Oaxaca-Blinder decomposition with dummy variables

[Fangxin Liu]

Hello all,

I'm tring to use the Oaxaca command. My code is:

oaxaca logtime age age2 i.i301 i.i18 i.i14 i.sni, by(i10)

It says "factor variables and time-series operators not allowed"

So what should I do if I want to do an Oaxaca decomposition with some dummy variables?

Thank you!

[Carlo Lazzaro]

Fangxin:
welcome to the list.
As the user-written programme -oaxaca- (as per FAQ you should have specified that -oaxaca- is not a built-in Stata command) dates back to 2008 (http://www.stata-journal.com/article...article=st0151), it does not support -fvvarlist- notation.
The usual fix is to prefix your code with -xi-:

xi: oaxaca logtime age age2 i.i301 i.i18 i.i14 i.sni, by(i10)

As an aside, you should avoid crating interactions yourself, as -fvvarlist- can handle them, too.

xi: oaxaca logtime c.age##c.age i.i301 i.i18 i.i14 i.sni, by(i10)

.

More than this, -fvvarlist- can introduce you to two wonderful Stata commands, such as -margins- and marginsplot-.

[Min Zhang]

Dear Fangxin,

Yes you can use dummy variables with oaxaca.

Your code can be represented:

oaxaca y i.x1 i.x2, by(group)

we assume that xi and x2 are three-level categorical variables.
This would not work with oaxaca.

First you need to create dummy variables.

tab x1, gen(x1_)

tab x2, gen(x2_)

You now have dummy variables x1_1 x1_2 x1_3. x2_1 x2_2 x2_3.

Then run oaxaca

oaxaca y normalize(x1_1 b.x1_2 x1_3 ) normalize(x2_1 x2_2 b.x2_3) , by(group)

Please note that no space between normalize and brackets.
Prefix b. indicates the base category.

Also note the oaxaca version.

which oaxaca
c:\ado\plus\o\oaxaca.ado
*! version 4.0.5  24aug2011  Ben Jann

I suggest Ben Jann's oaxaca paper. Or you may want to check the oaxaca help file "Normalization of categorical variables".

Hope it helps.


Regards,
Min