Monday, April 25, 2022

Seeing Double Standards

A new article co-authored by Albert Cheng, Jay P. Greene, and Ian Kingsbury was just published in Antisemitism Studies purporting to show that highly-educated people are more prone to adopting antisemitic double-standards

The authors had previously publicized their findings in a Tablet Magazine article, and at that time I actually shared critical feedback with one of the co-authors explaining some serious methodological flaws that severely undermined their conclusions. I had hoped that these concerns would rectified or at least addressed once the article was submitted for peer-reviewed publication. Unfortunately, the final published piece continues to ignore the issues outright -- not even acknowledging their existence, much less meaningfully responding to them. So I'll raise them publicly here, so unwary readers are not led astray.

The basic thrust of the article relies on the notion of the double-standard -- treating Jews differently (worse) than similarly-situated non-Jews should be viewed as a form of antisemitism. Unlike some, I am not instinctively averse to the idea that a double-standard is evidence of antisemitism or other discrimination -- indeed, it seems to me that the core of what discrimination is treating likes unalike. However, the devil in such analysis always is figuring out what counts as a "similarly-situated" case (in anti-discrimination law, we call this a valid "comparator"). If there is a different characteristic than the one being measured that distinguishes the comparators from the base (Jewish) case, then it's entirely possible that that characteristic is what drives any differences in resulting treatment.

For example, suppose one wanted to measure if baseball teams performed better against opponents based in the eastern United States. One thing you might do is look at 2020 MLB records and compare how teams fared against Pittsburgh Pirates compared to the Los Angeles Dodgers, where you would find that teams typically did much better against the Pirates. Hypothesis confirmed? Obviously not -- the Dodgers won the World Series that year, and the Pirates were awful. If we had instead compared the Atlanta Braves (top of the NL East) against the Arizona Diamondbacks (bottom of the NL West), we'd have reached the opposite conclusion -- but the real conclusion is that "it's not geography, it's the quality of the team". The choice of comparators ends up driving the conclusion.

The article constructs four pairs of "cases" which it claims are similar enough such that, if the Jewish-coded option is treated worse, we should infer a double-standard is at work. The pairings are:

  1. Whether "the government should set minimum requirements for what is taught in private schools," with either Orthodox Jewish or Montessori schools given as the illustrating example. 
  2. Whether "a person's attachment to another country creates a conflict of interest when advocating in support of certain U.S. foreign policy positions," with either Israel or Mexico offered as illustrating examples. 
  3. Whether "the U.S. military should be allowed to forbid" the wearing of religious headgear as part of the uniform, with either a Jewish yarmulke (kepah) or Sikh turban offered as illustrating examples.
  4. Whether public gatherings during the pandemic "posed a threat to public health and should have been prevented," with either Orthodox Jewish funerals or Black Lives Matter protests offered as illustrating examples.
Highly-educated respondents tended to be less favorably disposed towards the "Jewish" examples in these pairings, which the authors claims is proof of an antisemitic double-standard in that cohort.

When I saw these pairings, though, what jumped out at me is that each of the "comparator" examples was culturally-coded as to the "left" of its Jewish partner. Montessori schools are associated with wooly progressivism, attachment to Mexico most associated with immigrant communities under siege by Trumpists, Sikhism being a predominantly non-White and non-European religious community, and BLM ... well, that one's obvious.

Hence, it struck me as plausible -- if not likely -- that what we really were seeing in these results is favorability towards perceived cultural-ideological compatriots, given that highly-educated Americans tend to have left-of-center cultural-ideological attachments. If we used comparators which were culturally-coded as to the "right" of the Jewish example, I suspect the double-standard would disappear (or even reverse). Another commenter similarly observed that the "Jewish" examples are consciously stacked towards Orthodox Judaism in at least two and arguably three of the four examples, which also means we might be picking up on views about Orthodox Judaism rather than Judaism writ large.  That's definitely still meaningful, but it is also a more specific finding and one that arguably fits well within my own theory about of cultural-ideological attachment.

Indeed, it was striking and dare I say suspicious how one-sided the choice of comparators was -- even in circumstances where there was a far more obvious comparator that would have coded as more conservative. The clearest example is in Orthodox Jewish funerals versus BLM protests -- Evangelical megachurch services is right there, is far more obviously parallel (as a religious rite) to an Orthodox Jewish funeral than a political protest, and was very much in the news as a substantial area where COVID restrictions were being deliberately flouted under claims of religious practice. Choosing to "compare" the Orthodox Jewish funerals to BLM in those circumstances is, well, it's a choice, and one that feels very clearly motivated by a desire to come to the "correct" result.

Imagine, instead, that the study used the following as its comparator pairs:

  1. Whether "the government should set minimum requirements for what is taught in private education," with either Jewish day schools or Christian home-schooling given as the illustrating example. 
  2. Whether "a person's attachment to another country creates a conflict of interest when advocating in support of certain U.S. foreign policy positions," with either Israel or Russia offered as illustrating examples. 
  3. Whether "the U.S. military should be allowed to forbid" the wearing of religious apparel as part of the uniform, with either a Jewish yarmulke (kepah) or a large crucifix offered as illustrating examples.
  4. Whether public gatherings during the pandemic "posed a threat to public health and should have been prevented," with either Orthodox Jewish funerals or Evangelical megachurch services offered as illustrating examples.
  5. Whether medical professionals should be able to receive exemptions from laws which conflict with their religious commitments, with either "Reform Jewish doctor asking for a religious exemption from laws preventing him from prescribing contraceptives" or "Catholic doctor asking for an exemption from laws requiring him to prescribe contraceptives" as illustrating examples.
In those pairings, the non-Jewish comparator is coded to the "right" of the Jewish example, and my hypothesis is that highly-educated respondents would be more favorably disposed to the Jewish examples compared to the non-Jewish alternative.

