Huma Lecture 5 PDF

**An Introduction to Contemporary Theories of Discrimination (2)** Readings : - - (Statistical discrimination; when is discrimination wrong?) **I- A Reminder on the Concept of Discrimination** **A) The pre-conditions of discrimination** - - - - **B) The (descriptive, i.e. non-ev...

**An Introduction to Contemporary Theories of Discrimination (2)** Readings : - - (Statistical discrimination; when is discrimination wrong?) **I- A Reminder on the Concept of Discrimination** **A) The pre-conditions of discrimination** - - - - **B) The (descriptive, i.e. non-evaluative) definition of discrimination:** a practice imposing a relative disadvantage on individuals regarding the distribution of scarce goods based on perceived membership in a salient social group **C) The typology of discrimination:** direct/indirect; intentional/non-intentional; the four-cell matrix **D) The puzzle:** the gap between the decline in the expression of racist beliefs and attitudes and the decline in the occurrence of discriminatory practices The three methods have discussed a significant amount of racial discrimination in the post-civil rights era. If racism is seen as an ideology, it has gone down dramatically. However, **racial discrimination** has not. The notion of **statistical discrimination** is a specific kind of direct discrimination that is introduced to provide an answer to the basic question - to make sense of the gap between the decline of discrimination and decline of racism. The two economists responsible for this are Edmund Phelps and Kenneth Arrow **II- The Economic Approach of Discrimination** **A) Gary Becker's Model - the earlier model** **1) Presentation:** discrimination as a taste-driven, economically irrational behaviour bound to disappear as a result of competitive pressures in a free-market economy Discrimination is a purely taste-driven, economically irrational behavior that is not connected to the goal of profit maximization. According to Becker, white employers dislike black employees to such an extent that they would not be hired regardless of their qualifications. They're willing to bracket their financial interest to satisfy this aversion to black people. However, the racist taste is costly. If black people and white people have similar qualifications, but some employers discriminate against black employees, a racial wage gap is bound to appear. This is because, by definition, some employers would accept a wage increase designed to hire white applicants, rather than hiring black applicants at a lower cost. The firms that do discriminate based on race will be bound to pay higher wages, so their productivity will go down - ultimately leading to bankruptcy. Discrimination based on race will disappear due to pressures intrinsic to a free market economy. The only notion is that some employers are not prejudiced. However, if every employer is prejudiced, the market is open for new employers to enter, pay less money to those equally qualified black applicants, and they will win the competition over the racist employers who will discriminate based on the economically irrelevant criteria. There are several critiques of this model **2) A preliminary critique:** the issue of **'reaction qualifications'** (Alan Wertheimer, "Jobs, Qualifications, and Preferences", Ethics, 99 (1), 1983, p. 99-112) The more the supply of labor exceeds the demand for labor, the lower the cost of discrimination for the employer. If there are plenty of qualified white applicants around, it won't be costly for employers to discriminate based on economically irrelevant criteria. The whole model also rests on the notion that race is economically irrelevant. Even if an employer is not prejudiced, he may be driven to be prejudiced if enough people around him are prejudiced - like a restaurant owner not hiring black waiters if most of his customers are white racists who would stop patronizing his business. It also may be rational to hire based on race (i.e to hire a black police officer in a predominantly black neighborhood). This is a non-arbitrary form of discrimination that feeds on a form of discrimination that is arbitrary. This could create a discriminatory equilibrium that only a legal provision could unsettle. This model also uses the assumption that black people and white people are equally qualified. There are differences between the two groups in involvement in criminal behavior, education level, etc. Similarly, women are more likely than men to be the caretakers of children and to leave the labor market earlier; this fact could make employers less likely to hire women. Decisions based upon race or sex-based discrimination may be economically rational. There are cases in which acquiring more individualized data could be costly, making this the more economically rational choice. **B) The issue of statistical discrimination** **1) Definition:** a discrimination based on the existence of a correlation between (for instance) race (visible, intrinsically irrelevant, instrumentally relevant) and some