Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Social Judgment Biases in Comparable Worm Analysis Leslie Z,ebrowitz McArthur The pros and cons of a policy of remunerating jobs according to their comparable worth have been hotly debated in recent months. Most of the arguments in this debate have been economic ones. An issue that has received less attention, but that is equally problematic, is the difficulty of generating accurate job evaluations. Determining the comparable worth of jobs involves three majorjudgmental tasks: (1) to choose the set of compen- sable factors that contribute to the value of jobs as well as the weight that should be accorded to each factor, (2) to describe the jobs, and (3) to evaluate how much of each compensable factor is contained in each job. Because these tasks are difficult and inherently subjective, they can overtax people's ability to render accurate, unbiased judgments. In particular, considerable research has established that when people are faced with a complex inferen- tial task, such as comparable worth assessment, theirjudgments are vulnera- ble to a number of biases resulting from unequal attention to the various elements of information provided. Such selective attention can favor infor- mation that is most readily recalled, an "availability bias." It can also favor evaluatively consistent information, a "halo bias." And it can favor informa- tion that is congruent with existing beliefs, an "expectancy bias." It is important to note that these biases do not necessarily indicate any desire to distort the information provided; rather, it is people's cognitive limitations that produce biased judgments. The purpose of this paper is to discuss evidence documenting social judgment biases as it pertains to comparable worth analyses and to suggest research that could elucidate the least-biased procedures for establishing the comparable worth of jobs. 53
54 MCARTHUR CHOOSING AN INSTRUMENT FOR JOB DESCRIPTION In choosing a job description instrument for comparable worth analyses of men's and women's jobs, the goal is to select one that (1) provides job descriptions that are not biased in favor of either men's jobs or women's jobs, (2) differentiates equally well among women's jobs and men's jobs, and (3) measures all job elements that are deemed pertinent to wages. Although considerable research will be necessary to determine whether these condi- tions are met by any of the existing job description procedures, it is clear that certain methods are much more vulnerable to bias than others. On one hand there are qualitative, narrative descriptions of jobs. Unconstrained by partic- ular rating dimensions, such descriptions are likely to be heavily influenced by biases resulting from unequal attention to various aspects of the informa- tion provided. Structured job descriptions, on the other hand, are likely to minimize bias that results from a failure to even think about job elements that are not already salient. Such descriptions have the further advantage of allowing one to quantitatively compare a variety of different jobs on a set of common dimensions. In addition, they permit one to assess the consensual validity of job descriptions by comparing those provided by a variety of sources, something that is less feasible with narrative descriptions, which cannot readily be provided by incumbents, supervisors, and outside experts alike. Despite their advantages, structured job descriptions are not without prob- lems. All the biases discussed in this paper may enter into job descriptions even when they are provided through ratings on a set of structured dimen- sions. Another problem is that existing measures may be more applicable to some jobs than to others. For example, one structured job analysis measure, the Position Analysis Questionnaire (PAQ) (McCormick et al., 1969) has been reported to be better at differentiating among manual occupations, which are typically held by men, than among clerical ones, which are typically held by women (Frieling, 19771. To be useful for purposes of establishing comparable worth, a structured job analysis must be equally applicable to a wide variety of jobs. The development of a structured instrument that provides descriptions of men's and women's jobs that are not obviously sex-biased and that differenti- ates equally well among various occupations is a necessary but not sufficient condition for establishing the comparable worth of jobs. One must also establish the validity of the instrument for measuring job elements that are deemed pertinent to wages. An instrument that is useful for personnel selec- tion e.g., determining the kinds of abilities that are needed for a particular job may or may not be useful for setting wages. However, at least one instrument does have proven validity in measuring job elements that are
SOCIAL JUDGMENT BIASES 55 relevant to both. The Position Analysis Questionnaire has been shown to predict not only aptitude requirements for a wide variety of jobs but also their going rates of pay (McCormick et al., 19721. Some advocates of the structured job analysis procedure argue that the ability of an instrument like the PAQ to predict existing wages renders it sufficient to determine job worth. In particular, they argue that one should employ statistical procedures to determine what elements of job descriptions predict existing wages and from these data determine the worth of various jobs (e.g., McCormick et al., 1972~. Others (e.g., Treiman and Hartmann, 1981), however, have aptly pointed out that this "policy capturing" approach yields comparable worth scores for jobs that perpetuate any ineq- uities in existing wages. For this reason, although structured job analysis procedures may be useful for describing jobs, evaluating the comparable worth of jobs requires determining which job elements should be compen- sated, rather than relying on market indicators of which job elements are currently compensated. Given that this is the ultimate goal ofthe job descrip- tion, it might make more sense to begin by defining the compensable factors and then generating a descriptive instrument that has been specifically designed to describe job elements that are likely to be related to the compen- sable factors. Such an approach could aid in difficult decisions regarding the content of the job descriptions, such as how much coverage should be given to job elements (what the worker does), worker attributes (skills, training), and the job environment. POTENTIAL BIASES IN JOB DESCRIPTIONS AND EVALUATIONS The Availability Bias The availability heuristic (Tversky and Kahneman, 1974) is the tendency of people to assess the frequency or probability of events by the ease with which instances or occurrences can be brought to mind. In the realm of job descriptions it would be manifested in a tendency to describe the frequency of certain job-related activities according to their ease of retrieval from memory. However, ease of retrieval may not bear a one-to-one relationship to the actual frequency of the activity. Rather, it may be influenced by variables such as the familiarity of the activity, its recency, or its perceptual salience. One example of bias resulting from the availability heuristic is provided in research that has demonstrated that people tend to overestimate their own contribution to products and outcomes that are produced jointly with others (Ross, 19811. This tendency seems to be produced by the greater availability in memory of one's own efforts. Such a bias in job analysis could cause
