Contents
Summary
I provide a brief overview of Y Combinator’s non-profit program, comparing its work to that of GiveWell and the Open Philanthropy Project in the context of cause selection and prioritization. GiveWell and Y Combinator have significantly different approaches to selecting non-profits to fund. Whereas GiveWell seeks the most effective charities and causes, Y Combinator uses a portfolio approach of funding several projects that satisfy certain criteria (e.g. solving real problems and commitment to growth). In general, Y Combinator seems to take the approach they use for for-profits and apply it to non-profits. In particular this means they focus on non-profits that center around information technology, where they have the most experience.
Why look at Y Combinator?
My motivation for looking into Y Combinator’s non-profit program was in the context of my work on cause prioritization. I’ve been interested in different approaches to cause selection and prioritization, so looking at how a successful for-profit organization like Y Combinator approaches the problem is prima facie likely to provide insight that is distinct from that provided by looking at non-profit organizations like GiveWell. I also treated this investigation as a “test case” to see if this sort of overview of different approaches and comparison is useful.
To elaborate, there are several reasons looking into something like Y Combinator might be important for effective altruism (and especially cause prioritization) and why I decided to look into it:
-
As a general principle, within reasonable bounds, trying to gather the thinking of as diverse a collection of groups as possible seems like a good idea, in terms of having access to a larger pool of approaches with which to compare one’s own approach and expanding one’s imagination of what’s possible. At the very least, it shows that one has “done one’s homework”. Moreover, looking beyond self-professed effective altruists and their favorite go-to people and projects can help to keep effective altruism from seeming too insular.
-
Y Combinator has spent a lot of time thinking about which for-profits to fund. While this is not directly related to the non-profits that effective altruists tend to look into, Y Combinator’s for-profit work has clearly been valuable1, so we can expect to gain insight in terms of the general task of “deciding which projects to fund”.
-
As Y Combinator has now begun evaluating non-profits, it’s now possible to attempt a more direct comparison between, say, how Y Combinator chooses its non-profits versus how GiveWell does.
-
It’s possible to expand the existing discussion of for-profits versus non-profits (or more generally, just talking about non-profits in the context of for-profits or using a “for-profit mindset” of sorts).
-
There seems to be a class of people who are receptive to Y Combinator’s work (and the tech-sector/startup approach to improving the world) yet unwilling to embrace “effective altruism” (see for instance shminux’s comment in a LessWrong open thread and also possibly Effective Altruism from XYZ perspective). Figuring out how effective altruist approaches differ from that of Y Combinator can elucidate this difference in response by this class to some extent.
-
There is already interest in this topic. For instance GiveWell has been highly interested in different approaches to philanthropy. They have even begun their History of Philanthropy project, which “aims to better understand the role that philanthropy, as opposed to other factors, played in various historical episodes that some people have identified as ‘philanthropic successes.’ ”
Of course, it’s possible that looking into Y Combinator’s non-profit program, or the approaches of others, isn’t all that important. It’s also unclear whether this sort of investigation is tractable. As GiveWell has often noted, organizations don’t do a lot of explicit prioritization work,2 and even if they do, they don’t release a lot of their actual work/thought process.3 It’s also unclear whether groups hold themselves to a high enough epistemic standard, and even if they do, they may not be very good at explaining their reasoning.
Background on Y Combinator’s non-profit program
Y Combinator began funding non-profits in 2013 when it included Watsi, a crowdfunding non-profit for medical treatments, in its funding batch. Y Combinator co-founder Paul Graham wrote:
After we met the founders of Watsi, we realized they’d be the perfect [non-profit] to start with.
We owe Hacker News thanks for introducing us to Watsi. When Watsi launched, they posted about themselves on HN. They were a hit with HN users and raised a significant amount of money.
Since then, they’ve also funded Zidisha, Bayes Impact, CodeNow, CareMessage, Noora Health, One Degree, the Immunity Project, the Detroit Water Project, 80,000 Hours, DemocracyOS, and possibly more4. Note that this is still quite a small number compared to all the startups they’ve funded so far.5
Y Combinator’s non-profit program is distinct from YC Research, which is newer and targets “fundamental research” and innovation that takes a long time to develop. I won’t consider YC Research in this post.
For the rest of this post, I want to go into how Y Combinator chooses which non-profits to fund, by looking at both what it says on the topic and also its actual choices.
