I'm sympathetic to the idea that categories and abstractions are real in some sense. It's a weird view, but I think it has merit. The right question to ask, though, is "what makes your proposed abstract object worth treating as if it were real?" The utility of things like this comes from the predictive power you get from treating them as real: "the economy" is an abstract composite object in much the same way seems to think "dragon" is, but we get a lot of predictive and explanatory mileage out of tracking it as an individual thing. What predictive utility does talking about dragons get us beyond what we get from talking about, say, a predator? I can make up all sorts of objects--let's track the object that consists of the union of all pennies minted after 1982 and the left half of . That's a "legal" object in about the same sense that the economy is, but there's nothing interesting about it: picking it out doesn't enhance our ability to predict or understand the world. "Dragon" in the sense that he's trying to use it is like that.