A reply to “not so meritorious”

Co-blogger Ruchira blogs at length about caste, reservations and “merit” — or the purported illusion thereof.

This brings to my mind long email threads that I got into four years ago on the ever-so-volatile issue of reservations (and a pretty terrible blog post I wrote on the subject — I have learned a few things about writing style in the last four years). It is always interesting to me how much passion this excites on both sides of the debate. This may well be one of the topics that draws a lot of heat and little light.

I have made some passing observations about affirmative action and reservations/quota policies in the past here (see also a somewhat related post on privileges versus incentives) and would also recommend Thomas Sowell’s book and Thomas Weisskopf’s book on the subject for those who are interested. For this blog post, though, I will avoid the subject of actual policies (the prescriptive aspect) and instead critique the descriptive claims that Ruchira makes in her blog post.

Merit sherit

Ruchira writes:

Of course the nature of this oppression has changed. Earlier it was untouchability, being denied education and being looked upon as unclean as a consequence of the caste division of labour. Now it is merit. Oppression works in insidious ways. It prevents inter caste marriages, denies education to one’s forefathers and then declares that all are equal and one must compete on the basis of merit. What is this merit, this talent, this intelligence?

The sentiments expressed here are widely shared. A Google search for “so called merit” reservations returns about 12,000 results, such as this, this, and this. Since “merit” is the common buzzword in arguments against quota policies, this is unsurprising. But what does all this really mean? And are people just using the wrong English word?

I distinguish between three somewhat different notions. The first is the notion of desert, which is the noun form of deserve. Somehow, the term isn’t used very commonly in India. Just deserts refers to what we think people deserve based on our notions of justice and fairness. Since different people have different conceptions of what is fair and what is unfair, notions of desert vary. But here’s one example where there would be reasonable agreement among large classes of people — somebody who has been through a lot of humiliation and beating deserves special care and treatment to help him/her recover from that oppression. Of course, there are further questions of who owes this person that special care and treatment — particularly if it is not possible to coerce that person’s past oppressors into providing that special care and treatment. But I think it’s a fair statement to say that many would feel that, practical considerations notwithstanding, such desert is important. Another application of the concept of desert may be where we say that a person who shows exemplary qualities of character such as being hardworking, conscientious, kind, compassionate, and responsible, deserves to live a life of dignity with at least a basic minimum standard of living. Again, there may be the question of whether this intuitive desert translates to actual obligations on other people to provide this person with that basic standard of living — questions that I’ll set aside for now.

There is then the concept of productivity (primer on productivity at EconLib). In the context of work, productivity refers to the output per unit time (e.g., how many widgets do you produce per hour?) or per month worked — which would give a somewhat different measure since some people work more hours than others. Clearly, different people have different levels of productivity in different occupations, depending on their personality, their experience, and their abilities. And when comparing people across different occupations, productivity comparisons would invoke the value (monetary or other) placed on the various types of activity — and these comparisons can change depending on what kinds of goods other people are interested in using or buying.

What then is merit? It means different things to different people, but for now I’ll treat it as meaning potential productivity. In admissions to colleges, for instance, a merit criterion could be how productive you will be on average in your college years at the main task of college — learning. Another related criterion might be your productivity in tasks related to the discipline you’re learning about at the end of college. Yet another variant would measure the gain in productivity you would get through the college education relative to your current level of productivity.

The first question then is: does merit so defined make sense? The second question is: do college entrance tests measure this merit with reasonable accuracy? I think the answer to both questions is roughly yes. People who do poorly on a test such as the IIT-JEE are generally not prepared for studying engineering or other related disciplines at the IIT. Although I haven’t myself studied at the IIT, I did take the JEE and I know that ranks at the IIT-JEE correlate reasonably well with other rough measures of cognitive ability that I could use. In particular, many of the brightest people I knew (based on independent personal assessments) did well in the JEE, and many of these same people are now in cognitively more challenging and higher end jobs than those who scraped through at the lower end of the JEE. Exceptions abound, and I don’t expect the correlation to be strong within a restricted sample, but it’s there. (I don’t know too many quantitative studies done in India. But similar studies regarding the SAT in the United States have yielded similar results — see this for instance. There’s a huge literature that you can start getting a peek of with this Google Scholar search on the SAT’s predictive validity for college grades).