Y Combinator’s explicit reasoning
Compared to an organization like GiveWell, with its commitment to transparency, there is relatively little that Y Combinator has said about its reasoning in deciding which non-profits to fund.6 That isn’t to say they’ve been completely reticent though. For instance, in one of the original announcements about non-profits, Paul Graham writes:
We don’t know how many nonprofits we’ll fund, or what type of ideas we’ll like. We like Watsi a lot, because they help people who really need help, and do it in an efficient and transparent way. But fundamentally this is an experiment, just like YC itself was at first. So we don’t want to tell people too precisely what we’re looking for, because we’re not sure yet what we’re looking for.
More substantive is an interview done with Y Combinator’s Kate Courteau (who is Director of Y Combinator Nonprofits). Courteau explains how funding Watsi led to them considering more non-profits:
WATSI told us how much the YC program impacted their progress so we realized that a lot of the benefit of YC could be universally applied to all startups; for-profit and nonprofit. We also recognized that the WATSI team operated in the way that most great for-profit startups do. They were very determined and focused and moved quickly; launching early and iterating based on watching how their users interacted with their site. They had a vision for the future of WATSI. It’s evolved a bit but they continue to be a strong, mission-based company that works in the nonprofit world in a very transparent way. This experiment seemed to work so well that in January 2014, we decided to officially launch our nonprofit program.
Moreover, she says the following, which might be the closest thing to understanding how they “select causes”. The essential idea seems to be that Y Combinator is using the same framework with non-profits that they already use in determining which for-profits to fund:
YC’s motto is: “Make something people want.” In the for-profit space, we found that the best startup ideas come from founders solving problems they have encountered themselves. This works much better than thinking, “I’d like to start a company. What should I work on?” This concept also applies to nonprofits. It’s really important to work on something that you are passionate about and have some history with because startups are really hard.
The rhetoric is definitely different from what one hears from effective altruist types: instead of “proven, cost-effective, and scalable” (GiveWell) or “the most good with your donations” (Giving What We Can), it’s ‘being passionate about the work you do’, and also ‘understanding other people’s cultures’7 and ‘building a userbase’8.
In part 2 of the interview, Courteau elaborates on another part of Y Combinator’s thought process:
One of the biggest problems in the nonprofit world is the lack of early financial support and mentorship for young entrepreneurs who have innovative ideas about how to solve some of the world’s largest problems. This lack of early support is discouraging to really smart people who might have great potential to tackle these big problems. At YC we invest in people and we’re not afraid of teams with ambitious or potentially polarizing ideas. We recognize that a lot of startups fail but we believe that it’s worth the investment even if there’s only a small chance that an ambitious approach can solve a really big problem.
In other words, unlike GiveWell’s traditional work, which tries to find the “very best” charities, Y Combinator takes a portfolio investment approach, where it “doesn’t matter” whether any particular non-profit succeeds or fails, as long as on average they do a lot of good. So the Y Combinator mentality is something more like “try a bunch and see what works”, not “research a bunch and fund what works”.
Of course, given that Y Combinator’s main work focuses on for-profits, it’s possible to take whatever explicit reasoning they’ve made public regarding for-profits as a proxy for their non-profits reasoning, especially since they acknowledge this influence. I won’t elaborate on this, but some places to look at are Paul Graham’s essays like “How to Start a Startup” and “Do Things that Don’t Scale”.
Overview of actual non-profits Y Combinator has funded
In this section I want to list the non-profits Y Combinator has decided to fund so far, then try to see what might be some reasons they were chosen. At first I just want to list these and give brief descriptions of what they do because I think it’s illustrative; I’ll give my thoughts after the list. As mentioned near the beginning of the post, the ones I’ll look at are Watsi, Zidisha, Bayes Impact, CodeNow, CareMessage, Noora Health, One Degree, the Immunity Project, the Detroit Water Project, 80,000 Hours, and DemocracyOS.9
-
Watsi is a website that helps people crowdfund medical treatments for people.
-
Zidisha is a “peer-to-peer microlending service” that “allows people to lend small amounts of money directly to entrepreneurs in developing countries”.10
-
Bayes Impact is a bit vague; their website describes themselves as “a group of practical idealists who believe that applied properly, data can be used to solve the world’s biggest problems”.11 Y Combinator explains Bayes Impact in its announcement as follows:
Bayes Impact is a nonprofit that deploys teams of data scientists to create data-driven solutions for challenging social problems. The organization runs full-time fellowship programs that bring together domain experts and data scientists from top technology companies and academic institutions.
