Dynamic quantile panel with gmm

[LEE CHAN]

My data characteristics include N=63 and T=4:
dependence variable: gini
independence variables: financial inclusion (fii), life expectation (life), aging (% pop above 65 years old), government expenditure (gov), lag of gini (l.gini), grdp per capita (grdp_pc) and interaction variable = aging * fii.
In which, potential endogenous variables are l(-1).gini, fii, grdp_pc and gov. instrumental variables are l(-1).prdp_pc, l(-1) fii, distance (distance to banking institutions) and aging.
I've already completed DGMM estimator and trying to conduct Dynamic quantile panel estimator (like the previous research below) but i cannot find suitable Stata command to estimate, almost of them not for lag variable or not for endogeneity phenomenon. Can you give me suggestion potential Command.
Many thanks.

[LEE CHAN]

Can anyone help me, please?

[Joao Santos Silva]

Can you give the reference of the paper from where you obtained that table?

[LEE CHAN]

Can you give the reference of the paper from where you obtained that table?

Joao Santos Silva here is my reference paper 10.1016/j.jimonfin.2012.04.006

[Joao Santos Silva]

Thank you for this. Apparently they use a method presented in a working paper by one of the authors (Hsin-Yi Lin), but I could not find that document and therefore cannot comment on its adequacy. I suggest you contact the author.

[LEE CHAN]

Joao Santos Silva, I tried to use these commands
gen l_rp90p10=l.rp90p10 (also for grdp_pc and fii)
qregpd rp90p10 l_rp90p10 fii1 grdp_pc gov life aging, quantile (0.5) id(id) fix(year) optimize(mcmc) noisy draws(1000) burn(500) arate(.5) instruments(l_rp90p10 fii1 grdp_pc gov l_grdp_pc l_fii aging distance)
I'm not sure if it can solve the endogeneity problem and why the same command gives many different results (as below). How to fix that?

qregpd rp90p10 l_rp90p10 fii1 grdp_pc gov life aging,quantile (0.50) id(id) fix(year) optimize(mcmc) noisy draws(1000) burn(500) arate(.5) instruments(l_rp90p10 fii1 grdp_pc gov l_grdp_pc l_fii aging distance)
Adaptive MCMC optimization
................................................. 50: f(x) = -4.06010887
................................................. 100: f(x) = -5.68701796
................................................. 150: f(x) = -4.28468105
................................................. 200: f(x) = -4.88320876
................................................. 250: f(x) = -1.63606038
................................................. 300: f(x) = -4.81047711
................................................. 350: f(x) = -.71299085
................................................. 400: f(x) = -1.94310026
................................................. 450: f(x) = -2.79680232
................................................. 500: f(x) = -.618481789
................................................. 550: f(x) = -2.5250e-26
................................................. 600: f(x) = -1.83519297
Quantile Regression for Panel Data (QRPD)
Number of obs: 189
Number of groups: 63
Min obs per group: 3
Max obs per group: 3
------------------------------------------------------------------------------
rp90p10 | Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
l_rp90p10 | -.0799095 .3491866 -0.23 0.819 -.7643027 .6044838
fii1 | .8143741 1.421435 0.57 0.567 -1.971586 3.600335
grdp_pc | 7.016767 2.683872 2.61 0.009 1.756475 12.27706
gov | .4609649 1.010419 0.46 0.648 -1.51942 2.441349
life | -157.2481 60.15768 -2.61 0.009 -275.155 -39.34121
aging | 4.30844 1.443006 2.99 0.003 1.480201 7.136679
------------------------------------------------------------------------------

qregpd rp90p10 l_rp90p10 fii1 grdp_pc gov life aging,quantile (0.50) id(id) fix(year) optimize(mcmc) noisy draws(1000) burn(500) arate(.5) instruments(l_rp90p10 fii1 grdp_pc gov l_grdp_pc l_fii aging distance)
Adaptive MCMC optimization
................................................. 50: f(x) = -2.68325984
................................................. 100: f(x) = -.922401285
................................................. 150: f(x) = -1.66940218
................................................. 200: f(x) = -1.52082186
................................................. 250: f(x) = -1.02889808
................................................. 300: f(x) = -1.9560406
................................................. 350: f(x) = -2.12334952
................................................. 400: f(x) = -1.37927087
................................................. 450: f(x) = -3.38216991
................................................. 500: f(x) = -1.30946883
................................................. 550: f(x) = -1.96205492
................................................. 600: f(x) = -1.4333009
Quantile Regression for Panel Data (QRPD)
Number of obs: 189
Number of groups: 63
Min obs per group: 3
Max obs per group: 3
------------------------------------------------------------------------------
rp90p10 | Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
l_rp90p10 | -.4874328 1.275709 -0.38 0.702 -2.987776 2.012911
fii1 | 69.58705 120.8059 0.58 0.565 -167.1881 306.3622
grdp_pc | 12.27815 19.25332 0.64 0.524 -25.45767 50.01396
gov | -43.07006 79.03985 -0.54 0.586 -197.9853 111.8452
life | -2425.099 3922.333 -0.62 0.536 -10112.73 5262.532
aging | 329.9499 536.9262 0.61 0.539 -722.406 1382.306
------------------------------------------------------------------------------

[Joao Santos Silva]

Dear LEE CHAN,

I am afraid I am not very familiar it that command and I have concerns about the interpretability of the results of the method it implements.

Best wishes,

Joao

[LEE CHAN]

Joao Santos Silva thanks for your help. one last question, can you suggest me any command that can solve the endogeneity problem (with T=4 and N=63) in panel quantile regression?

[Joao Santos Silva]

Unfortunately, I cannot.