A systematic difference between the expected value of a statistical estimate, and the quantity it estimates.