Cross-post: Think twice before talking about ‘talent gaps’ – clarifying nine misconceptions, by 80,000 Hours.

post by 80000_Hours · 2018-11-12T20:52:08.539Z · score: 48 (26 votes) · EA · GW · 7 comments

Contents

  1. Problem areas are constrained by specific skills, not ‘talent’
  2. Skill bottlenecks are a matter of degree
  3. Skill and funding constraints are not opposites
  4. Skill bottlenecks vary by problem area
  5. Skill bottlenecks aren’t only about effective altruism organisations
  6. High value of recent employees doesn’t necessarily entail that hiring is the top priority
  7. The existence of skill bottlenecks doesn’t mean that it’s easy to get a job in the area
  8. Skill bottlenecks are not always solvable by raising salaries
  9. Skill bottlenecks can change rapidly
  Conclusion
None
7 comments

If you have a 'tool gap', it really matters which tool.

Cross-posted from the 80,000 Hours blog - see the original here. This is a partial follow-up to our previous post here Many EA orgs say they place a lot of financial value on their previous hire. What does that mean, if anything? And why aren't they hiring faster? [EA · GW].

--

After pushing the idea of ‘talent gaps’ in 2015, we’ve noticed increasing confusion about the term.

This is partly our fault. So, here’s a quick list of common misconceptions about talent gaps and how they can be fixed. This is all pretty rough and we’re still refining our own views, but we hope this might start to clarify this issue, while we work on better explaining the idea in our key content.

1. Problem areas are constrained by specific skills, not ‘talent’

Problem areas are rarely generically ‘talent constrained’. They’re instead constrained by specific skills and abilities. It’s nearly always clearer to talk about the specific needs of the field, ideally down to the level of specific profiles of people, rather than talent and funding in general.

For instance, work to positively shape the development of AI is highly constrained by the following:

Often when people talk about a field being ‘talent constrained’, they’re using that as shorthand for a more specific need.

Besides being less precise, talking about ‘talent constraints’ is confusing for a couple of reasons.

One reason is that it makes it sound like anyone who’s generally talented can contribute, whereas usually what’s most needed is a person with a specific set of characteristics.

Indeed, the more specific the profile, the harder it is to hire, and so the greater the bottleneck. The specificity of the need is often why there’s a ‘talent gap’ in the first place. So, the existence of a ‘talent gap’ may be evidence that most people can’t easily contribute.

Talking about ‘talent constraints’ also makes it sound like people who can’t get a job in the area aren’t talented, whereas it actually means they don’t have the specific characteristics needed. They might be extremely talented in other ways.

Rather than funding vs. talent gaps, we propose that people aim to identify specific bottlenecks facing the field and the skills needed to resolve them. A ‘bottleneck’ is the resource that a field most needs in order to make progress. In our recent article, we propose the following list of categories of bottleneck:

These bottlenecks can be made much more specific again, such as with the AI examples above. The ideal is to be as precise as possible about the characteristics needed.

There is a definition of ‘talent constrained’ that we and other organisations sometimes use:

An organisation is talent constrained when, for someone who could take (a reasonably important) job at that organisation, they would typically contribute more to that organisation by taking the job than earning to give.[fn 1]

This definition can be useful, but we suggest that when this more precise claim is what’s meant, people should just say that.

2. Skill bottlenecks are a matter of degree

As the above list of bottlenecks also shows, there are many ways to contribute to a field. We try to highlight the key bottlenecks facing a field because these are the most urgent to address, but most fields are somewhat constrained by many different inputs.

In particular, even in fields that are most constrained by a certain skill set, additional donations are usually still valuable, as well as other contributions.

This presents a difficult communication challenge, which we have not yet solved. If there’s a valuable direct work opportunity, we want to highlight that those who might be able to take it should strongly consider pursuing it over earning to give. But that doesn’t mean that earning to give is low impact, or that earning to give won’t be the top option for many people.

In general, we’re keen to see people ambitiously pursuing top direct work roles, but only if they might be a good fit, and have a good backup plan.

3. Skill and funding constraints are not opposites

All else equal, increasing funding generally increases the extent to which hiring is the bottleneck, and vice versa. So, there is some inverse relationship between funding and skill bottlenecks.

