How to create reference groups

[Sofia Nyström]

Hi,

We have a question regarding how to construct reference groups based on age, gender, country and education. We would like every individual in our sample to belong to only one reference group. The idea is to compare the net worth (nw) of one individual to the average net worth in the reference group of the individual. However, we are not interested in every reference group, only if relative comparisons matter. Thus, creating one dummy for each reference group will not solve this problem as we know it.

A basic version of the model we would like to estimate looks as follows: yit = a + b(nw_it - nw_ReferenceGroupAverage) + controls

Any help would be appreciated.

Best,
Sofia Nyström and Andrea Berggren

[Nick Cox]

I am not clear what you are asking for but http://www.stata-journal.com/article...article=dm0055 may help.

[Sofia Nyström]

Thank you, we solved the problem first by using the bysort and egen commands to create the reference groups and the corresponding average income variables. Then we created new variables for the difference between each individual and his or hers reference groups average income. This way, Stata compares each individual with his or hers reference group.

[Katarzyna Kucaba]

Hi Sofia,
I'm currently constructing similar reference groups following Senik (2010) and Caporale et al. (2009). I am using country, employment status, income , gender and age as reference categories. I wonder if you could help me by explaining a bit more your process when creating the new groups. Especially the part about age as in my case the groups should be 5 years older and younger while all the other categories stay the same, i.e. the reference income category for working females in Spain that are 5 years older and younger than the chosen individual. I hope that it is clear what I am trying to do. Any help would be much appreciated.

Very best,
Kat