(4)= 63.57 P-val=0.0000 Stock-Wright LM S statistic Chi-sq(4)= 7.73 P-val=0.1021 NB: Underidentification, weak identification and weak-identification-robust test statistics cluster-robust Number of clusters N_clust = 210 Number of observations N = 987 Number of regressors K = 15 Number of endogenous regressors K1 = 3 Number of instruments L = 16 Number of excluded instruments L1 = 4 CUE estimation -------------- Estimates efficient for arbitrary heteroskedasticity and clustering on rssd9348 Statistics robust to heteroskedasticity and clustering on rssd9348 Number of clusters (rssd9348) = 210 Number of obs = 987 F( 15, 209) = 8.49 Prob > F = 0.0000 Total (centered) SS = .7346479492 Centered R2 = 0.0823 Total (uncentered) SS = .7346479492 Uncentered R2 = 0.0823 Residual SS = .6741731309 Root MSE = .02946 ------------------------------------------------------------------------------ | Robust loan_ratio~h | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- managed~o_ch | -.1320594 .1425158 -0.93 0.354 -.4113853 .1472664 taf_rat~h_av | 1.667921 .5665844 2.94 0.003 .5574363 2.778406 tarp_ratio~h | .5547087 .2799986 1.98 0.048 .0059214 1.103496 core_ratio_1 | .1950304 .0801773 2.43 0.015 .0378859 .352175 core_ratio~h | .3758764 .1066273 3.53 0.000 .1668907 .5848621 leverage_1 | .0066694 .0017859 3.73 0.000 .0031692 .0101696 ln_asset_1 | -.0893153 .036743 -2.43 0.015 -.1613302 -.0173004 commit_ra~_1 | .0226384 .1109812 0.20 0.838 -.1948808 .2401575 illiquid_r~1 | -.0034237 .000788 -4.35 0.000 -.0049681 -.0018793 house_1 | -.0001013 .000468 -0.22 0.829 -.0010187 .000816 delin_1 | -.0026814 .0006613 -4.05 0.000 -.0039777 -.0013852 qr6 | .0181987 .0087961 2.07 0.039 .0009588 .0354387 qr7 | .021555 .0065191 3.31 0.001 .0087779 .0343321 qr8 | .010135 .0047161 2.15 0.032 .0008917 .0193783 qr9 | .0048483 .0031053 1.56 0.118 -.0012379 .0109345 ------------------------------------------------------------------------------ Underidentification test (Kleibergen-Paap rk LM statistic): 3.255 Chi-sq(2) P-val = 0.1964 ------------------------------------------------------------------------------ Weak identification test (Cragg-Donald Wald F statistic): 24.310 (Kleibergen-Paap rk Wald F statistic): 3.078 Stock-Yogo weak ID test critical values: <not available> ------------------------------------------------------------------------------ Hansen J statistic (overidentification test of all instruments): 0.971 Chi-sq(1) P-val = 0.3244 -endog- option: Endogeneity test of endogenous regressors: 3.742 Chi-sq(1) P-val = 0.0531 Regressors tested: taf_ratio_ch_av ------------------------------------------------------------------------------ Instrumented: managed_ratio_ch taf_ratio_ch_av tarp_ratio_ch Included instruments: core_ratio_1 core_ratio_ch leverage_1 ln_asset_1 commit_ratio_1 illiquid_ratio_1 house_1 delin_1 qr6 qr7 qr8 qr9 Excluded instruments: managed_ratio_1 tarp_ratio_1 risky_1 secur_1 Dropped collinear: qr1 qr2 qr3 qr4 qr5 qr10 qr11 qr12 qr13 qr14 qr15 qr16 qr17 ------------------------------------------------------------------------------ -- View this message in context: http://statalist.1588530.n2.nabble.com/ivreg2-weak-instruments-tp4979513p7456242.html Sent from the Statalist mailing list archive at Nabble.com. * * 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/