However, it’s possible for an organisation to be neither funding constrained nor skill constrained, and several organisations in our 2018 survey reported this.

This is because they might instead be mainly constrained by management capacity, coordination overheads, insights into their plan, etc.

An organisation can also be both constrained by funding and by making certain types of hires.

For example, imagine an organisation that has identified a simple, scalable model that requires little management overhead but lots of expensive staff who are hard to hire because they have a rare and difficult-to-measure skill set. This organisation might be constrained by both funding and ‘talent.’ By contrast, management capacity and insights do not seem like important bottlenecks.

Overall, it’s possible for an organisation to be mainly constrained by a certain skill, or funding, or both, or neither.

4. Skill bottlenecks vary by problem area

Each area has different needs, so faces different bottlenecks.

This is another reason why it can be misleading to talk about ‘the community being skill constrained’. It’s better to talk about specific problem areas or projects.

We try to highlight how our views depend on problem selection in our recent article and the survey. For instance, global health is significantly more funding constrained than global catastrophic risks, so earning to give is a relatively more attractive path if you’re focused on health -- though as per point 2, additional funding is useful in both.

5. Skill bottlenecks aren’t only about effective altruism organisations

People sometimes act as if the main alternative to earning to give is working at an ‘effective altruism non-profit’. However, this misses many types of high impact roles including those in academia, policy and relevant companies, which could absorb far more people. Our recent survey showed that roles in policy are highly valued, as are research positions that could be done within academia.

When people talk about how the community is ‘talent constrained’, they don’t just mean that the top priority is more people working at EA organisations. They often have these other roles in mind too.

6. High value of recent employees doesn’t necessarily entail that hiring is the top priority

Our recent survey showed that organisations were willing to pay a lot to retain recent hires. It’s natural to think that this implies the organisations are constrained by that type of talent, and that their key priority should be to hire more people. But this doesn’t follow.

Running a hiring round requires large costs, delays and risk, so the net benefit of hiring could be lower than the benefit of retaining existing staff.

It’s easy to conflate ‘skill bottlenecks’, ‘recent hires were worth a lot’ and ‘hiring is the top priority’, and these concepts are all closely connected, but they can diverge, so it’s clearer to separate them.

Read more.

7. The existence of skill bottlenecks doesn’t mean that it’s easy to get a job in the area

Even if hiring is the key bottleneck facing an organisation, it may still not be easy to get a job there. Skill constraints usually exist precisely because it’s very difficult for an organisation to find people to fill the skill set it needs. One reason this might happen is that they’re looking for a very specific skill set. This makes it unlikely that any given person will be a good fit. If people with the necessary skills were common, the organisation would probably be having an easier time hiring.

For instance, development of clean meat seems to be bottlenecked by outstanding scientists, but very few people can realistically take those jobs.

Again, it would be helpful if people communicated more specific skill profiles and indicators of fit that potential hires can use to gauge their chances.

Another way to see the problem is that a typical job application process only accepts 1-10% of applicants. This means that even if an organisation is 3-times keener to hire than average, its acceptance rate would still only be 3-30%, and most applicants will still not get the job.

In general, fit is both very important but hard to predict, so any given hiring process has a low chance of working out.

To help communicate this, it may be helpful if organisations published typical ratios of applicants to hires to let people plan accordingly.

However, we also often see people rule themselves out of jobs prematurely, and fail to apply even when they were actually a good fit. It’s unclear whether we and organisations in the community should be more or less encouraging of applications.

One way to solve this is to do more to encourage people to test out their fit. If a role might have a lot of impact, it’s often worth investigating more to see if it might work out for you. This is especially true if you can do it cheaply, such as by talking to a few people or making an application.

However, you should always do this while maintaining a good backup plan, or accumulating widely useful career capital. Because the most likely scenario is that it doesn’t work out, probably don’t quit your job or bet the farm.

8. Skill bottlenecks are not always solvable by raising salaries

A low salary is one reason why you might struggle to hire, but there are many other issues you can face, such as search costs or a long training process.