To be clear, this would not demonstrate that actually highly-educated Americans harbor a philosemitic bias, for the same reason why the published study does not show highly-educated Americans harbor an antisemitic bias. Rather, we'd have a situation where highly-educated persons treat Jews worse than some cases and better than others, and we'd need an explanation why -- and my proposed explanation is "perceived ideological-cultural compatibility".

It is unfortunate that the study did not do a better job picking its comparators (to be charitable -- the uncharitable take is that the authors did exactly the job they were intending to when picking their comparators), and it's still more unfortunate that the authors did not even address this criticism. The closest they come is in the following passage in the conclusion:
Of course, critics of this research might object that our four paired items are not truly comparable. They might claim that the public health dangers of BLM protests and public funeral gatherings are not the same, or that a yarmulke and turban are not the same, or that dual loyalty concerns with respect to Mexico and Israel are not the same. However, we are all limited by the lack of a clear objective standard for defining bigotry and for making defensible distinctions based on different circumstances. In fact, every bigot claims that the groups they treat more harshly are deserving of that treatment. All we can do is appeal to the critic to consider whether public health is jeopardized more when Jews gather than when protestors gather, whether yarmulkes interfere with military necessity more than turbans, or whether dual loyalty is more problematic if it involves one country over another. Like the average individual in the general public, we do not see differential treatment in any of these scenarios as defensible. That said, we do encourage the creation of new item pairs that closely correspond to each other and cover a variety of other scenarios so that any observed differences in response patterns can be attributable to whether the item concerns Jews or non-Jews as opposed to any other potential differences across the items.

The first part does not actually confront the criticism I offered. Certainly, one could argue that the paired items are not actually "the same" in relevant respects. But the criticism I'm leveling relates not to how the proposed non-Jewish comparators differ from theirs Jewish counterparts, but between how the proposed non-Jewish comparators differ from potential alternatives that have a more conservative as opposed to liberal cultural-ideological valence. Perhaps it's still the case that, for example, the public health dangers of BLM protests and Orthodox funerals and Evangelical megachurch services are the same, such that the truly virtuous (and public health minded) deliberator should treat them all identically. But if they don't, the reason why most likely relates to a different sort of in-group affinity than antisemitism -- particularly if they are relatively less favorable towards the Jewish case compared to the BLM example but more favorable compared to the Evangelical example. So while I'm heartened to see the authors encourage the "creation of new item pairs" that can drill down on these questions, that doesn't excuse the authors' own failure to construct their study with appropriate care and rigor.


David Lev said...

I dunno, I would say"I think a US soldier should be allowed to wear a turban but not a yarmulke" or "I think asserting that a Mexican-American legislator's opinions on issues related to Mexico are influenced by their race is outrageous, but a Jewish legislator's opinions on issues related to Israel should be fair game" seem to me to be pretty obviously and unambiguously antisemitic (the other two I would agree have enough confounding factors that it's hard to really compare)

David Schraub said...

I don't necessarily disagree. But the broader conclusion that highly educated people are more likely to be antisemitic than their less educated peers probably depends entirely on whether our comparator is Russia or Mexico, or turban or crucifix.

Put differently, is the respondent who is most willing to accommodate the turban, then the kipah, then the crucifix more or less antimsemitic than the one whose order is crucifix, then kipah, then turban? I don't know, and I actually don't think we've revealed anything.

Erl said...

Thanks for the post—I remember this paper the first time it came around, and its stacking of the deck was indeed rather obvious. Good job laying it all out.

Erl said...

Re: the turban question—I can't find the data directly but it's enormously suspicious the media coverage did not publicize the underlying levels. They report the comparison of the comparators as the dependent variable: "highly educated people 10 percentage points more likely to prefer Sikhs over Jews" or whatever. But that's perfectly consistent with a world where the less-educated population is 10% yes for kippot, 0% yes for turbans and the highly-educated population is 100% for both. It's a meaningless statistic without the ground reality.

The Tablet article does show one underlying graph, though, and it's the BLM/Orthodox wedding one!

Which comparison is as tendentious as they admit. Source: I was living in Crown Heights in 2020, tracking the COVID numbers and going to BLM protests. The BLM protests were universally masked, spaced out, outdoors, attended by young people, and unaccompanied by COVID spikes. The Orthodox weddings were inside, unmasked, full of the elderly, and at least correlated with that community experiencing greater COVID spread.

I wanted the weddings to stop, because, Jew that I am, I like to see fewer dead bubbes and zaydes. So sue me.

David Schraub said...

Erl -- I had not even picked up on that. It fits with behavior from a different study that came from this cadre, where they try to present DEI professionals as having an antisemitism problem because of their twitter accounts' disproportionate focus on Israel compared to China -- conveniently obscuring the problem that their data suggests the average DEI professional with a twitter account has tweeted on Israel ... less than once. Ever. Over the lifetime of their account.

I asked them for median/modal data (I was able to identify the mean as around .8 tweets/account), and they just flat refused to provide it, but both were almost certainly either "0" or "1" anti-Israel tweet.