What problem does multicollinearity cause in regression analyses?

• A. Decreases R-squared.
• B. Inflates standard errors and destabilizes coefficients.
• C. Increases model interpretability.
• D. Reduces sample size.

Answer: (B)

When predictors in a regression are highly correlated, they don’t provide independent information about the outcome. That makes it hard to tease apart each predictor’s unique effect. As a result, the estimated coefficients become unstable across different samples and their standard errors grow larger. Larger standard errors mean wider confidence intervals and less reliable t-tests, so we can be unsure about whether a predictor truly has an effect. The model may still fit the data well overall (R-squared can stay high or even rise because of redundancy among predictors), but the precision of each coefficient estimate collapses and interpretation becomes ambiguous. Multicollinearity doesn’t inherently reduce the sample size, and while it can complicate interpretation, it’s the inflated standard errors and unstable estimates that are the hallmark problem.