Raising salaries also sometimes has a limited ability to attract new staff, since people are motivated by many different things, of which salary is just one. If your potential employees are not highly money-motivated, you can increase salaries a lot for a small gain.

What’s more, the costs of raising salaries to attract new staff are often large because in order to be fair, you may also need to raise salaries for existing staff. For instance, if you have 10 equally-paid staff, and raise salaries 10% to attract one extra person, that final staff member effectively costs double the average previous salary.

Higher salaries can also attract people who do the job for the wrong reasons, can create PR issues, and so on.

This point is a particularly complicated debate. In general, if an organisation thinks that it’s constrained by a particular role, we think they should seriously consider raising salaries for that role. However, we don’t think this will fully solve the problem, and it isn’t always the right thing to do.

We asked organisations about why they don’t just raise salaries in our 2017 talent gaps survey.

9. Skill bottlenecks can change rapidly

We recently saw some positions that would have been hard to fill a couple of years ago get so many good applicants they couldn’t differentiate between the top ones. This suggests we may be making progress on finding the right people.

Organisations working on building effective altruism and reducing GCRs are also increasing their capacity to absorb new staff as they better work out their plans and establish management teams.

These factors could lead to an easing up of skill constraints in some key areas.

These recent changes also show how it can be difficult to make these kinds of assessments.

For instance, organisations often talk about their immediate constraints over the next year or so. But if you’re considering investing in your career capital, then your time horizon is probably more like 5-20 years. So, the key question is which bottlenecks will be most pressing over that entire period, which could be significantly different.

To help with this, in our last survey we also asked respondents to assess skill needs over 5 years.

The fact that key bottlenecks can change is also another reason why it’s important to always have a backup plan.

Conclusion

While we haven't successfully communicated it to this point, this more precise notion of skill constraints is a big part of the reasoning behind our current core career advice.

We believe many of our top problem areas are highly constrained by specific skill sets, as we outlined for AI safety earlier. What’s more, there was an important shift in this direction from 2014 as additional large funders entered these areas, especially the Open Philanthropy Project. This increase in funding created a spike in demand for certain key positions that couldn’t be quickly matched by an equivalent increase in people able to fill them. This led to a bottleneck of people with these types of skills, which persists today.

Today we usually recommend that people who are a good fit for filling these bottlenecks treat them as their first priority. This usually means initially considering relevant jobs in research, top non-profits and policy, and if you’re willing to consider something especially competitive, our list of priority paths.

In contrast, we rarely think that earning to give should be the top priority for people who could be a good fit for these other roles. This is another idea we hoped to highlight by talking about ‘talent constraints’.

However, we also recognize that our priority problems aren't ‘talent constrained' in general, and our priority paths require a fairly narrow set of skills. So, we continue to recommend building career capital and earning to give as a high impact option for people whose skills don't match the particular constraints currently faced by our priority problems.

In the future, we’ll aim to replace ‘talent constraints’ with a more specific phrase, which will usually be a claim about which specific skills are most needed.

What should people looking to choose a career actually do based on this? Here’s a quick summary of our process in practical terms:

  1. Make a list of potentially high-impact options: jobs that help relieve key bottlenecks in top problem areas (but not just at ‘EA orgs’).
  2. Work out if any specific jobs in these paths have the potential to be a good fit. If not, expand your options by considering different problem areas, different bottlenecks, or earning to give.
  3. Do cheap tests to learn more about fit. Even when a skill bottleneck exists, fit is both very important and hard to predict.
  4. If, after these tests, an option seems promising, try it out for a number of years.
  5. If it goes well, keep going. Otherwise try something else.
  6. Each option could easily fail to work out, so do all of this while maintaining a good backup plan.

7 comments

Comments sorted by top scores.

comment by Jon_Behar · 2019-05-13T15:43:56.347Z · score: 9 (3 votes) · EA(p) · GW(p)

I think this definition of talent constraints (found in clarification 1 of OP) lends itself to ambiguity and confusion:

“An organization is talent constrained when, for someone who could take (a reasonably important) job at that organization, they would typically contribute more to that organization by taking the job than earning to give… While we think this framing can sometimes be useful, it also has some problems. For example, this definition seems less useful when an organization’s best potential hires don’t have very high earning potential and wouldn’t be very good funders.”

