How is banding much different from setting a selection ratio for a given stage of a selection process? I ask this because it seems like if bands are used they are going to be relatively large (given problems with test reliability, etc.) and won’t do much more than separate out qualified from unqualified applicants. Even if the bands are more specific, the commentators supporting banding supported the use of secondary criteria to select individuals for the job. What if applicants who didn’t make it in the top band or two had higher scores on the secondary criteria than those individuals who were in the top band or two? Can we justifiably say that an individual who doesn’t make it into a top band wouldn’t have actually been a top candidate if considered with multiple selection criteria? So why really even use banding in the first place if in most cases it is much different from using a selection ratio or a cutoff score in a multistage selection system?
How do companies overcome the lack of statistical power when their applicant size (i.e., sample size) is small because the SED-banding method may not detect score differences?
The topic of banding is more complex than I realized from the Bobko article. I appreciated one of the authors emphasizing the importance of context of personnel decisions and the appropriate balance between type 1 and type II errors in looking at an organization’s needs (info tech needs narrower bands, grad schools could appropriately use wider bands). It seems that the practice of banding is widespread, even if not called by that name, and it is especially used in public sector employment. Is one take-away from this discussion (article) that traditional banding is used more in private employment and SED banding in public? If banding without minority preferences does little to reduce AI, why do some persist in using it?
After reading about banding, it appears that this technique is similar to doing a median split to form two different groups. From what I understand, you are turning a continuous variable into a categorical one, which eliminates the variance of scores. Why would you not just use multiple predictors, keep them continuous, and find the people that have the right "fit" for the job? One would hope that by using multiple predictors, the discriminatory predictors (i.e., the bad kind of discrimination) would have less impact on the final selection decision. In this way, you are not arbitrarily throwing out talent because they did not fit into a band on one particular predictor.
Schmidt in the Campion et al. (2001) article states that, "...banding without minority preferences does little to reduce adverse impact. So the current stance of the courts appears to block achievement of what is perhaps the major objective of SED banding" (p. 167.3) and later that, "Complete elimination [of group differences in hiring rates] will require wider social changes" p. 176.3. Do these statements place some limits on the ability and thus the responsibility of selection specialists to ameliorate these impacts? Where does our responsibility end and that of society as a whole begin?
Do you think using banding increases or decreases the number of predictors that need to be used? How do you think we should go about determining the order in which predictors should be used?
Interesting that the general consensus for the use of banding is to increase the chances of selecting a minority candidate, but the approach does not hold well in court when there is explicit minority preference. Doesn’t this seem overly contradictory to you?
How is banding much different from setting a selection ratio for a given stage of a selection process? I ask this because it seems like if bands are used they are going to be relatively large (given problems with test reliability, etc.) and won’t do much more than separate out qualified from unqualified applicants. Even if the bands are more specific, the commentators supporting banding supported the use of secondary criteria to select individuals for the job. What if applicants who didn’t make it in the top band or two had higher scores on the secondary criteria than those individuals who were in the top band or two? Can we justifiably say that an individual who doesn’t make it into a top band wouldn’t have actually been a top candidate if considered with multiple selection criteria? So why really even use banding in the first place if in most cases it is much different from using a selection ratio or a cutoff score in a multistage selection system?
ReplyDeleteHow do companies overcome the lack of statistical power when their applicant size (i.e., sample size) is small because the SED-banding method may not detect score differences?
ReplyDeleteThe topic of banding is more complex than I realized from the Bobko article.
ReplyDeleteI appreciated one of the authors emphasizing the importance of context of personnel decisions and the appropriate balance between type 1 and type II errors in looking at an organization’s needs (info tech needs narrower bands, grad schools could appropriately use wider bands).
It seems that the practice of banding is widespread, even if not called by that name, and it is especially used in public sector employment. Is one take-away from this discussion (article) that traditional banding is used more in private employment and SED banding in public? If banding without minority preferences does little to reduce AI, why do some persist in using it?
After reading about banding, it appears that this technique is similar to doing a median split to form two different groups. From what I understand, you are turning a continuous variable into a categorical one, which eliminates the variance of scores. Why would you not just use multiple predictors, keep them continuous, and find the people that have the right "fit" for the job? One would hope that by using multiple predictors, the discriminatory predictors (i.e., the bad kind of discrimination) would have less impact on the final selection decision. In this way, you are not arbitrarily throwing out talent because they did not fit into a band on one particular predictor.
ReplyDeleteSchmidt in the Campion et al. (2001) article states that, "...banding without minority preferences does little to reduce adverse impact. So the current stance of the courts appears to block achievement of what is perhaps the major objective of SED banding" (p. 167.3) and later that, "Complete elimination [of group differences in hiring rates] will require wider social changes" p. 176.3. Do these statements place some limits on the ability and thus the responsibility of selection specialists to ameliorate these impacts? Where does our responsibility end and that of society as a whole begin?
ReplyDeleteDo you think using banding increases or decreases the number of predictors that need to be used? How do you think we should go about determining the order in which predictors should be used?
ReplyDeleteInteresting that the general consensus for the use of banding is to increase the chances of selecting a minority candidate, but the approach does not hold well in court when there is explicit minority preference. Doesn’t this seem overly contradictory to you?
ReplyDelete