Home > Misc

Effective altruism survey results

The results of the 2014 Survey of Effective Altruists were released a few days ago. There's discussion at the link as to the worthiness of this exercise; we have no way of knowing how representative the survey takers are of the EA community as a whole. Whether or not the time cost involved is worth the sort of indicative-only results that we get, I don't know. But since we do have those results, I think we may as well have a look at them, always keeping in mind the sampling caveats.

Below is an interface for plotting bar charts and cross-tables from the raw data on Github. (Well, almost all the data – see details below.) Samples may be unrepresentative and sample sizes of subgroups can be tiny, so interpret with caution. Money amounts are in US dollars. Several questions allowed multiple responses – you can check if this is the case by making both primary and secondary categories the same: if it was a single-response question, then there will only be one bar in each group.

I've extracted most of the data that I think is interesting from respondents who said that they could, "however loosely, be described as 'an EA'". My sample is N=809; there are four more self-described EA's from the ACE survey that I've ignored out of laziness (the ACE file has some different columns in it). The first 17 entries of imdata.csv have some columns mixed up; I've removed almost all data in the columns that are deifnitely mixed up (one cell was easy enough to classify). It's possible that more columns are wrong than I realise, and that I'm introducing a bit of extra unknown noise into the results.

I haven't included lifetime donations (sorry – I know that this is an interesting part of the survey to study); income and donations refer to 2013 only. I tidied up the donation and income fields, and converted all currencies to USD, independently of the survey team. As a sanity check on this work (which I did quickly), I get a median donation of $450 and a mean of $9054; the survey analysis has $450 and $8905.90 – close enough!

I apologise for all the abbreviations; hopefully most will be clear from context (earlier this weekend I thought I'd be displaying the x-axis labels horizontally, and it's too late on Sunday to change them all now). LR is Leverage Research; PHC is Project Healthy Children.

The bands I've created for some variables are inclusive at the lower bound and exclusive at the upper bound. So, e.g., someone who donated 10% of their income goes in the "≥10%" band and not in the "5-10%" band.

The sample sizes won't always be known from the interface. You can get the total count for primary and secondary categories, but if you ask to plot, say, the median amount donated, then the sample size will likely go down: not all survey respondents entered a value for their donations.

Posted 2015-03-22, updated 2015-04-17 (changed colour wheel).

Home > Misc