56 McARTHUR incumbents to overestimate their own job's contribution to various endeav- ors. Similarly, it could cause supervisors to underestimate the contributions made by the jobs of their subordinates insofar as their own contributions are more available in memory. In addition to the potential of the availability heuristic to produce this general egocentric bias in job descriptions, it could have more specific effects. For example, secretarial job descriptions pro- vided by supervisors might overestimate the job element of typing because this noisy activity is more perceptually salient than is the quieter activity of composing literate and tactful letters. Similarly, job descriptions for physi- cians might overestimate the job element of examining patients because this is a more familiar and hence more available element of the job than is bool~eeping, which may be underestimated. Generally speaking, the operation of the availability heuristic could pro- duce different job descriptions by incumbents, supervisors, and outside experts. More specifically, outside experts may overestimate the frequency ofthose activities that are the most familiar to them; incumbents may overes- timate the frequency of those activities that they have performed most recently; and supervisors may overestimate the frequency ofthose activities that are the most perceptually salient- i.e., those activities that supervisors can see or hear. The potential influence of the availability heuristic suggests that in implementing any job evaluation procedure, one very important decision to make is who will describe the jobs . There are certain advantages to using job descriptions prepared by incum- bents. For one thing, incumbents have the potential to provide the best- informed description of the job tasks and the working environment, since they are the most intimately involved with the jobs. Incumbents' job descrip- tions should also be less costly to obtain than descriptions from outside experts. In addition, one might expect that incumbents' job descriptions would be the most satisfactory to employees involved in litigation to raise their wages. Of course, the acceptability of these descriptions to employers must also be ensured. More specifically, it must be demonstrated that job descriptions by incumbents are reliable and valid. Unfortunately, this is not an easy task. There is some evidence for the reliability of incumbents' job descriptions. For example, a longitudinal study found that individuals' perceptions of task characteristics were relatively stable over a 3-month period (Griffen, 19811. In addition to evidence for the reliability of job descriptions across time, there is also evidence for inter-rater reliability (e.g., Desmond and Weiss, 1973, 1975; Fischer and Sobkow, 19791. More specifically, when incum- bents from a wide variety of occupations provided questionnaire ratings of the degree of various abilities that their own jobs required, the results
SOCIAL JUDGMENT BIASES 57 revealed high inter-rater agreement among workers in a given job regarding the requisite abilities. Moreover, there was more agreement in the ratings of workers on the job than among workers in general. It thus appears that incumbents can provide reliable information about the ability requirements of their own jobs. Establishing the validity of incumbents' job descriptions is a more diff~- cult task than establishing their reliability. However, there is some evidence regarding convergent validity. Generally speaking, the ability requirements described by incumbents, supervisors, and experts in the foregoing research were all very similar to one another, but there were some discrepancies that is, incumbents would rate a particular ability as vein important for job performance, while experts would not, or vice versa. Discrepancies in the job descriptions provided by different sources have been found for job elements other than requisite abilities. In particular, researchers found dif- ferences in perceptions of the work climate by workers and employers (Narayanan and Venkatachalam, 19821. Further research is needed to deter- minewhetherthere is convergent validity in various sources' descriptions of still other job elements, such as the responsibility or the effort that a job entails. Job descriptions on these dimensions seem more vulnerable to the availability bias, and it is possible that incumbents will show less agreement with supervisors or experts than they do when describing requisite abilities. It should be noted that the availability bias can affect evaluations of job worth on various compensable factors just as it can affect job descriptions. If an analyst reads an entire job description and then makes ratings of job worth on a number of dimensions, these ratings may be overly influenced by the descriptive elements that are most available in memos. Moreover, if the jobs have been labeled or are otherwise identifiable, the descriptive elements that are most available in memory may well be those that are most familiar by virtue of preconceptions about the jobs. If so, the resulting evaluations of worth will be biased toward the status quo. One way to deal with this problem is to have each element in the job description rated separately as to its worth on the set of compensable factors. Another is to have this task performed by a computer programmed to allocate some predetermined "worth" or "job points" for each element in the job description. The Halo Bias Considerable research has demonstrated that most social judgments tend to be evaluatively consistent. Thus, for example, people who are labeled as "good" on some important dimension are surrounded with a positive aura, or "halo," which causes other positive qualities to be ascribed to them.
