Perfect Multicollinearity

An exact linear relation among regressors, for example x1i=5x2ix_{1i} = 5x_{2i} for all ii. The first-order conditions become redundant, there is no unique solution, and OLS cannot be computed.

When to use

It usually signals a mistake in the model, such as the Dummy Variable Trap, so treat it as a Model Misspecification to fix rather than a data problem; the milder case is Multicollinearity.

See Lecture 10 and Lecture 8.

Appears in