Epistemic Status: This is a subject I've been casually thinking about for a while, but I wrote this document fairly quickly. Take this with a big grain of salt. This is written in a personal capacity.
A lot of EA, especially in meta and longtermism, is made up of small organizations and independent researchers. This provides some benefits, but I think the downsides are substantial and often unappreciated.
More clearly:
- EA funding mostly comes from a very few funders, but it goes to a mass of small organizations. My impression is that this is an unusual combination.
- I think that there are a lot of important downsides to having things split up into a bunch of small nonprofits.
- I'm suspicious of many of the reasons for having small organizations that I've come across. There might well still be good reasons I haven't heard or that haven't been suggested.
- I suggest some potential changes we could make to try to get some of the best incremental tradeoffs.
Downsides
Low Management Flexibility
If you want to quickly create a new project in a sizeable organization, you can pull people from existing teams. This requires upper management but is normal for said management. On the other hand, if you instead have a bunch of tiny independent organizations, your options are much more limited. Managers of tiny organizations can be near-impossible to move around because many of them own key funding relationships. Pulling together employees from different organizations is a pain, as no one has the authority to directly do this. The best you can do is slowly encourage people to join said new project.
Moving people around is crucial for startups and tech firms. The first version Amazon Prime was made in under two months, in large part because Jeff Bezos was able to rapidly deploy the right people to it. At other tech companies, some amounts of regularly rotating team members is considered healthy. Strong software engineers get to work on many projects and people.
Small nonprofit teams with locked-in mission statements are the opposite of this. This rigidness could be good for donors with little trust, but it comes at a substantial cost of flexibility.
I’ve seen several projects in EA come up that could use rapid labor. Funding rounds seem particularly labor-intensive. It often seems to me like it should be possible to pull trusted people from existing organizations for a few weeks or months, but doing so is awkward because they’re formally part of separate organizations with specific mission statements and funding agreements.
A major thing that managers at sizeable organizations do is size up requests for labor changes. The really good requests (with good managers) get quickly moved forward, and the bad ones are shot down. This is hard to do without clear, able, and available authorities.
Low Employee Flexibility
Employees that join small organizations with narrow missions can be assured that they will work on those missions. But if they ever want to try working with a different team or project, even just for a few months, the only option is often that they formally quit and formally apply for new roles. This is often a massive pain that could take 2-6+ months. The employee might be nervous about resentment or retaliation from other team members.
My org, QURI, is doing fairly niche work and has a small team (2 people now!). I’d be excited to have some team members try rotating on and off for a few months here and there. It’s very valuable for us to have people with certain long-term skills “available” for long periods of time, but they don’t need to be active for long portions. I think this project makes sense to keep small, but I also feel awkward about asking others to join a very small organization for a lengthy time.
There could be many reasons for an employee to want to work with a different team, at least for a while.
- They might think their career/skills would progress faster somewhere else.
- Their team might require a physical location where they don’t want to be.
- They might not along well with other team members, or maybe there’s a toxic work environment.
When I look back on my career, I learned a whole lot at larger organizations that I would never have otherwise. I was a decent independent hacker for a while, but it wasn’t until I worked on a sizeable engineering team that I felt like I really leveled up.
Now I’m working on a small project where I feel good about the work, but I also feel very isolated. I expect my professional development to be slower because of it. In my case, I think this is workable, but I’d have much more trouble recommending it to junior employees.
I think we can mimic some of these benefits with things like shared coworking places and conferences, but it’s difficult. In practice, right now, I don’t think we’re doing a great job here.
Quality Control
Good organizations often need solid infrastructure for HR, PR, legal, management, IT, security, and more. These are difficult to get right. It’s very hard to do a consistently amazing job in these areas in a large organization, but it’s possible to mostly ensure that the really awful situations are rare.
It would be great if we could promise people, “If you work in an EA organization, you can expect a decent level of competence in key organizational aspects.”
Right now, I’m not sure if we can.
On the management level, making a new EA project often means making a new org, and then figuring out many things you’re not great at or interested in. I spent months with QURI figuring out things like which accountants to hire and how to set up our internal payment processes.
Sponsorship programs help. In the last year QURI joined RP as a sponsee, which I feel has been a solid win. But sponsorship only provides a subset of the full organizaitonal benefit, and RP has a limited capacity to accept new sponsees.
Boards, not Managers
Nonprofits are controlled by boards. Boards have a lot of problems, as Holden has written here. I’ve written about this briefly here.
In sizeable organizations, the vast majority of management is done by professional managers, not boards. But when you have a large collection of tiny nonprofits, you’re stuck with a large collection of boards.
Management is typically a full-time paid position. Managers are accountable to superiors. If it’s a decent organization, they get training and oversight. There’s often a management career track. Management hierarchies are portrayed in clear diagrams and are iterated on to be pragmatic and straightforward.
