Adding a mediator variable

[Kyle Patel]

Hello,

This may be slightly less of a stata question and more of an econometrics based question. I am looking at the impact of unemployment on mental health. I am using a mental health measure that uses a scale of 0-100. I am using fixed effects on panel data on individuals, in which I am looking at an unemployment transition on my mental health proxy. I have decided to add in a mediator control of a physical healthy proxy. Without my physical health control, I get an insignificant coefficient on unemployment transition variable (i.e. there is no mental health effect when an individual becomes unemployed). Once i add my mediator, I get a significant negative effective (i.e. an individual losing their job, reduces their mental health score by 3 points). I was just wondering if this has anything to do with the mediator I added - does this mean the mediator captures some of the mental health effect when an individual loses their job?
Thank you,
Kyle

[Carlo Lazzaro]

Kyle:
1)interested listers cannot say anything useful without taking a look at what you typed and what Stata gave you back (as recommended by the FAQ);
2) have you already ruled out reverse causation-led endogeneity (ie, that mental health can impact on employment)?;
3) as far as your mediator is concerned, in their valuable https://www.stata.com/bookstore/heal...cs-using-stata, page 201, authors state that observed health state usually comes with an unreported/untested counterparts that remains unobserved. It this part is time-invariant, -fe- will wipe it out, otherwise it may be an other cause of endogeneity.

I'd recommend to discuss all these issues (and probably many other points) with your professor/mentor/supervisor, if you're researching this topic for your dissertation/master thesis.