#49 - Prestige and YCombinator
How do we think about YC's increasing batch size?
A few years ago, Yale University finished building two new “residential colleges,” increasing the school’s total undergrad capacity from 5,400 students to 6,200. I went to Yale, I think it’s a fine school, and I’m happy that more students can take advantage of a Yale education. But I remember talking with my friend about the expansion, and he was actually mildly peeved about it! The reasoning was something along the lines of, “If more students are admitted to Yale, then that dilutes the value of my degree.”
Hm, well, to be fair, at some point this has to be true. If Yale increased its class size by 10,000%, then a Yale degree would intuitively begin to mean less than it does now. The reason is obvious: The Yale degree carries some level of prestige, but college admissions prestige is a zero-sum game. One student is admitted at the expense of another. To get in, you have to beat someone out. So if you open the floodgates and let in significantly more applicants, then it’s less meaningful of an accomplishment to get in.
But hold on—what if it’s not a zero-sum game? What if we can grow the prestige pie by admitting more students? Consider two worlds that are equal in every respect, except in World A, Yale admits 2,000 students per year, and in World B, Yale admits 4,000 students per year. Everything else about Yale is exactly the same, in particular (1) the student experience and (2) the quality of admitted students. Yes, in world B, Yale admits more students, so it’s less “prestigious” to get in, but World B Yale produces more graduates (on an absolute basis) that could presumably really, really, really change the world. Doesn’t the wild success of alumni increase the prestige of the school, which might offset any loss in prestige from the increased number of admits? In other words, does prestige have network effects?
World B is completely hypothetical, but I honestly don’t think the assumptions are too wild. For example, I feel like Yale could more than afford to double its undergraduate student population while preserving the student experience by constructing more residential colleges, hiring more professors, building new classrooms, etc. with its $32bn endowment. I also feel like Yale could afford to increase its undergraduate student population while preserving high intellectual standards for its students. How different is the applicant ranked #2,000 from the applicant ranked #2,001? #3,000? #4,000? I don’t know—these elite institutions are always saying that they have way more qualified applicants than they have room for, and sure, maybe it’s a lie to preserve applicant ego, but I’m inclined to take their word for it.
Look, we’ve talked about prestige before. In order to win, someone else must lose. The desire for prestige is an outgrowth of a scarcity mindset.
Prestige is usually associated with some form of picking: You’ve been picked to win an award, to get a good grade, to get into a college, etc. But I think of picking as consisting of two specific types: (1) Picking those who meet a threshold and (2) Picking those to meet a fixed quota. The first type is benign. If you’re good enough, you’re picked. If you score 90% on a multiple choice exam, you get an A. If you pass the bar, you can practice law. There is no cap on the number of As or the number of lawyers. The second type, however, is nefarious. This is essentially how college admissions work. Even if you’re good enough, you’re still not good enough because others are better. There are a fixed number of winners and an uncapped number of losers.
Some industries are so monastic and prestigious that their entire structure is built around quota-picking. Let’s look at the legal profession, which I am admittedly somewhat a part of. I mentioned that there is no bar on the total number of lawyers in the U.S. (and this is true!), but there is certainly a de facto bar on the total number of big corporate lawyers (whose jobs are very prestigious). According to some of my classmates, the top law firms have a quota for how many Stanford Law School students they’re allowed to hire. Also, district court judges have a fixed number of clerks, appellate court judges have a fixed number of clerks, and Supreme Court Justices have a fixed number of clerks, and district court clerkships are less prestigious than appellate court clerkships, which are less prestigious that Supreme Court clerkships, so it’s all just a large rat-race to get into the highest-prestige quota possible.
So if you’re a typical law school in the United States trying to fit this rat-race mold, you’re naturally going to grade students on a strict curve. You place a quota on the number of students that get a good grade in a class! You want a nice gradient of grades so you can say that Alice is definitely better than Bob, but Bob is wayyyy better than Chris, so Alice, Bob, and Chris can succeed in the stratified legal profession according to their “skills.” It’s quotas all the way down!
Well, actually, there is an interesting exception. Harvard and Yale apparently don’t use quota-picking for grades, they use threshold-picking! In other words, any number of students can get a good grade in a course. My take on this is that Harvard and Yale are saying that their students are so damn good that future law firms, judges, government employers, etc. don’t need to see where their students land in the internal rat race for grades. If a student has great grades, then he meets our standards for greatness, so shouldn’t he meet yours as well, regardless of what grades other students have? If an institution is just that good, then perhaps it can afford to relax a bit on the quota-picking.
At this point I’ve buried the lede a bit, but let’s move on to talking about YCombinator (“YC” for short). For those unfamiliar, here’s how a startup accelerator like YC works:
An early-stage startup needs help growing its business, so it applies to a startup accelerator.
A startup accelerator accepts only a handful of companies to participate in a so-called “batch.” The accelerator hands each accepted startup a small amount of capital (in return for a small amount of equity), gives it access to a network of seasoned mentors and other similarly situated startups, and provides it with some amount of prestige for being associated with the accelerator.
After the accelerator program for the current batch is over, the accelerator hosts a “demo day,” where the current batch’s startups present to venture capital investors.
If the startup company does well and the startup accelerator is able to increase the value of its equity in the startup, then the accelerator will have an easier time attracting new startups to apply, and the cycle starts over again.
