Home> Science & Math> Table of Contents> Background for Teachers>

Background for Teachers The issue of discrimination is of fundamental importance from the moral point of view. The country in which we live has as one of its core values equal opportunity for all social and ethnic groups and it is not only considered to be illegal to bias opportunities in favor of one group over another, but morally reprehensible as it goes against that core value. We can understand cases of individual discrimination quite well, but when we would like to determine whether or not an employer, university, or other organization is practicing hiring or admittance in a non-discriminatory way we have to investigate the practices of the employer or organization on the basis of statistical inference from a population of applicants to those jobs or organizations. A common way that this is done is to take the total number of applicants from various groups, e.g. males and females, and measure the proportion of applicants that were accepted to those that applied for each group considered individually. Discrimination is then determined by seeing whether or not the proportion of accepted applicants of one group is significantly greater than the proportion of those accepted from the other group. However, this method is not reliable and has proved misleading in several instances of discrimination investigations for the reason that population bias is not the only factor that should be taken into consideration in cases of discrimination (Bickel, et.al.). The method does in fact seem to be a reliable indicator of discrimination so it is a good exercise to see that it is flawed and how it is flawed when trying to determine whether or not an organization’s practices are discriminatory. Moreover, seeing how the statistical inference is flawed allows one to see the intricacies of the concept of discrimination and how better to detect it. Engaging in the critical thinking exercise involved in the cases below, which at first looks to be a mere homework problem in elementary statistics, should end up illuminating the issue of organizational discrimination. The first case involves mortality rates of African Americans and Caucasians from tuberculosis infection in New York City and Richmond, VA that was taken from data from 1910. The statistics entail that though the death rate per 100,000 of African Americans was lower in Richmond than in New York and that the death rate of Caucasians was lower in Richmond than in New York, that the total death rate of individuals, the combined population of African Americans and Caucasians, in Richmond was higher than that in New York. Such statistics bring up interesting questions in medical ethics, which will be raised and discussed below. Though the data in this case does not necessarily involve the issue of discrimination, it could be used fallaciously to persuade others that they are more likely to dies of tuberculosis in Richmond than in New York. Diagnosing what is wrong in this case is a good exercise in critical thinking and an evaluation of the concept of a death rate. (Ken, I can add more here if we can discuss some of the other issues from medical ethics that could arise) The second case involves a discrimination suit that was brought against the University of California-Berkeley in the early 1970’s. It was said that the university discriminated against women by disproportionately accepting into graduate programs men over women. The claim was based on data that was taken from the total set of applicants. The proportion of males that were accepted to a graduate program in Berkeley given the total number of males that applied was 10% greater than the proportion of females that were accepted to a graduate program to the total number of females that applied. The bias in favor of males that the data seemed to indicate was used to support the claim that UC-Berkeley was in fact discriminating against females in its graduate acceptance practices. What must be noticed though is that in order to make this claim from the data it must be assumed that the applicants, male and female, to any given department did not differ with respect to any factor, e.g. GRE scores, that may be legitimately used to decide their acceptance into a program (Bickel, et.al.). Moreover, it must be assumed that the ratio of men to women that apply to any given department is not influenced significantly by any other factors involved in their admission. These two assumptions ensure that men and women have equal chance of getting into the university. Now, despite the data for the total applicant pool, when the acceptance practices of each department were examined the data indicated that most departments had an acceptance bias that was in favor of women applicants. When the case is investigated in the exercises below it turns out that the second assumption above turned out to be false, a conclusion that comes right out of the data. Certain intricacies of the concept of discrimination are then brought out as a result. The conclusion to be drawn of course is that one should be wary of data that is meant to test for discrimination when the total pool of applicants is involved. These cases present just two instances, but many more may be out there. The important point to notice is that the data could indicate evidence for the opposite claim as well, namely that discrimination is not being practiced when it in fact is. For example, it could turn out that the data from the total pool of applicants to a company indicates that a higher proportion of the females from the pool were hired than males, but each department in the company exhibits practices that favor the hiring of men over women. When we investigate organizational discrimination we should be aware of this fact as well as when someone makes discrimination or non-discrimination claims with regard to an organization.