Next, the mean job scores for male- and female-dominated jobs were converted to standard scores (mean = 50, standard deviation = 10) to determine which weighting scheme most reduced the disadvantage of females relative to males. This transformation also made it possible to compare mean job scores on the five and the nine factor scales. Table 22 provides these transformations for female- and male-dominated jobs for each of the five and nine factor scales. The transformed mean job scores of female-dominated jobs were then subtracted from the mean job scores of male- dominated jobs and converted to standard deviation units to measure the difference for each scale.
As can be seen in Table 22, the difference in standard deviation units is approximately equal for all scales. This finding indicates that, as expected, none of these scales were able to reduce the gap in job scores between male- and female-dominated jobs because of the original difference in factor scores between men and women. Also none of these weighting schemes emphasized the factors that had the smallest difference in factor scores between male- and female-dominated jobs.
As Treiman ( 1984) suggested, changing weights can make a difference in the relative pay of females as compared with males if two conditions are met. First, job content must vary, that is, female jobs must score higher than male jobs on some factors; second, the factors must not be highly intercorrelated.
This chapter tested these assumptions using actual job evaluation data. Examining the raw data from the FES study ( Anderson and Corts, 1973) showed that the raw factor scores for both the five and the nine factors were consistently lower for female-dominated jobs. The correlation coefficients from both the five and the nine factor methodologies indicated very high intercorrelations among the main contributing factors. Thus these data did not meet the conditions suggested by Treiman, and I concluded that it would be impossible for any weighting scale based upon these factor scores to overcome the relative disadvantage of female-dominated jobs. I then proceeded to review the weighting scales generated by Anderson and Corts and others generated by this research to