Board membership is typically a minor side-project. In the UK, I believe many board members can’t be paid. Even in the US, pay for board duties is rare, and few organizations seem to budget for significant board spending. I’m barely sure what board membership accountability or recognition would look like. Board decisions and activity are often kept secretive, and there’s no clear “manager” or “meta-board” of the board. Boards are typically chosen by team members, which is a very obvious bias.
Painful or Partial Downsizing
When a sizeable organization decides it needs to end a project, it might be painful to the people on the project, but it can be swiftly done. Many team members are likely worth keeping, and can be reallocated.
If the organization decides it needs to do layoffs, the standard advice is to do them swiftly and then be done with it.
Meanwhile, consider the case of a small nonprofit tied a specific mission. If this mission is no longer great, a good outcome is that the nonprofit realizes this and closes up operations, laying off its team in the process. A more common outcome is that this action seems so severe that they get really good at rationalizing.
If funders need to significantly cut spending, this will take months or years to carry out. I think what happens is something like:
- The usual funding rounds begin.
- After 2-6 months of applications and evaluations, responses are given out.
- Many teams that don’t get their funding needs met then think there’s a good chance another funder will be interested, so go through one or two more funding rounds before cutting staff. There’s a lot of stress in the meantime.
- Sometimes, the org lives in some “zombie” state, where many employees are fired or lose confidence, but there’s still a trickle of funding large enough to keep a shell of the organization around.
The latter process is long, tedious, painful, and nerve-racking.
Funders often don’t want to take responsibility for closing down a project or laying off nonprofit staff. I haven’t heard much “you should shut down your project.” I have heard, “this project isn’t a fit for us, you should try other funders.” This is polite, but I think often the result is that a short and hard shutdown is replaced with a gradual, drawn-out shutdown. The latter is seems much more painful and a huge waste of time.
Specific Counter-Arguments I Disagree With
“Aren’t startups far more efficient than large companies?”
I believe that large tech companies are, on average, more efficient at converting talent into market cap value than small companies or startups are. They typically offer higher salaries, for one.
In terms of “market cap growth”, my impression is that big companies feature a lot more growth small ones. In the period of 2005 to now, all YC companies combined grew to a net worth of $600B. Meanwhile, Apple Inc alone grew from $30B to $2.7T.
I think it often feels like startups do things more efficiently. There’s less process or bureaucracy. It’s easier to implement features in small startups than big ones.
But I think these indicators are deceiving. In small startups it might feel like things are moving quickly, but most of these efforts either have tiny markets or end up as failures. There’s a lack of legal oversight, but some of that is because the startup doesn’t have to worry about downside risks. There’s fairly little risk of hurting the brand, but at the same time, you don’t have the benefits of a brand worth protecting.
“Effective Ventures had severe downsides from the FTX fiasco. This proves it was too large.”
I heard that the FTX fiasco caused significant problems at EV, which created a mess that impacted all of the organizations under the EV umbrella, even ones not otherwise associated with FTX.
I’m sure there’s a lot we can learn from this. I’m not convinced that one takeaway is that “we need to resign ourselves to very small organizations.”
The Gates Foundation has ~1600 employees. The ACLU, to pick a known nonprofit at random, has 1155 employees. Of course, the famous tech companies are much larger. Apple has over 160k employees.
At a quick glance, maybe all of EV has 100 employees? We’re really far from the size that other groups can manage.
I wouldn’t be surprised if it makes sense for there to be different substructures. Perhaps there’s a certain team or two that really make sense to be independent. But I think that broadly, it should be feasible to have groups much larger than EV and have them function adequately.
“The existing EA organizations aren’t good enough. We should fund new ones to discover teams worth growing.”
If you believe that a lot of value comes from large organizations, there is one legitimate strategy of spreading out your money to many small organizations, to see which ones will be worth growing later. However, I don’t get the impression that this strategy is really being done.
A good attempt at this strategy would involve:
- Be clear about your strategy upfront.
- Tell teams that the main value comes from them possibly becoming sizeable, and encouraging them to do so.
- Quickly stop funding groups that don’t prove themselves viable.
I don’t think the EA funders do this.
Personally, I don’t think we need this intense of an approach. I think that existing leadership teams are at least good enough to oversee projects that might later spin off.
But if as a funder, you want new organizations because you have a high bar and dislike the existing organizations, I think it’s important that you make your stance very clear and follow through with it.
“Middle management is toxic, we should avoid it.”
Having larger EA organizations would require more official middle management. A lot of people really don’t like middle management.
But in EA, I’m not convinced we’ve removed the importance of middle management. It seems to me like we’ve just converted middle management from a regular management class, into a large set of boards and funders.
I think the thing that makes middle management hard is not the managers, but the circumstances. It’s brutally difficult to coordinate a bunch of humans into aligned outputs. Replacing official managers with an ecosystem of boards and funders doesn’t remove the problem, it just moves it around.
“More organizations means more diversity of projects”
I prefer sizeable organizations to many tiny ones. But there are many ways sizeable organizations can be structured and run.
