Error when applying Heckman two-stage correction

[MariaElena Filippin]

Dear all,

I want to apply the Heckman correction to the Oaxaca-Blinder decomposition analysis.
My dataset contains the following variables:

 age           qualificat~n  qual_6        honors        area          sect_1        sect_7        ydate
byage         qual_1        groupqual     homeowner     exper         sect_2        sect_8        monthlywage
female        qual_2        groupq_1      it            exper2        sect_3        sect_9        lnwage
married       qual_3        groupq_2      city_inhab    private       sect_4        overtime
fam_size      qual_4        groupq_3      city_size     temporary     sect_5        condition
child         qual_5        schooling     region        sector        sect_6        lfp

I have created global set of variables to use in the analysis:

global y "lnwage" // dependent variable
global x "age qual_1 qual_2 qual_3 qual_5 schooling honors city_size exper exper2 private temporary sect_1 sect_2 sect_3 sect_4 sect_5 sect_6 sect_7 sect_8 overtime" // independent variable
global z "age married fam_size child qual_1 qual_2 qual_3 qual_5 schooling honors homeowner it city_size exper exper2 private temporary sect_1 sect_2 sect_3 sect_4 sect_5 sect_6 sect_7 sect_8 overtime" //set of covariates for the labour force participation equation (lfp), where lfp is equal to 0 if housewives, equal to 1 if working people

The problem comes when I execute the command:

oaxaca $y $x, by(female) model1(heckman, twostep select(lfp=$z)) model2(heckman, twostep select(lfp=$z))

and I get the following error:

Model for group 1
Dependent variable never censored because of selection:
model would simplify to OLS regression
r(498);

Do you have suggestions on how to overcome this issue?

Thank you in advance!

[Phil Bromiley]

You'll increase your chances of a helpful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. Also, simplify what you post to the minimum needed to generate the problem - all the macro substitutions and labels just make it harder to know what you did.

This message means just what it says. Of the observations for which you have values on all the specified variables, none of the values on the dv are censored (missing). If you id all the usable observations (I think this would do the id: regress $y $x $z), list $y if e(sample)) you'll see lnwage isn't missing any values where all the other variables are present. Without missing on y in observations where all the other variables are non-missing, there is nothing to do selection on.