One form of adjustment named for this property adds half of the variance and one to the lambda = 0 term of a Box-Cox transformation, which otherwise provides the median value. When this property is not present, a Mincer-Zarnowitz test gives an intercept of approximately 0 and a coefficient close to 1. Hetero·skedasticity does not introduce (-5[1])this property (15[1])for (15[1])OLS, but does for non-linear (15[1])models. (15[2])Per the Gauss-Markov theorem, the OLS estimator (15[1])has (15[1])the (-5[1])(*) lowest variance (10[3]-5[2])among (10[2])linear estimators (-5[1])that lack this (10[1])property, (10[7])as indicated by the expression (10[1]-5[1])BLUE. (10[1]-5[1])Forecasting models face a tradeoff between variance and (10[1])this property, (10[6])which includes an (10[1])“omitted-variable” form. For 10 points, (10[1])name this average difference between the expected and (10[1])true values of an estimator, (0[1]-5[2])a measure of distortion. ■END■ (10[6]0[9])