If you want highly-independent teams, you can still do that. Many corporate lab/R&D divisions are run with a lot of autonomy.
I heard that “Focused Research Organizations” created by Convergent Research are functionally set up as projects within a larger organization. That way they can be provided good legal and operations support. These organizations are essentially projects with a lot of independence.
If you’re creative, making an organization for each idea can be a restriction, not a benefit.
“We don’t have enough managers to oversee new projects. So it’s better to let these groups spin off into new organizations.”
As you might expect at this point, I think that the “solution” of getting around a lack of management capacity with the creation of new organizations, is often just moving one problem around into an even worse problem. When you do this, you replace the need for a good manager for the need for a good board. And the result is left with the extra burdens of setting up and maintaining a tiny organization.
In some cases where I’ve seen this trade happen, it has felt like it was because the institution making it, itself, would be the one on the hook for the manager but not the eventual board. It’s harder for an organization to set up a new internal project than it is for it to encourage someone to make a new organization and have them deal with all the resulting responsibilities and challenges.
So I suspect that “making new organizations because you don’t have enough management capacity” is often effectively just kicking the can down the road. You’re converting a straightforward problem that organizations need to deal with soon, to a harder and more nebulous one that others will have to deal with later.
I went back and forth with Max Dalton about this issue a few years ago, here.
Potential Proposals for Next Steps
I think that EA has gone too far along the axis of “having lots of tiny organizations with poor ability for coordination.” It feels like we have a bunch of independent fiefdoms with mediocre quality control. I think we’re getting many of the disadvantages of small orgs (mentioned above), without actually getting many of the hypothetical advantages of small orgs.
I would better understand the desire for many organizations in an environment with a diverse funding ecosystem. But now, a huge amount of funding comes from a few very similar donors.
I don’t think we should change to a completely centralized organization. I do think that there are probably a lot of incremental and pragmatic changes we can enact to get better tradeoffs.
Don’t Resign Ourselves to Tiny Orgs
Some of the discussion I’ve seen around EA seems to assume that we just need to have tiny (<20 person) organizations for reasons of liability or similar. I really don’t think this is a hard limit. It’s possible that in many situations small organizations are a good fit of costs and benefits, but I don’t at all think it’s essential. Running 300+ person organizations should be a much easier job than other challenges we’re facing.
Instead of making incubators that encourage new organizations immediately, make “Labs Divisions” or “Startup Studios”
Allow new teams to experiment with different ideas and team compositions for a while. Provide them with operations support and council. If things go poorly with the projects, but the team members are strong, you can then move them to other places in the organization. If things both go well and there are strong reasons for the project to later be fully independent, they can spin out then.
More Fiscal Sponsorship Programs
I like the model of Rethink Priorities and Effective Ventures of providing a fairly “heavy-weight” charity sponsorship.
More Mergers and Acquisitions
I see these frequently in the corporate world, but barely in the nonprofit world. I’d bet that we’re not doing these frequently enough.
Mergers and acquisitions lead to some executives losing power. In the corporate world, this is counterbalanced with substantial payouts. This is messier to do around nonprofits, so there would have to be some thinking here.
Invest in Management and Operations
A lot of the reason to make new orgs is because of a lack of internal management and operations resources. Building out management and operations can take a lot of time and money. You need spare capacity to support new projects.
Right now it feels to me like a bunch of EA problems come from management debt - areas where we haven’t properly invested previously. This is not a big surprise, because I’d claim that this is often a key problem in organizations and ecosystems. One great thing is that as the EA community ages and grows, we naturally have access to more experienced candidates.
Try to Mimic Organizational Strengths
Even if we keep with “lots of independent organizations”, there are steps we can take can do to minimize the typical disadvantages of these situations.
- We could have strong consultant businesses/organizations for operations/management tasks. (Antientropy is an example here)
- We could have coordinated efforts to help ensure that boards are run well and have good members on them. (The EA Good Governance Project is doing some work here). Maybe we can find legal ways to identity and compensate good board members, even in the UK.
- Organizations could set up and normalize programs to allow employees to work at other related organizations for different lengths of time.
- Funders could take on more management duties. They could be responsible for telling organizations to make large changes or not.
- We could normalize hiring practices more between certain organizations. Ideally, we’d have setups so that strong employees that do well in one organization and can be expected to do so in a different one, don’t have to go through another lengthy hiring process to move.
- We could have cross-organization roles like a cross-organization “COO”, who could have some powers to assist with coordination and quality control. This has been recommended for the US government, which also has power awkwardly split between a bunch of heterogenous turfs.
When is it good to make new nonprofits?
I think that making or spinning out nonprofits is often reasonable, I just think that EA tends to do it too frequently.
Some situations where it’s a particularly good idea include:
- A large part of the organization’s job is to oversee/evaluated senior people or organizations.
- The organization would present a severe brand or legal liability.
