Abi -

If you need multiple regression, but only use simple regression, then the intercept tries to make up for that. The question is, if your independent variables are zero, would you expect y to be zero?

"Significance" is nebulous. If you are trying to find the best prediction model, you could try a few reasonable models, and compare performances using graphical residual analyses, and if you could save out some test data, you could compare predictions, which may help avoid overfitting.

Cheers - Jim

PS - By "constant," I assumed you were referring to an intercept term in linear regression.

What is "JS?"

Sorry job satisfaction

Is that a continuous variable?  Could regression be nonlinear?  There are different kinds of regression you can use.  

You might look at the following, and click on the "lessons" at the bottom that are of interest to you:

 https://onlinecourses.science.psu.edu/stat501/,

Pennsylvania State University, Eberly College of Science,

Including: https://onlinecourses.science.psu.edu/stat501/node/369/,

Lesson 15: Logistic, Poisson & Nonlinear Regression

Also note subsections of lessons, such as the following:

 https://onlinecourses.science.psu.edu/stat501/node/324/,

9.7 - Polynomial Regression

Graph your data. Decide if your goal is prediction or explanation and then come back with your question. Best, David Booth