Time Dummy Variable for xtabond2

[Zahin Chowdhury]

Hi

A little background first. I am trying to test how democracy levels impact GDP. I am using the data of 105 countries spanning a period of 26 years (1990-2016) using the Arellano-Bond Estimator (xtabond2). I am trying to put in a time dummy variable to take into account time fixed effect into account. For example the Global Financial Crisis of 2007 and others.

This is my initial code

xtabond2 loggdp l.loggdp polrisk l.popgrow logtrade govsp democracy, gmm (l.loggdp logtrade) iv(polrisk l.popgrow govsp democracy)

H1: Democracy leads to economic development.

xtabond2 loggdp l.loggdp polrisk l.popgrow logtrade govsp democracy, gmm (l.loggdp logtrade) iv(polrisk l.popgrow govsp democracy)

Variables
loggdp - log of GDP
polrisk - Political Risk
l.popgrow - lag of Population Growth
logtrade - log of Trade
govsp - Government Spending
democracy - signifies if the country was democratic or not in a particular year.

What is the command to create time dummy variables and where do I insert it in my formula.
Thank You

[Dario Maimone Ansaldo Patti]

I am not sure if this is what you want, but you can add a set of time dummies by using i.time just before the comma.

Be sure that you add i.time within iv( ).

Few other things:

1) xtabond2 can implement both the Arellano-Bond (difference gmm) and the blundell-bond estimator (system gmm). The routine you presented above implements the second one. If you want the Arellano-Bond estimator you need to add the option nol (no level);

2) Apparently you are going to use all the available lags as instruments. This could be dangerous as it weakens the validity of your diagnostic tests. Have a look at a paper by Roodman (2009) in Oxford bulletin of economics and statistics about the drawback of using too many lags.

3) Democracy is treated as a predermined variable. I am not sure whether this is the correct approach, given the long-lasting debate in the literature about the direction of causality about development and democracy. Anyway, on the relationship between democracy and development, I would look at the recent paper by Acemoglu et al. (2018) entitled does democracy cause growth.

4) I would use a more refined measure of development as the per capital or per worker gdp

[Zahin Chowdhury]

Thanks Dario..
1) I am using system gmm
2) I read that paper.... Honestly I am super weak when it comes to stata.... everything sort of went above my head
3) I am actually using Acemoglu et al (2018) as one of my main source paper... It was a very good read.
4) I have been thinking about the per capital aspect of it for awhile now.. might do that later.

Thanks again
Cheers

[Sebastian Kripfganz]

If you are planning to use time dummies as suggested by Dario Maimone Ansaldo Patti, please be aware that there is a bug in xtabond2. See slide 78 of my recent presentation at the 2019 London Stata Conference. Slide 80 shows you an easy way to add time dummies with my xtdpdgmm command as an alternative to xtabond2. The presentation might provide further useful information on other questions regarding GMM for dynamic panel data models.

[Zahin Chowdhury]

Thank you Sebastian Kripfganz
will take a look

[Dario Maimone Ansaldo Patti]

Sebastian Kripfganz great presentation. Really helpful!