- The organization has important donors who either don’t like other organizations or would only donate with significant legal charitable limitations.
- The organization has a significant chance of becoming large and would do so adversarially with existing organizations.
- The organization is a student/local/social group, and it would be very weird for it to seem controlled by another organization. (Note: If funding mostly comes from a narrow set of funders, then arguably this group might de-facto be controlled by a central entity anyway, so this is a strange circumstance.)
It’s often the case that a certain founder has little other choice but to create a new organization. However, the flip side of this can easily be, “The ecosystem has not done a good job of enabling new groups to develop within existing organizations.” So in some of these cases, I think it might be a reasonable choice for the founders, but in the bigger picture, a mistake or weakness of the ecosystem.
Thanks to Angelina Li for providing comments
A useful post, Ozzie, and definitely food for thought.
I would just like to point out one fairly significant consideration in favour of small organizations that isn't factored in here - ownership and motivation (i.e. the founder and other early-stage employees
slave awaywork far harder because we feel a sense that the organization is yours - you are the organization; you don't work for it). This has been my own experience, and I imagine it's much the same for you. I believe Joey Savoie talks about this fairly often, when asked why Charity Entrepreneurship doesn't just hire people to implement effective global health & animal ideas in-house, rather than using these people to incubate new orgsFor me, it really depends on what I'm comparing it to.
If I were trying to make QURI within a mediocre organization, I'd both expect to be able to produce less useful output and I'd be less motivated.
If I could run it with a fair bit of flexibility, in an org that were able to give me various resources and integrate well with our work, that could be great.
I started QURI-related work at FHI and really liked a lot of things about that environment. I left, in large part because Oxford university specifically is a poor fit for programming projects - but otherwise I could have imagined something like that working well.
After FHI, I've been working with a small team, often doing remote work. I've found it fairly isolating and lonely, although I do like some elements of the independence.
As a manager of a small "EA organisation" I strongly agree with this post.
To expand slightly on the points you made under the flexibility headings, I focus on two things more significant organisations can offer that smaller ones can't:
There's also a point about weight and influence. We greatly reduce our policy influence by fracturing ourselves so much. We are a significant movement that ought to have a weight in the policy conversation equal to our significance. But we present ourselves poorly for that purpose and are weaker as a result. I understand there are risks we manage through the current approach, but I think we are very much on the wrong point of the spectrum.
In my specific case, I would much prefer a world where funders who wanted activities to occur in Australia were able to pool money into one or a small number of top-level organisations and communicate to the board their top-down view of prioritisation and the degree of autonomy + trust they are or aren't happy with through funding agreements. And then the organisation can do fundraising and community building and policy development and advocacy and organising conferences and coordinating volunteers and all the other things. Presumably, it would also be easier for donors to interface with one or a few organisations, rather than dozens.
I'd push back on the idea that certain "funding models and employment models . . . follow" from organizational size. I don't see why donors couldn't use different funding models, which would allow for different employment models. They choose to deploy short-term grants for their own reasons . . . but much of the fix is changing the funding model. I'm not sure how consolidating orgs is either necessary or sufficient for that. In the non-EA world, larger orgs often have more diverse and stable funding streams than smaller ones . . . but here, the bulk of funding is coming from the same few sources.
I think I mostly agree with what you're saying. Perhaps the difference at the margins is that if 10 organisations get 10 grants for 10 researchers for 6 months, what they'll each do is find a person for a short-term contract. If someone leaves or gets sick etc, a stream of work might not be completed and an organisation will be in a tight spot.
If that was instead one organisation, maybe it brings on 5-6 researchers for a year (or ongoing), makes some assumptions about staff turnover and part-time arrangements, and when someone leaves or gets sick it's not a big deal because work can be reassigned between the team and project timelines can be shuffled (surge people onto the more urgent work). Basically bigger numbers (dollars; staff; projects) gives managers more wiggle room to find ways to make things work.
But I agree with you, that if funders were hard over on smaller organisations, there are ways they could ease the employment model concerns in other ways. I just think the ideal situation from an employment and management perspective would be longer term and more centralised.
My rough understanding is that funders seem to like taking bets on early orgs, which tend to be small, and funders tend to be more reluctant to give large amounts of funding to large orgs.
This is probably sensible -- there indeed ought to be a higher bar for higher amounts of funding and you will actually have a track record to analyze which is probably not great as the initial bets even if just due to mean reversion -- but I think implicitly pushes towards smaller orgs.
Another issue is that most funders -- notably though not OpenPhil -- don't seem interested in entering into the enduring multi-year funding commitments that are necessary to sustain a large org. Which is totally fine as it is their money after all, but it also tilts the landscape towards smaller orgs as for a larger org it can be very difficult to handle the uncertainty of getting $1M one year but then having no idea if or when that funding could be expected to recur.
There's a lot of management and organizational culture things to fix within EA to sustain larger orgs but I personally think if we wanted more larger orgs the key place to start would be to make the funding ecosystem better equipped to support larger orgs.