other individual trait, X (intrinsically relevant but invisible: productivity, dangerosity...), that the decision maker reasonably cares about given his/her legitimate goals This discrimination is not based on ideology or attitude. The model assumes that decision makers are utility maximizers. They are imperfectly informed about the relevant characteristics of the individuals that they must choose and acquiring more information is costly. Discrimination is statistical when it comes from a calculation itself based on a correlation between race and some other individual feature (x) that the decision maker cares about. Definitions of x depend on decision makers. Police officers care about involvement in crime, managers want to improve productivity, etc. Though race is intrinsically irrelevant, it's instrumentally relevant because it can be used as a proxy (a functional substitute) for x. There is a correlation between race and x. The decision maker does not care about race; race is visible, however x is not. Race serves as a proxy for x. The underlying goal of the discrimination (the reason why he cares about x) is legitimate and commonly acknowledged as being legitimate. Employers legitimately aim at profit maximization, police officers legitimately aim at curbing crime, etc. There are cases in which employers are not completely aware of statistical data, but they are aware of the correlation between race and x. He has a practical knowledge derived from his or her prior experience that may or may not be formalized into statistics. **2) Some preliminary illustrations** We all engage in statistical discrimination. Studies by Wilson and Tilly have demonstrated this. The studies have been validated by courts. In the U.S. Supreme Court decision **United States v. Martinez-Fuerte**, it was found that border patrol agents treated those of Mexican heritage more harshly. However, this was an exception because courts usually do not validate using race as a proxy. **3) Five preliminary comments** **a) 'Forbidden grounds' as bases of statistical discrimination: some contrasts** Forbidden grounds (race, sex, color, religion, etc) vary in their capacity to become a useful proxy for decision-makers. Sex is essentially a binary criteria, and age (on the other hand) is not. Therefore, statistical discrimination based on sex is bound to be much cruder and less reliable as an instrument than statistical discrimination based on age. Statistical discrimination based on sex would involve sweeping generalizations, while statistical discrimination based on age would be more nuanced. The correlation between being under the age of 8 and needing parental care and the correlation between being over 100 and not being able to fly a plane will be much higher than any correlation based on age. The wide range of ages allows for more precise statistical discrimination. **b) The proxy may be (and generally is) sub-optimal** The proxy does not have to be perfect. The inferences based on it do not have to be more reliable than that of any other proxy. However, using the proxy needs to be more reliable than not using it at all. But, the fact that there are better proxies available doesn't mean that statistical discrimination is not present. For discrimination to be statistical, the proxy only needs to be minimally efficient so that the costs of using it are outweighed by the benefits. **c) Distinguishing racist from statistical discrimination:** the example of race-based discrimination by cab drivers It may be difficult to distinguish statistical discrimination from racist discrimination. Most measures of it will not be that helpful. Audit studies, like sending many resumes to the same employers, will not tell you why the discriminator is discriminating. You need qualitative interviews with decision makers to find out why. Decision makers are often willing to admit that they discriminate based on a certain ground and view it as a risk-minimizing tactic - even if it is illegal. There is solid evidence that, in large American cities, cab drivers are less likely to stop for a young black man hailing a cab at night. Cab drivers are particularly vulnerable to robbery. Sex and race are highly-correlated with robbery. The race of the driver has no impact on the driver's potential to discriminate based on race. Black drivers hardly stop to pick up black customers in comparison to drivers of other races. All individuals have a compelling interest in not being robbed. The time to make the decision is very limited with nothing to go on than race, sex, clothing, etc **d) Statistical discrimination as a (potential) by-product of non-statistical discrimination:** the case of race and robbery There are cases where the very correlation that is being used by the decision maker to statistically discriminate based on race is the outcome of past or present discrimination of another kind. There is a correlation between sex and life expectancy. The existence of this correlation is why women were charged higher insurance premiums until they were banned