Meng Hu retweeted
There’s no “fight against affirmative action”. There’s “fight against affirmative action against whites” and “fight against affirmative action against Asians”. Contrary to what these “colorblind conservatives” would like you to believe, those are distinct issues in practice.
I will just leave you with what I have written on this topic:
For years, Republicans denounced it (rightfully) as discriminatory.[1] They framed it as a moral outrage against meritocracy. But when they made their case publicly, they rarely spoke about discrimination against white applicants. Instead, they foregrounded Asians as the primary victims. The charitable interpretation is that this was strategic: argue on universal principles, avoid language that sounds like racial self-interest, and once race-based preferences are dismantled, everyone benefits equally. A colorblind system emerges. Merit wins. But in the end, the outcomes didn’t follow.
After Students for Fair Admissions v. Harvard, elite admissions did not suddenly become purely race-neutral in spirit. The Asian share rose dramatically. Black and Hispanic shares remained stable. White representation declined. As @herandrews noted:
“Since the ruling was issued, the share of Asians in Harvard’s freshmen class has gone up, from 26 percent of the class of 2025 to 41 percent of the class of 2029, but the share of whites has gone down. White students were 47 percent of freshmen in the class of 2025 and approximately 31 percent of the classes of 2028 and 2029 (Harvard published percentages for every race but white for those years, and 8 percent of each class reported no race, making the white share a matter of inference). Black and Latino shares have stayed roughly the same, with small fluctuations.
So Harvard did not stop discriminating by race, it simply stopped doing so against Asians. Affirmative action continues, but now it is entirely at the expense of one race instead of two.”
compactmag.com/article/ameri…
Harvard’s own internal modeling had suggested that in a truly preference-free system, whites would’ve comprised a slight majority of admits.
nitter.gq/cremieuxrecueil/status…
Instead of the colorblind meritocracy Republicans so desired, what we got instead was selective recalibration.
There’s another uncomfortable data point: nonblack Democrats[2] show ambivalence and reluctance toward affirmative action when told it disadvantages Asian Americans. But when the same policy is framed as disadvantaging white Americans, support remains strong.
manhattan.institute/article/…
In other words, the coalition that supports race-based preferences does not oppose it as a general principle when framed as disadvantaging other groups. It adjusts its moral rhetoric based on the group affected. But worse yet, this strategy of using Asians as a political shield to achieve policy reforms is a bizarre internalization of the notion that Asians are “white-adjacent” on the part of Republicans. The reality is that Asians themselves want nothing to do with being associated with white people, considering that they see themselves as being more similar to people of color than to white Americans.
brookings.edu/articles/are-a…
Given this then, the Republican strategy of shielding behind Asian plaintiffs to restore universal colorblindness was always naïve. It assumed that the other side shared the same end goal. It assumed good faith symmetry in a system built on asymmetry. That assumption failed, and the cost was borne by the party’s own base.
------------------------------
[1] And besides that, it did not even achieve its desired outcome of uplifting underperforming groups (see this post by @MengHu13: humanvarieties.org/2023/11/1…)
[2] Unsurprisingly, black Democrats are truly dedicated fans of affirmative action. From the appendix of the report, we can obtain roughly what the results for black Democrats are. Now, a few of the people in the “All Democrats” sample who are not in the “Non-Hispanic White, Asian, and Latino Democrats” sample will not be black, but the vast majority nonetheless are, so it probably won’t change the results too much. Anyways, to do this, see the “yes” row with Asian frame, and calculate the difference between “All Democrats” versus “Non-Hispanic White, Asian, and Latino Democrats”:
No: 84 – 74 = 10
Yes: 71 – 45 = 26
Total: 155 – 119 = 36
Percentage against admissions preferences if it hurts Asians: 10/36 ≈ 27.78%
Percentage for admissions preferences if it hurts Asians: 26/36 ≈ 72.22%
Meng Hu retweeted
This study found that, in Germany, a 1% increase in immigration inflow leads to an 3.2% increase in flat prices and 1% in flat rents, as short-term effects, consistent with results from the US.
researchgate.net/profile/Umu…
Meng Hu retweeted
Several studies found that mother's educational gain after childbirth is not causally related with child's cognitive scores. In terms of effect size, it's very, very small.
