Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: EA Survey 2020 Series: Donation Data, published by david reinstein on October 26, 2021 on The Effective Altruism Forum.
Introduction and summary
Charitable donation (and earning-to-give) has been, and continues to be a prominent, prevalent, and impactful component of the Effective Altruism movement. The EA Survey has been distributed between 2014 and 2020, at roughly 15 month intervals. As a result, surveys were released at various points in the year, ranging from April to August, and no survey was released in 2016. In each survey we asked EAs about their charitable donations in the previous year, and their predicted donations for the year of the survey. Our work in this post/section reports on the 2020 survey (2019 giving), but our analysis extends to all the years of the EA survey.
In this post (and the accompanying bookdown supplement chapter), we consider donation responses, presenting both raw numbers, and descriptive, predictive, and causally-suggestive analysis. We present simple numbers, statistical comparisons, vizualisations, and descriptive and 'predictive' (machine learning) models. We cover a range of topics and concerns, including:
the total magnitude of EA giving and its relationship to non-EA giving,
career paths and 'earning to give',
the broad relationship between EA giving and individual characteristics (such as employment status and country, and income),
donations versus income trends across recent years,
which causes EAs are donating to, and
EA's donation plans versus realized donations (and future plans).
Our modeling work work considers how donations (total, share-of-income, and 'donated over 1000 USD') jointly relates to a range of characteristics. We first present 'descriptive' results focusing on a key set of observable features of interest, particularly demographics, employment and careers, and the 'continuous features' age, time-in-EA, income, and year of survey. We next fit 'predictive', allowing the 'machine learning' models themselves to choose which features seem to be most important for predicting donations.
Note: A 'dynamic version' of this document (an R-markdown/Bookdown), with folded code, margin notes, some interactive graphs and tables, and some additional details, can be found here. This may be helpful for anyone that wants to dig into this more deeply, and perhaps for those who are data, code, and statistics-inclined.
Note: In the narrative below, we simply refer to "donations" rather than "reported donations" for brevity. Unless otherwise mentioned, all figures simply add, average, or otherwise summarize individual responses from the EA Survey years mentioned.[1]
Summary (some key results and numbers)
55.5% of EAs in the 2020 survey reported making a charitable donation in 2019, 13.7% reported making zero donations, and 30.8% did not respond to this question. (Thus, of those who responded, 80.3% reported making a donation in the prior year.)
Participants reported total donations of 10,695,926 USD in 2019 (cf 16.1M USD in 2018).
However, the number of survey participants has declined somewhat, from 2509 in 2019 (1704 of whom answered the donation question) to 2056 (1423 answering the donation question) in 2020.
Over the past years, we see no strong trend in median or mean donation amounts reported.[2]
The median annual donation in 2019 was 528 USD (cf 683.92 USD in 2018).
The mean (reported) annual donation for 2019 was 7,516 USD (cf 9,370 for 2018) or 8,607 USD excluding those who joined in 2020 (cf 10,246 USD for 2018 excluding those who joined in 2019).
The median annual donation in 2019 excluding those who joined EA in 2020 was 761 USD (cf. 990 USD for the comparable median for 2018/2019 and 832 USD for 2017/2018). (See 'donation and income trends in EA' for more details).
In 2019 1.3% of donors accounted for $6,437,404 in...