The Philosophy and Economics of Measuring Discrimination and Inequality: A Roundtable Discussion

This is an interview by the Erasmus Journal for Philosophy and Economics with Kasper Lippert-Rasmussen, Xavier Ramos, and Dirk Van de gaer, conducted as part of a roundtable on the philosophy and economics of discrimination and inequality. The interview covers the concepts of discrimination and inequality; the current state of the literature on measuring discrimination and inequality; the relevance of measuring discrimination and inequality for policymaking; and the future of measuring discrimination and inequality.


I. THE CONCEPTS OF DISCRIMINATION AND INEQUALITY
EJPE: Let us begin with some preliminary questions to define the terms.Let's start with discrimination.Some authors argue that discrimination is best understood as a moralized concept, as something that's inherently wrongful.Others argue that discrimination is best captured as a non-moralized concept, according to which discrimination amounts to differential treatment, but is not necessarily wrongful.What do you think about these two approaches?KASPER LIPPERT-RASMUSSEN: I think my general approach to this question is that discrimination is a term that's used in different senses.Some people on some occasions use it in a moralized sense and on other occasions people use it in a non-moralized sense.So, I guess that the main thing is to make clear what sense of discrimination one has in mind.In some of my own work, I've adopted a non-moralized definition of discrimination. 1 One advantage of doing so-for the specific purposes I have in that book-would be that, in terms of how discrimination is used in ordinary language, there are at least a number of practices of which people tend to agree they are discriminatory; and yet, they have different views (or perhaps no views) on whether the relevant differential treatment is wrongful.
One example that I discuss is the case of age discrimination, in cases of setting priorities in relation to the distribution of scarce life-saving organs.Many people would say that it is age discrimination when younger patients are prioritized over older patients, but only some of these people will say that this is necessarily unjust.Some of them might even think that it is actually required by justice to give priority to younger patients, in light of the fact that younger patients typically have enjoyed fewer good life shares than others.I think there are other examples, such as various forms of statistical discrimination, where people tend to agree that there is discrimination, but do not necessarily agree on whether it is unjust or whether it is all-things-considered morally justified.One advantage of adopting a non-moralized version of discrimination is that people who disagree about the moral qualities of various kinds of differential treatment can, so to speak, describe their topic as one of discrimination.110 Could it be a disadvantage, then, that we do not have a separate word to describe discrimination that's inherently wrongful?KASPER LIPPERT-RASMUSSEN: This could be a disadvantage that, I think, in much theoretical literature can be easily remedied simply by adopting a specialized term-discrimination with a subscript or something like that-or simply stipulating that by discrimination, you will meet the following (e.g., differential treatment of members of different socially salient groups which is based on irrelevant reasons, or which is unfair), so I think for theoretical purposes it is less of an issue.However, I think that in much ordinary discussions about discrimination sometimes it's a serious obstacle that people do not clarify that they actually have different notions of discrimination in mind.
How would the economists in the room define discrimination?XAVIER RAMOS: I think that economists would subscribe to everything that Kasper just said.We can, however, try to add further comments on some of the issues.Certainly, what one can think of moralized or nonmoralized concepts from our point of view is that concepts that are moralized allow one to guide policy action in a better manner.But it is often difficult to understand why some forms of discrimination are socially accepted and some are not, which is what Kasper was referring to.And discrimination is a great example.
I can think of two more relevant cases that may serve us to think about these issues.On the one hand, for instance, the price that men and women pay for car insurance is different ultimately because car accident rates are different.We (or our society) agree that, since risks are different, insurance companies are entitled to set different prices in order to make their business viable.On the other hand, social security contributions of men and women are usually the same across gender even though we know that the life expectancy of men and women differ.So, in this case, we do not allow discrimination by gender, whereas in the first case we do.I don't know whether this has anything to do with people finding this treatment just or unjust or something else.But I also don't know whether, in accepting this differential treatment, the fact that people conceive it as moralized or non-moralized or just or unjust is relevant.
