When we started 80,000 Hours, one of the key ideas we presented was the replaceability argument:
Suppose you become a surgeon and perform 100 life saving operations. Naively it seems like your impact is to save 100 people’s lives. If you hadn’t taken the job, however, someone else likely would have taken it instead. So your true (counterfactual) impact is less than the good you do directly.
I still think this is a good argument, but I’m not sure how relevant it is when comparing real career options.
In particular, I see the argument often being used incorrectly in the following two ways:
Ignoring direct harm: Suppose you’re considering taking a job that some people think is harmful (e.g. certain parts of the financial sector) in order to donate, do advocacy or build skills. You reason “if I don’t take the job, someone else will instead, so the potential harm I’ll do directly doesn’t matter”.
Ignoring direct impact: Suppose you’re considering working at a high-impact nonprofit. You reason “if I don’t take the job, someone else will instead, so I won’t have much impact.”
I disagree with both of these claims in most circumstances. Why?
The replaceability argument only establishes that your true impact in a job is less than what you appear to do directly. But we don’t know how much less. You can only ignore your true impact if it’s a lot less than what you appear to do directly, and we don’t know that it will be. Suppose your true impact is 50% of what you appear to do directly. 50% of a big positive impact is still a lot of impact; 50% of a lot of harm is still a lot of harm.
We don’t know that true impact is a lot less than what you appear to do directly because it’s very hard to know what would have happened if you hadn’t taken the job, and therefore to quantify the importance of replaceability. For instance:
If you don’t take the job, there’s some chance they can’t hire anyone else instead at the same salary. So you’re not replaceable at all.
If you take the job, we’d expect you to perform better than the person who would have been hired otherwise because you’re the top candidate, and again this could be a significant difference. Real organisations often say they find it really hard to hire, so the top candidate is often a much better fit than the next best.
If you take the job, the person who would have taken it otherwise is now free to go and work at another socially positive organisation, so there are complicated ripple effects.
In the case of doing harm, there’s additional moral reasons that you shouldn’t ignore the harm. Most non-consequentialist moral theories think you shouldn’t harm someone even if they’ll get harmed otherwise: if someone is about to get shot by a hitman, it’s still wrong to shoot them just beforehand. And there are additional consequentialist reasons too against taking harmful jobs: doing something harmful could damage your character and reputation, or encourage others to do the same.
Overall, the typical importance of replaceability is highly uncertain, and this means you either need to do lots of situation specific research (e.g. ask your potential employer what they’d do if they couldn’t hire you), or focus on robustly important considerations such as:
Will you get good career capital?
Is the organisation doing high-impact work?
Do you have good personal fit for the position? (If you do, then you’re probably not especially replaceable).
If someone uses replaceability to justify the idea they’ll have a little impact, ask them to elaborate, taking account of the effects listed above. In general, to be confident in replaceability-grounded arguments for low impact, you’d need to do much more detailed analysis, as we did for doctors.
What are some good applications of the replaceability argument?
The key reason we introduced the replaceability argument is to show there’s a difference between what your tangible, direct impact and your true (counterfactual) impact. The point was that taking a job which does good directly doesn’t mean you’ll have a large positive true impact – it could be near zero or even negative if you perform less well than the person who would have replaced you. Existing advice on social impact careers, however, seems to mainly focus on direct impact, which looks like shallow analysis at best, and perhaps a major mistake. We stand by this argument. However, it doesn’t mean you can ignore your direct impact, it just means you need to do more careful analysis.
The other main way we used replaceability was as an argument for earning to give, and I think that still stands too. If you donate 30%+ of your income to charity, that money very likely wouldn’t have been donated otherwise, because so few people give such a large share of their income. However, your direct impact is indeed less than it first looks, so your donations are comparatively more important than they first look: they’re not at all replaceable whereas your direct impact is somewhat replaceable.
You can also make a similar argument in favor of other forms of indirect impact, such as impact through advocacy. For instance, if you become a journalist and write about neglected social issues that are different from what journalists normally write about, those stories very likely wouldn’t have been written about otherwise.
However, I see the replaceability argument as a relatively weak argument in favour of earning to give. The stronger reasons are:
Some people have an unusually strong comparative advantage for earning money, and it makes sense for them to specialise in that, while letting others specialise in direct work.
We have good research about which uses of money are most effective.
High-earning careers often provide good career capital.
We mainly used replaceability as an illustration of the difference between direct and true impact and as an argument in favor of earning to give, and we still think both of those uses were valid.
We’ve also always thought that the importance of replaceabilty is highly uncertain and varies from situation to situation – I wrote a Master’s Thesis that said as much in 2012, the year 80,000 Hours was founded. And as we learned more, we became more cautious (e.g. see this post and this one). Due to this uncertainty, replaceability never played a big part in our online career guide – instead we recommend people focus on other factors such as personal fit, cause effectiveness and career capital.
However, in the early days of 80,000 Hours, we presented a misleadingly simple view of replaceability on occasions, for instance, suggesting that you could just think about the impact of yourself compared to who would have taken the job instead and that your direct impact is typically near zero (e.g. this post and some media appearances) and used it as a partial defence of taking potentially harmful jobs. The dilemma was that (i) people were thinking incorrectly about their impact (as tangible impact rather than counterfactual impact) but (ii) it was hard to spell out the full implications of thinking about counterfactuals, so we could either give people (a) a conceptually weaker but sometimes more accurate view or (b) a conceptually stronger but sometimes misleading view. Now we think it may have been better not to draw attention to it.
We also failed to clarify our views on replaceability soon enough, even when we became aware people were misusing it and making it out to be a stronger consideration than it is. I’m concerned that due to this some people may have been excessively discouraged from taking jobs in high-impact organisations (ironically, including at 80,000 Hours).
For this I and the others at 80,000 Hours would like to apologise. I didn’t post sooner because I wanted to write an updated analysis of replaceability, but struggled to make it a top priority. I’d still like to write an explanation of how to analyse replaceability in typical decisions, but for now I hope this post will serve.
Update 2021: To learn more about how to analyse replaceability in practice, see this interview with me and the attached reading list.