- Since 2020, the percentage of women in our sample has increased (26.5% vs 29.3%) and the percentage of men decreased (70.5% vs 66.2%).
- More recent cohorts of EAs have lower percentages of men than earlier cohorts. This pattern is compatible with either increased recruitment of women, non-binary or other non-male EAs in more recent years and/or men being more likely to drop out of EA.
- We examine differences between cohorts across years and find no evidence of significant differences in dropout between men and women.
- The percentage of white respondents in our sample (76.26%) has remained fairly flat over time.
- More recent cohorts contain lower percentages of white respondents (compatible with either increased recruitment and/or lower dropout of non-white respondents).
- We also examine differences between cohorts across years for race/ethnicity, but do not find a consistent pattern.
- The average age at which people first get involved in EA (26) has continued to increase.
- Education and employment
- The percentage of students in the movement has decreased since 2020 and the percentage in employment has increased. However, just over 40% of those who joined EA in the last year were students.
- 11.8% of respondents attended the top 10 (QS) ranked universities globally.
- Career strategies
- The most commonly cited strategy for impact in one’s career was ‘research’ (20.61%) followed by ‘still deciding’ (19.63%).
- More than twice as many respondents selected research as selected ‘earning to give’ (10.24%), organization-building skills (ops, management), government and policy, entrepreneurship or community building (<10% each).
- Men were significantly more likely to select research and significantly less likely to select organization-building skills. We found no significant differences by race/ethnicity.
- Highly engaged EAs were much more likely to select research (25.0% vs 15.1%) and much less likely to select earning to give (5.7% vs 15.7%).
- Respondents continue to be strongly left-leaning politically (76.6% vs 2.9% right-leaning).
- Our 2022 sample was slightly more left-leaning than in 2019.
- A large majority of respondents (79.81%) were atheist, agnostic or non-religious (similar to 2019).
3567 respondents completed the 2022 EA Survey.
A recurring observation in previous surveys is that the community is relatively lacking in demographic diversity on the dimensions of gender, age, race/ethnicity, and nationality. In this report, we examine the demographic composition of the community, how it has changed over time, and how this is related to different outcomes.
In a forthcoming follow-up survey, we will also be examining experiences related to gender and community satisfaction in more detail.
The percentage of women has slightly increased since 2020 (26.5% to 29.3%), while the percentage of men has slightly decreased (70.5% to 66.2%).
Gender across survey years
Looking across different survey years, we can see that there is now a higher percentage of women in our sample than in the earliest years. In the earliest EA Surveys, we saw just over 75% men, whereas in the most recent survey, we see just over 65%.
Gender across cohorts
Looking across cohorts (EAs who reported first getting involved in a given year), we see that more recent cohorts contain more women than men. This is compatible with either/both increased recruitment of women (or decreased recruitment of men) or disproportionate attrition of women over time (see below for more analysis on this).
Gender across cohort by year
We can also look at the gender composition of different cohorts (years people joined EA) within surveys across different years.
This shows us the same pattern across cohorts as mentioned above within previous years’ surveys (i.e. there are more women and fewer men in more recent cohorts).
However, it also allows us to compare cohorts across different surveys. This could potentially point to signs of differential attrition. If, for example, non-male respondents were more likely to drop out of EA than men, then within cohorts we should see an increase in the percentage of men, i.e. the line for 2020 would be lower than the line for 2022 and so on.
It may be tempting to see signs of a gap between earlier and later surveys in the more recent cohorts on the right hand side (the cohorts on the left hand sides are particularly noisy due to smaller numbers of respondents). However, examining the differences between cohorts in more detail in the plot below, there seems no consistent or significant trend in this direction.
Our sample remains majority white, with relatively little variation among the other subgroups compared to previous years.
Race/ethnicity by year
We observe a slight decrease in the proportion of respondents identifying as white compared to the earliest years, though the percentages are relatively flat after 2017-2019. As in previous years, in our forthcoming post in this series, we will highlight positive and negative experiences of the community across different groups within the community.
Race/ethnicity across cohorts
The trend across cohorts is less clear for race/ethnicity than for gender, however, there appears to be a slight reduction in the percentage of white respondents among those who joined EA more recently. As noted above, this is compatible either with increasing recruitment of non-white respondents in these cohorts or with higher attrition of non-white respondents.
Race/ethnicity by cohort by year
Comparing each cohort across survey years, we see relatively little difference in the percentage of white respondents (this is confirmed by examining the second plot below) and no significant differences aside from EAS 2017 (the earliest survey analysed here) having more white respondents.
The age of respondents remains disproportionately young, though, as shown in the section below, the average age is increasing with time (reaching 29 in 2022 up from 25 in 2014).
Changes in age over time
Age when first getting involved in EA
The median age when respondents reported first getting involved in EA was 24 (with a slightly higher mean of 26.49).
Age of first getting involved by cohort
The average age at which people first got involved in EA has also continued to increase (now at median 26). As we commented last year, this is somewhat older than the age of the typical college student (though a large minority (41.13%) of those who joined EA within the last year were students).
Careers and Education
Despite being composed of a relatively large number of students, a majority of respondents (just over half) are in full time employment.
Employment/student status by year
Here we can only compare the results to those from EAS 2020 (so that the categories are comparable). We observe that the percentage of students in the movement has decreased and the percentage in employment has increased.
