Why Ivies Should Exist
In an ideal world, students would take a final exam in addition to an IQ test at the age of 16 and their scores would be posted online. Given that undergrad primarily serves to signal intelligence and conscientiousness, this would decrease the need for undergrad and hopefully individuals could comfortably enter high-paying jobs in their late teens or early twenties after taking on the necessary training and apprentice work. Unfortunately, until supreme leader Bryan Caplan takes total control of the United States, this does not seem like a plausible reality. For now, I will discuss educational policy within the constraints of the university system.
Elite universities offer a loophole to individuals who want to signal intelligence in addition to a university degree. Given that the average SAT/ACT score at these colleges is about 1500, the average student attending will have an average IQ of about 130. Actual IQ tests given to Harvard students suggest that the average IQ there is about ~125, close enough to the conversion from my table. According to signalling theory, if you have a wider range of potential signals, then the market is more efficient. However, if one of these signals is overvalued, then the effect of that signal’s value on the efficiency of the market decreases. If this overvaluation is large enough, then it could actually make the market less efficient.
Dynomight has recently published a post against the existence of elite universities, arguing that the causal effect of attending on success is too large to justify their existence, and they reduce meritocracy by impeding the natural flow of talent. He also raises the following problems with the existence of Ivy Leagues:
Gating mechanisms are bad
Gating mechanisms should come as late as possible
Having gating mechanisms be the college admissions process is bad
To these points, I propose the following counterarguments:
1. Gating mechanisms are great
In society, there are a limited number of opporitunities. The thiel fellowship is handed out to only 20-25 people per year, there are only 21,766 Harvard graduate students, and so on. The providers of these positions, for the purposes of self-interest, must screen out the least intelligent and capable individuals. Fortunately, this self-interest also complies with society’s interest in having the most capable individuals for a position be in that position. The author finds this to be dystopic, but this doesn’t matter, all that matters is if it works.
The author’s alternative to gating is to have job stratification be fluid and to be determined informally by the job market. In his words, “Your ability to become a famous journalist should be determined only by how good you are at journalism”. While an intuitively plausible idea, the problem with this is that employers are not that good at screening for talent or recognizing it. The average job tryout procedure correlates at .44 with job performance, and the interrater reliability of supervisor ratings is only .52. Because of this, metrics like test scores and Ivy attendance that correlate with performance will be incremental predictors of job performance beyond these job tryouts or supervisor ratings.
This is not intuitive, so allow me to use the classic ‘ice cream cones sold’ and ‘swimming deaths’ analogy. The two measures do correlate, but it’s because heat causes people to swim and buy ice cream cones. If you control for subjective heat, there is still a correlation between ice cream cones and swimming deaths per day. However, if you control for recorded temperature, the relationship disappears. This is because subjective heat and recorded temperature positively correlate, but don’t perfectly correlate because of factors such as humidity and sunlight.
This example is also analogous to the subject of the article:
Deserved status: swimming pool deaths
Talent: Ivy League admission
Subjective heat: subjective assessments of talent
Recorded temperature: true talent
Given that true talent and subjective assessments of talent do not perfectly correlate, Ivy League admission and deserved status will still correlate in a highly meritocratic society independent of talent assessments.
Dynomight also notes that the wealthy can hijack the admissions process of elite universities in favor of them. This is good, as long as the the predictive validity of the university rank does not drop, for 3 reasons:
SAT/grades are not perfect predictors of competence, and the competence of parents (based on legacy admissions or wealth) will be incrementally valid over these standard predictors. This is validated by the fact that John Hopkins legacy students get better grades in college, despite being slightly weaker applicants across the board.
Grouping the most wealthy and most talented individuals in the same social circles should theoretically enhance the amount of entrepreneurship that happens in these universities.
Universities that are dependent on donations would be wise to recruit more wealthy students, as many donations to universities come from students who attended that institution.
2. Gating mechanisms should not come as late as possible.
Age affects the value and efficienct of gating mechanisms in two ways. The first is that predictors sampled at later ages tend to be more predictive of the future because of the existence of developmental curves. However, beyond the age of 15 or so, age at testing becomes increasingly irrelevant, as the humans have already undergone most of their biological development. This can be observed from the fact that child IQ becomes increasingly predictive of IQ at 17 as age at testing rises, and that the heritability of psychological traits increases with age.
The second effect of delaying gating/signaling mechanisms is a reduction in the rate at which individuals acquire status throughout their lives. Because education delays fertility (particularly in women), it would be socially beneficial for these gating mechanisms to be employed when humans are in their late teens.
3. Causal effect of Ivy attendance on status is overstated
The author also finds evidence that elite universities causally impact success by comparing accepted and rejected students who were put on the waitlists. In comparison to the rejected students, the accepted students were more successful than their rejected peers:
One of my followers noted that a lot of these outcome measurements used in the paper he cited are circular, as they are define elite firms and graduate schools based on where Ivy+ league students are going:
Elite and Prestigious Employers. We construct measures of “elite” and “prestigious” employers that expand upon conventional lists of high-status jobs based on the revealed preferences of Ivy-Plus graduates. In particular, we define elite firms as those that disproportionately employ students from Ivy-Plus colleges. We first calculate the share of all Ivy-Plus attendees in the 1979 to 1996 birth cohorts that work at each firm when they are age 25. We then calculate the same share for the highly selective public colleges, and compute a ratio of those shares, restricting the sample to firms that employ at least 25 college attendees from the 1979-96 birth cohorts and excluding each individual’s own college from the ratio. We rank firms using this metric and define a firm as “elite” by pulling firms from the top of the list until we have accounted for 25% of Ivy-Plus attendee employment (see Appendix D for further details).
Graduate Schools. We use data from 1098-T forms to measure graduate school attendance at various ages. The 1098-T forms include a flag for graduate school attendance, but they do not include information on the type of graduate school attended (e.g. medicine, law, business, etc). We define “elite” graduate schools as Ivy-Plus institutions, as well as UC-Berkeley, UCLA, UCSF, University of Michigan, and University of Virginia, all of which have multiple programs that are consistently ranked in the top 10 or 15 by U.S. News and World Report.
There are ways firms with lots of Ivy+ graduates could be different from regular high status firm - Ivy+ graduates are probably more likely to attend firms that are close to the universities. Equally high status firms may put different weights on attendance, making them seem more elite than they actually are.
The author also argues that the journalists ignored that using ‘income rank’ is uninformative, as income is highly skewed by high earners:
First, note—as most journalists did not—that “mean earnings rank” contains the word rank. Your rank is your position on a list of people sorted by incomes. This—unlike mean income—is insensitive to high earners. Roughly speaking, what this says is that for the median person, the causal effect of Ivy+ admissions on income is small.
Fortunately, the paper he cited does actually test whether Ivy+ attendance disproportionately affects high earners. The effect of attendance on outlier income is small - admitted waitlisted students are only about 2.5% more likely in comparison to rejects to join the top 1% in income.
There are, however, positions of power where the proportion of people who were Ivy+ atendees is extraordinarily high, beyond what would be expected from a confound effect. Individuals who attend Ivy+ schools are overrepresented by a factor of 31 in the senate, 33 in elite jouralism, 52 in the presidency, and 89 in the supreme court. However, this represents a rather small proportion of the status pie in the United States - Ivy+ students are over-represented by only a factor of 12.5 in more relevant positions of power (CEO, high earner).
