How do I diagnose multicollinearity using the variance inflation factor?

Question

AcadiFi · Accepted Answer

Multicollinearity occurs when predictors are highly correlated with each other. The consequence is inflated standard errors, unstable coefficient estimates, and counterintuitive signs.

Unlock with Scholar — $19/month

Get full access to all Q&A answers, practice question explanations, and progress tracking.

Start 7-Day Free Trial Compare Plans

No credit card required for free trial

To compute VIF for variable X_j, regress X_j on all other predictors and obtain the auxiliary R-squared, call it R_j^2. Then VIF_j = 1 / (1 - R_j^2). A VIF of 1 means no collinearity. Rules of thumb: VIF above 5 warrants attention, VIF above 10 signals severe collinearity.

Suppose Chenara Vasquez is modeling Quillmark Holdings earnings using GDP growth, industrial production, and capacity utilization. The VIFs come in at 1.3, 9.8, and 11.2 respectively. Industrial production and capacity utilization are clearly duplicative because they capture similar macro signals.

Remedies include dropping one of the redundant variables, combining them into an index, using principal components regression, or collecting more data to reduce estimation noise. Ridge regression also handles collinearity by shrinking coefficients.

Before reacting, confirm the problem is harming inference. If your goal is prediction rather than coefficient interpretation, multicollinearity may be tolerable. But for structural or causal analysis, inflated standard errors undermine every conclusion you want to draw.

How do I diagnose multicollinearity using the variance inflation factor?

Master Level II with our CFA Course

Related Questions

Practice Questions