According to this meta post, it is community consensus that theoretical answers should be deleted as off-topic. The reasoning is that
[o]ne of the premises of skepticism is the application of the scientific method: empirical proof validates (or disqualifies) theoretical models. All questions we allow here are empirical in nature, thus answering via a purely theoretical model is inappropriate.
Where does an answer start to become "theoretical" in this sense? Obviously, no answer can be "empirical all the way through". Even the application of concepts from everyday language to describe an empirical phenomenon isn't purely empirical, but this is clearly not meant by "theoretical answer".
Perhaps an undisputedly theoretical answer would
- apply some model on a high level of abstraction
- that makes a lot of assumptions
- the model itself and the application are very hard to evaluate by laypeople
- the scientific community considers the model just as one plausible contestant among many.
The list is just an example for a clearly theoretical answer. The more difficult question is whether there exists a list of criteria that can be used to distinguish theoretical from non-theoretical answers. The list above is not it.
What about questions like this, which asks whether burning 5 pounds of fuel can produce more than 5 pounds of CO2?
The "theoretical" answer comes down to: Yes, when two reactants are combined, the mass is greater than that of one reactant only. The argument is very basic, it does not even require an understanding of chemistry beyond knowing that "burning" has something to do with "adding" oxygen. But more importantly, there is no empirical study on this question that we could cite, because it would be too trivial to be scientifically useful. (It might be useful as a primary or middle school experiment.)
Any answer to this question can only be theoretical, if the "theoretical answer" yardstick is too narrow. Lacking "empirical" references, there could be no acceptable answer, which is a pity because it's not a bad question.
Of course, an answer could cite some textbook that explains the basics of chemical reactions, but applying these basics to the question at hand would still be considered "theoretical". That's nonsense!
So how could a tentative yardstick for "theoretical" answers look like?