In a post proposing we look at the EA labor market using traditional models, I suggest an alternative definition [EA · GW]:

An organization is talent constrained when it doesn’t have (and/or can’t hire reasonably easily) the people it needs despite offering competitive compensation.

I argue that while this definition...

“may not be perfect[1] [? · GW], but it has the advantages of being simple and applicable across organizations, causes, roles, and geographies. Perhaps most importantly, it’s an intuitive definition: if someone hears the term “talent constrained” it will likely conjure up images of an organization that doesn’t have the talent it needs, rather than an organization where (some) potential supporters would be more valuable as employees than donors.”
comment by David_Moss · 2018-11-14T19:17:58.866Z · score: 8 (5 votes) · EA(p) · GW(p)
What’s more, the costs of raising salaries to attract new staff are often large because in order to be fair, you may also need to raise salaries for existing staff. For instance, if you have 10 equally-paid staff, and raise salaries 10% to attract one extra person, that final staff member effectively costs double the average previous salary.

This also seems to have an impact on other orgs. I have lost count of the number of times I have heard people refer to "starting salaries at [large EA org with a >$1,000,000 annual budget]" as a baseline for salary expectations. This clearly has a disproportionately negative effect on smaller EA orgs or those trying to run more cheaply.

comment by Benjamin_Todd · 2018-11-20T02:19:56.961Z · score: 5 (3 votes) · EA(p) · GW(p)

That's a good point. We have worried about this in setting our own salary policy, but I forgot to mention it in the post. I've added to the original version. (Edit: Although we worried about this consideration, we didn't end up using it in setting our policy, and don't recommend that other orgs do. I've also removed the new sentence from the original version.)

comment by Gregory_Lewis · 2018-11-20T08:29:15.217Z · score: 11 (3 votes) · EA(p) · GW(p)

Although orgs tacitly colluding with one another to pay their staff less than they otherwise would may also have an adverse effect on recruitment and retention...

comment by Benjamin_Todd · 2018-11-20T09:06:42.432Z · score: 6 (3 votes) · EA(p) · GW(p)

Also agree, though presumably some fraction is zero-sum.

comment by SiebeRozendal · 2018-11-13T10:01:46.157Z · score: 6 (4 votes) · EA(p) · GW(p)
To help communicate this, it may be helpful if organisations published typical ratios of applicants to hires to let people plan accordingly.

This would be helpful indeed! Although I suspect organizations would be reluctant to do this as it might put off applications. On the other hand, it might turn off applications that would not be good anyway and save a lot of work. This data would also be useful for job seekers to decide whether they should apply mostly within EA, or focus efforts on impactful opportunities outside EA (like me). If there are 1,000 applicants for 50 positions yearly, I should not bet too much on getting into an EA org. I think this should especially encourage applicants who don't have the right skills or knowledge yet to apply outside EA to skill up, instead of hanging around in EA limbo.

I think it is worth thinking more about how the application processes of EA orgs can create more value (for the community). I think it is an opportunity for the organization to help the wider community: reviewing hundreds of applicants is a unique data set. One concrete action is publishing or sending general feedback about what differentiated top applicants from non-top applicants, which skills/backgrounds were in wide supply and which were rare and useful. These can also be referred to for later hiring processes for similar roles. There's more discussion about this topic in this Facebook post.

comment by Linch · 2018-11-18T00:01:19.582Z · score: 4 (3 votes) · EA(p) · GW(p)
it may be helpful if organisations published typical ratios of applicants to hires to let people plan accordingly

While in general I'm a big fan of more data, I worry that in this particular case it will shed more heat than light. I suspect that "ratios of applicants to hires" will be a really poor proxy for how competitive a position is. For example, Walmart in DC has something like a 2.6% acceptance rate.

Further, I don't think there's a good way to provide actually useful statistics on applicant quality in a privacy-conscious way. Eg, you can't just use the mean or the median. It doesn't matter if the median candidate has <1 year of ops job experience if the top 10 candidates all have 15+.