Auctions are a popular way to raise money for charities, but relatively little is known, either theoretically or empirically, about the properties of charity auctions. The small theoretical literature suggests that the all-pay auction should garner more money than winner-pay auctions. We conduct field experiments to test which sealed bid format, first price, second price or all-pay raises the most money. Our experiment suggests that both the all-pay and second price formats are dominated by the first price auction. Our design also allows us to identify differential participation as the source of the difference between existing theory and the field. To conclude, we show that a model of charity auctions augmented by an endogenous participation decision predicts the revenue ordering that we see in the field.