Extended Data Fig. 8: An overview of the inferred countries of email addresses to which we sent invitations, comparing those who responded with those who did not.

Our survey did not ask about participant’s geographic locations, and generic email domains such as ‘gmail.com’ do not license inferences about location. However, for university (or other institutional) email domains we are generally able to infer what countries those institutions are based in. We were able to do so for 6421 respondents so, even if this is not full coverage, it represents the bulk of the survey data. (a) A comparison of the log counts of those who responded to the survey and those who did not, with a linear fit (predicted log mean, with 95% CIs); (b) A comparison of the proportion represented by each country (as a percentage), either as a proportion of our survey respondents (blue) or as a proportion of invitation emails that did not receive a response (red); (c) A histogram of the difference between each country’s proportions from panel b, subtracting the proportion who did not respond from the proportion that did. For example, the largest value in the histogram is a difference of 8.36% percentage points representing the USA because 48.28% of our survey respondents had email domains in the US whereas 39.92% of emails that were invited but did not respond had the same. Given these indications that there is a good match in respondent vs non-respondent countries apart from the USA, which is somewhat over-represented, we believe our sampling strategy reached a wide range of participants (even if it is not a representative sample of the global academic psychology community).