Going from standardized tests to recommendations will compound inequality in the long-term

The New York Times and The Wall Street Journal are both reporting that a coalition is suing the University of California system to remove standardized tests. This is ostensibly in the interests of equity.

But really the issue is ethnicity and race. The UC system does not represent the students of California very well. This may not be sustainable for a public university system. I say the issue is ethnicity and race because UC does a reasonable job with economic diversity. A few years ago The New York Times reported that UCLA was the elite university with the highest proportion of students from the lower and middle class in the nation. If you use the interface at the story you see quite clearly that the UC schools have a very large proportion of students (30-40%) from the bottom 60% economically.

But, for example, 35% of the undergraduates at UC-Berkeley are of Asian ethnicity. 15% of California’s population is. This is causing problems.

Most of the people reading this weblog know all the facts (you have surely read Intelligence: All That Matters). Rather, I would like to remind readers that Chinese society has utilized testing to select officials for thousands of years.

What can we learn from this? First, the tests were not fair. You often needed resources to engage in the study of “classics.” They clearly favored those with privilege. But, they were invariably fairer than other paths to an official position. If someone was bright, but without resources, a village or a whole lineage might pool resources to subsidize their learning. There were cases of people from very poor backgrounds who obtained high office, even if it was irregular.

In contrast, other paths toward an official position, such as recommendation, tended to be monopolized by those with connections.

The “natural experiment” has happened many times. The plaintiffs in the above case assert that “teacher recommendations” “would provide a fairer way of judging students.” They are either ignorant or being disingenuous. Teachers are human, and there is a fair amount of evidence that they naturally bias toward believing children with polish and who “look the part” are smarter and more competent.

In the near term, I think the UC system can tune their demographic problem of nonrepresentation by removing standardized tests, since Asian students, in particular, do so well on these. But over the long run the more you move toward subjective and intangible methods of judgment, the more people with connections and privilege will manipulate the system and benefit from it.

Note: Just to be clear, I think American society will get rid of “objective” measures that have a “disparate impact.” But, ironically or not, I think in the long-term it will probably just solidify class privileges even more than is currently the case.

0

17 thoughts on “Going from standardized tests to recommendations will compound inequality in the long-term

  1. I don’t think eliminating standardized testing is as much of an issue for smaller schools and private liberal art colleges, because those aren’t the big vehicles for social mobility that state flagship schools and public school systems like UCLA and the UC schools in general are. Sure there are problems with the SAT and ACT’s ability to predict success, but we don’t have a good alternative atm.

    A big reason why these public schools are so disproportionately Asian American is because they’re relatively open to out-of-state Asians- who pay significantly higher tuition for undergrad. This also makes them less socioeconomically representative of the in-state population.

    1+
  2. @khuzifenq.2: One of the things that came out of the suit against Harvard in this calendar year is that roughly 40% of applications are from people of Asian ancestry, the current student body is about 20%-25% of Asian ancestry and the share with Asian ancestry in the US population is about 4%-5%.* I mentioned this to a recent Ivy graduate of my acquaintance who observed that many of the students are from abroad, spending perhaps HS or only a year of HS in the US before applying (and matriculating), with the result that this overstates the share of Asian-Americans at Harvard.

    *I am pulling these figures from memory. I think they are correct within an order of magnitude, base 2. Also, I think the term Asian ancestrycovers all parts of Asia except for west Asia, thus conflating East Asia, South Asia and SE Asia, the parts that by US norms would still be considered non-white after a generation in this country.

    0
  3. IIRC, Texas guarantees a spot at one of its public universities or colleges to anyone who graduates from a public HS in the top 10% of the class. Any idea how that is working out with respect to matriculation, retention, graduation and all the other criteria of interest?

    0
  4. As a short, bald, and ugly male, I thank God that I didn’t have to interview or get recommendations. I’m in a RN program and half the students think interviews should be part of the admission process.

    0
  5. An interesting intersection of testing and recommendations comes with gifted programs. A number of school districts have found that switching to universal testing for gifted programs led to more balanced ethnic representation of previously underrepresented ethnic groups than their old system of relying on teacher recommendations. If Card’s research on this subject isn’t used to dispute the claim that recommendations are fairer, then the state isn’t doing its job.

    2+
  6. acquaintance who observed that many of the students are from abroad, spending perhaps HS or only a year of HS in the US before applying (and matriculating), with the result that this overstates the share of Asian-Americans at Harvard.

    this makes zero sense. the overseas students are in the ‘international’ category. they’re evaluated differently. but if this person went to harvard it makes sense. they can talk, but not count 😉

    1+
  7. A number of school districts have found that switching to universal testing for gifted programs led to more balanced ethnic representation of previously underrepresented ethnic groups than their old system of relying on teacher recommendations.

    stop introducing well known facts into the argument.

