Accumulated time dummies in GMM

[Marcos Gonzalez]

I have usually used ordinary time dummies to take into account time effects in GMM but recently I have heard that it would be more appropiate to use accumulated time dummies. Is it correct? If it is so, what would be the explanation? Is there any paper where it is recommended?

Thank you very much for your help

[Sebastian Kripfganz]

Could you please elaborate a bit more on what you mean by "accumulated time dummies". Do you mean something like the following?

time	dummy1	dummy2	dummy3	dummy4
1	1	1	1	1
2	0	1	1	1
3	0	0	1	1
4	0	0	0	1

That would just be a linear transformation of your "ordinary time dummies" without any effect on the estimates of the remaining regressors. The only thing that changes is the interpretation of your time dummy variables.

[Marcos Gonzalez]

Thank you very much for your answer

Actually I thought that it meant what you have tabulated, but it is more something like that

time	dummy1	dummy2	dummy3	dummy4
1	1	0	0	0
2	1	1	0	0
3	1	1	1	0
4	1	1	1	1

the command I have been told to apply is:

forvalues y=year1/yearN {gen yr `y' a = year >=`y'}

So it is generating dummies with value 1 if the year is equal or higher than the reference year

instead of the ordinary dummies command:

forvalues y=year1/yearN {gen yr `y' c = year ==`y'}

I do not know if it has any sense. Becasue until my knowledge ordinary time dummies are ok in GMM.