Am having problem using Blinder-Oaxaca decomposition command as used by Ben Jann

[olorunfemi sola]

Hi All
Am trying to use the commands use by Ben Jann in his article ‘The Blinder-Oaxaca decomposition for linear regression models” on his own data but I have the following errors for each of the commands after using them as stated below:

1. svyset [pw=wt]

pweight: wt
VCE: linearized
Single unit: missing
Strata 1: <one>
SU 1: <observations>
FPC 1: <zero>

. pweight: wt
unrecognized command:pweight
r(199);

end of do-file

r(199);


(ii) tabulate isco, nofreq generate(isco)
.oaxaca lnwage educ exper tenure isco2-isco9, by(female) pooled ///
> detail(exp_ten: exper tenure, isco: isco?) categorical(isco?)
option > not allowed
r(198);

end of do-file

r(198);
Can anybody assist me to fix it?
Thanks in anticipation
Sola Olorunfemi

[Zuhumnan Dapel]

To have the first situation squared out please go for

help svy

[Clyde Schechter]

It appears you are copying and pasting from Ben Jann's article directly into Stata without separating commands from output, and dealing with line breaks properly.

So, in the first situation, -svyset [pw = wt]- is a command, but everything after that, including pweight: wt is Stata's output in response to that command. So you can't feed that back in as if it were commands.

In the second situation,

oaxaca lnwage educ exper tenure isco2-isco9, by(female) pooled ///
> detail(exp_ten: exper tenure, isco: isco?) categorical(isco?)

is how Stata logs will record a long command that is split over two lines. In particular, the > character is not part of the command, but is inserted as Stata logs it to call visual attention to the continuation of the command on a second line. But, not being part of the command, if you try to feed it into Stata as if it were, Stata complains. The correct way to issue this command to Stata is:

oaxaca lnwage educ exper tenure isco2-isco9, by(female) pooled detail(exp_ten: exper tenure, isco: isco?) categorical(isco?)

More generally, just copying and pasting log-file output into Stata as commands will not work. It requires a bit of editing.

[olorunfemi sola]

I want to thank Clyde Schechter for been able to address my ignorance. Your suggestions work perfectly. I appreciate Zuhumnan Dapel too. Please one more thing, why is Ben Jann using " isco2-isco9" as one of his variables as this is not in the table showing his data set? Thanks in anticipation.

[Nkanyiso Maphanga]

Good day

I'm having a problem running the oaxaca command I get the "command oaxaca is unrecognized " error, this is the code i sued: "oaxaca lfps yrschool age agesq maristat exprnc coloured indian white , by(male) noisily if qtr==1"

What can I do differently?

[daniel klein]

I get the "command oaxaca is unrecognized " error, this is the code i sued: "oaxaca lfps yrschool age agesq maristat exprnc coloured indian white , by(male) noisily if qtr==1"
What can I do differently?

oaxaca is community-contributed software and not part of official Stata; you will have to install the command, first. I am not sure this is the latest version, but you can install the command by typing

ssc install oaxaca

If this does not solve the problem, you probably need to change your ado paths (we will deal with this if needed).

Best
Daniel

[Nkanyiso Maphanga]

Thank yo that was helpful, i was able to run my code.

[Bijoyata Yonzon]

Hi,
I am using Stata 14 and am trying to use oaxaca command for decomposition of earnings (labour market) for different groups. I have installed package oaxaca from

http://fmwww.bc.edu/repec/bocode/o: 'OAXACA': module to compute the Blinder-Oaxaca decomposition

Thhe command I tried was

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq normalize(SocialGrp1 SocialGrp2 SocialGrp3 b.SocialGrp4, Relign1 Relign2 b.Relign3), by(Female1) detail pooled
Result
b.Relign3 not allowed
r(198);

where lgwg is usual log of earnings; Rural1 is binary variable (1-rural/0-urban0); Nevermarried1- binary variable (1-not married/0-married); TechEdu1- binrary (1-has technical education/0-does not have); YearsofSchooling- continuous variable (no.of years taken in school); Age and age sq are continuous variables; Socialgrp1-4 are dummies for 4 category variable Socialgroup with 4th group as the base; Relign1-3- 3 dummies for 3 category variable Religion with 3rd category as the base.
I used normalize, as I had read in Ben Jann's (https://www.stata-journal.com/sjpdf....iclenum=st0151) article and a chapter by Fortin. N., et.al (https://www.sciencedirect.com/handbo.../vol/4/part/PA) that the choice of base group has serious effect on the results.