58 McARTHUR Conversely, people who are labeled as "bad" on some important dimension are perceived as having a number of other negative qualities the "negative halo" or "forked tail" effect (e.g., Kelley, 19501. The operation of halo biases in social judgments may have important implications for comparable worth analyses. In particular, job descriptions and evaluations of worth may be strongly influenced by a job's prestige or its assumed salary. In keeping with this suggestion, Grams and Schwab (1985) found that when a job was presented as having low pay, it was evaluated as less worthy one series of compensable factors then the same job presented as having higher pay. The power of the halo bias in assessments of job worth was even more strikingly revealed in the results of a study by McArthur and Obrant (19841. These authors found that analysts' judgments about the traditional compensable factors of skill and responsibility did not predict their evaluations of a job's monetary worth, while their judgments about the job's salary and its desirability (ignoring salary) were strongly predictive of rated worth. A third variable that independently predicted ratings of the job's monetary worm was the gender of the incumbent: female incumbents were associated with lower worth. Since it has been found that rising numbers of female incumbents may lower the prestige of an occupation while rising numbers of male incumbents may raise occupational prestige (Touhey, 1974a, 1974b), this result can also be interpreted as a halo bias. The tendency for a woman incumbent to depress judgments of a job's monetary worth suggests that a halo deriving from the sex composition of a job may also bias job analysis. More specifically, occupations that are dominated by men may be evaluated more favorably than those that are dominated by women. However, Grams and Schwab (1985) found that the manipulated percentage of women in a job had no consistent impact on evaluations of the job on three compensable factors: education required, experience required, and job complexity. Similarly, whereas the gender of the incumbent had strong effects on job descriptions and evaluations of worth in the research of McArthur and Obrant (1984), the manipulated sex composition of workers in the occupation had no significant effects at all. 1 While these findings do not reveal the predicted halo bias, they are actually consistent with another social judgment bias the "vividness bias." That is, it has been found that people's social judgments are relatively insensitive to abstract information concerning population "base rates" at the same time that they are overly sensitive to the more vivid "target case" information 1 In keeping with the findings of Grams and Schwab (1985) and McArthur and Obrant (1984), recent studies have failed to replicate Touhey's (1974a) finding that the rated prestige of an occupa- tion is influenced by the percentage of women workers (e.g., Suchner, 1979; White et al., 1981).
SOCIAL JUDGMENT BIASES 59 concerning a particularindividual (Nisbett and Ross, 19801. Consistent with this principle, the general information provided to analysts about the sex composition of the occupation seems to have no impact on job evaluations, while the specific information about the incumbent's gender has a strong effect. Although the foregoing discussion provides some reason to be optimistic that a job's sex composition will not bias descriptions, caution must be exercised in extrapolating from the existing research to job evaluations in the real world. In particular, the experimental studies have manipulated infor- mation about sex-neutral jobs in order to create the impression that the percentage of women workers was high, moderate, orlow. Sex composition effects may be much more potent when job descriptions are provided for occupations that are in fact filled predominately by workers of one sex or the other. Since many job titles can reveal to people both the sex composition and the salary for the job that is being evaluated, the documented halo effects indi- cate the desirability of withholding such information from job analysts. Unfortunately, this is impossible when incumbents or their supervisors are providing job descriptions. Even when outside experts provide such descrip- tions, the information about job activities that they are given will often be sufficient to infer the job title and thus its salary and probable sex composi- tion. Indeed, there is some evidence for the parallel effects of job activity descriptions and job titles on evaluations. Crowley (1981) found that sub- jects showed as much agreement regarding the status of various job activities as they did regarding the status of various job titles, and the activities elicited individual interest profiles that were just as sex-biased as those elicited by titles. While it thus seems unlikely that job descriptions could ever tee gener- ated by people who are unbiased by knowledge of the job's sex composition, prestige, and salary, it may be possible to conceal such information from those who use the job descriptions to evaluate the worth of the jobs on various compensable factors. As noted earlier, this phase of comparable worth analysis could even be performed by a computer, thereby ensuring that there is no halo bias. It should be noted that a halo can be produced by more minorjob attributes than salary or prestige. In particular, there is evidence to indicate that if the first piece of information provided in a social judgment task is positive, then evaluations of subsequent information will be more positively biased than those made when the first piece of information is negative. For example, evaluations of an individual's task performance were more positive when that person's initial performance was successful than when it was unsuccess- ful, even if the success rate was identical to those of other individuals (Jones and Goethals, 1972~. This kind of primacy effect has important implications