-
CodeNow runs workshops designed to teach people programming. TechCrunch explains:
CodeNow aims to teach programming basics to high schoolers, particularly girls, ethnic minorities, and other underrepresented groups. It launched in Washington, D.C. in 2011 before expanding to New York City and San Francisco last year.
-
CareMessage; TechCrunch on CareMessage (March 2014):
Considering that 70 percent of older patients have difficulty using and understanding printed materials, CareMessage thinks this is where technology can really help, and is likely needed most. The Y Combinator-backed non-profit startup is launching today to provide clinics and healthcare organizations with software and interactive mobile programs that help them stay in touch with (and engage) at risk patients.
It’s worth noting that “where technology […] is likely needed most” is a bold claim.
-
Noora Health; TechCrunch article on Noora Health (March 2014):
The nonprofit graduating from Y Combinator is building hospital education platforms for patients and their family members to teach them the skills needed to recover from a major medical event (like a surgery) or manage chronic conditions like diabetes and palliative care. Using an iPad app, Noora Health works with hospitals to offer patients and their families a combination of videos, quizzes and interactive content to teach skills to aid in their recovery at home.
-
One Degree is a website that helps people find community resources. The TechCrunch article on One Degree says:
One Degree, a new non-profit that helps people find social services like affordable housing and job training.
[…]
“We are here to revolutionize the way that people access social services,” Faustino said. “We believe that we can do this because the non-profit sector has been stuck in the Internet dark ages. People still use binders and still rely on information that’s stuck in their heads. That means people aren’t getting the resources they need quickly and easily.”
-
The Immunity Project (Wikipedia) aims to develop an HIV/AIDS vaccine.12
-
The Detroit Water Project helps people pay their water bill. A TechCrunch article on the Detroit Water Project explains:
The organization behind the Detroit Water Project attempts to throw a life-preserver to those drowning in unpaid water bills by connecting donors to those in need. This allows homeowners to, in a sense, crowdsource their water bill. Donors can either pay the entire unpaid balance, which can be several thousand dollars, or just part of a bill. Since its launch, the company has expanded to Baltimore with the Baltimore Water Project.
-
80,000 Hours is a website and service that gives people career advice so they can maximize their social impact. There is a TechCrunch article (“Want To Make An Impact With Your Work? Try Some Advice From 80,000 Hours”), and Benjamin Todd of 80,000 Hours has also announced their involvement with Y Combinator in “What’s it like being a non-profit in Y Combinator?”. 80,000 Hours is also an explicitly effective altruist organization. Since its involvement with Y Combinator, 80,000 Hours has focused more on growth; see for instance “Take the growth approach to evaluating startup non-profits, not the marginal approach”.
-
DemocracyOS is a platform that allows people to discuss policy ideas.13 The same people are involved in the Argentine Net Party.
It should be noted that the vast majority of the non-profits in the list have to do with information technology. The Immunity Project might be an exception in this regard (it still involves technology, but does not aim to provide an app or website). This is probably the case because Y Combinator doesn’t just want to give money to the startups, but rather also wants to provide assistance, connections, etc. Y Combinator has the most experience with information technology, so this makes sense. And because they have the most experience here, they also have more experience reviewing ideas here, which helps them pick good ideas.
Some of the startups (like CodeNow) may come across to effective altruists as being particularly “ineffective”. Indeed, their results section seems to consist of a single graphic and testimonials from students. Here it might be worth noting that Y Combinator has a particular audience they might cater to (consisting of programmers and other tech people).
Comparison to GiveWell and the Open Philanthropy Project
Besides the rhetorical differences noted above, there are several other important differences between Y Combinator’s non-profit program and GiveWell/Open Phil.
-
Unlike GiveWell’s traditional work, which funds existing charities that have produced significant results, Y Combinator’s non-profit program (with a few exceptions) targets non-profits that are just starting.
-
Y Combinator also hasn’t formulated their selection process explicitly as cause selection. Rather, they seem to be paying attention to a lower level of selection (compared to the high (cause) level selection) in a particular area (namely those interventions involving information technology). In this sense, Y Combinator’s approach might be similar to investigating all interventions that involve leafleting (whatever the cause).
