Statistical and econometric techniques of correcting for supervisor bias in models of individual performance appraisal were developed, using a variant of the classical linear regression model. Location bias occurs when individual performance is systematically overestimated or underestimated, while scale bias results when raters either exaggerate or minimize the differences between individuals. A multi-scale model was designed to deal with subjective supervisory ratings, and it is described in detail. Several specific multi-scale estimating techniques were established, including equal total variance, equal residual variance, maximum likelihood, and least squares. Asymptotic results were derived for each of the four techniques. However, because of the difficulty in deriving small sample properties analytically, Monte Carlo experiments were also conducted. Maximum likelihood and equal residual variance techniques gave consistent parameter estimates. However, for small sample sizes and configurations for cases with large random errors, the equal total variance residual estimator was preferred. (Author/MV)