Nightly Business Report did a short piece on Barack Obama's tax cut plan that featured Moody's estimates that the cuts would raise the US Gross Domestic Product by some percentage. I like tax cuts rather than big government payouts because tax cuts let us control more directly how our money is spent. But Moody's estimates have to be bulls--t. That's doesn't mean they are going to be wrong. It just means that the estimates are founded on inadequate models.
Now I'm no economist, but I am a modeler of complex physical systems, so here is what I think is going on.
Consider that the global economy is a giant network of people, corporations, and governments all interacting with each other in a highly complex and dynamic way. Mathematically we can represent each economic actor as a point or node or vertex, and each interaction between any two actors as a line segment or or edge connecting them. This collection of nodes and edges is called a graph. The graph theory literature is large, even though the subject is relatively new. It is new because large graphs can't be analyzed by continuum mathematics (like calculus). They have to be analyzed by large computers.
But the simplest questions one can ask about large graphs can be very hard even for a computer to answer. NP-hard in fact, which means that in practice you can't get the answer in a useful time. Even searching a large graph can be hard, let alone trying to compute how a graph of interacting agents will evolve in time. But that is exactly what you have to do in order to predict how the economy will react to a given stimulus. Such a model is currently beyond the capability of anyone, Moody's included.
Moody's, like all the other predictors, must be using a much simpler, and thus over-simplified model. It has to be over-simplified because, if it weren't, they would have predicted the current economic slowdown and done their investment rating much differently.
The modelers are getting a clue, however. I noticed that an Economics and Math professor is getting geared up for these kinds of problems, a book has been published, and another one is on the way. And while you're at it, check out the blog of Valdis Krebs, one the the world's go-to people on the subject of graphs and networks.
Don't hold your breath waiting for them to do real predictive economic modeling, though. Manipulating these large graphs may require quantum computers, and we don't quite have any of those yet.