Note that these thoughts are my own personal ones and certainly shouldn't be taken as complaints or comments on behalf of other RP execs tasked with running and sustaining a relatively large org.
This feels similar to me to how EA invests so much in very early career individuals, but they have a fairly limited window before being dropped or at least finding their options much more limited. It's more like buying lottery tickets than making investments.
Is the "like taking bets on early orgs" thing true of OpenPhil? I think CEA is one of their larger grantees, and I don't feel like their reluctance has been superlinear in the size of our budget, but I don't have a ton of insight.
In my personal opinion I think so but it's hard to say.
Thanks!
I appreciate this sort of pushback / liked this post! :)
For proponents of more consolidation in EA, I'd be excited for more people to try this exercise:
Reasoning: I worry that high level arguments for or against community restructuring sound much nicer in theory than in practice (e.g. it wouldn't be that surprising to me if the vast majority of small orgs all had good reasons for not starting off as projects). I'll share my 10 min version in the comments.
.
Also to push back slightly on this narrative, I think there is more coordination in EA than you might expect if you only looked at the raw number of small organizations associated with EA. Some examples:
Here is my babble from the above exercise:
A reminder that I spent only 10 minutes on this, I have no special inside knowledge / am not affiliated with any of these orgs, and my main goal was to check if there is any low hanging fruit for consolidation! I'm not sure how good either of the above ideas are.
.
Some observations:
In some cases, regional EG groups need to exist as separate entities at a national level to offer tax-deductible options to residents of that country. In many countries, a contribution is only tax advantaged if made to a charity incorporated in that country.
I feel like if I were to call out any specific organizations, people from those organizations would get upset, so I'm hesitant to do that exactly.
Looking at this list, I think a bunch can be consolidated somehow. I get the impression that many of these:
- Have employees would be good fits at the another's org
- Share very similar funding sources
- Don't have severe branding/community risks
- Aren't part of Universities
Groups that are very friendly with EA and OP, or some AI safety groups, seem like the easiest to semi-merge. (I say semi because they could still have some levels of isolation)
If two organizations get funded by the same funding org, is it really that different than them being projects within the same org? It seems almost equally awkward to have one funder fund two competitors, than it is for one org to have two competing sub-groups.
Fyi, the list you linked doesn't contain most of what I would consider the "small" orgs in AI, e.g. off the top of my head I'd name ARC, Redwood Research, Conjecture, Ought, FAR AI, Aligned AI, Apart, Apollo, Epoch, Center for AI Safety, Bluedot, Ashgro, AI Safety Support and Orthogonal. (Some of these aren't even that small.) Those are the ones I'd be thinking about if I were to talk about merging orgs.
Maybe the non-AI parts of that list are more comprehensive, but my guess is that it's just missing most of the tiny orgs that OP is talking about (e.g. OP's own org, QURI, isn't on the list).
(EDIT: Tbc I'm really keen on actually doing the exercise of naming concrete examples -- great suggestion!)
Yeah, fair! It's frustratingly hard to get comprehensive lists of EA orgs (it's hard to be in the business of gatekeeping what 'EA-affiliated' is). I did a 5 min search for the best publicly available list and then gave up; sometimes I use the list of organizations with representatives at the last EAG for this use case.
Maybe within AI specifically, someone could repeat this exercise with something like this list. If someone knows of a better public list of EA orgs, I'd love to know about it :)
I thought about this briefly a few months ago and came up with these ideas.
Also looking at local groups, there is some coordination on the groups slack and some retreats but there is still a lot of duplication and a high rate of turnover which limits any sustained institutional knowledge.
I directionally agree with this, but am generally averse to putting medium-risk-or-above projects inside a big organization without a sufficiently clear upside to the risk.
As relevant here, turning CBG grantees into CEA employees could potentially create a lot of exposure for CEA for various things that happen in the groups that these people lead. I'd be much more comfortable with spinning off community building into relatively asset-light, special-purpose organizations, e.g., "EA Community Builders of the Bay Area / UK / Etc." I think you get many of the benefits of centralization that way without exposing the balance sheets of projects that need significant operating capital (and creating a potentially more attractive target for that reason).
A separate organisation just for CBGs would have been useful too rather than a lot of one and two person teams with constant turnover.
I would be concerned about the power-centralizing effect of organizational consolidation. Although there is some material here about how to mitigate that effect (e.g., forming things like “Focused Research Organizations”), the overall effect of following this perspective would seem to be power-centralizing.
In legislative appropriations talk, you'll hear the term "revenue-neutral" thrown about a lot; it means that any tax cuts and spending increases are offset elswhere in the package so that the total effect on budgetary balance is neutral. I'm curious about whether/how some of the advantages described in this article could be achieved in a power concentration-neutral manner.
My best guess would be to merge a bunch of small organizations into a larger org that is legally set up as a society with a membership of 50-500 (not open membership), whose members elect the board (and new members if they see fit), and whose bylaws prohibit granting membership to anyone recently affiliated with one of the major power centers.
