"Finally, at the end of the survey I had a question offering respondents a chance to cooperate (raising the value of a potential monetary prize to be given out by raffle to a random respondent) or defect (decreasing the value of the prize, but increasing their own chance of winning the raffle). 73% of effective altruists cooperated compared to 70% of others - an insignificant difference."

Assuming an EA thinks they will use the money better than the typical other winner, the most altruistic thing to do could be to increase their chances of winning, even at the cost of a lower prize. Or maybe they like the person putting up the prize, in which case they would prefer it to be smaller.