On top of that, if we adopt a moralized concept, we tend to align moralized concepts with wrongful treatments or wrongful consequences, and this is not necessarily the case.Think, for instance, of affirmative action.Affirmative action or positive discrimination amounts to differential treatment that aims to correct previous differential treatments and it is meant for a good cause.It is thus not supposed to be wrongful.So, it is not all together clear that the concepts of just or unjust serve to explain why people (do not) accept differential treatment.Moreover, it is definitely not clear that moralized equals wrongful.
Let's go to a deeper question about discrimination.According to some traditional accounts of wrongful discrimination, discrimination involves an objectionable mental state, which is rooted in prejudice or animosity.Many have argued that policies can be discriminatory without there being such mental states so long as they produce effects that are particularly bad for some group.Should we adopt such a concept of direct discrimination in our moral language?And wouldn't this definition run the risk of being too inclusive?DIRK VAN DE GAER: It is an interesting question, but Xavier and I think that the concepts of objectionable mental states like prejudice or animosity are pretty useless.We don't think they have much normative appeal.
The first reason is because we think about these kinds of objectionable mental states as a form of mechanism of approval or disapproval of certain behavior that serves to solve some coordination problems between people or to induce them to do something that is morally good.So, using these kinds of concepts (objectionable mental states) has no added value.
The second reason is that we also believe that there are forms of discrimination that do not involve any objectionable mental state.This is the case, for instance, of statistical discrimination for efficiency reasons, as opposed to self-confirming stereotypes.Statistical discrimination does not stem from prejudices or anything like that, and still, it is plausibly a form of discrimination, and that it is not good for society.In a nutshell, we see neither the value added nor the normative attractiveness of concepts like objectionable mental states.
Kasper, do you agree with this statement?KASPER LIPPERT-RASMUSSEN: In my view, it all comes down, at least for certain purposes, to why one thinks discrimination is wrong.If, for instance, you have the view that discrimination is wrong because it is disrespectful and disrespect is tied to having a certain kind of mental state representing those people who may be discriminated against (as, say, having a lower moral status or being deficient in some other relevant way), then, as far as I can see, it's hard to get around talking about objectionable mental states-mental states that having those towards others involve disrespect.Of course, it's true that it is a big issue when exactly a mental state is objectionable.But I think that's simply an issue that is unavoidable, assuming that discrimination is wrong because it's disrespectful.
However, I certainly agree with Dirk about there being forms of discrimination which are not disrespectful and, insofar as they are wrong, they are not wrong because of the objectionable mental states.So, for instance, if you think that certain kinds of sex discrimination are wrong because they bring about unequal opportunities for men and women, then that's a fact about those forms of discrimination which does not depend on mental states.
One final point that connects to the previous issue is that if you define direct discrimination appealing to the notion of objectionable mental states, then that doesn't make your concept of direct discrimination moralized.This is so because what defines a moralized notion of direct discrimination is that one can infer from the fact that something is direct discrimination that it's morally unjustified.But the mere fact that one acts on the basis of an objectionable mental state doesn't entail that the act in itself is morally unjustified, so it will still be a non-moralized conception of discrimination.I don't think that goes against anything Dirk said.
This is a nice bridge to the next question, which is whether or not objectionable mental states are involved in the definition of discrimination.Could it be the case that any form of discrimination is wrong because of its inequality producing effects?So, for example, because it increases group inequality or because it increases inequality of opportunity?KASPER LIPPERT-RASMUSSEN: My view is that there's not a very tight connection between discrimination and equality.So, you can have forms of discrimination which most people would say are morally objectionable but nevertheless don't result in any inequality such as, for instance, racial discrimination against the privileged racial minority.Most people would find that objectionable, but it would be racial discrimination that actually decreases inequality.Similarly, it is possible that there are forms of inequality which are not connected to what people would normally consider discrimination-say, inequalities between people who are naturally talented and people who are not, which would not normally be considered a case of discrimination.But people might think that such forms of inequality are nevertheless objectionable inequality.Hence, I'd say that there is not a very tight connection between concerns for anti-discrimination and concerns for anti-inequality, even though, of course, one of the main objections to certain forms of discrimination is that they increase inequality.
