Many times I hear about theories that seem really good, but further examination shows they don't work. There are lots of feedback loops that are hard to see, and so, many times a theory is really a partial derivative, while reality is full of only total derivatives.
For example, young mammals often play, and countless times I have heard the nature show narrator note that such play is practice fighting when the bear cub/lion cub/etc. grows up. But scientists actually tested this theory by recording the frequency of play fights by squirrel monkeys and meerkats, noted their success fighting as an adult, and found no correlation between either the number of play fights as an infant, or the successes at infant play fighting, and the adult fighting prowess. They also found that play fighting does not address key tactics in real fighting. Infants learn from fighting, but mainly things like coordination, how to deal with surprise, and socialization. Thus, at a high enough level of abstraction, it seemed perfectly reasonable theory, but the closer one looked, it was plain wrong. Animal play is not practice for adult fighting, except in some very abstract sense.
So too, the government multiplier, import quotas, unionizing industries, and all sorts of other great top-down ideas. They work in some very hazy, abstract sense. But you factor in all the opportunity costs, the disincentive effects, the costs spread among many to help a few, and they are all worse than doing nothing. Unfortunately, such theories are so consistent with other objectives (read: redistribution), they are really too good to check.