I'm also be curious about this line of thinking.
On my side, I like the idea of a lot of EA power being decentralized, but I also would very much like us to be able to quickly coordinate on major changes, as seems best.
One way of doing this is having a small group in charge, but have that group voted on by a much larger group in regular intervals.
I really like this post; it updated me in favor of some types of consolidation in EA (although I was already a bit sympathetic to this view). One nitpick on this:
I think it probably makes sense to look at percentage growth, rather than the growth in absolute value of assets. In this post I found that YCombinator companies grew substantially faster than the S&P 500. My methodology is not super comparable with what you are doing here, but e.g. Airbnb is worth $83B compared to the $1.8M it was valued at when it received the YCombinator standard deal, for a 46,000x ROI, which compares quite favorably to Apple's ~90x.
As an alternative approach: this review claims that angel investors average 35% IRR, which compares favorably to the returns of investing in larger companies. Numbers from angel investors are dubious for a bunch of reasons, including some of the ones listed in that paper, but my impression is that angel investing is not that much worse than public market investing.
Anyway, sorry for the long comment. I agree with the overall claim that startups are not "far more efficient" but think they might still be a bit more efficient, on average (with a bunch of caveats).
This may be true for market cap, but let's be careful when translating this to do-goodery. E.g., wages don't necessarily relate to productivity. Higher wages could also reflect higher rents, which seems plausibly self-reinforcing by drawing (and shelving) innovative talent from smaller firms. A quote from a recent paper by Akcigit and Goldschlag (2023) is suggestive:
I don't have a good grasp of the literature. Still, the impression I got hanging around economists interested in innovation during my PhD led me to believe the opposite: that smaller firms were more innovative than larger firms, and the increasing size of firms over the past few decades is a leading candidate for explaining the decline in productivity and innovation.
Speaking from my own experience working in a tiny research organisation, I wish I could have started as a researcher with the structure and guidance of a larger organization, but I really doubt I'd have pursued as important research if we hadn't tried to challenge other, larger organizations. Do you feel differently with QURI?
I'm sure that large companies do monopolistic techniques to get unfair advantages.
I remember hearing about this literature before. I can't help but be a bit suspicious.
If decentralization were so productive, wouldn't huge companies notice this and take advantage of it? It would be very easy for large firms to reorganize so that there's very little centralization. I'd be skeptical of the specific hypothesis, "large firms intentionally want their employees to be unproductive," if that's being suggested.
I'm similarly suspicious of the "6 to 11 percent" figure. It's very easy for me to imagine that their innovations do better on some measures, like, "more likely to influence a lot of people later on, because of integrations."
> I wish I could have started as a researcher with the structure and guidance of a larger organization, but I really doubt I'd have pursued as important research if we hadn't tried to challenge other, larger organizations.
From the standpoint of funders, I'd expect that they would want the "really promising people" to be running and reforming the big institutions, instead of creating new institutions. My main points are for those in charge of things and at the margin. From the standpoint of an individual facing an organization that clearly doesn't seem good to them - I agree it can easily make sense for them to figure its best making their own external organization.
(I'd also flag that many people I know who left companies to do independent work, did projects that really didn't seem that great/scalable to me. I think it's hard to tell.)
Like many problems in EA, this one is 100% at the door of OpenPhil. There is absolutely nothing to stop them funding fewer organizations with more money per organization, rather than overseeing this sprawling empire of countless tiny orgs. The real question is why OpenPhil prefers it this way?
Yea, I think the funders are the ones with the main powers here. I'm also curious.
I assume it would take a fair bit of effortful work and coordination to set things up large organizations that people feel good about. My impression is that the funders in our ecosystem have had pretty low staff levels and didn't have spare senior management capacity or experience that would have made this sort of thing straightforward.
The Bill & Melinda Gates Foundation and the Chan Zuckerberg Initiative both seem to have much larger headcounts than OP, for instance.
The cynical explanation is that because all of these orgs are small ,they're also weak and very likely to remain entirely dependent on the funder. Larger orgs would be much more likely to start their own fundraising teams and seek to diversify their revenue sources.
For what it's worth, with what I know, this specific thing doesn't seem very likely to me.
My impression is that one reason for having these orgs be independent is the hope that other funders would come in to help fund them. (But that doesn't seem to have happened as we might have liked)
Great post, I agree with a lot of it. There is definitely a kind of large org sclerosis that can develop, but IME it's generally associated with much larger orgs unless you have very dysfunctional management (which is a risk!).
One missing factor I think is fungibility. It's hard for donations to be funged between organizations, but it's very easy for them to be funged between protects within an organisation. So we might expect donors to have some preference for separate orgs for separate projects.
I agree this could be a benefit, but I see it as being fairly minor. I think that the cost that larger organizations would need to incur to make trustable promises of separation, can easily be less than the costs of having totally different small organizations for each of these projects.