XAVIER RAMOS: I agree with Kasper.The example of the ethnic minority illustrates well how these two concepts can be different and separate.Moreover, one exception to the thought that discrimination is wrong because it prevents at least some people from reaching their full potential (by, for instance, worsening the outcomes of the worst-off) is the one that Kasper was pointing out before.
Group inequalities, however, are different because they arise when people with exactly the same characteristics face the same hurdles to achieve their potential.So, group inequalities are an indicator that, because of certain characteristics, people are treated differently.Segall2 claims that discrimination is wrong only because it undermines equality of opportunity, which is a type of inequality.The reason is that it's certainly unfair, which is what lies behind this relationship with inequality of opportunity.
But there are two further points.The first one is that some kinds of discrimination, for instance affirmative action, can actually equalize opportunities.These are different examples from the one that Kasper was referring to but go in the same direction.Second, discrimination attributes or grounds-characteristics on which discrimination is based, such as gender, race, ethnicity and so on-tend to be circumstances, i.e., characteristics that are beyond the control of individuals.However, if these discrimination grounds are responsibility characteristics (think for instance of obesity, which is the best example we can think of), they actually increase inequality of outcomes (for instance, when speaking of health) but not inequality of opportunity.In sum, when based on a characteristic for which individuals are deemed responsible, discriminatory treatment may increase inequality of outcomes but not inequality of opportunity.This example is in conflict with Segall's argument.

II. MEASURING DISCRIMINATION AND INEQUALITY
Let's discuss the current state of the literature on measuring discrimination and inequality.How do we start thinking about this project of measuring inequality and discrimination?How do we operationalize these concepts, and what kind of tools are being used in your fields?DIRK VAN DE GAER: In economics we have a very well-established tradition in measuring inequality and measuring poverty in which we formulate very precise properties that we want these measures to have.And then, based on those properties, we derive conditions under which we can order, let's say, income distributions.So, we start from axiomatic properties and we combine them into a precise measure.This means that we try to define the concept as precisely as we possibly can in order to make clear statements about when one, let's say, vector of income is more unequal than another one.
The economics literature on discrimination measurement relies on observational or correspondence studies to infer whether discrimination exists or not.This entails that one can only measure discrimination in situations like, for instance, the differences in the shares of individuals that are invited for a job interview or for visiting a flat or a house, between groups of individuals that differ in one characteristic.As a result, in terms of discrimination, we use empirical methods that do not really have a welfare basis: usually we look only at one particular relevant outcome (which is, in case of the example above for instance, whether a person gets a flat or not, whether one gets the job interview or not).This is very different from the theory of inequality measurement, including the theory of inequality of opportunity, where very often we're not only concerned with one particular outcome but we're more concerned with something that is an overall measure of well-being, which may be income, but it could also be something else.In this sense, the way in which we measure discrimination nowadays is much less encompassing than the way in which we measure inequality.
