I want to analyse determinants of corporate leverage in the past twenty years (with regard to recent low interest rates). Therefore, I perform a system GMM with xtabond2 as instructed by Roodman 2009 ("How to do xtabond"). I have read the paper first, of course.
My core firm-specific factors (endogenous) are: profitability, tangibility, market-to-book ratio, size and median industry leverage. I also use year dummies, which I assume to be strictly exogenous. As Roodman says, all regressors should enter the instruments matrix set up by gmmstyle() and ivstyle(), depending on the classification of regressors (predetermined, endogenous, strictly exogenous). However, even though I am strictly following Roodman, my Hansen test results are quite disappointing. I tried different sets of instruments, but the best I got was 0.05. I also use the collapse and laglimits option, but the null of instrument validity keeps getting rejected. I also reduced the number of years to ten. The null keeps getting rejected. What am I missing here?
Additionally, since I want to estimate coefficients for interest rate factors, how can I run the analysis without the year dummies? As Roodman says, one should always include the dummies. But that makes all of my interest rate factors omitted due to collinearity.
I run the command:
Code:
xtabond2 tdm_w l1.tdm_w profitablty_w mtb_w tangiblty_w size indlevm_w yr*, gmmstyle(L.(tdm_w profitablty_w mtb_w tangiblty_w size), lagl > imits (2 3) collapse) ivstyle( yr*, equation(level)) twostep robust small orthogonal
Code:
Dynamic panel-data estimation, two-step system GMM ------------------------------------------------------------------------------ Group variable: id Number of obs = 89422 Time variable : year Number of groups = 9135 Number of instruments = 34 Obs per group: min = 1 F(26, 9134) = 693.94 avg = 9.79 Prob > F = 0.000 max = 19 ------------------------------------------------------------------------------- | Corrected tdm_w | Coef. Std. Err. t P>|t| [95% Conf. Interval] --------------+---------------------------------------------------------------- tdm_w | L1. | .7361274 .0219671 33.51 0.000 .6930671 .7791878 | profitablty_w | .0238564 .01837 1.30 0.194 -.012153 .0598657 mtb_w | -.002019 .0023063 -0.88 0.381 -.0065398 .0025018 tangiblty_w | -.0092912 .0222226 -0.42 0.676 -.0528525 .03427 size | -.0086231 .003725 -2.31 0.021 -.0159249 -.0013212 indlevm_w | .3406029 .1060849 3.21 0.001 .1326528 .5485531 yr1 | 0 (omitted) yr2 | -.077951 .0105912 -7.36 0.000 -.098712 -.05719 yr3 | -.0692224 .0075676 -9.15 0.000 -.0840565 -.0543883 yr4 | -.1041507 .0098989 -10.52 0.000 -.1235548 -.0847466 yr5 | -.0846216 .0070196 -12.06 0.000 -.0983815 -.0708616 yr6 | -.1429814 .0148919 -9.60 0.000 -.1721728 -.11379 yr7 | -.1097363 .0157845 -6.95 0.000 -.1406775 -.078795 yr8 | -.0866864 .0154577 -5.61 0.000 -.1169869 -.0563859 yr9 | -.0960668 .0166473 -5.77 0.000 -.1286993 -.0634344 yr10 | -.0625026 .0130212 -4.80 0.000 -.0880272 -.036978 yr11 | 0 (omitted) yr12 | -.1575229 .0121805 -12.93 0.000 -.1813993 -.1336465 yr13 | -.1154279 .0142346 -8.11 0.000 -.1433309 -.087525 yr14 | -.0661134 .0105783 -6.25 0.000 -.0868493 -.0453775 yr15 | -.0929222 .0118196 -7.86 0.000 -.1160912 -.0697532 yr16 | -.1137083 .0150861 -7.54 0.000 -.1432805 -.0841361 yr17 | -.0761268 .012738 -5.98 0.000 -.1010962 -.0511575 yr18 | -.0620756 .0095116 -6.53 0.000 -.0807205 -.0434308 yr19 | -.1078423 .0130321 -8.28 0.000 -.1333882 -.0822964 yr20 | -.097287 .0142452 -6.83 0.000 -.1252107 -.0693633 _cons | .1296548 .0311623 4.16 0.000 .0685697 .1907398 ------------------------------------------------------------------------------- Instruments for orthogonal deviations equation GMM-type (missing=0, separate instruments for each period unless collapsed) L(2/3).(L.tdm_w L.profitablty_w L.mtb_w L.tangiblty_w L.size) collapsed Instruments for levels equation Standard yr1 yr2 yr3 yr4 yr5 yr6 yr7 yr8 yr9 yr10 yr11 yr12 yr13 yr14 yr15 yr16 yr17 yr18 yr19 yr20 _cons GMM-type (missing=0, separate instruments for each period unless collapsed) DL.(L.tdm_w L.profitablty_w L.mtb_w L.tangiblty_w L.size) collapsed ------------------------------------------------------------------------------ Arellano-Bond test for AR(1) in first differences: z = -27.69 Pr > z = 0.000 Arellano-Bond test for AR(2) in first differences: z = -1.74 Pr > z = 0.082 ------------------------------------------------------------------------------ Sargan test of overid. restrictions: chi2(7) = 109.97 Prob > chi2 = 0.000 (Not robust, but not weakened by many instruments.) Hansen test of overid. restrictions: chi2(7) = 56.49 Prob > chi2 = 0.000 (Robust, but weakened by many instruments.) Difference-in-Hansen tests of exogeneity of instrument subsets: GMM instruments for levels Hansen test excluding group: chi2(2) = 18.67 Prob > chi2 = 0.000 Difference (null H = exogenous): chi2(5) = 37.82 Prob > chi2 = 0.000
Thanks in advance and merry christmas to everybody!
0 Response to SYS-GMM: Problem of overidentifying restrictions
Post a Comment