I am regressing stock market participation (binary variable equal to 1 if an individual owns equities and 0 otherwise) on debt and a bunch of control variables using a logistic regression. debt is not significant. But, if I add risk aversion as an additional control variable, debt becomes significant.

I tested, and did not find any evidence for multicollinearity.

Any ideas why the significance of debt changes when adding risk tolerance?

Similar questions and discussions