To quote from the article,

> We polled our users and asked them what they would be willing do donate and the average was twenty seven dollars, and of course we have thousands of users

Which is based on the illusion that you get an unbiased sample when you ask users to respond to a poll. You don't - the people most likely to sacrifice a moment of their time are also those more likely to come forward with money, conversely those who didn't fill out the questionnaire probably have a majority of people who won't give any money.