-
So far, Y Combinator’s non-profit program has been similar to the Open Philanthropy Project and distinct from GiveWell in that rather than recommending non-profits for individual donors, they provide the funding themselves. Indeed, for its part, the Open Philanthropy Project has explicitly stated that their “decisions about whether or not to make a grant to an organization should probably not be taken as a signal to donors about whether or not they should donate”14. Y Combinator is not as explicit as the Open Philanthropy Project in this regard, and from what I have seen, they do not explicitly suggest people to donate to the projects they fund, either (they just announce their own funding).
Document source and versions
The source files used to compile this document are available in a GitHub Gist. The Git repository of the Gist contains all versions of this document since its first publication.
Acknowledgements
Thanks to Vipul Naik, Jim Terry, and Benjamin Todd for their input while I was writing this post. All imperfections are my own.
The writing of this post was sponsored by Vipul Naik.
License
This post is released to the public domain.
-
Y Combinator (and Paul Graham in particular) has arguably had more impact than any explicit “effective altruist” group (or person). See for instance Vipul Naik in “Paul Graham on US immigration policy and high-tech programmers”:
I’m a great fan of Paul Graham, essayist, entrepreneur, and co-founder of startup accelerator Y Combinator (along with his wife Jessica Livingston, whom I also admire greatly). Through Y Combinator, Graham has changed the startup and tech company landscape and profoundly affected the world. (Some Y Combinator-funded companies you’ve probably heard of are Reddit, Airbnb, Dropbox, Scribd, Disqus, and Stripe). Graham also started Hacker News, a Reddit-of-sorts for the programmer/startup crowd. In the world of letters, Graham is better known for his long-form essays that include incisive social commentary. If you haven’t yet read his pieces, I encourage you to check them all out (I particularly like this one, that might be somewhat relevant here). He’s done more for the world than most people, including me, could dream of. And he knows a lot more about how the world works than I do.
↩
-
For example in Strategic Cause Selection:
[Cause selection] is where existing philanthropists seem to be least thoughtful and to ask the fewest critical questions; yet this is where we’d guess the bulk of variation in “how much good a philanthropist accomplishes” comes from.
↩
-
See for instance Paul Christiano:
Unfortunately, [cause prioritization] seems to be done primarily in private; to the extent there is high-quality thinking on these questions, it is very rarely shared, and public discussions rarely move forward our overall state of knowledge or reveal important new arguments.
↩
-
Y Combinator doesn’t seem to keep a comprehensive list of startups they’ve funded, and when announcing startups they often don’t mention whether something is a for-profit or a non-profit, which makes the task of compiling a list of YC-funded non-profits difficult. See Y Combinator Companies for maybe more that I didn’t manage to find.↩
-
There’s another list that contains fewer companies.↩
-
Y Combinator does seem to value transparency though; see for instance Kate Courteau:
Our hopes are to make some headway on some of the biggest problems that still exist in the world. […] I get frustrated that there are still so many people working on the same problems and not sharing their findings. I realize that some of that has to do with competition for scant philanthropic resources, but if there was a way to overcome this and get people to share, I think we could work a lot faster and have greater impact.
↩
-
Further on in the interview, Courteau explains:
Sometimes in the nonprofit world, people do a lot of planning in isolation because their users might geographically be far away. I learned this lesson early when I worked for a nonprofit, teaching a new building technique in Sudan. We were very surprised that our concept and methods were not universally well received in Africa. Our theory had been developed in the US and we had not spent much time in this corner of the world before we rolled it out. This was the root of our problem — we planned too much without deeply engaging with the people who were to use our product.
↩
-
Courteau explains:
Build a community that really values your product. This initial core user base will be the foundation of your company and a spring point for growth.
↩
-
There is in addition Giveffect, which is a for-profit that helps non-profits. I discuss Giveffect in this footnote, since it’s not actually a non-profit.
The main source of information that discusses the startup is a TechCrunch article, “Y Combinator’s Giveffect Has Built A Shopify-Meets-Salesforce For Non-Profits”:
Giveffect, part of [Y Combinator’s] current cohort, has built a suite of cloud-based software that focuses specifically on the needs of non-profit businesses, covering services like accounting and CRM (including donor tracking), through to fundraising and crowdfunding platforms, all built from the ground up with its target customers in mind.
It’s also noteworthy that Giveffect, unlike many others that pass through YC, is not actually a very young startup. It was founded in 2012 and already has more than 300 customers.
