Carlos,
you are understanding the output properly, as far as I can see. I
cannot say much about the interpretation because of lack of background
information. However, I would be highly suspicious against your
results because the selection effects are so strong. I have never seen
anything like that. What also makes me suspicious is that the number
of obs from the two commands is the same. I suggest you replicate the
models outside oaxaca and check whether the results for the separate
models are the same as the model results obtained by oaxaca. Use
-noisily- or -xb- to make oaxaca's model estimates visible.
ben
On Fri, Jun 13, 2008 at 6:16 PM, Carlos Eduardo Hernandez Castillo
<[email protected]> wrote:
> Hello everyone. "Oaxaca" by Ben Jann (ssc install oaxaca; http://repec.ethz.ch/rsc/ets/wpaper/jann_oaxaca.pdf) computes the the Blinder (1973)-Oaxaca (1973) decomposition and allows the use of Heckit models. If both equations are estimated using "heckman", it deducts "the selection effects from the overall differential and then apply the standard decomposition formulas to this adjusted differential" (Jann, 2008; http://repec.ethz.ch/rsc/ets/wpaper/jann_oaxaca.pdf). As far as I understand, this is equivalent to equation (7) in Neuman & Oaxaca (2004; http://www.springerlink.com/content/x0403312676k0042/).
>
>
> I am not sure if I am interpreting my results in a proper way. I am estimating a wage decomposition. Could you please tell me if I am interpreting my (partly unexpected) results properly?
>
>
> If I do not take selection into account, I get:
>
> . oaxaca ln_saltotal_hora escolaridad experpotencial experpotencial2 mujer, by(dummy_inm_lejos) svy weight(0)
>
>
> Blinder-Oaxaca decomposition
>
>
> Number of strata = 1 Number of obs = 3258
> Number of PSUs = 3227 Population size = 2408211
> Design df = 3226
>
>
> 1: dummy_inm_lejos = 0
> 2: dummy_inm_lejos = 1
>
>
> ------------------------------------------------------------------------------
> | Linearized
> | Coef. Std. Err. t P>|t| [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> Differential |
> Prediction_1 | 8.157024 .0332043 245.66 0.000 8.09192 8.222128
> Prediction_2 | 8.384579 .0794506 105.53 0.000 8.2288 8.540358
> Difference | -.2275553 .08611 -2.64 0.008 -.396391 -.0587195
> -------------+----------------------------------------------------------------
> Decomposit~n |
> Explained | -.0631618 .0686883 -0.92 0.358 -.197839 .0715153
> Unexplained | -.1643935 .069404 -2.37 0.018 -.3004739 -.028313
> ------------------------------------------------------------------------------
>
>
> If I take selection into account, I get:
>
> . oaxaca ln_saltotal_hora escolaridad experpotencial experpotencial2 mujer, by(dummy_inm_lejos) svy weight(0) model1(heckman, select(unionli
>> breocasado nummen6 mujconmen6)) model2(heckman, select(unionlibreocasado nummen6 mujconmen6))
>
>
> Blinder-Oaxaca decomposition
>
>
> Number of strata = 1 Number of obs = 3258
> Number of PSUs = 3227 Population size = 2408211
> Design df = 3226
>
>
> 1: dummy_inm_lejos = 0
> 2: dummy_inm_lejos = 1
>
>
> ------------------------------------------------------------------------------
> | Linearized
> | Coef. Std. Err. t P>|t| [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> Differential |
> Prediction_1 | 7.343391 .1449762 50.65 0.000 7.059137 7.627646
> Prediction_2 | 9.085916 .3827884 23.74 0.000 8.335383 9.836449
> Difference | -1.742525 .4093227 -4.26 0.000 -2.545083 -.9399657
> -------------+----------------------------------------------------------------
> Decomposit~n |
> Explained | -.0600035 .0693592 -0.87 0.387 -.195996 .075989
> Unexplained | -1.682521 .4012326 -4.19 0.000 -2.469218 -.8958244
> ------------------------------------------------------------------------------
>
>
> As can be seen, the differential grows from -.2275553 to -1.742525 (which is huge, given that wage is measured in logs). Notice also that prediction_1 decreases, while prediction_2 increases. This is mostly explained by growth in the unexplained part, which grows from -.1643935 to -1.682521.
>
>
> My interpretation is the following:
>
>
> Group 1 estimated wage was biased upwards, while group 2 wage was biased downwards. Bias in group 1 can be explained in the "traditional" way: People who would receive lower wages are unlikely to work, because those wages are lower than their reservation wages. Bias in group 2 could be explained via reservation wages: reservation wages are higher (more than proportionally) for people with high wages: It's like if the income effect of wages were more important than the substitution effect for group 2, assuming everyone in that group has the same preferences.
>
>
> Am I understanding the output properly?
> What do you think about my interpretation?
>
>
> Thanks in advance for your help.
>
>
> Carlos Eduardo Hernandez
> Colombia
> _________________________________________________________________
> News, entertainment and everything you care about at Live.com. Get it now!
> http://www.live.com/getstarted.aspx
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