60 MCARTHUR for the process of job analysis, since it is possible that early information will have disproportionate influence on job descriptions. If the initial informa- tion is highly positive, ratings on various descriptive dimensions may be positively inflated, whereas if the initial information is negative, these rat- ings may be negatively inflated. One way to deal with the problem is to instructjob analysts to attend equally to all pieces of information, since it has been found that such instructions can eliminate the primacy effect (Anderson and Hubert, 19631. Unfortunately, however, there is some evidence to suggest that with such instructions, a recency effect obtains that is, judg- ments are overly influenced by later information. A second way to deal with this problem is to have the jobs systematically described by a number of raters, each of whom receives information about the job in a different order. Another kind of primacy halo concerns the order of the ratings as opposed to the order of the information. If the job yields highly positive ratings on the first descriptive dimension e.g., mathematical skills required for the job- this positive impression may spill over onto subsequent dimensions e.g., responsibility required by the job. Once again, this problem can be handled by having the jobs described by a number of raters, each of whom receives a different order of the rating dimensions. The Expectancy Bias Considerable evidence indicates that people's expectancies can exert an important influence on their social judgments. More specifically, people's judgments about the characteristics of other people tend to be overly influ- enced by information that confirms what they expect these characteristics to be (see McArthur, 1981, for a review of this literature). In the realm of job descriptions, an expectancy bias would be manifested in a tendency to describe a job with whatever characteristics are culturally expected for that job. Some evidence for this effect is provided in a study by Weiss and Shaw (1979), who found that workers' judgments regarding the motivating poten- tial of various tasks were influenced not only by objective task differences but also by the attitudes of other workers toward the tasks. Such effects of social influence may also obtain for other judgments about jobs, such as the degree of various skills that they require or the negativity of the working conditions. The result may be that job descriptions reflect at least in part social stereotypes rather than the true nature of the jobs. The influence of cultural expectations on job descriptions is particularly problematic when the goal is to describe accurately jobs that are segregated by sex. This is because attitudes toward men and women may influence descriptions of male and female jobs. There are widespread stereotypes
SOCIAL JUDGMENT BIASES 61 concerning differences in ability and personality between men and women, stereotypes that are shared by members of both sexes. For example, in a now classic study, Broverman et al. (1972) found that a psychologically healthy woman, in comparison with a psychologically healthy man, was described by men and women alike as less independent, less competitive, less skilled in business, less of a leader, less ambitious, and less interested in math and science. On the positive side, a psychologically healthy woman was described as more tactful and more aware of others' feelings than was a psychologically healthy man. Such sex stereotypes may well create assumptions regarding what abilities or personality traits are needed for women's work as opposed to men's work, and job descriptions provided by incumbents, supervisors, and experts may reflect these culturally shared expectations rather than reflecting the true requirements of a job. Thus, for example, there may be a bias against describing a woman's job as requiring a high degree of leadership or mathe- matical skill, as well as a bias against describing a man's work as requiring a high level of supervision or interpersonal sensitivity. Differential expectan- cies regarding the skills and abilities of men and women can bias job descrip- tions in another, more subtle way namely, by influencing job analysts' causal attributions for successful task performance by men and women. Considerable evidence indicates that for many tasks, men tend to attribute their successes to ability and their failures to bad luck, while women tend to do just the opposite, a difference that is due to men's greater expectancy for success (e.g., Deaux and Farris, 1977; Feather, 1969; Frieze, 19771. More- over, when luck cannot reasonably be invoked to explain successful perfor- mance, women tend to attribute it to their high degree of effort, still eschewing ability as a cause. For example, a study of men and women in first-level management positions in a number of large organizations revealed that neither the women nor the men believed that luck had much to do with their success. Both sexes said that effort was responsible for their successful performance, but only men believed that ability was an equally important cause (Deaux, 19791. It is not difficult to imagine how the tendency toward self-enhancement among men and self-deprecation among women borne of their differential expectancies for success could yield more favorable descriptions of jobs held predominately by men than those held predominately by women. One might suppose that using job descriptions from supervisors or expert ana- lysts would circumvent the sex-linked biases that may enter into incumbent descriptions; unfortunately, this is not the case. Considerable research has revealed that the tendency for men to provide more self-enhancing explana- tions for their own successful performance than women do is paralleled in the explanations offered by outside observers, who also have stronger
62 McARTHUR expectations that men will succeed (e.g., Deaux and Emswiller, 1974; Feather and Simon, 1975; Feldman-Summers and Kiesler, 1974~. While the foregoing arguments suggest that job descriptions will vary as a function of the worker's sex, a study by Rosenbach et al. (1979) found that male and female incumbents rated their jobs similarly on several important job dimensions, including the degree to which the job requires a variety of skills and the degree to which it provides autonomy in carrying out the work. However, there were some sex differences in job descriptions e.g., women rated jobs lower in the degree to which they have a substantial impact on the lives or work of others. A major problem with this study is that the men and women from whom job descriptions were obtained did not occupy the same jobs. This is not so problematic for the authors' primary interest, which was relating job perceptions to job satisfaction. But it makes it very difficult to interpret similarities and differences in the actual content of job descriptions provided by men and women. A study by Arvey et al. (1977) employed a methodology more appropriate to ascertaining the influence of sex-biased expectancies on job descriptions. These authors compared analysts' ratings of the same job when it was depicted with male and female incumbents in a verbal narrative accompa- nied by color slides. Although the results of this study revealed no effects of incumbents' sex on job descriptions, another study employing a videotaped depiction