We can assess this empirically, to some extent, by looking at changes in respondents who we can track over time (i.e. those who logged in to their EA account to pull their previous responses or provided their email). This allows us both to compare both how many individuals with different political views in 2022 changed to different views in 2024 and how many people with different views in 2022 may have dropped out (because we can’t track them in 2024[1]).
TLDR:
Left respondents switch in quite large numbers to the Center-left, Center leftists show a smaller switch to the center, while Centrists showed no major shift.
We see no signs of Leftists or Center leftists being more likely to dropout, but some weak evidence of Right and Center rightists being more likely to dropout.
You can see the total flows between categories on this Sankey diagram.[2]
Due to the low numbers, it can be hard to compare the results for different categories though.
First, we can compare how many people switched to NA (i.e. were not tracked in 2024 who were in 2022).
We can see that there’s very little difference between the Left (77%), Center Left (73%) and Center (74%). We do, however, see that Center right (85%) and Right (83%) respondents seem more likely to not appear in the 2024 dataset, though these are small groups due to the very low number of right-of-center respondents.[3]
Switch to NA
Left
NA
785
77.49%
Center left
NA
788
72.76%
Center
NA
155
73.81%
Center right
NA
52
85.25%
Right
NA
15
83.33%
Libertarian
NA
109
71.71%
Other
NA
144
75.39%
Prefer not to answer
NA
165
83.76%
Did not answer
NA
74
85.06%
Did not view
NA
182
97.33%
Then, we can compare changes in political views, setting aside the large number of NA-in-2024 responses, to see the changes more clearly. We can see quite a strong shift (27%) among Left (in 2022) respondents to the Center left (this compares to the total drop in Left respondents being 28.8%). Among the center left, we see a smaller but still noticeable switch to the Center, and for the Center, we see most respondents staying the same.[4]
Non-NA switches
Left
Left
154
67.54%
Left
Center left
61
26.75%
Left
Center
4
1.75%
Left
Center right
1
0.44%
Left
Right
0
0.00%
Left
Libertarian
1
0.44%
Left
Other
7
3.07%
228
Center left
Left
7
2.40%
Center left
Center left
246
84.25%
Center left
Center
32
10.96%
Center left
Center right
2
0.68%
Center left
Right
0
0.00%
Center left
Libertarian
3
1.03%
Center left
Other
2
0.68%
292
Center
Left
1
1.85%
Center
Center left
1
1.85%
Center
Center
46
85.19%
Center
Center right
3
5.56%
Center
Right
0
0.00%
Center
Libertarian
3
5.56%
Center
Other
0
0.00%
54
These results aren’t suggestive of leftists being particularly likely to drop out. But there is some evidence of Left respondents in 2022 switching in quite large numbers (similar to the total change in % Left) to the Center left in 2024.
Caveat: a respondent might not appear in 2024 because they dropped out of EA or they didn’t take the survey or because they took the survey but didn’t log in or provide the same email address as last time. Differential disappearance across groups might still be suggestive, however, since seem to be few innocent explanations for why leftists/rightists would systematically become less likely to take the survey or provide their email address across time.
It’s also important to bear in mind that the sub-sample of respondents who logged in or provided their email address might differ from the total sample (e.g. they might be more engaged or more satisfied). But just over 90% of respondents in both 2022 and 2024 fit into this category, so it does not seem a very selective sub-sample.
NAs in 2022 include only people who did not reach the politics question of the survey, because this sub-sample includes only people who were tracked in the 2022 survey.
It makes sense that the various other categories for not answering the question would also show high rates of not answering the email question or not logging in.
Center right and Right not shown because, excluding NAs in 2024, this was only 9 and 3 respondents respectively (though most stayed in the their prior category).
We can assess this empirically, to some extent, by looking at changes in respondents who we can track over time (i.e. those who logged in to their EA account to pull their previous responses or provided their email). This allows us both to compare both how many individuals with different political views in 2022 changed to different views in 2024 and how many people with different views in 2022 may have dropped out (because we can’t track them in 2024[1]).
TLDR:
Left respondents switch in quite large numbers to the Center-left, Center leftists show a smaller switch to the center, while Centrists showed no major shift.
We see no signs of Leftists or Center leftists being more likely to dropout, but some weak evidence of Right and Center rightists being more likely to dropout.
You can see the total flows between categories on this Sankey diagram.[2]
Due to the low numbers, it can be hard to compare the results for different categories though.
First, we can compare how many people switched to NA (i.e. were not tracked in 2024 who were in 2022).
We can see that there’s very little difference between the Left (77%), Center Left (73%) and Center (74%). We do, however, see that Center right (85%) and Right (83%) respondents seem more likely to not appear in the 2024 dataset, though these are small groups due to the very low number of right-of-center respondents.[3]
Then, we can compare changes in political views, setting aside the large number of NA-in-2024 responses, to see the changes more clearly. We can see quite a strong shift (27%) among Left (in 2022) respondents to the Center left (this compares to the total drop in Left respondents being 28.8%). Among the center left, we see a smaller but still noticeable switch to the Center, and for the Center, we see most respondents staying the same.[4]
These results aren’t suggestive of leftists being particularly likely to drop out. But there is some evidence of Left respondents in 2022 switching in quite large numbers (similar to the total change in % Left) to the Center left in 2024.
Caveat: a respondent might not appear in 2024 because they dropped out of EA or they didn’t take the survey or because they took the survey but didn’t log in or provide the same email address as last time. Differential disappearance across groups might still be suggestive, however, since seem to be few innocent explanations for why leftists/rightists would systematically become less likely to take the survey or provide their email address across time.
It’s also important to bear in mind that the sub-sample of respondents who logged in or provided their email address might differ from the total sample (e.g. they might be more engaged or more satisfied). But just over 90% of respondents in both 2022 and 2024 fit into this category, so it does not seem a very selective sub-sample.
NAs in 2022 include only people who did not reach the politics question of the survey, because this sub-sample includes only people who were tracked in the 2022 survey.
It makes sense that the various other categories for not answering the question would also show high rates of not answering the email question or not logging in.
Center right and Right not shown because, excluding NAs in 2024, this was only 9 and 3 respondents respectively (though most stayed in the their prior category).