Kasper, would you like to expand on this answer?KASPER LIPPERT-RASMUSSEN: I'm beginning to move outside my field of expertise, so I just want to add that I think it might be interesting to look to other issues comparable to discrimination (say issues about altruistic versus non-altruistic behavior) and ask whether there's any kind of overall measure for the degree to which people act altruistically or non-altruistically in society.I would suspect that there's very little precise measurement of that sort, and I think that if we want some kind of measure of how much discrimination exists in a certain society, many of the same obstacles that economists have come up against would come up again in an instance like that.So, one can measure the degree to which discrimination takes place in a particular sphere, against a particular group, but the sort of overall measurement about how much discrimination goes on in the society, I suspect that that's a hugely complicated thing to measure.Do philosophers have an important role in formalizing a concept of discrimination that is easier to measure by practicing scientists?KASPER LIPPERT-RASMUSSEN: I don't have a good overall answer to that question.I'm involved in two experimental studies at the moment where we try to measure the degree to which people's inclination to classify something as discrimination depends on the existence of a comparator who's been treated better and, secondly, whether or not people object to discrimination based on whether it involves disrespect or harm.And in both studies it's very hard to ask the sort of questions that philosophers would like to ask in order to test people's intuitions.I suspect that similar issues would come up in a lot of attempts to measure the degree to which discrimination takes place.
I think the main contribution of philosophers in this respect would probably lie in trying to provide some more analysis of the concept of discrimination and not so much in the effort to operationalize the concept of discrimination, which I took was what described in relation to having these very precise concepts of discrimination that economists measure.
When it comes to measuring discrimination and inequality, there might be some trade-offs.One trade-off we thought economists might deal with is the following: on the hand, they want to gain society-wide information on the extent of discrimination and inequality, even if the quality of research might be really limited but, on the other hand, they might want to conduct smaller scale studies that allow for precise insights into the structure and nature of discrimination of inequality, though this may be less useful to measure the extent, or societal impacts, of discrimination and inequality.So, do you think that there is such a trade-off?And what are the benefits and limits of either option?XAVIER RAMOS: We realized that often, at least in inequality measurement, the trade-off is not between the width of the group we want to study (that is, for instance, whether we want to study a population or groupinequality in that group or in the entire population) and the quality or the validity of the research conclusions.This is different from the case of measuring discrimination, where there are a lot of case studies and we may think that often these are either informative or externally valid. 3The reason why, when we analyze income inequality, this is not so much the trade-off that we face is because studies suffer from different types of problems.Let me name just three of them.
First of all, we lack information about both tails of the income distribution.With the data we have at hand (basically survey and administrative data) it is difficult to collect data on very poor people (those, for instance, that are living under a bridge or in collective houses like prisons) and also on the top part of the distribution, although there are large efforts being expended by quite famous economists like Thomas Piketty or Tony Atkinson (who passed away shortly ago).So, this problem applies to these ends of the distribution and not to particular subgroups of the population defined by different socio-economic characteristics.
The second problem is that the degree to which using some income sources can be applied to the analysis of other classes of income is limited.One famous example is that of self-employed earnings.We know that self-reported earnings are not truthful, so we are capturing this type of earnings incorrectly.And we encounter self-employed individuals in each population subgroup.
Thirdly, there is also lack of information about important behaviors when we think of income inequality such as tax evasion or tax avoidance, which prevents us from measuring inequality properly.Once again, this is very much related to income but perhaps not so much to all other characteristics.In addition, we know that different sources of data that we can use in order to measure income inequality (I have named two of themsurvey data or administrative records-and currently there is huge work on tax records and on social security records) are captured at the individual level or micro level and they do not match with national account data.It is on these latter data that many policies are made.There is, again, a huge effort by the same group of scholars at the Paris School of 117 Economics that I mentioned earlier to make progress to match these tools, but these data problems apply to all population's subgroups.
These problems apply differently to different parts of the income distribution, but they may apply to different population subgroups.So, there is often no benefit in trying to focus on a population subgroup and analyze the subgroup with the expectation that we are going to have better data and therefore better measurement.The reason, as I have explained, has to do with the income data problems that we face.The take-home message is that the strategy of focusing on a group may be informative and useful for discrimination analysis but not so much for inequality analysis.