In my opinion, the most interesting step is (4): For startup accelerators, success begets even more success. If a startup accelerator does well, it’ll have more money to give to startups, it’ll have a wider network to help startups, it’ll increase in prestige, and as a result, it’ll have more startups banging on it’s door to participate in a batch.
But step (4) needs to be balanced against step (2). Just because you’ve accumulated tons of resources to invest in early-stage startups doesn’t mean you can pick startups like throwing darts. The picking function is important, and you still need to be selective in how you pick companies! If you start picking bad companies, then your returns decrease, and the flywheel in (4) begins to slow down.
But what if you start picking more companies? Here’s a chart of the number of YC companies over time:
Does relaxing the quota dilute the prestige of the brand? A few people on Hacker News seem to believe that it does. This is an important question to answer! If the prestige of YC goes down, then theoretically you might have two bad results: (1) Fewer startups want to apply to YC; (2) Fewer venture capitalists, who attend YC’s demo day to invest in companies in the batch, will be excited to invest.
How do we think about prestige and batch size?
One place to start is by intuiting, as we did for Yale undergrad admissions, whether a “World B YC,” which accepts more startups per batch, can (1) provide the same level of support for each startup and also (2) maintain the quality of startups accepted to each batch, such that the success of the larger batches outweighs any decrease in prestige from the larger batches. First, yes. We already went over this one above. Because YC is so successful, it’s cultivated an ever-wider network and more resources to devote to each batch. In fact, it’s quite possible that the larger the batch, the better the experience for each startup (again, network effects?). Second, sure. I don’t think the bar to get into YC is necessarily going down just because more startups are accepted. The acceptance rate into YC is still at ~2%—more startups are being formed, and more startups are applying, and if we assume some constant proportion of “good” to “bad” startups being formed, it’s only natural that YC can afford to increase its batch size while maintaining its threshold for “good.”
A 2016 empirical study of YC companies conducted by Andrew Barnes also seems to confirm these points. He found that batch size had a negligible impact on amount of subsequent funding obtained by each startup in the batch and a negligible impact on the time to successful exit (via acquisition). That is, even as YC has grown larger, the startups in the bigger batches had just as good of outcomes as startups in the smaller batches. Interestingly, Barnes hypothesizes that his findings are driven by the fact that YC is already renowned globally as the premier accelerator, so YC can afford to increase its batch size without compromising quality. As an aside, is this to Harvard and Yale’s threshold-picking grading scheme?
Another thing to consider is whether the startup / entrepreneurship market as a whole is monastic like the legal profession is. In other words, whether the creation of startups is a fixed-pie rat-race based on scarcity. The answer to this is, of course, no. Sure, fine, the dream for any startup is to dominate a market, but this doesn’t act as a complete bar for new market entrants as a quota would. For instance, there are like a bajillion enterprise software applications, many of which are occupying similar or adjacent markets. Venture capital funding is at all-time highs, and it’s easier than ever to become an entrepreneur. Silicon Valley has historically adopted a bias towards thinking that new value can always be created. Welcoming more entrepreneurs = good, Excluding entrepreneurs = bad.
Oh, and also, if YC has gotten less prestigious, then why hasn’t venture capital gotten less prestigious as well? You don’t need to squint too hard to see that venture capital investing is kinda’ similar to startup accelerators:
Venture capitalists raise a pool of money from what are called “limited partners.”
Venture capitalists deploy chunks of that pool of money into startups they think will earn them a huge return.
After the investment, the venture capitalists help to support that company along its lifecycle. More often than not, the company will fizzle out, but every so often, a company is a hit, generating the vast majority of returns for the pool of money that the VC investors had originally raised. A percentage of the proceeds from the IPO / acquisition gets returned to the limited partners.
If the venture capitalists did well with their pool of money, they’ll have an easier time raising their next pool of money, and the cycle starts over again.
A VC firm has a nice flywheel in step (4), just as a startup accelerator does: If a venture capital firm does well, it’ll have more startup banging on its doors for money, it’ll have more limited partners banging on its doors to give money, and it’ll have more talented investors banging on its doors for a job. Prestige is also very important in VC, just as prestige is important for startup accelerators. Entrepreneurs that are able to say “a16z led our round” will have an easier time attracting customers and subsequent capital.
Top VC firms have raised more and more capital to deploy, and their portfolios are getting larger and larger, yet I don’t hear as many people saying that a16z has lost its prestige as much as I hear people saying that YC has.
So it seems, at least to me, that the prestige of YC shouldn’t go down due to increasing batch size, but sure, I think reasonable minds can disagree here. For people who do think that YC prestige has dropped, though, Dan Gackle of YC has a great rebuttal:
YC has always sought to filter out the resume-padding sort of founder who sees YC as a step up the status ladder. Such founders are not up for the gruelling slog of really building a startup, and tend to bail not long after the batch since they've already gotten what they wanted out of it.
People should apply to YC because they believe it would [help] them them build a successful startup, not because they want prestige. If expanding YC leads to a decrease in prestige (alongside an increase in real value to startups), that's a win for everybody: more business-builders get access to YC, YC gets to fund more successful companies, and status-seekers can find something else that will look better on their resume.
In this sense YC is very different from the elite colleges whose brand depends on their exclusiveness. The upper bound on YC's expansion is how much startup opportunity exists in the world.
📚 What I’m reading
Ezra Klein interview with Sam Altman. (NYTimes)
Curtailing GPT-3 toxicity. (WIRED)
G7 leaders grapple over China rebuke. (Politico)
On synthetic data. (MIT Tech Review)