Removing Relign

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq normalize(SocialGrp1 SocialGrp2 SocialGrp3 b.SocialGrp4), by(Female1) detail pooled

oaxaca_normalize(): 3498 inconsistent dummy set: SocialGrp1 SocialGrp2 SocialGrp3 SocialGrp4
<istmt>: - function returned error
r(3498);

Using categorical instead of normalize

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq SocialGrp1 SocialGrp2 SocialGrp3, by(Female1) detail pooled categorical(SocialGrp1 SocialGrp2 SocialGrp3 b.SocialGrp4)

factor variables and time-series operators not allowed
r(101);

wihtout specifing which one to take as base group

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq SocialGrp1 SocialGrp2 SocialGrp3, by(Female1) detail pooled categorical(SocialGrp1 SocialGrp2 SocialGrp3 SocialGrp4)
oaxaca_normalize(): 3498 inconsistent dummy set: SocialGrp1 SocialGrp2 SocialGrp3 SocialGrp4
<istmt>: - function returned error

but if i run the command taking my specific base and without normalize/ categorical

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq SocialGrp1 SocialGrp2 SocialGrp3 Relign1 Relign2 ,by(Female1) detail pooled

Blinder-Oaxaca decomposition Number of obs = 84,279
Model = linear
Group 1: Female1 = 0 N of obs 1 = 63988
Group 2: Female1 = 1 N of obs 2 = 20291

	Robust
LgWg	Coef.	Std. Err.	z	P>z	[95% Conf.	Interval]
overall
group_1	4.526128	.0033489	1351.54	0.000	4.519564	4.532692
group_2	3.882628	.00634	612.41	0.000	3.870202	3.895054
difference	.6435	.0071701	89.75	0.000	.6294469	.6575531
explained	.2017233	.0049231	40.98	0.000	.1920742	.2113723
unexplained	.4417767	.0055427	79.70	0.000	.4309133	.4526402
explained
Rural1	.0101074	.0007199	14.04	0.000	.0086965	.0115183
NeverMarried1	-.0059328	.0005666	-10.47	0.000	-.0070434	-.0048222
TechEdu1	.0038111	.0005352	7.12	0.000	.0027621	.00486
YearsOfSchooling	.1812202	.0037093	48.86	0.000	.1739501	.1884902
Age	.0050753	.0055929	0.91	0.364	-.0058865	.0160371
AgeSq	-.0028538	.0039108	-0.73	0.466	-.0105188	.0048112
SocialGrp1	.0007689	.0002372	3.24	0.001	.0003039	.0012339
SocialGrp2	.0042137	.0005155	8.17	0.000	.0032033	.005224
SocialGrp3	.0036787	.0005894	6.24	0.000	.0025235	.0048338
Relign1	.0013467	.0004311	3.12	0.002	.0005017	.0021917
Relign2	.000288	.0006321	0.46	0.649	-.0009509	.0015268
unexplained
Rural1	-.0059295	.0072843	-0.81	0.416	-.0202065	.0083475
NeverMarried1	.0101336	.0037994	2.67	0.008	.0026869	.0175804
TechEdu1	-.0056356	.0017255	-3.27	0.001	-.0090176	-.0022536
YearsOfSchooling	-.0692977	.0065519	-10.58	0.000	-.0821392	-.0564563
Age	.7408358	.0959907	7.72	0.000	.5526975	.9289741
AgeSq	-.2237564	.0471882	-4.74	0.000	-.3162435	-.1312692
SocialGrp1	-.0206385	.0027832	-7.42	0.000	-.0260934	-.0151835
SocialGrp2	-.0138324	.0042121	-3.28	0.001	-.022088	-.0055769
SocialGrp3	-.0035054	.005742	-0.61	0.542	-.0147594	.0077487
Relign1	.0027664	.0016095	1.72	0.086	-.0003881	.005921
Relign2	.0054985	.0021415	2.57	0.010	.0013012	.0096958
_cons	.0251378	.05597	0.45	0.653	-.0845614	.1348371

the result obviously keeps changing if i change my base category. I tried uninstalling and re installing the package as well. Kindly help.

