xtdpdgmm command query

[Muhammad Ibrahim Shah]

Sebastian Kripfganz

Can anybody explain this command to me please, step by step?

xtdpdgmm L(0/1).n w k, gmm(L.n w k, l(1 4) c m(d)) iv(L.n w k, d) two vce(r)

1. What does L(0/1).n mean please? what would happen if we specified L.n only? or even L(1/1).n?

2. What is this equation: L(0/1).n w k?

3. What does c m(d) and l(1, 4) mean here?: gmm(L.n w k, l(1 4) c m(d))

4. Why include d here?: iv(L.n w k, d)

5. I understand that gmm is specifying internal instruments and iv is specifying external instruments? Internal instruments are the ones that come from the model itself and external instruments are the ones coming from outside such as other independent variables that we use in a model. Is my understanding correct at all? Because I believe i am getting something wrong here. If iv() only specifies external instruments, then what is L.n doing here? If gmm() really specifies internal instruments, what is w and k doing here? Are W and K not external instruments?

I know i am asking very naive questions and you will probably guide me to Roodman or the original articles of system GMM. But I would prefer it if someone can answer my question one by one so that before I go those documents, my understanding gets cleared. Thank you.

[Sebastian Kripfganz]

1. L(0/1).n is equivalent to n L.n.; see the following Stata help file: help tsvarlist. The first variable, n, is the dependent variable. The second variable, L.n, is an independent variable. You do not usually want L.n to be the dependent variable.

2. The dependent variable is n. The independent variables are L.n w k.

3. These are abbreviations for collapse model(difference) lagrange(1 4); see the help file for details: help xtdpdgmm.

4. d is an abbreviation for difference. This applies a first-difference transformation to the specified variables; see the help file for details.

5. You are describing one common use for the different types of instruments. However, iv() is just a special case of collapsed gmm(). iv(L.n w k) is equivalent to gmm(L.n w k, collapse lagrange(0 0)).

I recommend to have a look at the following presentation:

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.

[Muhammad Ibrahim Shah]

1. So please let me get this straight. When we use [n, lagrange(1 4)] or in other words [n,l(1 4)], it means we are using first(or current) and fourth lag of the dependent variable as instrument?

2. In journal publication, there is usually a note below the regression table. In the case of your specified model, what would you have said, is it something like this?

xtdpdgmm L(0/1).n w k, gmm(L.n w k, l(1 4) c m(d)) iv(L.n w k, d) two vce(r)


current and fourth lags of L.N, w and k are taken as instruments for the differenced equation [gmm(L.n w k, l(1 4) c m(d)) considering this differenced equaion] and first lag of n, current lag of explanatory variables (w,k) are used as instruments for the level equation? (this is from iv(L.n w k, d considering this is level equation))


I might be confused about the differenced equation and level equation. I have searched many articles including yours but I see different cases of definition. For example, in your command, what would you say the differenced and level equations are

[Sebastian Kripfganz]

1. n, l(1 4) means you are using the first, second, third, and fourth lag of n as instruments. This option specifies the two endpoints of a range.

2. Current to fourth lags of L.n, w, and k are used as instuments for the differenced equation. First differences of L.n, w, and k are used as instruments for the level equation.

[Muhammad Ibrahim Shah]

Thank you so much! I will ask additional questions if I come across any problems. But this is great for now! Thank you for replying, you are so kind.