Stata code for Dynamic Panel Threshold Model (with Endogeneity)

[River Huang]

1. Dear All, I just read the following thread: https://www.statalist.org/forums/for...nd-endogeneity,
which is about asking for a Stata code for estimating dynamic panel threshold model (with endogeneity). Please see the paper http://www.stata-journal.com/article...article=st0373 for a Stata code for (static, or non-dynaymic) panel threshold model (without considering endogeneity).

2. May I suggest/encourage anyone who has the (Stata, Gauss and econometric) expertises to translate the Gauss code into Stata counterparts. As far as I know, there are many persons.researchers who are interested in applying the approach (to many topics) using Stata code (rather than the original Gauss code), including myself.

3. Or, if possible, may I suggest the Editors of Stata Journal to invite someone to complete this task?

[fai ot]

Dear River Huang,

I will be so grateful if you can help me to suggest the Editors of Stata Journal to invite someone to complete this task. The stata code for this methodology is important for me to do my thesis research.

Many Thanks
Fairouz

[River Huang]

Dear Fairouz, As an empirical applicant, I understand how you feel at this moment. Unfortunately, I am nobody, and I do not think most of people care about what I proposed (although I truly believe a Stata command for estimating dynamic panel threshold regression with endogeneity will benefit many empiricalists, like you and me.) I told my student that if they can (but very unlikely) write such a command, they can influence (measured by, say, Google citations) many people. So, you better not hold your breath.

[Santiago_C]

Dear River Huang:

I am also using the xthreg command for a dynamic panel. Even though there is no command to apply I believe we can do it manually:

1) Estimate a DPD using the Arellano-Bover (forward orthogonal deviations) . That is : run a reduced form regression
of the endogenous variable on a set of instruments Then, extract the dependent "variable".

2) Estimate the Threshold Model using the new dependent variable regarding one/two thresholds.

3) If we find a significant threshold value (using the F test), we need to repeat step 1) and 2) to finally obtain the "clean thresholds".


Nonetheless, it is not simple, since we need to control the quantity of instruments and of course create a new scalar variable to interact with the regimes. What do you think?

Best Regards,

Santiago

[River Huang]

Hi, Santiago_C: 1. I understand the idea for estimating a dynamic panel threshold model but have problem in writing a handy Stat code for that (including possibly the bootstrapping procedure). 2. As far as I know, xthreg CANNOT be used to this purpose.

[Santiago_C]

Dear River,

I agree, the bootstrapping code is the key. Nonetheless, there is matlab code that perfectly does the gmm/ iv process along with the Arellano-Bover orthogonal deviation. It perfectly works for threshold panel. Indeed, it has been written in matlab, not in Stata, but really useful at the moment.

Regards,

Santiago

[sedki zn]

Dear all

please i need your help about the means of threshold variables ;
xtreg i q1 q2 q3 d1 qd1, rx(c1) qx(d1)
c1 and d1, especially

Best regards
SEDKI

[River Huang]

Hi sedki, What did you mean by "needing help about the means of threshold variables"?

[sedki zn]

Dear Huang

I’m working on the impact of CEO remuneration on the firm financing constraints.
My question is ; could i define the CEO remuneration variable as independent variable and the threshold variable?

because even on the stata help, i cannot understand the means of rx() and qx() and witch kind of such variables are allowed

Best regards

[River Huang]

Hi, sedki, Yes, you can.

[sedki zn]

dear Huang
thank you vert much

just a last question

can i threshold the Time variable ? ( in order to study the relation between Y and X in the time )
" xthreg Y X controle_var, rx(YEAR) qx(YEAR) ..... "

Best regards
Sedki

[River Huang]

Dear sedki, I don't think it is a good idea.

[SPANOS KONSTANTINOS]

Hi to all,
Given that dynamic panel threshold model (with endogeneity) is a new method,
the previous methods, all suffered from endogeneity problem? unless dependent regime-variables were strictly exogenous?

[Toukam hermann doel]

hi to all,
can someone help me with stata commands on panel smooth transition regresion model of Gonsalez et al ?

[Khethang Mokoena]

I hope everyone knows that the code for specifiying dynamic panel threshold with endogeneity is xtendothresdpd. But i am having a problem getting the results. I am estimating the threshold of government expenditure on inflation. My specification is like this.

. xtendothresdpd INF L.INF MSY GDP IQI EXR FPI INT, thresv(GOV) stub(enr) pivar(GOV) dgmmiv(INF) twostep vce(robus
> t) grid(100)

I get the following error:

variance–covariance matrix of the two-step estimator is not full rank
Two-step estimator is not available. One-step estimator is available and variance–covariance matrix provides
correct coverage.
r(198);
Any help?