Journal of Applied Psycholog 1 1982, Vol. 67, No. 2, 177-181 Copyright 1982 by the American Psychological Association, Inc. 002I-9010/82/6702-0177S00.75 Statistical Control of Halo: A Response Frank J. Landy Pennsylvania State University Robert J. Vance Ohio State University Janet L. Barnes-Farrell Purdue University Recently Landy, Vance, Barnes-Farrell, and Steele suggested a method for sta- tistically controlling unwanted variance in performance ratings. The method was based on partial correlation techniques through' which the variance in overall ratings was subtracted from variance in dimensional ratings. The technique.has been attacked on the basis of logic, methods, and conclusions. This article re- sponds to criticisms of the methodology and suggests methods by which more accurate estimates of expected values of dimensional intercorrelations might be obtained. A short time ago we published an article describing a technique for reducing the con- taminating effect of halo on peformance rat- ings (Landy, Vance, Barnes-Farrell, & Steele, 1980). The technique was based on the notion that it might be possible to reduce contamination statistically through the use of partial correlation. As can be seen in the previous three articles, there are those who disagree with our suggestions. In this re- sponse we will deal with the objections raised by those who are critical of the method. Since there are several common themes that transcend the comments of the individual critics, we will address the common criti- cisms for the most part. Where appropriate, we will deal with comments unique to a par- ticular critic. Criticisms The expected value of the average inter- correlations among performance ratings is not .00. It is clear from the introduction of our original article that an idealized goal of our method was to reduce the intercorrelations among dimensions to .00. As our critics point The preparation of this article was partially supported by Contract NOOI4-81-K-0197 from the Organization Effectiveness Branch of the Office of Naval Research to the first author. Requests for reprints should be sent to Frank J. Landy, Department of Psychology, Pennsylvania State University, 417 Bruce V. Moore Building, University Park, Pennsylvania 16802. out, this is unrealistic. Hulin (1982) suggests a general performance factor that results from hierarchically arranged human capac- ities and limitations. Murphy (1982) sug- gests a "general evaluative dimension" that seems to pervade human judgment. Cooper (1981) has recently published a well devel- oped case with respect to the role of true halo in performance ratings. At the outset, we must admit to being guilty as charged. The implicit goal of the technique was to reduce component inter- correlations to .00. As we indicated in dis- cussing the results of that study (Landy et al, 1980), there might be value in looking at an underlying set of relationships with more general influences removed. For ex- ample, we might be interested in developing specific criterion measures for the validation of specific abilities measures. Similarly, in the process of developing an instrument for the evaluation of performance, we might be interested in identifying the specific perfor- mance dimensions in the domain without the obscuring influence of a more general effec- tiveness factor, The reader can imagine other potential uses of performance ratings with general influences eliminated. We agree with our critics that the actual numbers that result from this technique (re- sidual performance scores) are not to be con- sidered in the same way as raw performance scores. In a sense, we may have eliminated the most important information with respect to absolute performance levels of employees. 177