It is still possible to justify reservations on merit grounds. It is possible to argue that although entrance tests generally do a reasonable job of assessing merit, they under-predict the merit of people from certain backgrounds. This under-prediction could take the form of lower predictive validity (i.e., the test score of a person from the under-privileged group reveals less about his/her “merit” than for a person from a privileged group) or a difference in intercept (i.e., a uniform under-prediction of the performance of one group, which can be corrected for by a suitable monotonic transformation on the scores). Roughly, if either story is true, then a student from an under-privileged background who scores the same as a student from a privileged background in the entrance test should have higher merit (because achieving that same score required a more uphill struggle), and hence, should perform better (and definitely not worse) by the end of college. (To take an analogy, suppose a person was severely ill and unable to concentrate on the day that she wrote the JEE, and she got the same score as another person who was feeling fit. Then, you might predict that the person who was severely ill has more inner talent and hence should do better through college, unless the illness will debilitate her throughout the rest of her college days.)

While I’m not sure if either an intercept difference or a difference in predictive validity have been observed in India, the story in the United States in white-black comparisons suggests that blacks with equivalent SAT scores do somewhat poorer than whites — the opposite of what the story would be if the entrance test were the main point/source of discrimination. There are again a whole host of studies in this regard, yielding somewhat different conclusions, but the general consensus seems to be that if the SAT is biased against blacks, it is less biased than other methods of evaluation, including college grades. See here for a start on exploring the studies. Again, things could be different in India, because test design in India is probably not such a well developed field.

So, the “merit” argument — as I use the term — could in principle be used to justify reservations, but this justification is empirically testable, and it seems to me (though I don’t have area specific evidence and am open to being convinced to the contrary) that the empirical tests would largely refute this merit argument.

Other arguments continue to exist — namely the “just deserts” argument. This argument would say that, hey, even if the person from the under-privileged background does not do as well as the person from the privileged background whom she displaces, i.e., the productivity acquired at the end of college is lower, it is still the right thing to admit this person, because the person has worked harder and suffered more discrimination and oppression. This matches with some intuitions of fairness, where people who work hard and are denied the opportunity to acquire certain skills through no fault of their own, deserve an extra leg up. These arguments can again be subject to empirical tests — are the people who benefit from reservations generally more oppressed and discriminated against than the ones who are left out? And, does reservation help them more than hurt them? Here, the answers seem to me to be (relatively) more in the affirmative, and hence, they provide a stronger rationale for reservations. However, as promised, I will not go into the detailed pros and cons.

My point here is simply to delineate the merit argument clearly and point out that this argument is perfectly valid and important, regardless of the way the evidence on it turns out, and regardless of how much other concerns (such as “just deserts”) matter. Moreover, the two arguments are distinct (though related) and hence one argument cannot be refuted with the other.

Head held high

Repeating this sentence from Ruchira:

Today, I can hold my head up and say that I am intelligent and my merit can take me anywhere.

Although she can certainly say that, is it true? Can her “merit” take her anywhere? Can she, if she so chose, become an outstanding person in any field of endeavour, or even in her chosen field of endeavour?

Sowing and reaping

Ruchira further writes:

I can speak well, write well and I can handle basic math. This is not because I have worked harder on these things than anyone else but because my parents have post graduate degrees. My mother’s father was also a post graduate and had in fact, topped both, BSc as well as MSc. My father’s father was a BA and he couldn’t have gone that far if his uncle hadn’t been in the academic profession. Both my grandfathers worked and struggled hard against phenomenal odds but though they weren’t Brahmin, they did not belong to backward castes. Of course, there are first generation literates who do get into higher education and learn to speak and write well but they are more the exception than the norm. Most people who get to do their MAs, leave alone their MPhils and PhDs, are people like me –children of educated parents. Don’t believe me? Check out the econometric models set by Amartya Sen and Jean Dreze in India: Development and Participation. The number of years of formal education of the mother have a positive, significant effect on that of the child. This is not just an empirical result but an intuitive argument. Having educated parents means that education means something in your family. No one will expect you to drop out of school and college to work. Also, educated parents can help you with homework, sympathise and work out ameliorative measures when you fail an exam and support and advise you on how to get ahead- something economists tend to ignore as the emotional side of the story.

