Halo error, although traditionally viewed as being detrimental to the quality of social cognition, has been purported to have potentially beneficial effects in some situations (Murphy, Jako, & Anhalt, 1993; Nathan & Tippins, 1990). Consequently, Murphy et al. called for investigations on the effects of inducing halo. The present experiment tested the effect of inducing halo on performance rating accuracy. Participants (N=160) were randomly assigned to either the induced halo or control condition. Participants observed videotaped performances then made job performance ratings after a 24-hour delay. In spite of a high degree of power, the effect of the halo induction on rating accuracy was not significant. Thus, it appears plausible that any potentially beneficial effects of halo on performance ratings may be limited to naturally occurring halo rather than to induced halo.
Halo error is a prominent but poorly understood aspect of social cognition that generations of researchers have considered (Barrette & Haccoun, 1995; Cooper, 1981; Murphy, Jako, & Anhalt, 1993; Nathan & Tippins, 1990). Reflecting recent thinking that has challenged long-held assumptions, this paper describes the results of a unique experimental approach to the investigation of the purported consequences of halo error.
Balzer and Sulsky (1992) defined general impression halo as a general impression bias whereby a rater's overall evaluation or impression of a ratee leads the rater to evaluate all aspects of performance in a manner consistent with this general evaluation or impression (p. 976). Balzer and Sulsky's definition would appear to be consistent with the main thrust of most definitions of halo error (Murphy et al., 1993) and was therefore adopted in this experiment.
For decades, halo was assumed to detract from the goodness of a set of ratings (Cooper, 1981). However, the use of typical halo error measures as indicators or the validity of performance appraisal scores has been questioned (e.g., Balzer & Sulsky, 1992; Cardy & Dobbins, 1994; Murphy & Balzer, 1989). In particular, a meta-analysis has found that commonly used measures of halo error do not covary in the expected direction with performance rating accuracy. Specifically, higher levels of halo do not tend to be associated with less accurate ratings (Murphy & Balzer).
Research in person perception and cognitive information processing suggests that performance raters will tend to form an overall evaluative impression of the ratee, and that forming such an impression facilitates memory of specific behaviors (Wyer & Carlston, 1994). Thus, there would appear to be a cognitive basis for the expectation that halo can contribute to greater accuracy in performance appraisal.
A variety of methods that have been developed to aid in the removal or control of halo (e.g., rater training, statistical control), have had generally disappointing results (Barrette & Haccoun, 1995; Borman, 1991; Murphy et al., 1993; Nathan & Tippins, 1990). Murphy et al. (1993) suggested that halo error may actually be advantageous in applied contexts. When the purpose of appraisal is to distinguish between ratees' overall performance levels (e.g., for promotional decisions) halo may cause raters to focus on the most important features being rated (Murphy et al., 1993) and not unduly weight negative incidents (Nathan & Tippins). Thus, the general impression that is formed of the performance level of each ratee, relative to the other ratees, may be reasonably accurate.
Forming a general impression of the ratee and allowing it to color one's ratings of all performance dimensions is likely to result in higher intercorrelations among performance dimensions. Higher intercorrelations among performance dimensions will tend to increase the reliability of aggregate ratings, which may, in turn, enhance the rater's accuracy in discriminating between ratees' overall performance levels (Murphy et al. …