insignificant variables- what to do?

#1
Hello, I'm running a few regressions for an econometrics paper and have a question:

should insignificant independent variables be omitted from the model?
meaning should variables with a p value over 5% be omitted?
If not, how many insignificant variables can be in the regression results without harming the model?

My dependent variable is CRIME, and my independent variables are:
UNEMPLOYMENT (significant)
POVERTY (insignificant)
EDUCATION (insignificant)
LAW ENFORCE (significant)
IMMIGRATION (insignificant)
DRUGS (significant)
GDP (insignificant)
PUNISHMENT (insignificant)

when omitting IMMIGRATION, EDUCATION and PUNISHMENT, all variables are significant except GDP, but the model only have 5 independent variables.


What should i do?
THANKS !~!~!
 

Mean Joe

TS Contributor
#2
You can leave insignificant variables in your model. To show you have adjusted for them. I don't really know your field, but I think EDUCATION would be a nice variable to keep in the model because it is often suspected to affect a thing like crime.

If you leave a variable out, then people may question why you did that, eg they may believe that the significant result you see from DRUGS is tied to the variable you left out.