While nothing here is wrong per se, it is a severely limited perspective. Psychologists, sociologists, and economists have for the past several decades been trying to figure out the extent of similarity between parental and child outcomes, and the causes behind this similarity. Just a few days ago, I came across yet another paper (PDF) (linked to from this blog post by Matthew Yglesias) that considered four possible sources of the correlation between parent and child outcomes. The paper, based on analysis of data in the United States, says:

We find positive correlations between specific characteristics of parents and children. But we also find that few parental characteristics predict characteristics of children other than the same one that is measured in parents. Four mechanisms might explain such correlations — socioeconomic resources, parenting practices, genetic inheritance, and role modeling. These four mechanisms make varying predictions about which parental traits will be correlated with which child traits; whether the traits of fathers or mothers should be more important to sons or daughters; and to what extent parental socioeconomic characteristics [SES], parenting behaviors, and children’s identification with their parents account for the observed correlations. Our evidence provides little support for the SES and parenting explanations, but more substantial support that role modeling may account for some of the intergenerational correlations, and genetic factors may account for others.

Since data differs from country to country and place to place, the authors’ conclusions may well be disputed in contexts different from those they used for their data. Each of the four explanations (SES, parenting, genetics, and role model/identification) are “obvious” and “intuitive” and empirical data can be gathered to support it.

If you just concentrate on one explanation, it is usually not hard to “explain” everything using that one explanation. If I were adopting a strictly genetic perspective, I could argue, for instance, that the fact that Ruchira’s personal educational outcomes matched those of her relatives were due to genetic similarities: in fact, even a correlation of 0.7 with parental outcomes could be consistent with a 100% genetic explanation because children share only about 50% of their segregating genes with their parents (modulo some caveats; I will not go into the technical details here). This of course is not to say that a genetic explanation is correct, but rather, that the fact of a correlation between parents and children does not say much either way about the causes of that correlation (duh!) — and some causal mechanisms that are important in some places and times may be relatively less important in other places and times. The tricky thing (which these authors and many others attempt to do) is devise approaches (such as controlling for various parameters) that allow them to judge the relative strength of each of the proposed causal mechanisms. Doing this in the social sciences is a tricky job — see this article by Jim Manzi — but that doesn’t make blanket declarations any more trustworthy.

Ruchira then writes:

Thus, this ‘merit argument’ is eyewash. A system has denied education and what Nivedita Menon calls ‘cultural capital’ to a family over generations.

This is an inaccurate caricature since the “merit argument” does not deny the existence or importance of cultural capital — it simply argues from the perspective of merit, not desert. By the merit argument, a person from an under-privileged background who has done poorly on an entrance test but can “overcome” the cultural under-privilege once that person has joined the college or organisation, should be given admission. The question then (from the perspective of the merit argument) is whether the cultural capital handicap is overcome, or narrowed significantly, once the person is admitted. The broader question may be whether “merit” should really be the criterion rather than “just desert” — but whatever Ruchira’s personal opinions on the relative weighting of merit and desert, she’s wrong to say that the merit argument is hogwash.

She further writes:

Going by the prime assumption within the merit argument itself –that there are no social differences between people –why are there so many Dalit construction workers, shoemakers, crematorium workers, cobblers and washermen and such few Dalit teachers, professors, experts or corporate sector employees? Is it that the Dalits do not have merit? Is that not a casteist conclusion?

Here, Ruchira uses a remarkably bizarre debating tactic. First, she sets up an argument that could potentially undermine her position. Then, she says that if you agree with that argument, then you are a “casteist” — and since nobody wants to be called casteist, you must oppose that argument, and hence support her position.

Ruchira writes:

Talking to people on reservations is generally frustrating. For one thing, they give you arguments like –“But you know, you wouldn’t want to be in a college or filling a post through a quota. You’d like to do it yourself. A quota is an insult. It implies you can’t do it on your own. Quotas are casteist.” This argument usually frustrates me so much that I am usually left agape, striving hard to pick up my scattered wits.

