test for inverse causality

[David Jones AA]

I need to test my dataset for inverse causality; currently I am trying to test the affect of income on happiness, however being happier may make you more productive, raising income.

I've tried help tsvarlist, and I've tried google, but ultimately I'm a student and I havent been able to figure this out, but I assume the stata command is relatively simple so would appreciate a nudge in the right direction!
Thanks,
David

[Carlo Lazzaro]

David:
welcome to the list.
I do not whether such a test exists in Stata or elsewhere: my take on your issue is that reverse causality (i.e., a form of endogeneity) seems self-evident (just like you diagnosed it).
You may consider replacing income with home ownership, if feasible (I would not see reverse causality with this predictor).
I'm confident that the literature on the economics of happines (which seems growing these days) can give you some hints.

[David Jones AA]

Dear Carlos, I have already researched this topic and its unfortunately quite mixed- I've referred to about six papers in my dissertation to highlight this divide!

In the case of home ownership, again if happier people are richer theyre more likely to be able to afford a home, meaning inverse causality would still be a problem! im currently using ordered logits to compare happiness to:
retirement
age
children
marriage
unemployment
disable
seperated
cohabitting
widow
age squared
ruralregion
london (I have used urban as a base group, but london deserves its own coefficient)
income decile nationally
income decile locally
logwealth
final income

with all these variables will it limit the ability for inverse causality between happiness and income? As I said, I'm not all that familiar with stata!

[Carlo Lazzaro]

David:
thanks for providing further calrifications.
The issue is not being familiar with Stata (or whichever other software) but forestalling the risk of reverse causality.
Perhaps, if you do not have better advice form you teacher/supervisor, you may run two different regression models (with and without predictors concerning income and wealth) or using adequate instrument(s) to tackle endogeneity issue.