Ryan Kaldari wrote:
... There's definitely a lot of work that we need help with, so any assistance is appreciated!
What can I help with to prepare for experimental measurements? Do you already have a way to collect arbitrary radio button and checkmark form responses from your PayPal donations? What format do those get loged in? I would love to write an R script for doing the regression.
Do you think this sort of thing would work better with radio buttons like the Red Cross uses, or a set of checkmarks with language specifying that the funds would be earmarked in equal proportions between all the checked options -- or is that another independent variable which should be tested?
Have you looked in to http://www.wepay.com? They are supposed to be offering a lower overhead rate than PayPal. I know you have an account with moneybookers.com -- have you asked them all for a better deal from each of them to get some competition between them going?
Aryeh Gregor wrote:
On Fri, Sep 3, 2010 at 6:11 PM, James Salsman <jsalsman at gmail.com> wrote:
Absolutely; a multivariate linear regression test to determine the extent to which each of the earmark options tends to maximize total contributions should be run in advance, with a sample size (assuming 30 earmark possibilities offered four at a time in a variety of different languages and locales) of between 5000 and 30,000 donations.
Is that a practical number?
According to http://wikimediafoundation.org/wiki/Special:FundraiserStatistics there were more than 3,600 contributions on the second day of the 2008 fundraiser. During the first few days the independent variables should be presented in random permutations. After you've collected enough data for your desired confidence level (I used 95%) then you can start sorting them. But if you want to use a lower level of confidence you can vastly reduce the number of initial observations. If you want to use a 90% confidence level for 30 independent variables then you would need less than 290 observations (donations.)
Ryan Kaldari wrote:
... There are a few potential problems with such a system:
- More overhead for managing donations
In most cases, could this be addressed by disclaimers? E.g., "the Foundation reserves the right to cancel earmarked projects for any reason, and to override donor selections if funds fall short in any essential areas"?
- The Foundation is trying to move away from any type of strings
attached to donations (including grants) so that resources can be managed optimally and flexibly
If that presents an actual problem with small donations, under a sufficiently flexible disclaimer, please let me know why.
I do think, however, that such an earmarking system would make donating more attractive to some people
Isn't the reason that the Red Cross does it because it substantially increases donations? Rand Montoya said that he had measured that, and although I forget the numbers, I remember that it was a very significant difference.
David Gerard wrote:
Unless the donation is really quite substantial, this may not be entirely worth the effort.
I know Rand said the effect was substantial, but it varies so much with all of the different permutations that there is literally only one way to find out the extent, and that is to measure it experimentally with actual donors. Merely discussing the possibilities can not arrive at even a vague idea of how much the presentation of each option serves to maximize donations.
Best regards, James Salsman