Let's return to discrimination and inequality of opportunity.The measurements of discrimination and equality of opportunity are currently seen as distinct research projects, even though we think there is considerable conceptual overlap between the concepts.So, should we move towards greater integration between these measurement projects or, if not, why not?DIRK VAN DE GAER: Concepts need to be defined very precisely.Kasper already said that in the context of discrimination there is a disagreement about what people mean by discrimination and the same is true with inequality of opportunity-there are very many ways in which these concepts can be defined.And I think that the main focus should be on getting very precise definitions of these concepts and all of their different variants, and then investigate their implications for ordering different possible societies (a 'society' is here understood as a complete description of society).
By taking each one of these different conceptualizations separately, one will get one particular conceptualization or one particular variant of a concept and, from that, one can deduce certain conditions under which one can determine whether a society is better than another one or not.It will not always be possible to order societies since the ordering will likely be incomplete.This latter problem, however, can be addressed by increasing the precision of the conceptualizations (or variants), which makes the ordering more complete.I think that this should be done for every possible conceptualization and every variant separately and then confront the conditions for ordering societies.This is the only way in which we can then establish whether the different variants or different conceptualizations will agree on the way they rank different societies.This looks like a much more fruitful way than starting from the concepts and trying to narrow down what is the common theme of the concepts, which I don't think to be a very fruitful exercise.I think it's better to try to work with as precise conceptualizations as possible, apply them consistently, and then then look at the conditions for ordering societies that one can derive from these very precise concepts, even if they do not determine a complete ordering of all societies.

III. THE RELEVANCE OF MEASURING DISCRIMINATION AND INEQUALITY
We have discussed many conceptual questions.Let us now bridge the gap between the conceptual questions and the policymaking applications.Why, if at all, is it important to measure discrimination and inequality?It seems a very hard and controversial process.Why is it important to have a quantitative understanding of this phenomenon, rather than rely on a purely qualitative one?XAVIER RAMOS: Both things are relevant, and the latter question is more relevant for policymakers.Detecting that there is discrimination or inequality is informative and helps, but what can policymaker do if they only have such coarse-grained pieces of information?Should they put lots of resources into addressing inequality or discrimination, or very little?What is the part of the government budget that they should want to devote to that?
Quite likely, this depends on many issues such as social relevance, or even social alarm, but also on whether one thinks that this is a huge or a small problem.This, in turn, will depend on the extent to which one believes that there is inequality or discrimination.The mere presence of discrimination or inequality can help one take some decisions, but certainly not design policies that specify the amount of resources or effort devoted to address the problem, especially knowing that all the resources and efforts that we have are scarce and devoting them to address inequality or discrimination implies subtracting them from something else (whenever resources are employed for something, there is always the opportunity cost of not doing something else).As such, measuring the extent of the problem is extremely relevant and we cannot do without that.I think we are going to be more helpful to policymakers if we can tell them that discrimination is large or small.And how large and how small this is.
Not all methods are able to say that.For instance, the correspondence analysis that Dirk was mentioning before, which economists use in order to detect discrimination, often cannot say much about the extent to which there is discrimination.By sending letters with invented CVs and names or other characteristics, it is possible to detect that there is discrimination because some get more job interviews than others, as Dirk was saying.And yet, it is not possible to gain a deeper knowledge of the issue, since it is not possible to fake the interview in a relevant and credible way.Measuring the extent of discrimination is therefore crucial.KASPER LIPPERT-RASMUSSEN: I guess the reason why the question is slightly provocative is that it somehow implies that it's an either-or question, but, of course, I suspect that the best thing would be to rely both on quantitative measures and qualitative measures of discrimination.In relation to this, a supplementary reason why it might be good to have quantitative measures is that often there can be a gap between, so to speak, experienced discrimination and actual discrimination, so if the qualitative judgments are judgments somehow reporting or being constructed on the basis of when the people themselves see themselves as being discriminated against or discriminating, then those judgments are likely to be quite misleading in a way which, say, making precise quantitative measurements of discrimination can actually show to be misleading.