In most jurisdictions, donors and charities can agree to legally restrict donations -- e.g., this money can only be used for the philosophy department unless the donor consents to removing the restriction (or in very limited cases, this is done by court order). Indeed, if my recollection of Catholic diocese bankrupcties holds, this sort of structure can provide some protection for the projects from exogenous legal shocks (in that case, child sex abuse claims against other projects) if the formalities and accounting are done correctly.
However, it seems there would be a tradeoff between anti-fungibility (which is also important for protecting project independence and minimizing centralization) and some of the flexibilities your post identifies. Suppose we created a big organization out of 20 existing micro-organizations, which became BigOrg projects. Those projects continue to obtain funding -- 75% of which is tightly locked to that project, and 25% of which is less restricted (i.e., it can be used to pay the project's fair share of BigOrg's legal/ops/management/etc costs). Under that model, BigOrg doesn't seem to have much if any fungible money -- unless it persuaded a funder to make additional donations for its unrestricted use.
I suspect that a number of your suggestions about management flexibility would be difficult under tight internal accounting controls. Most staff would be locked to work for specific cost centers to comply with anti-fungability/independence norms and legal requirements. You could reallocate part or all of their time to another cost center in order to use them for another project . . . but the funding has to be in place for them to bill that cost center.
(separate comment for agree/disagree voting purposes)
While much has been rightly said about the job insecurity created by small grant-dependent organizations, that model does provide some job security advantages. If your independent organization has two employees, a grant for this year, and six month's runaway after that . . . you have a decent bit of job security until mid-2024. The independent organization is constrained to use its funding to accomplish the organization's mission as stated in its bylaws, so you don't have much job insecurity from someone deciding the mission isn't that great anymore.
In contrast, if your org becomes a BigOrg project, you run the risk that BigOrg decides its mission isn't good anymore (and that you're not worth moving to another project), Even if funding is legally locked to your project, the restriction can be removed with the donor's consent. Thus, it seems likely that some additional protections for employees, such as employment contracts vs. at-will employment, would be necessary to avoid creating greater job insecurity than already exists.
I'm not sure that bigger orgs would lead to less job security.
Living grant-to-grant certainly can feel pretty scary. Sure, you can feel confident about ~12-16 months at the very start of getting grants, but those last 3-8 months can be stressful.
Of the people I know at OP/EV/RP, I can't think of many who got laid off.
If I were working in a larger organization on an uncertain initiative, I could imagine, and would hope for, organizations that we could trust with clear things, even if they aren't legal. For example, if my superior clearly states in writing that my project will get 1-2 years of funding, and I know a fair bit about the history and track record of that person, I might well trust them.
I think the crux here may be how the consolidated/bigger org would be funded and budgeted. If the funding that would be going to projects of the bigger org consists largely of the same short-term / specific-purpose grants that were going to the smaller orgs, it's probably going to be difficult for the bigger org to commit to longer-term funding for a project.
I have some concerns about switching from many small orgs to fewer bigger orgs without simultaneously changing the funding model for this work from the current style of grants to (quoting Peter's comment) "the enduring multi-year funding commitments that are necessary to sustain a large org." Trying to run a bigger org on small-org style funding arrangements feels like it could be the worst of both worlds.
Nice one Ozzie - I don't see why we should take this with a grain of salt, its an important issue and you have done a lot of thinking about it. I've got a few comments, as I'm quite interested in this as I run a small organisation like yours. My experience comes from the "NGO" world, and I'm sure doesn't apply so well to other kinds of orgs like AI safety or community building
If your organisation has to massively pivot from what it is currently doing or have a high degree of flexibility, than perhaps it shouldn't exist at all or at least think hard about its purpose.
I'm also probably in the minority in thinking that low donor trust is more good than bad. Because NGOs don't really operate within the market, supply and demand doesn't apply in the same way its up to the donor to carefully monitor whether what they are "doing good better" than other orgs and deserve more funding. I actually think donor trust is still too high, which is part of the reason why the "BINGOS" (world vision, oxfam, save the children etc.) continue to get stackloads of cash even while they do so little good per dollar spent.
2. I really agree with the benefits you listed of working in larger orgs, but this can also go both ways. Here in Northern Uganda the bigger NGOs are so inefficient and burocratic, and staff there learn bad habits. At OneDay Health I'm super hesitant to hire anyone who has worked in a larger NGO because they are so used to moving suuuuuuuper slowly and having low productivity expectations. So it depends on the quality of the larger org and the team within it, whether people will learn a lot or not within that org.
I agree with the potential downsides of "downsizing" and qualiyt control too some extent
I would love to hear Charity Entrepreneurship's response on this as well.
Wow I failed badly on the formatting front haha.
Thanks for the comments here!
I agree that there are situations where funders would prefer organizations to have significant restrictions. I'm hopeful that in our situation, we can find ways for funders to find sizeable organizations that they can have enough trust in that they don't need to go through many legal restrictions.