of jobs did find an effect of incumbents' sex (McArthur and Obrant, 19841. When jobs were depicted with male incumbents, analysts rated them as being relatively more critical to the company's assets and operations, as involving relatively more responsibility, and as having rela- tively less structure (i.e., fewer predetermined activities) than the same jobs depicted with female incumbents. These higher responsibility ratings for male-occupied jobs are consistent with sex-role stereotypes that paint the typical man as more skilled in business, more of a leader, more able to make decisions, and more independent than the typical woman (Broverman et al., 19721. Ratings of the skills required to perform the jobs provided less evidence of sex bias than did ratings of responsibility. While jobs depicted with male incumbents were perceived as requiring relatively more persua- sive ability than those depicted with female incumbentsa finding consis- tent with the stereotype that men are more dominant than women jobs depicted with male incumbents were not perceived to require more reason- ing ability or more mathematical ability, thus providing no evidence for biasing effects of the stereotype that men are more logical and mathemati- cally inclined than women. Although procedural differences argue for the greater ecological validity of McArthur and Obrant's (1984) evidence of sex bias than the null findings of Abbey et al. (1977), additional research is clearly needed to establish the
SOCIAL JUDGMENT BIASES 63 conditions under which job descriptions will be biased by sex-stereotyped cultural expectations. McArthur and Obrant's results suggest that such biases may be less pronounced in descriptions of a job's skill requirements than in descriptions of its responsibility. While the absence of gender bias in ratings of a job's skill is somewhat reassuring, it should be noted that respon- sibility has traditionally been a very important factor in job evaluations (Treiman, 1979), and the lesser responsibility ascribed to jobs depicted with female incumbents could yield an underestimation of the wages that wom- en's work is worth. Indeed, McArthur and Obrant (1984) found that jobs depicted with female incumbents were rated by analysts as meriting rela- tively lower wages than the same jobs depicted with male incumbents. While the absence of sex bias in analysts' ratings of the skills required by jobs clearly provides no assurance of unbiased ratings of comparable worth, it does provide some suggestions regarding how to generate job evaluations that are not biased by the incumbent's sex. More specifically, the absence of sex bias in most skill ratings may be due to the fact that people have become sensitized to sexism in judging abilities, with the consequence that analysts deliberately attended to information indicative of skilled work by female incumbents. This interpretation suggests that raising people's awareness of sexism in judging responsibility may reduce the sex bias in ratings on this dimension as well. The lesser sex bias in ratings of skill than in ratings of responsibility may also be due to the fact that evaluating the skills required of a job is a more objective task and thus less subject to the expectancy bias. This possibility suggests that providing raters with more explicit criteria regarding what constitutes a job with high responsibility, low structure, or high criticality could reduce sex bias in these ratings by virtue of guiding attention to the relevant as opposed to the expected information. While careful instructions to job analysts may help to reduce sex bias, the ideal solution to the expectancy bias would be to conceal the incumbent's sex during the process of job analysis. As noted earlier, the incumbent's sex could tee concealed curing evaluations of ajob's worth on various compensa- ble factors by employing a computer programmed to allocate some predeter- mined worth orjob points for each element in the job description. However, to conceal the incumbent's sex during the process of job description is more problematic. One possibility would be to provide written information about the jobs for analysts to use in generating job descriptions on a structured instrument like the PAQ. The question is, who will provide the written information and will the information be the same for male and female incumbents? Even if it is the same, it is likely that the analysts will envision incumbents of one or another gender when rating a job and, according to the availability bias, the gender of the incumbent they envision will be which- everone is the most frequent in that particular occupation. Anotherpossibil-
64 McARTHUR ity would be for the incumbents to describe their own jobs on a structured questionnaire. Unfortunately, as noted earlier, there is considerable evi- dence to indicate that the sex biases manifested in observers' judgments about male and female workers are paralleled in the judgments offered by the workers themselves. Given the difficulty of concealing the sex of incumbents, it may be advisa- bleto obtain structured ratings of all jobs from incumbents of a single sex. In the absence of information regarding whose descriptions are the most accu- rate, either male or female workers could be selected so long as workers of the same sex are used for all jobs whose worth is to be compared. A problem arises, however, when descriptions are sought for the large majority of occupations that are filled predominately by workers of one sex or the other. It is possible that the descriptions provided by a male secretary, for example, would not be representative of secretaries in general. Despite the null effect of sex composition in past research, it is also possible that the sex composi- tion of highly segregated occupations could bias job descriptions even when the sex of the incumbents who are viewed is held constant. DEFINING THE COMPENSABLE FACTORS As noted earlier, evaluating the comparable worth of jobs requires deter- mining which job elements should be compensated rather than relying on market indicators of which job elements are currently compensated. Like job descriptions and evaluations, the choice of compensable factors and their weights is necessarily a subjective judgment that is vulnerable to the judg- mental biases discussed above. Most job analysis procedures have employed a relatively small set of factors deemed as legitimate bases for pay differentials: skills, experience, responsibility, effort, and working conditions. While these factors have some face validity, they nevertheless represent value judgments and should not be accepted without question. One question is: Who should make the judgment? When employees are organized, the decision could be made through collective bargaining. However, it is not clear what should be done when employees are not organized. It is entirely possible that having the employer make the decision could yield weighted compensable factors that ultimately serve to justify lower wages for certain categories of "women's work" than for "men's work." This may result not so much from deliberate discrimination on the part of the employer as from the likelihood that the employer's choices will be vulnerable to judgmental biases. The availability heuristic could bias the choice of compensable factors, since the bases for pay differentials that most readily come to mind may be those that are the most frequent in the work world. Halo effects could also