Perhaps unsurprisingly, the career area that is the focus for the largest number of EAs is research, although a large number are still deciding. However, it may still be striking that there are more than twice as many EAs interested in a career in research as in any one of earning to give, government and policy, entrepreneurship or community building.
Career strategy by gender
These results show significant differences in the proportions of respondents of different genders following career paths in research and organization building, with more men in the former and fewer men in the latter.
Career strategy by race/ethnicity
We observe no statistically significant differences in career paths of white and non-white respondents.
Career strategy by engagement level
There were large differences in the career strategy of low/high engagement EAs. Many more highly engaged EAs were focused on research and community building, while many more less engaged EAs were focusing on earning to give or not focusing their career on impact right now.
A disproportionate number of EAs attend highly (QS) ranked universities. This first plot shows the number of EAs having attended the most commonly attended universities in our sample which are mostly quite highly ranked.
This second plot shows the number of EAs attending the top 25 ranked universities. Notably a number of the globally top-ranked universities which have low numbers of EAs are not in majority English-speaking countries.
This final plot shows the number of EAs having attended each university plotted in order of university ranking, showing a clear general trend towards more EAs at higher ranked universities.
Finally, we show the number of EAs having attended universities of different ranks in broader bins. We can see that, though the EA community does have a disproportionate number of members who have attended highly ranked universities, this is far from a majority. While 11.8% have attended top 10 universities, 18.4% have attended top 25 ranked universities and 38% have attended top 100 ranked universities globally, 61.3% attend universities ranked outside the global top 250.
2022 is the first year that we have included politics and religion in the EA Survey, since 2019, after skipping them to save space.
The EA community remains majority left-leaning (76.8%) with very small numbers of respondents identifying as right-leaning (2.9%).
The political composition of the EA movement has changed little since 2019, though is slightly more left-leaning.
Politics by year
Our sample remains highly atheist, agnostic or non-religious and this has not changed significantly since 2019.
Religion by year
We changed a number of plots (as described in the comments) for greater clarity and consistency and updated the plots for race/ethnicity to correct an unrelated error.
This post was written by Willem Sleegers and David Moss. We would also like to thank Rachel Norman and Peter Wildeford for comments.
The gender question was based on self-selection from categories listed in the plot below, with an option to self-describe. As Slade et al (2020) note, there are advantages and disadvantages to employing fixed categories for respondents to select from and to exclusively providing an open comment box to allow respondents to write in their own identification. The latter approach provides maximum freedom for respondents to provide the answer they see fit. However, providing an open comment box also means that analysis of these responses will be determined by the researchers’ own categorization of these responses into fixed categories which the respondents will be unable to access. Providing fixed categories upfront, in contrast, has the advantage of transparency, with respondents knowing how their answers will be categorized. Here, we opted to allow a ‘self-describe’ option, but do not attempt to reclassify these responses into other categories (as in previous years, this category accounts for a very small number of responses, and many of the responses would be very hard to re-classify).
As there are very few responses in either the ‘Non-binary’ or ‘Prefer to self-describe’ categories, when analyzing group differences, we simply analyzed differences between respondents who selected ‘Man’ and those who selected anything other than ‘Man’, as the low number of respondents in the other categories would preclude meaningful analysis. Although this approach means that we cannot assess differences between those who selected ‘Woman’, ‘Non-binary’ or a self-description, this seems like a practically useful categorisation, given that the EA community is predominantly male and given that prior community-building efforts have focused on both women and non-binary people under-represented groups. We opted to include those who selected ‘Prefer to self-describe’ in the ‘selected something other than “Man”’ category in order to avoid simply excluding these respondents. It’s important to note that inclusion in the “Selected something other than “Man”” category does not necessarily mean that the respondent would not identify as a “Man” if they selected “Prefer to self-describe”.
Suppose 50 men and 50 women join EA in 2018, 2019 and 2020. If 10 women and 5 men drop out each year, then when we survey the community in 2020, then we would observe relatively fewer women in 2018 and 2019 than in 2020.
Surveying race and ethnicity in a cross-cultural international context is fraught due to a number of theoretical complications (Roth, 2017). Not only do the specific categories utilized vary internationally, but whether these are understood as about race, ethnicity, nationality or some other superordinate category can itself vary (Morning, 2015). In this case, similar to the gender question, respondents were able to select multiple different categories out of a fixed list or to self-describe. The categories in this survey are designed to match those used by CEA in their surveys (which are based on the US/UK censuses). As with the gender analyses, for those analyses which look at differences between groups, due to the low number of non-white respondents, we combine these into a single category.
The one obvious exception is that there is a lower percentage of white respondents in EAS 2017 across most cohorts, compared to all later years. This pattern is compatible with differential attrition of non-white respondents between 2017 and 2018, and for no later years. That said, we think it is more likely explained by EAS 2017 being a strange outlier year. We’ve found EAS 2017 data to behave strangely compared to other survey years in a variety of cases e.g. here (note: this author was not involved in running EAS 2017. And it seems relatively unlikely that there would be a large difference in attrition between 2017 and 2018 and not for any year between 2018 and 2022.
We categorized as “high” engagement those who were 4-5/5 on this scale and “low” engagement, anyone who selected 1-3/5 on that scale.