It seems that the connection between scholarly work on both the concepts of discrimination and inequality and its measurement and policymaking can be complex, or at least not straightforward.Let us thus ask two questions.On the one hand, if there were no connection between policymaking and scholarly work, would it still be relevant for scholars to try to pursue this project of measuring discrimination?On the other hand, what would be the role of scholars in cases in which the connection between their work and policymaking were noisy (their research outputs were used superficially or misunderstood by policymakers)?DIRK VAN DE GAER: I think that the problem is that the task of detecting discrimination and inequality is not only to inform policymaking, it also allows us as economists, for instance, to study how markets work and how inequality arises in the economy.It's important to understand human behavior, and this has a much wider relevance than just trying to inform policymaking, so its task is wider than just informing policymakers.What happens when policymakers misunderstand core concepts?
Well, I think that our gut feeling is that society will be better off if policymaking is evidence-based on scholarly work.And I believe that there's a quite a bit of progress nowadays: scientific communicators or properly trained technicians and advisors can help policymakers to gain a correct interpretation or conclusion.With the right kind of communication, much can be obtained.Moreover, sometimes policymakers might not even want to understand what we're saying.In some cases, they do so because it's a way to keep a free hand and to implement a policy that they like for ideological or other reasons, rather than implementing a good scientifically based policy.This notwithstanding, we should try to communicate as clearly as possible, not just to policymakers, but also to the general public.KASPER LIPPERT-RASMUSSEN: I think that perhaps the relevant comparison is whether it's better that policymakers act on the basis of various misunderstandings of research results, rather than act on the basis of no information about research results at all.In my view, it might well be better to act on some deficient understanding of research results than acting on no research results at all.
Before wrapping up the session, I'd like to ask whether there are questions from the audience.AUDIENCE: 4 One of the things Professor Van de gaer was talking about was that we want characterization results to understand the phenomena properly.Specifically, characterization results of distinct conceptualizations of inequality.But a problem is that these conceptualizations are welfarist: they rely only on vectors of well-being or income or some sort of understanding of preferences, which allow us to get the ranking of distinct vectors of different states of affairs.However, discrimination is relational.It goes away from the question of 'equality of what kind of thing' to a kind of game-form understanding of 'is someone else frustrating the opportunities that I confront?'.So, do we need to move to game-forms to understand discrimination, rather than focus on characterization results that we use in inequality measurement?DIRK VAN DE GAER: Do you want to use game-forms for the purpose of understanding discrimination or measuring discrimination?AUDIENCE: The purpose of it is to get some sort of conceptual grip on the phenomena, since we take discrimination to be some sort of opportunities sets that are not available for another person.Insofar as these opportunities sets are jointly produced, should we be using more game forms in economic theory to understand the phenomenon of discrimination, rather than welfare functions (which economists use for inequality)?DIRK VAN DE GAER: I'm not too sure what we should do that in terms of measuring.However, this is very important for understanding how discrimination arises.We still have a lot of problems in incorporating these kinds of phenomena which arise, of course, in game-forms, which can be understood as externalities.It is very difficult to deal with that from an evaluation formal point of view.As such, everything that we do in terms of measurement theory is very individualistic.We do not have a clear answer to that, at least yet, I believe.So, all the rankings of different states of affairs in terms of discrimination or inequality are based on individual outcomes, and therefore are 'stipulated' from one individual perspective or another.How the outcomes ranked arise is a positive question.However, I believe that some of the discrimination models that have to do with stereotypes can be given a game-theoretic interpretation, but there's a distinction between the normative side and the positive side.And so, in terms of evaluation purposes, we're still very individualistic.But on the positive side, there's lots of things that we can do, and game theory could indeed be very relevant as an approach to get things going.