    2+
    3
    1
  8. “I think American society will get rid of “objective” measures that have a “disparate impact.””

    Alexis De Tocqueville would nod in agreement. Book learning is an aristocratic enterprise. Democratic peoples (in his sense of the word) reject it as a qualification for anything.

    The only way around this problem is make college admissions a true lottery. Then it would be fair to everyone.

    0
  9. “… went to harvard it makes sense. they can talk, but not count”

    1) No not Harvard; was admitted but was ultimately unimpressed by the intellectual atmosphere and chose a different Ivy.

    2) Likely right about the inability to count. This individual ended up a mathematician. For counting, I would trust an experimental physicist or chemist much more than a mathematician. 😉

    1+
  10. I can’t see how “Teacher recommendations” can ever be practically used as an instrument for university entry. There are tens of thousands of teachers and they all have different experiences and criteria. How could universities sift between all of these, and how could they then not be immediately lawsuited for whatever decision they come to from it?

    Whether we think about “teacher recommendations” as having more equity and genuine ability to select excellence on a small scale, applied by a fairly uniform means to a small pool, they would quickly become either bureaucratized (and probably be gamed by exactly the families that do well on standarized testing) or fail, and be impractical to pursue on a large scale.

    On the other hand, on the question raised of how different regimes of educational selection ever impacted or will impact inequality, I’m not sure what evidence there is that testing can really be credited with increases in equality, whether income, wealth, social status, consumption, or political equality. Is there any evidence of before-after increases in equality in any of these that can securely be linked to the implementation of testing?

    Even less so with Chinese history using much more small scale and less culture-blind testing to select a small pool of officials and posting them remotely from their birthplace, rather than recruiting directly from local gentry (which probably lessened the willingness of local gentry to fund, support and cooperate with officials, a cost worth considering).

    0
  11. Razib: If someone was bright, but without resources, a village or a whole lineage might pool resources to subsidize their learning. There were cases of people from very poor backgrounds who obtained high office, even if it was irregular.

    This seems like a pretty undesirable example to focus on as a defense of standardized testing though.

    To unpack, the classic defense of standardized testing has always been that once you have a uniform level of basic primary education, you then select for inherent differences in ability that have not much to do with community support from your birth community (certainly not a large extended family), and this is then to the benefit of society as a whole, mainly through excellence in science (not monopolizing politics, which is in theory dominated by democratic representation).

    Whereas this kind of framing where the benefit of testing is that ambitious communities and subcultures can fund intensified studying by pre-selected members (the “Golden Boys” of particular nepotistic lineages), and then pilot those individuals into positions of power, you would seem to have people at large thinking “Why on earth would we want to support this? How is this better that elites that simply perpetuate themselves without regard to natural talent?”.

    So I don’t know if this framing is useful. The useful framing seems more like framing how standarized testing is exactly unlike the historical example you note (presuming that it actually is!).

    2+
  12. Let’s see, how could a parent get their child a better teacher recommendation? Bribing the teacher? Making a large donation to the school? Make a donation to a preferred charity?

    There is no way any rich parent would ever do things like that. I am sure it would all be honest and reflect the true potential of the student.

    0
  13. Cynically, I think the push to eliminate standardised testing in this case will have a much greater chance of succeeding simply because it is about pushing an over-represented minority out rather than letting an under-represented minority in.

    As to the value of testing itself: I think Goodhart’s law obviously applies in abundance whenever you standardise, as that gives ample opportunity to game the test.

    So really there’s no good solution here (which is Razib’s contention in any case, if his chosen example is anything to go by) and the proposed options seem to be along the lines of “allow a vaguely meritocratic racial/ethnic inequality (which happens to favour Asians in this case), or allow an inequality based on resources, connections and other social capital (which will almost certainly favour rich whites as the traditional elite).

    If I had to try to fix such a Gordian knot, my best approach would be to keep the testing but find ways to make it less easy cram or test-study your way to a good score. Which is a whole can of worms on its own.

    0
  14. FYI, poor Chinese needed a recommendation to even sit the exams. The selection of Chinese officials was a messy, multistage process and examination rarely exceeded how 25% of officials were selected and never exceeded 50%. These weren’t math tests either, they were more like persuasive writing exams with a heavy emphasis on allusions to the classics. Take that as “merit” if you will.

    0

Comments are closed.