ncbi.nlm.nih.gov/pmc/article…
strathprints.strath.ac.uk/85…
researchgate.net/publication…
gwern.net/doc/iq/ses/2021-ma…
Meng Hu retweeted
Probably a study many people didn't pay attention to. They found that after controlling for specific cognitive abilities, test motivation and time spent on homework do not predict SAT scores.
sciencedirect.com/science/ar…
Meng Hu retweeted
Under the Out-of-East-Asia theory, the emergence of modern humans around 50,000 years ago signifies a major leap in intelligence, which must have a corresponding genetic signature. Two key genetic indicators are described: (1) a marked reduction in the maximum genetic diversity (MGD), reflecting functional optimization, and (2) a greater genetic distance from chimpanzees, indicating accumulation of human-specific adaptive changes.
Populations that are direct descendants of those earliest modern humans (those least admixed with archaic homo) should therefore show the lowest MGD and the greatest genetic distance from chimpanzees today. This expectation is consistent with the East Asia model but entirely inexplicable under the Out-of-Africa model, which offers no coherent account of the genetic basis of intelligence progress. Indeed, if the earliest and most intelligent humans arose in Africa, it would be paradoxical that present-day Africans generally show the lowest levels of civilizational and cognitive development.
Meng Hu retweeted
The RIOT IQ Test is officially here in early access! We've developed the first professional, purely online IQ test.
First 2,000 users get 50% off with code: EARLYACCESSRIOT
Redeem at the Full RIOT test checkout: riotiq.com/assessments/Full
Meng Hu retweeted
Joseph Bronski appears to have been banned from Twitter.
Despite some of his recent controversial takes, I believe he shouldn’t be banned and contributes valuable work.
Free him.
Meng Hu retweeted
It is now known that neutrality is dead in Wikipedia. Fortunately, we have Rational Wiki, far more neutral.
Meng Hu retweeted
This study compiled ample evidence that measurement error (e.g., due to misreporting) is greater among Blacks & Hispanics, compared to Whites. (This could be due to lower IQ.) This has huge consequence for the study of race differences in life outcomes. docs.iza.org/dp17349.pdf
Meng Hu retweeted
They found a negative correlation between flaccid penile length and national IQ, adjusting for GDP/capita, BMI, education expenditure, max temperature, racial demographics. Interesting, the paper is withdrawn due to not being PC.
longdom.org/open-access/peni…
Meng Hu retweeted
Test development update:
We have finished preliminary data analysis on the subtests on the Fluid Reasoning index for the RIOT. This index consists of three subtests: Matrix Reasoning, Visual Puzzles, and Figure Weights.
The data look good so far. Here's the histogram for the Matrix Reasoning subtest:
The histogram is little bit boxy, but even with the small sample size in this item tryout, the data are already starting to take on the shape of a normal distribution. (The really short bar in the middle is an artifact of the number of bins selected for the histogram.)
Here's the histogram for the Visual Puzzles subtest:
We aren't seeing any ceiling or floor effects, which is a good sign. There is a slight tendency for the Visual Puzzles data to be a little lopsided towards the high end of the distribution. But this was the easiest of the three subtests, and this can be managed by dropping unnecessary or redundant easy items.
The Figure Weights subtest looks ideal for this stage of test development:
When we combine the data from all three subtests into one index score, we get a test that is challenging for very bright people and can still differentiate in the average and below-average ranges of ability:
We still have more analyses to do on these subtests, but they're looking good. We'll likely know within the week if we can proceed to final item selection and test norming or if we need to run some items through the wringer again.
Meng Hu retweeted
People ignore that the Flynn effect on the Raven's test is hollow (no g). Raven requires two abilities: 1) map dissimilar objects and 2) infer a greater number of rules. Recent cohorts do better at 1 but not 2. No transfer effect = no g loading.
shorturl.at/zE9N8
Meng Hu retweeted
Trends in cognitive/achievement gap since late 19th, between Blacks and Whites in Nova Scotia. The pattern is not too clear but the safest conclusion so far is that there is no convergence.
qeios.com/read/6KFZVL.2
Meng Hu retweeted
Blacks with low earnings overreport their income to a greater degree than comparable Whites; Blacks with high earnings underreport to a greater degree their income than comparable Whites. This could be an issue for B-W analyses of "self-report" income.
mh19871004.files.wordpress.c…




