SOCIAL JUDGMENT BIASES 65 bias the choice of compensable factors, since the factors that are designated may be those that are evaluatively consistent with financial rewards. The consequence of this will be that the positive aspects of jobs will be viewed as meriting compensation, while negative job elements will be given little weight. It should be noted that both the availability heuristic and the halo effect favor compensable factors that would legitimize the current wage structure. Consider, for example, the occupation of child care. The avail- ability heuristic will generate compensable factors that are the most frequent bases for pay differentials, such as special skills and training and responsibil- ity for other workers or money. Since these factors are not highly represented in child care, they will serve to justify its poor compensation. In addition, halo effects will militate against compensating "negative" factors, such as little opportunity for advancement, littlejob security, end a nonoptimal level of difficulty (relatively easy end boring), all of which are highly represented in child care. By traditional yardsticks, then, child care will not be viewed as comparable in worth to, say, dentistry. But if one considers the worth of these two jobs by another yardstick such as what would happen in our society if nobody engaged in child care versus what would happen if nobody engaged in dentistry then clearly the women's work of child care is of greater worth to society. The question remains as to what can be done to guard against biases in the choice and weighting of compensable factors that may devalue the worth of jobs held primarily by one sex or the other. Ideally, one would like to have some objective criterion for evaluating the legitimacy of a designated basis for pay differentials. The next best thing would be to establish a broad-based subjective criterion. For example, surveys could be conducted to ascertain employers' and employees' beliefs regarding the legitimate bases for pay differentials as well as how these bases should be weighted in fixing wages. One would hope to find a consensus regarding at least some compensable factors. However, the small amount of relevant research on this subject suggests that the choice of compensable factors will vale across jobs as well as across job analysts. McArthur and Obrant (1984) found that those job description measures that were significantly related to analysts' ratings of a job's monetary worth varied as a function of the job and as a function ofthe worker's gender. Most notably, there were no significant predictors of the monetary worth of jobs depicted with female incumbents. It should be noted that this study did not directly ask job analysts to indicate what job elements should be compen- sated. Rather, regression analyses were employed to determine empirically which job elements predicted ratings of monetary worth. One might hope that direct assessment of analysts' beliefs regarding what elements should be compensated would yield more thoughtful and thus consistent responses.
66 McARTHUR However, several surveys have revealed that people's responses to the ques- tions "What aspects of jobs do you think should be compensated?" and "How important should each factor be in determining salary?" varied as a function of the particular job they were considering as well as their own job (McArthuret al., 1984~. The apparent lack of consensus regarding what job factors should be compensated presents a very serious problem for comparable worth analy- ses, and additional research is clearly needed to determine whether the preliminary findings reported here are reliable. In conducting this research, it is important that judgments be obtained regarding the legitimacy of a large variety of bases for pay differentials, not just those that have been commonly employed in existing job analyses. In particular, potentially compensable factors that are embodied in "women's work" need to be identified and included along with the more standard factors to see how they stack up in people's judgments regarding the legitimate bases for pay differentials. Also, factors that have been shown to influence how worthwhile a job is to the employee should be rated. These include the opportunity for advance- ment, job security, andjob difficulty (Jurgensen, 1978; Walker et al., 19821. In a sense these factors represent an expansion of the standard factor of working conditions, and the implication of including them is that jobs may be worth higher wages not only if the physical working conditions are poor but also if the psychological conditions are poor.2 CONCLUSIONS The complexity and inherent subjectivity of job analysis coupled with the limitations of human judgment make it extremely unlikely that the objec- tively true worth of jobs can ever be established, even if some consensus could be reached regarding what job elements should be compensated. The inevitable biases in job descriptions would not be so problematic for compa- rable worth analysis if one could assume that they were not sex-linked: if judgmental biases simply added a constant error to evaluations of every job, the true relative standing of men's and women's jobs would be maintained. Unfortunately, however, He evidence reviewed in this paper provides con- siderable reason for concern that judgmental biases are weighted in the direction of underestimating the wages that women's work is worth relative to that of men. The halo bias will overestimate the worth of work that is 2 While one might expect that the allocation of higher wages in compensation for these poor working conditions would find considerable support, this does run contrary to halo effects. And it should be recalled that, consistent with a positive halo effect, McArthur and Obrant (1984) found that the more desirable a job was, the greater was its judged monetary worth.
