Quick BOTEC of person-hours spent on EA Job Applications per annum.
I created a Guesstimate model to estimate a total of ~14,000 to 100,000 person-hours or ~7 to 51 FTE are spent per year (90% CI). This comes to an estimated USD $ 320,000 to $3,200,000 unpaid labour time.
All assumptions for my calculations are in the Guesstimate
The distribution of effort spent by candidates is heavy-tailed; a small percentage of candidates may spend 3 to 10x more time than the median candidate.
I am not very good at interpreting the guesstimate, so if someone can state this better / more accurately than would be helpful
Keen to get feedback on whether I’ve over/underestimated any variables.
I’d expect this to grow at a rate of ~5-10% per year at least.
Sources: My own experience as a recruiter, applying to EA jobs and interviewing staff at some EA orgs.
Edited the unpaid labour time to reflect Linch’s suggestions.
I’ll adjust the estimate a bit higher.
In the Guesstimate I do discount the hours to say that 75% of the total hours are unpaid (trial week hours cone to 5% of the total hours).
I did not review the model, but only 75% of hours being unpaid seems much too low based on my experience having gone through the job hiring process (including later stages) with 10-15 EA orgs.
Okay, so I used a different method to estimate the total manhours and my new estimate is something like 60%. I basically assumed that 50% of Round 2 −4 in the application process is paid, and 100% of the work trial.
I expect that established / longtermist orgs are disproportionately likely to pay for work tests, compared to new or animal / GH&D orgs.
I think Josh was claiming that 75% was “too low”, as in the total % of unpaid hours being more like 90% or something.
When I applied to a bunch of jobs, I was paid for ~30 of the ~80 hours I spent (not counting a long CEA work trial — if you include that, it’s more like 80 out of 130 hours). If you average Josh and I, maybe you get back to an average of 75%?
*****
This isn’t part of your calculation, but I wonder what fraction of unique applicants to EA jobs have any connection to the EA community beyond applying for one job?
In my experience trying to hire for one role with ~200 applicants, ~1/3 of them neither had any connection to EA in their resumes nor provided further information in their applications about what drew them to EA. This doesn’t mean there wasn’t some connection, but a lot of people just seemed to be looking for any job they could find. (The role was more generic than some and required no prior EA experience, so maybe drew a higher fraction of outside applicants.)
Someone having no other connection to the EA community doesn’t mean we should ignore the value of their time, and the people who apply to the most jobs are likely to have the strongest connections, so this factor may not be too important, but it could bear consideration for a more in-depth analysis.
Quick BOTEC of person-hours spent on EA Job Applications per annum.
I created a Guesstimate model to estimate a total of ~14,000 to 100,000 person-hours or ~7 to 51 FTE are spent per year (90% CI). This comes to an estimated USD $ 320,000 to $3,200,000 unpaid labour time.
All assumptions for my calculations are in the Guesstimate
The distribution of effort spent by candidates is heavy-tailed; a small percentage of candidates may spend 3 to 10x more time than the median candidate.
I am not very good at interpreting the guesstimate, so if someone can state this better / more accurately than would be helpful
Keen to get feedback on whether I’ve over/underestimated any variables.
I’d expect this to grow at a rate of ~5-10% per year at least.
Sources: My own experience as a recruiter, applying to EA jobs and interviewing staff at some EA orgs.
Edited the unpaid labour time to reflect Linch’s suggestions.
I think
As a normal distribution between $20-30 is too low, many EA applicants counterfactually have upper middle class professional jobs in the US.
I also want to flag that you are assuming that the time is
but many EA orgs do in fact pay for work trials. “trial week” especially should almost always be paid.
Hi Linch, thanks for the input!
I’ll adjust the estimate a bit higher. In the Guesstimate I do discount the hours to say that 75% of the total hours are unpaid (trial week hours cone to 5% of the total hours).
I did not review the model, but only 75% of hours being unpaid seems much too low based on my experience having gone through the job hiring process (including later stages) with 10-15 EA orgs.
Okay, so I used a different method to estimate the total manhours and my new estimate is something like 60%. I basically assumed that 50% of Round 2 −4 in the application process is paid, and 100% of the work trial.
I expect that established / longtermist orgs are disproportionately likely to pay for work tests, compared to new or animal / GH&D orgs.
I think Josh was claiming that 75% was “too low”, as in the total % of unpaid hours being more like 90% or something.
When I applied to a bunch of jobs, I was paid for ~30 of the ~80 hours I spent (not counting a long CEA work trial — if you include that, it’s more like 80 out of 130 hours). If you average Josh and I, maybe you get back to an average of 75%?
*****
This isn’t part of your calculation, but I wonder what fraction of unique applicants to EA jobs have any connection to the EA community beyond applying for one job?
In my experience trying to hire for one role with ~200 applicants, ~1/3 of them neither had any connection to EA in their resumes nor provided further information in their applications about what drew them to EA. This doesn’t mean there wasn’t some connection, but a lot of people just seemed to be looking for any job they could find. (The role was more generic than some and required no prior EA experience, so maybe drew a higher fraction of outside applicants.)
Someone having no other connection to the EA community doesn’t mean we should ignore the value of their time, and the people who apply to the most jobs are likely to have the strongest connections, so this factor may not be too important, but it could bear consideration for a more in-depth analysis.