The productivity benefits of substantial compensation in effective altruism

At first glance, one might think that effective altruist salaries should be relatively low, because it means that more money could go to highly effective causes (e.g. global poverty charities). Against this, many effective altruists argue that higher salaries increase the productivity of direct work in effective altruism, and that it’s therefore better to pay higher salaries.* In this post, I’ll discuss this argument. (Note that it’s the converse of the costly signalling-argument and the lifestyle-causes-bias argument, which argue that indirect effects rather give us reason to make effective altruism more demanding.)

One oft-mentioned argument is that higher salaries will allow highly productive staff to spend a greater fraction of their time working, via direct time purchases (time-money tradeoffs). Here is an example from Caroline Ellison:

For instance, say A works at an EA org.

- His work produces $500/hour of value.

- He gets paid $50/hour by his employer.

- He has a fraudulent charge of $100 on a card that he could dispute.

— This requires him to spend 1 hour on the phone with customer service.

— He is indifferent between this and spending an hour doing a relatively unpleasant work task.

As things currently stand, he might spend the hour to recover the $100. But I think it would clearly be better if someone paid him $100 to spend an hour doing the unpleasant work task for his organization rather than trying to recover the money. It would keep his utility (and thus demandingness) constant, while resulting in $400 of surplus value created.

Another line of argument is more psychological, and has to do with the fact that higher salaries will make it more appealing, or less unsustainable, to work at an effective altruist organisation. They come in two forms. 

• The burnout argument: too low salaries can, especially when combined with a large workload, lead to burnout.

• The incentives argument: higher salaries will make it more appealing to do high-impact direct work. This argument can be divided into three sub-arguments.

  • The recruitment argument: higher salaries will increase the number of competent applicants to effective altruist organisations. 

  • The substitution effect: higher salaries will make existing staff work more, since it increases the rewards.

  • The task-specific incentives argument: higher average compensation allows you to specifically reward high-impact actions (e.g. via prizes/bonuses) or high-impact roles, which will make people more likely to undertake those actions and take those roles.

***

It seems to me that many effective altruists discussing productivity benefits of higher salaries focus on direct time purchases. I agree that this is a factor, but my hunch is that the incentives-related effects are bigger.

It is true that a higher salary could allow you to buy time in various ways—e.g. it can allow you to hire a cleaner, or accommodation closer to work, thereby shortening the commute. I think these effects can be non-trivial, but usually not huge. (Though admittedly, there may be some service purchases that can make a very big difference; like child care.) Many people outside of effective altruism who earn a low salary work many hours a week, so in a sense it’s clearly possible. Moreover, if what you’re interested in is letting people buy more time with their money, then an alternative to paying a higher salary is paying for specific expenditures—e.g. meals, cab rides, child care, etc. 

Also, this argument assumes that paying effective altruists higher salaries will lead them to purchase more services and make other decisions that allow them to work more. But it could also lead to them consuming more leisure, since their improved financial situation could allow for that (the income effect, which works in the opposite direction of the previously mentioned substitution effect). For instance, some effective altruists take time off between jobs, which may have been harder had pay been lower. I also think that social norms make ~40 hours weeks an attraction point for many, meaning that service purchases freeing up five hours a week may lead to less than five hours additional work.

***

Turning to the psychological arguments, burnout is also relatively frequently mentioned as an argument for higher salaries in effective altruism. I think it is a factor, but potentially mostly at relatively low compensation levels, since very low compensation may make life unusually difficult.

The most important argument for higher salaries is, in my view, rather the incentives argument. I think that higher salaries facilitate recruitment to effective altruist organisations substantially. I also think that over the long run, effective altruism having a reputation of paying good salaries may help with attracting established and highly competent people to the effective altruism community. (There is a risk, of course, that some of those people wouldn’t be particularly value-aligned in key regards, but as previously discussed I think it’s better to test directly for that than to try to exclude them via paying lower salaries.) And I think that it may help with retention in the effective altruism community.

Using money to incentivise effective altruists to undertake specific high-impact actions, or to take particular high-impact roles, can probably also be highly impactful. There have been some initiatives and policies in that direction, but in my view it hasn’t been tried to the extent to which it deserves. I will discuss this strategy in a later post. As stated, it likely leads to higher average compensation levels.

Another important issue is that I think there’s a difference between the incentives argument, on the one hand, and the direct time purchases and the burnout arguments, on the other, in that the former can motivate higher salaries than the latter. The marginal benefits of higher salaries via direct time purchases may taper off more quickly than they do via greater incentives—since many (though not all) direct time purchases, like a cleaner or accommodation closer to work, would be available through a relatively moderate compensation. Similarly, the burnout argument may only motivate a moderate compensation, as we saw. By contrast, my sense is that incentive effects can continue to provide more substantial benefits even in quite high compensation ranges. As such, the other two arguments may provide less of an additional reason for higher salaries than it may seem.

It’s worth noticing that on the regular labour market, companies are under no illusion that the incentives argument is by far the most important out of these arguments. Of course, effective altruists are different in some ways, but I think that people sometimes overrate how different we are in this respect. There may be a whiff of self-congratulation in the notion that for effective altruists, literally buying time with money is a more important mechanism than that between economic incentives and labour supply.

***

For these reasons, I think effective altruist organisations should pay substantially higher salaries than the default mentioned in the first sentence of this post. I’m deliberately vague about what the exact level should be, though (and don’t have any particular level in mind by expressions like “substantial”, “high”, or “higher”.) In general I support the move towards higher salaries that we’ve seen in recent years, and think that it at least in some cases can go further. 

However, there are two other counter-arguments worth taking seriously that I should comment on. The first is that high salaries harm effective altruism’s reputation. I will address that argument in a later post, but briefly my view is that that is something that should be monitored continuously. I don’t think that it’s a reason to lower salary levels, or against increasing them; but that it is a reason for moving at a moderate pace and see how things develop.

The second is that arguments for substantial compensation in effective altruism are self-serving and based on motivated reasoning. I think that’s definitely possible. At the same time, many who notice this possibility refuse to engage with the object-level arguments, resting content with merely claiming they’re self-serving. I don’t think that’s the right attitude either. We should be aware that our judgement may be clouded, but even so, we need to address the object-level arguments, as I’ve been doing.

***

I recently criticised effective altruists who give indirect psychological arguments for us undertaking particular actions. In this post, I am, however, effectively giving psychology-based arguments for the notion that effective altruist salaries should be higher than you might think at first glance.

This might seem like an inconsistency, but I’d argue it isn’t. The fact that people react strongly to incentives is a very well-established phenomenon, which we know both from everyday experience and from economic and psychological research. Whereas the arguments I referred to in my recent post seem to have a “just so”-character, the notion that we react to economic incentives does not.

* For ease of expression, I talk about salaries at effective altruist organisations in most of this post. However, the reasoning is also supposed to cover other forms of compensation (e.g. prizes and grants) and work outside effective altruist organisations (e.g. independent work).

Thanks to Ryan Carey, Daniel Eth, and Pablo Stafforini for incisive comments.