How can I tell whether my fundraising project is harmfully cannibalizing existing donations v.s. actually inspiring new people to donate?
Currently, I am organizing a speaker series to raise money for Peter Singer’s charity The Life You Can Save. To make an experiment out of it, I structured the event as a blind auction. People donate to the charity and email me the receipt. On the day of the event, I release invitations to the top-thousand most generous donors.
So far, I’m really pleased with the response! I feel like this approach is inducing people to donate, which is exciting. On the other hand, it has led me to ask myself (1) to what extent do unique fundraisers induce more aggregate donating, as opposed to simply shifting funds from alternative charities; and (2) to the extent the donations I’m receiving receipts for are simply people shifting a donation they would have already made to another charity, is there a way to approximate which charities are being negatively affected? These seem like important questions for assessing the effectiveness of fundraising.
I’m sure I’m not the first person to ask these questions. I found a few relevant studies, including:
This randomized controlled trial done by Facebook that found that Save the Children ads increased Save the Children donations, but reduced donations to other charities.
This study that found zip codes near tornado events donated just as much to charity as other zip codes, suggesting minimal crowdout.
Several experimental lab studies that seem to suggest crowdout (here and here).
I take from this that crowdout is a very real concern. But from my quick reading, I had trouble deriving good rules-of-thumb for quantifying crowd-out. So I thought I’d ask here for anyone had any suggestions.
It is worth noting that we are interested in maximising the positive impact achieved by the donations, which is the product between the amount donated and cost-effectiveness of the donation. Even if the amount donated is constant, differences in cost-effectiveness could imply that a given donation is much better than other. So, it is quite important to understand not only how much people would donate otherwise, but also to which organisations they would give.
In practice, if the counterfactual donation would go to the median charity, I would say it is safe to assume that directing that donation to organisations aligned with effective altruism is quite good (especially because the distribution of cost-effectiveness is heavy-tailed).
The above may suggest that, for fundraising projects which mainly attract people already engaged with effective altruism, the additional impact is small. However, I think organisations like Giving What We Can may still be quite effective via increasing awareness for effective altruism. That being said, I would be quite interested in seeing more (public) analyses of the multiplier effect of such organisation and other fundraising projects (or effective giving organisations like Ge Effektivt).
Answering quickly and informally for Ge Effektivt (sorry, I hadn’t seen this earlier).
When I used something like 1h to try to estimate the counterfactuality of our donations I ended up around 50%. This is removing users who say they found us through Effective Altruism Sweden, Giving What We Can pledgers (although some signed up because of us), people who found us through webpages of charities we donate to, people I know are in the EA Sweden network (some of who I think give more than they otherwise would), and then adding some on top of that. I’m open to suggestions on how to get even closer to the right answer, without spending too much time.
We’re raising a total of ~$700k on a ~$105k budget this year, so in the end, I think our counterfactual multiplier might be around 2-3x.
With that said, I expect the next 12 months to include a more in-depth analysis of counterfactual funds raised for us and similar sites. I am looking forward to sharing the results with you!
Thanks for the feedback.
Spending 1 h on assessing the counterfactual seems very little for such an important component of the theory of change of multiplier organisations. Some models you may find useful for inspiration are this one from the Effective Altruism Foundation, and this one from RC Forward (mentioned here).
I had/have an academic research agenda on this very question. I’d summarize the evidence as
We expect crowding-out, theoretically and intuitively
The evidence is not solid, and it is very hard to measure
I have a more detailed writeup as part of the ‘synthesis of barriers to effective giving’; chapter ‘Are charities in competition?’
Will try to return to this discussion in more detail if I have a chance. (Also had plans/progress on a synthesis and meta-analysis paper on this, but not sure when if ever I’ll have a chance to return to it.)
My guess is that there will be a chance of crowd-out and for givers who are at their hard limits on their charitable giving the crowd-out will be 100%. I would not expect a very good rule of thumb to exist.
That said, there may be things you can do to minimize it (just off the top of my head):
Make your request unique so that it doesn’t bring to mind other places people are donating to
Make a case that giving to your cause can make other giving more effective
Make your request at times where people get fewer requests (e.g., not around the holidays)
One time requests tied to a specific event like a natural disaster are more easily absorbed without offsetting reductions
Giving small perks people might value a lot might help them slot this giving separately from other forms of giving and thus not feel the need to offset it
A very good question, and a nice and thoughtful answer.
I think this has been an ongoing debate on donation matching and giving multipliers
I’d just add that this is one of the reasons why neglectedness is my favorite (and neglected) ITN metric—it’s hard to crowd-out resources for neglected causes.