Region and year fixed effects model for pooled cross sectional data

[Conor Murphy]

Dear all,

I have a dataset of pooled cross sectional data for individuals across a number of different countries, regions and years. I am using a difference in difference design to compare the effect of foreign aid on citizens' attitudes by comparing both posttreatment citizens (aid active) and pretreatment individuals (aid inactive) with control individuals (non-aid area). In order to ensure that I am comparing citizens who are similar except for their aid status I want to run models which include region and year fixed effects.

In terms of the command required in stata for this my understanding is that because my data is not panel data I would not need to use "xtset" or "xtreg" to run such a model. Would it be acceptable to simply run a regression with dummy variables assigned for region and year ? Would the following be correct in such a case:

xi:reg depvar1 depvar2 control1 control2 i.year i.region

[Carlo Lazzaro]

Conor:
see http://www.princeton.edu/~otorres/DID101.pdf.
As an aside, please note that -xi:- is redundant, as -regress- supports -fvvarlist- notation.

[Conor Murphy]

Thanks for the useful guide Carlo.

My design is somewhat of a pseudo diff in diff in that I only have one control group of individuals and make comparisons using that group as a reference so I'm not actually using the "diff" command.

My concern I'd like to clear up is whether the code I suggest in post 1 is an appropriate means of running a fixed effects (for region and year) model for a regression of pooled cross sectional data.

[Carlo Lazzaro]

Conor:
thanks for further clarifications.
Yes, your code is correct for your research goal.
There's a probable copy and paste typo in your code, as you cannot have two regressands with -regress-.