Multicollinearity

• The inclusion of independent variables which are correlated with the existing independent variables causes the multicollinearity:
• standard errors: biased upward
• t-statistics: biased downward (deflated)
• If multicolinearity is present in a model, the interpretation of the individual regression coefficients become problematic.
• The existence of multicolinearity is generally signaled by
• High R2 value (e.g. 81%)
• Low t-statistics on the regression coefficients (<2; approximate critical value = 2)