What effect does multicollinearity have on regression analysis?

• A. It improves statistical power
• B. It has no effect on estimates
• C. It makes coefficient estimates unstable and hard to interpret
• D. It eliminates confounding

Answer: (C)

Multicollinearity occurs when two or more predictors in a regression model carry overlapping information because they are highly correlated. This doesn’t bias the estimated relationship on average, but it does inflate the standard errors of the affected coefficients. With larger standard errors, the estimates become unstable: small changes in the data or in the model can lead to big swings in the coefficient values or even change their signs. That makes it hard to interpret the independent effect of each predictor, since the precision of those estimates is reduced and the associated p-values and confidence intervals become unreliable. It also lowers the model’s effective power to detect true effects for the collinear predictors. This is not a method for removing confounding; multicollinearity mainly worsens precision and interpretability. A practical check is the variance inflation factor—high values signal problematic multicollinearity.