from doing so as a result of the U.S. Supreme Court decision **City of Los Angeles v. Manhart** (1978) and European Union Court of Justice decision Association Belge des Consommateurs Test-Achats (2011). There is a plausible causal connection between race and crime in the US. Discrimination tends to preclude access to legal sources of income; more widespread discrimination in the job market leads to a higher likelihood of turning to illegal behavior as a source of income. Black people and hispanic people are more likely to commit crime; however, this does not buy into any racialist explanation of why that is the case. Black people made up 13% of the population, but they were also 60% of those arrested and 16 times as likely to be incarcerated as compared to non-hispanic whites. However, discrimination cannot account for the entire gap. According to Rajiv Sethi and Brendan O'Flaherty ("Racial stereotypes and robbery", Journal of Economic Behavior and Organization, 68 (3-4), 2008, p. 511-524), robberies involve dynamic interactions under conditions of incomplete information. Victims of attempted robberies have the choice to comply or resist. Offenders can either respond to resistance by fleeing or by attempting to force compliance through violence. The likelihood of victim resistance can depend on the perceived race of the offender and make robbery more profitable for offenders of races where the victims are less likely to resist. If white victims find black offenders threatening, they will be less likely to resist; thus making robbery more profitable for black offenders. There is an enduring racist stereotype associating blackness with violence, and there is also a correlation between race and class. The underlying assumption could be that the probabilities of victim resistance and offender violence could be associated with income and wealth. There is a systematic and well-known variation in regards to income and wealth across racial groups. Income can be used as a proxy as well (i.e if poorer people are more likely to resist). One can combine the relationship between race and income to form a sort of double statistical discrimination. **e) Statistical discrimination as a self-fulfilling prophecy** Statistical discrimination operates so as to make the stereotypes involved self-sustaining. This point has been made effectively by Glenn Loury. When decision makers make a decision, they can set off a chain of behavior that violates initial expectations. Many cab drivers will not stop for young black men after a certain hour in fear of being robbed. For the young black men who are not robbers, anticipating a wait for cab drivers will discourage taxi transportation. For the young black men who are robbers, they will stick around; they have much greater incentive to wait until there's one cab driver that must stop. Stopping less frequently for black people and that robbers are less easily deterred means that cab drivers will just pick up more robbers - which they are expecting in the first place. The negative expectations of employers in some context may also work like a self-fulfilling prophecy. **4) The legal prohibition of race-based statistical discrimination:** illustrations The analytical distinction between statistical and racial discrimination is important for anti-discrimination legislation and strategy. Convincing people that they have racist thoughts (when they are practicing statistical discrimination) will not work. In U.S. Supreme Court decision **Batson v. Kentucky** (1986), it was stated that a lawyer or prosecutor can never use race as grounds to challenge a jury. The fear was, if there is a black juror and if the defendant is black, there is increased risk of the juror deciding to acquit the defendant because of previous experience with institutional racism. This risk may exist, but this does not allow a race-conscious decision to dismiss a juror. Race-based statistical discrimination is banned regardless of whether race is a good proxy. **5) Justifying the legal prohibition of race-based statistical discrimination** **a) Bernard Harcourt's consequentialist, holistic, and incentives-focused critique of racial profiling as a law enforcement strategy** According to Bernard Harcourt (Against Prediction. Profiling, Policing, and Punishing in an Actuarial Age, University of Chicago Press, 2007), there is a focus on a specific kind of race-based statistical discrimination - **racial profiling** (using race as a proxy for potential involvement in crime) by police officers. Regardless of any moral concern, racial profiling is not a rational law enforcement strategy. Any positive, direct effect that it might have in the short term is outweighed by negative, indirect consequences in the medium and long term. It works as a self-fulfilling prophecy. If officers stop more blacks than white, then more black criminals will be found. The whole process is also self-sustaining and cumulative. Decisions of racial profiling are made based on rates of arrest and incarceration. Police departments