Logit Model...Selecting the right dependant variable for my thesis, Advice needed!!!!

[Enzo Mesanovic]

Hello

First of excuse me for my english it`s been a while and I`m new in Stata so I would appreciate your help

For my thesis I am doing a research about "If foreigners have a higher risk of unemployment than locals" in my work it`s the swiss labour
So I have a few problems, should my binary dependant variable be "Foreigners(1)/Swiss(0)" or "employed(1)/unemployed(0)"
I have a given dataset which was taken as survey of several inhabitans from 2003 - 2013 I have a several dataset for every year

At first I tried to do a probit regression with dependant variable employed(1)/unemployed(0)" and several Independent variables : sex, age, health, civil status, stay (foreingers),
which was succesfull , next step I added some more variables: fulltime or part-time, length of service, profession, thats were stata gave me an error beacause the outcome does not vary and
I realised when I checked with my dependant variable as an example "fulltime or part-time" it only vary with (1) and for the unemployed it`s just 0.

Should I do two different probit regressions where once my dependant variable is (Foreigner/swiss) and the second probit with dependant (employed/unemployed)??



So as conclusion I`m not really sure which dependant variable to choice and how to measure the risk of being unemployed, I clearly have to derivate the risk in fact of the data of unemployed Person or am I wrong? is there maybe another model or another way of thinking?

Thank you a lot in advance
Enzo

[Clyde Schechter]

Well, it sounds like you need to create a new three-level variable taking on values 0 for unemployed, 1 for part-time employed, and 2 for full-time employed. This variable is not dichotomous, so you cannot use it as the dependent variable in a logistic or probit regression, but it is suitable for use as the dependent variable for -ologit-, -oprobit- or -mlogit-.

[Enzo Mesanovic]

Thank you for your response Sir Schechter

In my data set there`s just this one variable TBD1 which means "Employment Status" the outcome is defined as:

1= indpendently 2= Employee 3=apprentice 4= unemployed 5=houswifes 6=pensioner


so what I did was to generate a new variable and gave them the values :

gen employment =.
replace employment =0 if TBD1 >=1 | TBD1 <=4
replace employment =1 if TBD1 ==5

0= employed
1= unemployed

Maybe I am wrong at all because I thought in case of my work which is to measure "if foreigners have a higher risk of unemployment than locals" my dependant variable should be the "Employment Status" or should my dependant variable be if the individual is a Foreigner or a locoal citizen?

I have another variable which is IJ02 and means "nationality" and the outcome is:

50. Swiss
51. Foreigner
52. Dual citizen
-9. no comment

what I did is to generate antother dummy variable

gen nationality =.
replace nationality= 0 if IJ02 == 50
replace nationality= 1 if IJ02 == 51

0= swiss
1= Foreigner

So for my Topic one of this two variable must be my dependant variable or not?
I`m sorry if these questions are maybe ordinary but I'm not sure anymore how to measure if the risk for unemployment is higher for foreigner than for swiss citizen?


I hope that I explained it better than in the first quote, if you have any understanidg question because I didn`t explain it enough please ask me.

thanks a lot
Enzo

[Clyde Schechter]

1= indpendently 2= Employee 3=apprentice 4= unemployed 5=houswifes 6=pensioner


so what I did was to generate a new variable and gave them the values :

gen employment =.
replace employment =0 if TBD1 >=1 | TBD1 <=4
replace employment =1 if TBD1 ==5

0= employed
1= unemployed

Wait, this doesn't look right to me. Your code says employment = 1 if and only if you are a housewife. That can't be right, can it?

It seems to me you should be saying

gen employed = inlist(TBD, 1, 2, 3)

Maybe you want to exclude housewives and pensioners altogether, in which case adding -replace employed = . if inlist(TBD, 5, 6)- would accomplish that.

I also note that your nationality variable leaves missing values for dual citizens. If that's what you want, it's OK. But are you sure that's what you want?

Anyway, what I was suggesting in #2 is something to combine your full-time/parttime variable (let's call it fp--you don't say what its name is) as follows:

gen new_outcome = employed // SEE ABOVE FOR DEFINITION OF employed
replace new_outcome = 2 if fp == "PART TIME" // OR WHATEVER THE CODING FOR PART TIME EMPLOYMENT IS

and then using new_outcome as the dependent variable in an -ologit-, -oprobit-, or -mlogit- analysis.

[Enzo Mesanovic]

Yes you are right, I have forgot about the missing values, and my notation is a little bit amateur so excuse me for that.
thanks for your help I just recoded it Stata

about the ologit, oprobit analysis I will try to run it with my undependant variables hope it works well