what is your best guess for the large discrepancy between how many people say they are influenced by the match and the numbers you were finding?
Couple hypotheses:
Personally, if information is available to me, my default assumption is that it influenced my behavior.
Also, if you view yourself as a rational agent, then the donation match should clearly affect your behavior, so the ego-preserving answer is “yes.”
Also, a ~20% increase in revenue is consistent with 50% of people responding to the match, if half of those (25% of all donors) donated a small amount but would not have donated at all otherwise, and the other half went from the average to 1.5x the average.
A secondary thought is that I think I would expect the effect of the match to be stronger on social media due to the eye-catching effect.
Yeah, unfortunately I couldn’t find any field experiments on social media campaigns. It seems pretty hard to study. Actually, Charity Science might be in an ideal place to look at this with their birthday/Christmas fundraisers...
I couldn’t find any field experiments on social media campaigns. It seems pretty hard to study. Actually, Charity Science might be in an ideal place to look at this with their birthday/Christmas fundraisers...
We haven’t directly tested matching scientifically yet, though we’d definitely like to. Our closest is in the write-up you mention “The Power of Donation Matching” where we anecdotally observed a time-ordered effect of introducing “live” donation matching (though this could be due to other factors, and isn’t evidence of matching overall).
Yeah. If it turns out that the true effect is smaller than you observed in that post (which it may well not be—lots of heterogeneity), my guess would be that the spike was partly caused by some interaction with people promoting it at the same time. But again, it’s super hard to tell without randomization, because the effects vary so much.
Yeah, although you’d have to be careful to account for whether people who get matching promote their campaign more, and you’d probably need a fairly large sample of fundraisers because there’s a lot of variance in amount raised. It seems like a fairly tricky study to do.
(Out of curiosity, do you have a list of fundraiser totals anywhere? That would help calculate the study power.)
Couple hypotheses:
Personally, if information is available to me, my default assumption is that it influenced my behavior.
Also, if you view yourself as a rational agent, then the donation match should clearly affect your behavior, so the ego-preserving answer is “yes.”
Also, a ~20% increase in revenue is consistent with 50% of people responding to the match, if half of those (25% of all donors) donated a small amount but would not have donated at all otherwise, and the other half went from the average to 1.5x the average.
Yeah, unfortunately I couldn’t find any field experiments on social media campaigns. It seems pretty hard to study. Actually, Charity Science might be in an ideal place to look at this with their birthday/Christmas fundraisers...
We haven’t directly tested matching scientifically yet, though we’d definitely like to. Our closest is in the write-up you mention “The Power of Donation Matching” where we anecdotally observed a time-ordered effect of introducing “live” donation matching (though this could be due to other factors, and isn’t evidence of matching overall).
Yeah. If it turns out that the true effect is smaller than you observed in that post (which it may well not be—lots of heterogeneity), my guess would be that the spike was partly caused by some interaction with people promoting it at the same time. But again, it’s super hard to tell without randomization, because the effects vary so much.
I quite like the idea of randomizing who gets matching funding on birthdays and seeing if we can pick up any differences between the two groups.
Yeah, although you’d have to be careful to account for whether people who get matching promote their campaign more, and you’d probably need a fairly large sample of fundraisers because there’s a lot of variance in amount raised. It seems like a fairly tricky study to do.
(Out of curiosity, do you have a list of fundraiser totals anywhere? That would help calculate the study power.)