have a structural incentive to focus on groups among which it is more likely to find offenders; they use statistics based on arrest and incarceration, which are side effects of their own behavior. Harcourt also makes 2 bolder claims about the kind of incentives triggered by race-based profiling. Claim number one is that criminal and terrorist networks will strategically adapt to racial-profiling by adapting their recruitment strategies and deliberately including more of those from non-profiled groups. Claim number two is that those of non-profiled groups will be more likely to break the law because they know that they are being surveilled less. Consequently, the enabling effect on the non-profiled group will outweigh the surveillance effect on the profiled group. **b) The limitations of Harcourt's critique** **i. The potential unavailability of substitution strategies** Harcourt assumes that adaptive strategies of would-be lawbreakers are available. In the situation of crossing the southern US border, the large majority of immigrants crossing the border would be Mexican and visually identifiable as such. Members of the profiled group simply cannot be replaced by members of other groups. **ii. The potential inelasticity of the behavior of members of non-profiled groups** The elasticity of the behavior of non-profiled groups will depend on the nature of the crime involved. It may be true that most of those stopped and frisked by the police are black or hispanic; this may make whites more likely to profile. It doesn't work in other situations. More Europeans won't cross the southern US border just because people are not looking for them. Non-muslims won't be more likely to engage in terrorism because of racial profiling toward muslims. **iii. The potential amendment of the initial profile by the profiling entity** The critique is a critique of the illusion that the same initial profile will keep being valid over time. There's no critique towards a profile that continually changes. Law enforcement institutions are part of a dynamic interaction process and are capable of strategic adaptation; they can revise a profile based on strategies of criminals. **iiii. Ignoring the officious character of racial profiling in the computation of incentives** Harcourt's reasoning assumes that the existence of the practice of racial profiling is universally known. **Profiling** is either illegal or broadly considered as illegitimate, meaning that it is broadly officious and concealed. **iiiii. The virtues of a political, contextualist approach** Harcourt can only demonstrate that racial profiling is irrational in a subset of cases. It's necessary to understand the grounds on which racial profiling can be rational, but can still be dismissed. According to Randall Kennedy (Race, Crime, and the Law, Pantheon, 2003), racial profiling should be forbidden because, despite its efficiency, the social costs are larger than the benefits. **6) Toward a general critique of race-based statistical discrimination** **a) The issue of social costs** Statistical discrimination may be efficient from the perspective of the decision maker, but may be seen as unacceptable once all of its costs are taken into account. If we broad the entity of reference to cover society as a whole, it will yield a negative assessment of statistical discrimination. It may be efficient for an insurance company to discriminate against insurance purchasers based on genetic features when they are strongly predicted to result in disability or death. However, if insurance purchasers don't pursue genetic testing in fear and if genetic testing can create better healthcare, the use of genetic testing could not be efficient for society as a whole because public health would be compromised. **b) The risk of overusing the racial proxy** A ban on racial profiling can be grounded on a historically informed judgement that the proxy may be used beyond its statistical value. Race, as a proxy, does have a statistical value. However, there are specific historical and sociological reasons to fear why race as a proxy could be overused. Individuals will jump at the ability to use race, without any pretext for doing so, because the US is a racist society. This is why it's a better idea to use a blanket prohibition on the usage of race as a proxy because enforcement could be too difficult or too costly. **c) The specific unfairness of statistical discrimination arising from non-statistical (unjust) discrimination** It's not because the proxy is reliable that it is acceptable to use it. If that statistical accuracy of race as a proxy comes from some morally objectionable pattern of behavior, there's a reason to balance statistical discrimination - despite it being an effective instrument. **III- The Moral Approach of Discrimination: When is Discrimination Wrong?