[Sven-Kristjan Bormann]

As far as I understand the package, then the package behaves as expected and re-installing won't save the problem.
According to my understanding of the package, I would change the line

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq normalize(SocialGrp1 SocialGrp2 SocialGrp3 b.SocialGrp4, Relign1 Relign2 b.Relign3), by(Female1) detail pooled

to

oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq normalize(SocialGrp1 SocialGrp2 SocialGrp3 b.SocialGrp4) normalize(Relign1 Relign2 b.Relign3), by(Female1) detail pooled

Using two normalize groups instead of one. That's at least how I deal with multiple categorical variables in my decomposition.
As far as I understand the help file of the package, then the normalize "option" can only deal with one group of categorical or interacted variables at a time.

[Bijoyata Yonzon]

Hi Sven-Kristjan,
Thanks you for responding to my query. I tried your suggestion by using normalize separately for two groups but got an error.

. oaxaca LgWg Rural1 NeverMarried1 TechEdu1 YearsOfSchooling Age AgeSq normalize(SocialGrp1 SocialGrp2 SocialGrp3 b.SocialGrp4) normalize(Relign1 Relign2 b.Relign3), by(Female1) detail pooled
oaxaca_normalize(): 3498 inconsistent dummy set: SocialGrp1 SocialGrp2 SocialGrp3 SocialGrp4
<istmt>: - function returned error

[Sven-Kristjan Bormann]

Could you show an example of your dataset? Just run the dataex command in Stata and post the results here. If you do not have the dataex command installed, then run

ssc install dataex

.
My guess is that there is something wrong with your SocialGrp variable. As far as I understand the code of the oaxaca_normalize function, then you should have somewhere the case that an individual belongs to two social groups. If my understanding is correct, then oaxaca command points you to a problem in your dataset.

[Bijoyata Yonzon]

Hi Sven,
Thank you for the response.
I ran the said command dataex for the categorical variables that I want to use in my decomposition

dataex SocialGrp1 SocialGrp2 SocialGrp3 SocialGrp4 Relign1 Relign2 Relign3
copy starting from the next line -----------------------
* Example generated by -dataex-. To install: ssc install dataex
clear
input double(SocialGrp1 SocialGrp2 SocialGrp3 SocialGrp4 Relign1 Relign2 Relign3)
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 1 0 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 1 0 0 0 1 0
0 1 0 0 0 1 0
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
1 0 0 0 0 0 1
1 0 0 0 0 0 1
1 0 0 0 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 1 0 0 0 1
1 0 0 0 0 0 1
1 0 0 0 0 0 1
0 0 1 0 0 0 1
0 0 1 0 0 0 1
0 0 1 0 0 0 1
0 0 0 1 0 0 1
0 0 1 0 1 0 0
0 0 1 0 1 0 0
0 0 1 0 1 0 0
0 0 1 0 1 0 0
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 1 0 0
0 0 0 1 1 0 0
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 1 0
0 0 0 1 0 0 1
0 1 0 0 0 0 1
1 0 0 0 0 0 1
0 0 1 0 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 1 0 0 0 1
0 0 1 0 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 1 0 0
0 0 1 0 0 0 1
0 0 1 0 0 0 1
0 0 1 0 0 0 1
0 0 0 1 1 0 0
0 0 1 0 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
0 0 0 1 0 0 1
end
copy up to and including the previous line ------------------
Listed 100 out of 84322 observations
Use the count() option to list more
.

[Sven-Kristjan Bormann]

Your data example looks fine for me, so the problem might be in a part that you did not show.
You could run something like

egen SocialGrp = rowtotal(SocialGrp*) //Calculate in how many social groups a person is.
count if SocialGrp!=1 //Gives you the number of observations which are in more than one social group or in no Social Grp
list if SocialGrp!=1 // Show the faulty observations

This code reproduces somewhat the behaviour of the oaxaca command when it fails.

[Bijoyata Yonzon]

Thank you for your response.
I have not run the above command will try that, though I tried to work around by making only two categories for my variable social group. Am I right in not using the normalize option for variables with only two categories?
Thanks lot for your help

[Sven-Kristjan Bormann]

Yes, you are correct that you do not need the normalize option for variables with two categories. That's at least my understanding.