I’m surprised Ruchira had a hard time knocking this (weak) argument down. Some people hate the idea of filling a quota position. Others don’t mind it that much. Some people don’t mind filling a quota position, but don’t like it when others take note of the fact. This is all natural. Of course, if everybody greatly hated the idea of filling quota positions, people would not eagerly be re-classifying themselves into castes and tribes that were eligible for quotas. So, on balance, at least for some individuals, their desire for college positions, jobs, and the other perks that reservations entitle them to overcomes whatever self-image they may be sacrificing.

The more correct version of this is that quotas generate over-resentment (because more people think they’ve missed out on an admission because of quotas than those who actually have), as many people have argued, and the stigma of being a quota beneficiary also extends to individuals from the oppressed castes who did not benefit from quotas, or who would have done equally well without quotas.

Living examples

She then writes:

A cogent argument that often proves to be hard to refute is that reservations should not be given at the level of higher education –that a Dalit graduate is akin to any other graduate. But no one quite understands how very important it is to have more Dalits in teaching posts. It can break the Brahmin hegemony in education! It can even break the all powerful notion that Dalits are not meritorious. More Dalits in professors’ posts will go to show that levelling the playing field can put a Dalit where a Brahmin considers it her natural right to be. More Dalits in influential posts can help lobby against ‘merit’-based arguments and ensure reservations elsewhere aren’t scuttled. They can also make policy recommendations to ensure social justice more effectively -redistribution of land, reservations in primary schools and can politically engage with and beat the merit argument. It is also psychological. For all the Dalits who are afraid to chase their dreams, they can be told, “Looks at XYZ. He’s a Dalit like us.” As a woman, I feel that way about our few women professors. Whether they are feminists or not, seeing them there gives me hope.

These are all plausible arguments, but the problem is that opposing arguments are equally, and sometimes more, plausible. Ruchira says that more Dalits in teaching posts can break the all powerful notion that Dalits are not meritorious. It could, if the Dalits in teaching posts are just as good teachers as non-Dalits in corresponding teaching posts. (In fact, in so far as existing prejudice biases people against Dalits, the Dalits in teaching posts would need to be better than corresponding non-Dalits to break even in public perceptions). But selecting Dalits to fill quotas if they don’t have the requisite qualifications and skills could undermine this very purpose. People who formerly only had a hearsay prejudice that Dalits are incompetent teachers now get living confirmation of the fact. Of course, it’s possible that in the type of jobs that Ruchira is referring to, selection for teaching posts is anyway so far from meritorious that reservations will not cause any significant decline. I’m just pointing out that the line of argument I’m making is at least as plausible as Ruchira’s. If public perceptions are key, it may make sense to be even more selective when hiring Dalits so that students see only the best Dalits and hence form even more positive impressions of them.

The rest of her arguments involve the idea that larger Dalit presence would lead to greater endorsement of specific policy positions that Ruchira supports (and this is a good thing in her view). I’m not exactly sure why Ruchira thinks that people from Dalit backgrounds would share her particular policy positions, but since I don’t think that teaching posts are or should be a place for political activism, I’ll refrain from commenting on this. Perhaps she thinks that the self-interest of Dalits will lead them to support positions that she (as a non-Dalit) supports for selfless ideological reasons? While identity politics is certainly strong and influential, this by no means implies that most or even all Dalits who reach high positions would or should be supportive of reservations policies. Again, taking an analogy with Blacks in the United States, people ranging from Thomas Sowell to John McWhorter to Walter Williams have written out in varying degrees against affirmative action for blacks.

The anti-climax

Lastly, reservations are a right. No one has the right to look down on someone who has got a post or a seat via reservation. She is only exercising her democratic right.

I’m not sure in what sense Ruchira is using the term “right.” Let’s assume that reservations are a right (at least for the SC/ST groups for which it was originally intended), on the grounds that the Indian constitution makes an explicit provision for it. But a person having a right to reservations does not deprive aonther person of the right to look down on that person for having got a seat for reservation. My having a “right” to express views that Ruchira considers obnoxious, or to beg on the street, or to praise the virtues of patriarchy or free love or fundamentalist Islam, does not deprive Ruchira of the “right” to disapprove of these activities, or look down upon me for performing them.