** **A) The moral status of discrimination: a 'choice-insensitive' issue** According to Rawls, every theory of justice has to be compatible with the idea that race-based discrimination is wrong. The question of why discrimination is wrong when it is wrong has become more complex. Ronald Dworkin introduces a distinction between choice-sensitive (the solution to which depends on the existing preferences of the political community; a utilitarian answer is adequate here (i.e whether or not to build a playground)) and choice-insensitive (the right answer doesn't depend on how many people believe that something is wrong. Slavery was wrong both before and after the 13th amendment) issues. Whether discrimination is wrong is a **choice-insensitive** issue. **B) The weaknesses of the immutability/uncontrollability-focused account** There is a popular view that discrimination is wrong when it operates based on individual features that are either immutable, unchangeable, or not under the control of the individual possessing them. There are two reasons why this will not happen. Some instances of discrimination are legally banned, despite being based on features that are mutable (age) or controllable (religion). Discrimination can be wrong, even if it is not based on immutable or uncontrollable features. Denying a drivers' license to blind people or denying a short person a space on the basketball team are acceptable for utilitarian reasons. **C) The weaknesses of the effects-focused account** Discrimination is wrong whenever it has harmful effects on individuals. By definition, all forms of discrimination will have some form of consequence on affected individuals. If we consider only the immediate consequences of discrimination, this entails that discrimination is always wrong and that there's no morally distinction between straightforward racism and race-based affirmative action. If we take the view that we should consider the long-term consequences, these consequences could be positive - despite the initial discrimination being unfair. If a Jewish applicant is not hired due to being Jewish and leaves 1939 Germany as a result, she is better off than if she had stayed in Germany. Still, the refusal to hire her seems to be a pragmatic case of unjust discrimination. **D) Deborah Hellman's meaning-focused account:** discrimination is wrong whenever it expresses a demeaning message denying the discriminatee's equal moral worth In *When is Discrimination Wrong?*, Hellman states that discrimination is wrong because of its expressive value. Whenever it expresses a meaning that is intrinsically anti-egalitarian, it is wrong. You must ask yourself "what kind of meaning does the discrimination convey?" It is wrong when it sends a demeaning, degrading, stigmatizing message about those being affected by it. However, it can do this without intending to do so. If a customer at a fancy hotel hands his keys to the first well-dressed black man he sees under the assumption that the man is a valet, it sends an objectively demeaning message; however, it was not intentional. In addition, the target may not feel demeaned or stigmatized. Lastly, the judgment about something demeaning will shift over time. **E) Confirming illustrations from U.S. case law** The U.S. **1964 Civil Rights Act**, **Title VII** (banning discrimination in employment based on race, color, religion, sex, and national origin) has exceptions - especially for sex, national origin, and religion. There are cases where those are bona fide occupational qualifications. When are those generally forbidden grounds of discrimination considered as legitimate criteria? In **Fesel v. Masonic Home of Delaware** (U.S. district court decision, 1978), a residential retirement home refused to hire male nurses. Responsibilities of nurses involved intimate personal care; 22 of the guests in the retirement home were women and did not want their needs attended to by male persons. The courts accepted the preferences of the customers as a bona fide occupational qualification. The argument was that those preferences did not rely on any demeaning, degrading, stigmatizing view of any group. They could be rooted in religious doctrine or could also be rooted in non-religious, social norms of privacy. However, if the claim was that residents did not want to be touched by black, female nurses, this would not be acceptable. Race is not acceptable as a bona fide occupational qualification. The deeper reason is that the preference for white nurses cannot be disentangled from an objectively demeaning view of black people, and no court would be willing to validate this. The outcome is that legitimate and illegitimate customer preferences depend on the equal dignity of all people. **Conclusion: social, 'relational' equality (Elizabeth Anderson) as the underlying purpose of the antidiscrimination project** We have a broad conception of equality that shows that discrimination is wrong when it conflicts with the equal dignity of all people. We learned about "equal respect and concern" (Dworkin) that only applied to state actors. However, this definition of whether discrimination is wrong is even broader. The underlying purpose of the antidiscrimination project is about the creation of a society of equals.