The quickie answer -- double since it is much more precise than float at minimal cost to you the programmer. The real-world answer is to go with BigDecimal since you almost never want to represent...