[…]
So why go to YC? Mirza admits that the trio was hesitant at first about applying. “We did wonder if it was suitable given that so many startups here are at a much earlier stage,” she said. “But based on the traction that we’d seen other YC startups get, and the success stories we heard, we decided that we could build a good company without YC, but we could build a great enterprise with it.”
The Giveffect website lists more “traditional” charities. Giveffect, while not a new startup (as noted above), still fits the pattern of the non-profits discussed in this post by being software-oriented.↩
-
TechCrunch has an article on Zidisha that compares it to other micro-finance programs.
There has also been some discussion of Zidisha within the effective altruism community.↩
-
There is also a TechCrunch article on the non-profit, with the following quote:
“I was volunteering at homeless shelters for a little while,” says Duan. “I was shocked by the difference between this manual labor, this one-to-one exchange, and the idea that improving the accuracy of a [programming] model by 1% can affect millions of users and bring in hundreds of millions of dollars in revenue.”
The ah-ha moment was determining that the same science can be leveraged for social good. “I had been working with the City of New York implementing civic solutions and recognized that the chasm between the technology that industry and tech companies are using and what’s available to civic organizations is huge,” says Jiang.
↩
-
There is a Wall Street Journal article on the Immunity Project:
Immunity Project will seek to succeed where other efforts have failed, in creating a vaccine that can prevent healthy people from contracting HIV, the company’s director of strategy, Ian Cinnamon, said.
Somewhat surprising (emphasis mine):
The program has not only given Immunity Project $20,000 in non-dilutive funding, but Y Combinator Partner Sam Altman, who is also the founder of mobile-application company Loopt Inc., said in a statement that he would personally donate blood as the company moves its technology further through clinical trials.
↩
-
See “A Conversation With DemocracyOS, The YC Non-Profit That Built A Latin American Political Party”, “Why Y Combinator Funded a Radical Political Party in Argentina”, and “Out in the Open: An Open Source Website That Gives Voters a Platform to Influence Politicians”.↩
-
Quote from a Facebook post by Claire Zabel. See also Holden Karnofsky’s “Why the Open Philanthropy Project isn’t currently funding organizations focused on promoting effective altruism”.↩
I think many of the differences between how traditional GiveWell makes funding decisions and how YCNP makes them is that GiveWell is aiming to fund larger, more established organisations, while YC is aiming at startups. If GiveWell switched to funding startups, then they'd use similar criteria to YC, except (i) put more emphasis on funding evidence-backed interventions (ii) not target organisations that use technology (which is just a pragmatic decision by YC given their expertise in and enthusiasm for that area).
Outside of international development (i.e. in Open Phil), then difference (i) would likely be reduced too, since in most causes it's much harder to fund evidence-backed interventions that in international development. Though, Open Phil is still putting significant emphasis on cause selection, which YCNP puts much less emphasis on. (Though YC recently started doing something a bit like cause selection: https://www.ycombinator.com/rfs/).
In short, if you take Open Phil + focus on startups - cause selection + tech focus, then you get pretty similar criteria to YCNP.
This is non-obvious to me; GiveWell seems to recommend startups as well as more established organizations in my opinion. Even the more established ones are quite new in relative terms. One of GiveWell's newer issues seems to be that these startups aren't experienced at scaling and thus (in GiveWell's opinion) don't have enough room for more funding.
AMF is a two-man show operating in a limited area, there's three full-timers on Deworm the World and they operate in just one geography, DMI is quite new and in one geography, etc.
With regards to YC, I'd say '(i) put more emphasis on funding evidence-backed interventions' is a very significant difference. I'd also say another major one is considerations of cost-effectiveness, which don't seem to factor in for YC nearly to the extent of GW.
It's a good point the existing GiveWell nonprofits are very small in terms of staff. They don't seem like startups at this point though because they've already had millions of dollars of funding (well over $10m for the top three) - I think revenue is more relevant than staff number - and they already have a clearly established model. Being focused on an already proven intervention is also a significant difference to most startups.
What I meant is that if the Open Phil team started funding brand new projects doing new interventions, then I guess they'd use pretty similar criteria; except Open Phil would put more emphasis on cause selection; and YC would put more emphasis on tech.
I'm lumping these together.
It is a significant difference, I agree. Though like I say, it's most relevant in international development, which Open Phil is expanding out of.
Also I should be clear that two groups could have similar criteria but apply them in very different ways e.g. GiveWell seem more sceptical, so would need more evidence that a program works.