OLS Regression - explanation of coefficients with control variables

[Beatriz Gomes]

Good morning,

I´m doing an OLS regression with the dependent variable having a health number (

Health_Number

- a dummy variable that takes the value of 1 if migrants have the health number and 0 if not - and the independent variable that is years since migrating (

YSM

. To see if as the years since they arrive in the destination country affect their possession of a health number that allows them access to healthcare.
For that I used the controls of Destination Network, having a health visa, being a female, age, having completed at least 12 years of schooling and being employed.

regress Health_Number YSM  
outreg2 using Regression0, excel append ctitle(Basic) dec(3)
regress Health_Number YSM Dest_Network
outreg2 using Regression0, excel append ctitle(Network) dec(3)
regress Health_Number YSM Dest_Network Health_Visa
outreg2 using Regression0, excel append ctitle(Having a Health Visa) dec(3)
regress Health_Number YSM Dest_Network Health_Visa Female Age AtLeast_CompletedSecondaryEduc Employed
outreg2 using Regression0, excel append ctitle(Migrant Controls) dec(3)

I am however having a hard time to understand the main coefficient since it does not change much across specifications (0.070, 0.071, 0.072, 0.071).
Does this means that the controls variables are not explaining much of what´s going on?
Or does it mean the controls are cancelling each other?
Besides, what does it mean if in some of the columns when I added the different specifications, the main coefficient (so the relation between YSM and the dependent variable) loses significance? For instance, instead of a 1% significance it has 5% significance, although it has similar values?

Thank you

[Beatriz Gomes]

Besides, what does it mean if in some of the columns when I added the different specifications, the main coefficient (so the relation between YSM and the dependent variable) loses significance? For instance, instead of a 1% significance it has 5% significance, although similar values.

[Carlo Lazzaro]

Beatriz:
your R_sq (that is really low) does not change that much across your four specifications either.
Provided that a simple OLS (as your Basic specification) can be hardly informative, I'd check the goodness of the functional form of the regressand via -linktest-.
In addition, is a good habit to test whether -age- has a non-linear relation with the regressand (-c.age##c.age-).

As far as your #2 is concerned, the short answer is that different specifications give back differen results (and no wonder about that )!.