Dear all,
I am new to the xtabond2 command and I have been struggling with this for a while so I really appreciate any assistance and help in this matter.
My panel data consists of 28 observations (countries), 16 years, dependent variable (log_depo_ratio) and 14 explanatory variables - L.depo_ratio, ir, log_hh_debt, L.log_income, log_m2, log_convergence, log_wealth, log_tot, lab_product, govsaving_ratio, log_inflindex, log_old, log_urb_rate, log_unrate, out of which only the last three variables are exogenous while all the rest are endogenous. Namely, I am trying to analyse the determinants of deposit rates.
This is the code I used:
And this is the output that I got:
According to my research and what I read in other working papers and some topics on Statalist, both my Hansen and AR(1) are good. There was a discussion on shoulds we look at Hansen or Sargan, but as I understood we can just focus on one of the tests because they both show the validation of our instruments. Therefore, is my logic flawed if I just focus on AR(1) being 0.000 and Hansen being not too big?
Lastly, my biggest concern are the p-values. I tried a lot of things, but I can't get them at the conventional 5% significance level. I know that this may sound ridicoulous, but is there a statistical theory that might back up the idea of p-values not being so important?As I understood, the Fisher was not so strict about using them strictly at the alpha significance?
If not, what else could I do to make them smaller and acceptable?
I am new to the xtabond2 command and I have been struggling with this for a while so I really appreciate any assistance and help in this matter.
My panel data consists of 28 observations (countries), 16 years, dependent variable (log_depo_ratio) and 14 explanatory variables - L.depo_ratio, ir, log_hh_debt, L.log_income, log_m2, log_convergence, log_wealth, log_tot, lab_product, govsaving_ratio, log_inflindex, log_old, log_urb_rate, log_unrate, out of which only the last three variables are exogenous while all the rest are endogenous. Namely, I am trying to analyse the determinants of deposit rates.
This is the code I used:
Code:
xtabond2 log_depo_ratio L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence log_wealth log_tot lab_product govsaving_ratio log_inflindex log_old log_urb_rate log_unrate, gmm(L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence log_wealth log_tot lab_product govsaving_ratio log_inflindex, laglimits(0 0) eq(level) collapse) gmm(L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence log_wealth log_tot lab_product govsaving_ratio log_inflindex, laglimits(1 1) eq(diff) collapse) iv( log_old log_urb_rate log_unrate, eq(level)) robust nodiffsargan
Code:
xtabond2 log_depo_ratio L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence log_wealth log_tot lab_product govsaving_ratio log_inflindex log_old log_urb_rate
> log_unrate, gmm(L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence log_wealth log_tot lab_product govsaving_ratio log_inflindex, laglimits(0 0) eq(level) col
> lapse) gmm(L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence log_wealth log_tot lab_product govsaving_ratio log_inflindex, laglimits(1 1) eq(diff) collapse)
> iv( log_old log_urb_rate log_unrate, eq(level)) robust nodiffsargan
Favoring speed over space. To switch, type or click on mata: mata set matafavor space, perm.
Dynamic panel-data estimation, one-step system GMM
------------------------------------------------------------------------------
Group variable: ctry_dum Number of obs = 420
Time variable : year Number of groups = 28
Number of instruments = 26 Obs per group: min = 15
Wald chi2(14) = 1826.52 avg = 15.00
Prob > chi2 = 0.000 max = 15
---------------------------------------------------------------------------------
| Robust
log_depo_ratio | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
log_depo_ratio |
L1. | .6646304 .0640918 10.37 0.000 .5390128 .790248
|
ir | .0021721 .0050198 0.43 0.665 -.0076666 .0120108
log_hh_debt | .1005615 .0366328 2.75 0.006 .0287625 .1723605
|
log_income |
L1. | .3276255 .1364211 2.40 0.016 .0602451 .5950059
|
log_m2 | .0005251 .021366 0.02 0.980 -.0413516 .0424017
log_convergence | -.7696703 .2441639 -3.15 0.002 -1.248223 -.2911179
log_wealth | .1544913 .0517407 2.99 0.003 .0530813 .2559012
log_tot | -.0583764 .3766962 -0.15 0.877 -.7966874 .6799347
lab_product | -.0069043 .0026771 -2.58 0.010 -.0121514 -.0016573
govsaving_ratio | .0001679 .002976 0.06 0.955 -.0056648 .0060007
log_inflindex | -.2735538 .1781045 -1.54 0.125 -.6226323 .0755246
log_old | -.2359759 .0927227 -2.54 0.011 -.4177089 -.0542428
log_urb_rate | -.0201112 .1189348 -0.17 0.866 -.2532192 .2129967
log_unrate | .0458677 .0315098 1.46 0.145 -.0158904 .1076259
_cons | 2.579339 2.135743 1.21 0.227 -1.60664 6.765318
---------------------------------------------------------------------------------
Instruments for first differences equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L.(L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence
log_wealth log_tot lab_product govsaving_ratio log_inflindex) collapsed
Instruments for levels equation
Standard
log_old log_urb_rate log_unrate
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
D.(L.log_depo_ratio ir log_hh_debt L.log_income log_m2 log_convergence
log_wealth log_tot lab_product govsaving_ratio log_inflindex) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -3.49 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = -0.17 Pr > z = 0.863
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(11) = 100.87 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(11) = 15.44 Prob > chi2 = 0.163
(Robust, but weakened by many instruments.)
.
Lastly, my biggest concern are the p-values. I tried a lot of things, but I can't get them at the conventional 5% significance level. I know that this may sound ridicoulous, but is there a statistical theory that might back up the idea of p-values not being so important?As I understood, the Fisher was not so strict about using them strictly at the alpha significance?
If not, what else could I do to make them smaller and acceptable?