IV. THE FUTURE OF MEASURING DISCRIMINATION AND INEQUALITY
Let us wrap up the roundtable with a question that looks at the future of the field.How do you envisage the development of the field, both of measuring discrimination and the conceptual work on what inequality and discrimination are?In what ways could these projects improve?Is there room from increased collaboration between different fields?Is there something to be gained by interdisciplinarity? XAVIER RAMOS: It is a difficult question.And so, precise answers are difficult as well.Let us say that we would like to see the field going much in the same direction that Dirk mentioned in one of his answers.We believe that the best way forward is to acknowledge and identify these different ways to conceptualize concepts such as discrimination and inequality as precisely as we can, recognize them as different concepts because the more precise the definitions we're going to have the more social states we're going to be able to order.And then, if there are contradictions because the concepts are incompatible, we'll have to make a choice.After all, this is often what happens when one tries to use different tools to measure inequality.In such cases, different tools correspond to different sets of axioms that the measure should respect, which then define the basic properties that characterize the measure.But it often happens that using different tools (i.e., different axiomatizations) leads to measures that have basic properties that clash.Because the underlying properties are clashing, the economist using the measure has to make a choice.Since the clashes between measures arise from disagreements about the characterization of the concepts used, the primary issue that needs to be addressed is conceptual in nature.Whether economists can measure discrimination or inequality from these properties or not is a different problem, and it's more technical.
But the way forward is to try to identify as precisely as possible the different meanings of what is commonly known as the same concept.In this regard, the collaboration between different social scientists or philosophers sometimes is (and has proven to be) key and a good way to proceed forward.Equality of opportunity is a great example.It's a field that started in philosophy, from which economists drew.With their tools or interpretations, they attempt to push it forward.If we are at the stage we are now, in which we are able to characterize different possible concepts of equality of opportunity with different measurement tools, this is because this collaboration went through.It often helps that there are scholars with one foot in one field and one in the other, like Marc Fleurbaey or John Roemer, who serve to bridge the two.KASPER LIPPERT-RASMUSSEN: One area of conceptual work needed in the future is work that explores what makes differential treatment of certain groups into discrimination.Historically, sex discrimination, race discrimination, and religious discrimination were the, so to speak, paradigm cases.But it's clear that the list of groups that are considered victims of discrimination has expanded considerably since the civil rights movement in the '60s in the US (think about discrimination against smokers or appearance discrimination).So, a better understanding of the concept of discrimination in terms of what the relevant properties are such that differential treatment on the basis of having those properties qualify that as discrimination.
Besides that, I suspect that there are some interesting issues in relation to poverty.Generally, people do not think of poor people as a group that is subjected to discrimination in various forms.Compare for instance the preference that upper-middle class people have for avoiding living in areas where poor people live in part because of the socio-economic status of people who live there (in part for other reasons) with the preference many racial-majority people have for avoiding living in areas where a racial minority form the local majority in part because of the racial identity of most people who live there (in part for other reasons).I suspect acting on the former preference would generally not be considered engaging in poverty discrimination, whereas acting on the latter preference would generally be considered engaging in racial discrimination (albeit permissible racial discrimination perhaps).
I believe, however, that, under many understandings of what makes differential treatment discrimination, it's hard to explain why these two cases should be regarded as different from the perspective of discrimination.If they shouldn't, this can have quite radical implications.For instance, it would imply that we should adopt a view of standard social mechanisms of socio-economic segregation which is as critical as our views of racial segregation, since in the end both are forms of morally objectionable discrimination.
In terms of measurement, I think once one has a better understanding of the list of properties which are such that differential treatment of people who have these properties can cover discrimination, it might also be interesting to see if some kinds of discrimination that we classify in one way should at least to some degree be classified in a different way.Think about the relationship between race discrimination and appearance discrimination.Presumably there might be cases which are now classified as race discrimination which might be better conceived as cases of appearance discrimination.The kind of cases I have in mind here are cases where people treat people differently on the basis of their facial appearances differing in racially stereotypical ways irrespective of whether these people are considered to be members of the same race, whatever that is.So, I think that there are some interesting things here to explore.