SOCIAL JUDGMENT BIASES 67 relatively prestigious and well paid men's work. The expectancy bias will overestimate the skills, leadership, training, and effort that are required by work that is culturally expected to require such qualities men's work. And the availability bias will reinforce this tendency, inasmuch as it will overesti- mate the frequency of familiar i. e., expectedjob activities, in addition to those that are recent or perceptually salient. Given the difficulty of removing certain biases from job descriptions, it would seem advisable to conduct research designed to determine how to best reduce their influence. Providing raters with very explicit criteria regarding what aspects of a job merit what ratings on structured job description instru- ments might help to reduce bias. Alerting job analysts to their potential biases may also help. Finally, research comparing the job descriptions pro- vided by incumbents, supervisors, and experts can identify the source of job descriptions that are the least sex-biased. While research can determine the conditions that minimize sex biases, it can never prove that such biases will be absent from comparable worth analyses. This is because most jobs are sex-segregated. When the same job is occupied by men and by women, one can compare the job descriptions provided by different sources (incumbents, supervisors, and experts) under various instructional conditions and determine whether descriptions vale as a function ofthe incumbent's sex. If they do not, one can conclude that there is no sex bias given a particular source and a particular set of instructions. However, when a job is occupied primarily by incumbents of one sex, it is virtually impossible to determine whether descriptions vary as a function of the incumbent's sex, and it is thus impossible to ensure that the job descrip- tion is not sex-biased. In such instances, the best one can do is to choose the source of job descriptions and set of instructions that provide the most similar descriptions of jobs whether the incumbent is male or female. It must be noted, however, that the absence of sex bias in descriptions of jobs that are not sex-linked may not be generalizable to descriptions of jobs that are sex- linked. For example, even if female bank tellers describe their jobs as requiring the same amount of physical effort as male tellers do, this does not ensure that female typists' descriptions ofthe physical effort entailed in their work will not be underestimated compared with male truck drivers' descrip- tions of their own physical effort. An alternative to eliminating or reducing sex biases in job descriptions would be to attempt to compensate for them. For example, if one found that female incumbents underestimate the physical effort required by a particular job in comparison with male incumbents, then some constant factor could be added to female incumbents' estimates of effort required in otherjobs jobs for which there are few male incumbents. However, such a strategy assumes a constant sex bias in job descriptions, and this assumption may well be
68 MCARTHUR wrong. Another strategy for dealing win sex biases in job descriptions is to attempt to employ at least some objective measures. For example, the physi- cal effort required by a job could conceivably be assessed with some kind of mechanical device. Similarly, the working conditions could be objectively described via indices of temperature, square footage per worker, and so forth. Such measures are, of course, limited in scope as well as being more costly and cumbersome to employ than subjective judgments are. While research may provide methods for reducing judgmental biases in comparable worth analysis, it is likely that such biases will never be com- pletely eliminated. As a result, comparable worth analyses may well yield an underestimation of the wages that women's work is worth, thereby redress- ing only a portion of the existing wage gap. This in itself is not so bad. What is more worrisome is the likelihood that any remaining gender gap in wages would be certified as just. Whether or not the benefits of the first outcome outweigh the risks of the second is an important question to consider in deciding whether comparable worth analysis is better than alternative meth- ods, such as unionizing and collective bargaining, for reducing wage inequi- ties. REFERENCES Anderson, N.H., and S. Hubert 1983 Effects of concomitant verbal recall on order effects in personality impression forma- tion. Journal of Personality and Social Psychology 2:379-391. Abbey, R.D., E.M. Passino, andJ.W. Lounsbury 1977 Job analysis results as influenced by sex of incumbent and sex of analyst. Journal of Applied Psychology 62:411-416. Broverman, I.K., S.R. Vogel, D.M. Broverman, F.E. Clarkson, and P.S. Rosenkrantz 1972 Sex-role stereotypes and clinical judgments of mental health. Journal of Social issues 28:59-78. Crowley, A.D. 1981 The content of interest inventories: Job titles or job activities? Journal of Occupational Psychology 54: 135-140. Deaux, K. 1979 Self-evaluation of male and female managers. Sex Roles 5:571 -580. Deaux, K., and T. Emswiller 1974 Explanations of successful performance on sex-linked tasks: What's skill for the male is luck for the female. Journal of Personality and Social Psychology 29:80-85. Deaux, K., and E. Farris 1977 Attributing causes for one's own performance: The effects of sex, norms, and outcome. Journal of Research Personality 11 :59-72. Desmond, R., and D. Weiss 1973 Supervisor estimation of abilities required in jobs. Journal of Vocational Behavior 3: 181-194. 1975 Worker estimation of abilities required in jobs. Journal of Vocational Behavior 5: 145- 158.
