I'm an economist. A lot of economic theories aren't too many steps up from reading the tea-leaves. Though it depends what you mean by theories. The theories are rather high level (if the price goes up, all else being equal, demand goes down - that's about the level of a theory in economics - it doesn't produce a prediction to 6 decimal places like a theory in physics, indeed the best it usually does is predict the sign of a number). At a lower level I think we have methods rather than theories. And the main shortcoming of the methods is lack of information. (What is the sensitivity of demand for X to price, not just now at the current price, but at all times at all prices? Be nice to know that but you never do.)

At the largest scale, eg, the whole-economy models which central banks and national finance ministries use for setting their macro policy, the models typically used are known to have shortcomings, for the following reasons.

People's behaviour does, and rationally should, depend upon their best estimate of the future, what we usually call their expectations. Suppose you are a major player in the economy, say, a central bank, or government. You need a model in order to form your expectations, and set your policy. Other agents in the economy know you do this, and they know what the prediction of your model is, because you publish it. They will set their own expectations and choose their actions based upon that. It can be shown that their collective action in doing that will typically act to subvert the prediction made. So the model can't be right. The prediction of the model, used as a basis for policy formation, is self-defeating.

There is a method called "rational expectations", in which you try to explicitly model the expectation formation process, so that the prediction will not be self-defeating. People have tried to build rational expectations models of the economy, but they don't work. So banks and finance ministries carry on using models of a type known to produce self-defeating predictions.