I've written about this here: https://forum.effectivealtruism.org/posts/rqo9AswmH7KRF77ko/improve-delegation-abilities-today-delegate-heavily-tomorrow
Or, if we can develop important relationships with high levels of justified trust, then we can heavily delegate decision-making to sizeable organizations, giving them the ability to be very flexible as is best. If we can't get to these levels of trust, then the appropriate thing is to restrict them more.
In the corporate world, big businesses have to change a lot to do well. In the startup world, pivoting is a major deal. There are a lot of best practices in how to not do this too much or too little.
I'd also agree that there are some really bad large organizations out there. I'd guess that main reason they are bad is because they are in incentive landscapes that are poorly structured or noncompetitive. I'd be curious how effective the median small nonprofit is in the ecosystems you are pointing at - I'd expect that they're fairly mediocre as well.
If I were a strong team member in a bad organization, I could easily see myself expecting to be more effective outside, in a tiny org. But if I were a high-level decision-maker (board member or funder), and I knew that there was a very strong team member inside my big organization, I'd probably prefer to have/move this person to a top leadership role, then have them be a leader in a new tiny organization.
I think that the specific team and focus areas at Charity Entrepeneurship seem like arguably good fits for an incubator, instead of a larger organization. However, if there are other groups considering making incubators, I'd definitely encourage them to consider making startup studios as well. (I have made this recommendation to three other groups considering considering incubators.)
Re “Middle management is toxic, we should avoid it.”:
I want to flag that: your counterargument here does not properly address the points from Middle Manager Hell / the Immoral Mazes sequences. (Less constructively, "Middle management being toxic" seems like a quite weak version of the arguments against large orgs. Which suggests that your counterargument might not work against the stronger version. More constructively, one difference between current EA structure and large orgs is that small EA orgs are not married to a single funder. This imo reduces the "toxicity" you might otherwise get by the invectives structure in large companies. There might be other important differences; I just haven't thought about this enough.)
All that said, perhaps we can get the best of the both worlds by using larger orgs for some things but not all? And inventing some tools that make it easier to get the benefits you want without all of the costs? (Example: something that allows people to temporarily/tentatively switch jobs without having to deal with all the paperwork.)
I didn't mean for it to. I was just pointing at the general dislike for Middle Management.
I think one awkward thing now is that many of them are sort of married to a single funder. The EA funding community can be really narrow right now - just a few funders, and with similar/overlapping opinions and personnel.
I think we can do better about getting better sets of trade-offs. There's also "large orgs, with lots of team freedom" as a medium. I also listed some strategies in the "Try to Mimic Organizational Strengths" section. I'd be happy to see this area worked on further!
I mused about this yesterday and scribbled some thoughts on it on Twitter here.
Investing marginal resources (workers, in this case) into your single most promising approach might have diminishing returns due to A) limited low-hanging fruits for that approach, B) making it harder to coordinate, and C) making it harder to think original thoughts due to rapid internal communication & Zollman effects. But marginal investment may also have increasing returns due to D) various scale-economicsy effects.
There are many more factors here, including stuff you mention. The math below doesn't try to capture any of this, however. It's supposed to work as a conceptual thinking-aid, not something you'd use to calculate anything important with.
A toy-model heuristic is to split into N separate approaches iff the extra independent chances counterbalance the reduced probability of success for your top approach.
One observation is that the more dependent/serial steps (k) your plan has, the more it matters to maximise general efficiencies internally (c), since that gets exponentially amplified by k.[1]
You can view this as a special case of Ahmdal's argument. If you want. Because nobody can stop you, and all you need to worry about is whether it works profitably in your own head.
I think this was a cool post and I'm excited to see this kind of discussion here. (I think it misses a bunch of advantages of small orgs but seems fine to have a post that's mostly about the disadvantages. unfortunately don't have time to write out my object level thoughts here - just wanted to be clear that this comment is a "like" not a "(fully) agree." )
Thanks!
> I think it misses a bunch of advantages of small orgs
Yep. This topic could easily become a much larger post+project. My main priority was to represent this one side, as I've felt like this side of the discussion hasn't gotten as much attention.
One point I'll flag on the "advantages of small orgs" that I've come across:
I've come across some specific problems of specific large organizations, but I'm not sure why exactly they exist. I hear people complain about legal/PR restrictions that come from larger organizations, but it's not clear to me if funders should try to address that problem by encouraging small organizations, or by improving the big ones. Some "big organizations" challenges might be fundamental/inherent to big organizations, and some challenges exist for reasons we can fix.
Regarding the specific problems of specific large organizations, maybe there's something here to do with bureaucratic mazes? For instance, Raemon's post Recursive Middle Manager Hell has a section titled Implications for EA and AI that goes like so:
I'd agree with that the above quote in that I'm not suggesting we should hire a lot more.
The way I see it, we already have a lot of people - the main choice is how to best configure them.