SOCIAL JUDGMENT BIASES 69 Feather, N.T. 1969 Attribution of responsibility and valence of success and failure in relation to initial confidence and task performance. Journal of Personality and Social Psychology 13: 129-144. Feather, N.T., andJ.G. Simon 1975 Reactions to male and female success and failure in sex-linked occupations: Impressions of personality, causal attribution, and perceived likelihood of different consequences. Journal of Personality and Social Psychology 31 :20-31. Feldman-Summers, S., and S.B. Kiesler 1974 Those who are number two try harder: The effect of sex on attributions of causality. Journal of Personality and Social Psychology 30:846-855. Fischer, D.G., and J. Sobkow 1979 Workers' estimation of ability requirements of their jobs. Perceptual anal Motor Skills 48:519-531. Frieling, E. 1977 Occupational analysis: Some details of an illustrative German project. International Review of Applied Psychology 26:77-85. Frieze, I.H. 1977 Women's expectations for end causal attributions of success and failure. In M. Mednick, S.S. Tangri, and L. Hoffman, eds., Women and Achievement: Social and Motivational Analyses. Washington, D.C.: Hemisphere. Grams, R., and D.P. Schwab 1985 Systematic sex-related error in job evaluation. Academy of Management Journal (Forth- coming). Griffen, R.W. 1981 A longitudinal investigation of task characteristics relationships. Academy of Manage- ment Journal 24:99-113. Jones, E.E., and G.R. Goethals 1972 Order effects in impression formation: Attribution context and the nature of the entity. Pp. 79-94 in E.E. Jones, D.E. Kanouse, H.H. Kelley, R.E. Nisbett, S. Valins, and B. Weiner, eds., Attribution: Perceiving the Causes of Behavior. Morristown, N.J.: Gen- eral Learning Press. Jurgensen, C.E. 1978 Job preferences (What makes a job good or bad?). Journal of Applied Psychology 63:267-276. Kelley, H.H. 1950 The warm cold variable in first impressions of persons. Journal of Personality 18:431- 439. McArthur, L.Z. 1981 What grabs you: The role of attention in impression formation and causal attribution. In E.T. Higgins, C.P. Herman, and M.P. Zanna, eds., Social Cognition: The Ontario Symposium. Vol. 1. Hillsdale, N.J.: Erlbaum. McArthur, L.Z., and S. Obrant 1984 Sex Biases in Comparable Worth Analysis. Unpublished research, Department of Psy- chology, Brandeis University. McArthur, L.Z., J. Karas, J. Neri, A. Pressel, and D. Rudin 1984 Choosing Compensable Factors for Comparable Worth Analysis. Unpublished paper, Department of Psychology, Brandeis University. McCormick, E.J., P.R. Jeanneret, and R.C. Meacham 1969 Position Analysis Questionnaire. Purdue Research Foundation, West Lafayette, Ind.
70 MCARTHUR 1972 A study of job characteristics and job dimensions as based on the Position Analysis Questionnaire (PAQ). Journal of Applied Psychology Monograph 56:347-368. Narayanan, S., and R. Venkatachalam 1982 Perception of organizational climate. Perceptual and Motor Skills 55: 15- 18. Nisbett, R.E., and L. Ross 1980 Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, N.J.: Prentice-Hall. Rosenbach, W.E., R.C. Dailey, and C.P. Morgan 1979 Perceptions of job characteristics and affective work outcomes for women and men. Sex Roles 5:267-277. Ross, M. 1981 Self-centered biases in attributions of responsibility: Antecedents and consequences. In E.T. Higgins, C.P. Herman, and M.P. Zanna, eds., Social Cognition: The Ontario Symposium. Vol. 1. Hillsdale, N.J.: Erlbaum. Suchner, R.W. 1979 Sex ratios and occupational prestige: Three failures to replicate a sexist bias. Personality and Social Psychology Bulletin 5 :236-239. Touhey, J.C. 1974a Effects of additional women professionals on ratings of occupational prestige and desir- ability. Journal of Personality and Social Psychology 29:86-89. 1974b Effects of additional men on prestige and desirability of occupations typically performed by women. Journal of Applied Social Psychology4: 330-332. Treiman, D.J. 1979 Job Evaluation: An Analytic Review. Committee on Occupational Classification and Analysis. Washington, D.C.: National Academy of Sciences. Treiman, D.J., and H.I. Hartmann, eds. 1981 Women, Work, and Wages: Eq~ual Payfor Jobs of Equal Value. Committee on Occupa- tional Classification and Analysis. Washington, D.C.: National Academy Press. Tversly, A., and D. Kahneman 1974 Judgment under uncertain~: Heuristics and biases. Science 18S: 1124-1131. Walker, J.E., C. Tausly, and D. Oliver 1982 Men and women at work: Similarities and differences in work values within occupa- tional groupings. Journal of Vocational Behavior 21: 17-36. Weiss, H.M., andJ.B. Shaw 1979 Social influences on judgments about tasks. Organizational Behavior and Human Per- formance 24: 126-140. White, M.C., M.D. Crino, and G.L. DeSanctis 1981 Ratings of prestige and desirability: Effects of additional women entering selected business occupations. Personaliry and Social Psychology Bulletin 7:588-592.