Unfortunately, proofs aren't all they are made out to be. Proofs don't tell if the problem is worth solving, or if a theory will turn out to be fruitful, or if it will lead to solving the intended problem. Proofs are not analogs of experiments, they do not have discriminatory power except in preventing logical mistakes.

But I will cross over to your side, and say that there is a discriminating process in mathematics, albeit not a straightforward falsification. Unfruitful theories are not falsified, but they are abandoned, turned into niche curiosities.

Continued fractions were supplanted by positional fractions because arithmetical operations on the former are notoriously difficult. Of all the theories of integral Lebesgue's came to dominate the scene due to its generality and technical convenience.

Mathematical theories and methods are "tested" by mathematicians that apply them. If a theory fails to advance a problem it was expected to solve we get a Kuhn's "anomaly", and anomalies do accumulate. Classical invariant theory (Gordon, Kronecker, etc.) failed to classify higher order invariants, and was swept away by Hilbert's methods. Mathematics is not that different from empirical sciences.

And empirical sciences are not that different from mathematics. Experiments have to be properly designed, arguments may not be rigorous, but they have to be rational and sound, that would be the analog of proofs. Which is why "need something" doesn't imply "need falsification guided discrimination", especially all the time, or where experiments can't be staged, or where multiple theories can be designed to fit the same data.

In advanced sciences individual experiments are not as discriminatory as they seem, nor should they be. Established theories are worth holding on to, as are creative new ones because better theories are more likely to come out from bending than from starting from scratch. In economics and other human sciences whether models fail is inescapably unclear, Aristotle's rock and feather would "falsify" the law of gravity too if there was no way to remove the air.

With so many accidental factors in different markets and different countries there is no way to tell if the explored effect does not manifest, or is obscured by noise, or is overtaken by other effects. The only approach I see is massive and prolonged accumulation of data along with developing thousand models for thousand effects until the critical mass is reached. And that should be encouraged rather than slighted as sub par.

There would be reason for frustration and criticism only if economics had circumstances similar to physics or chemistry, and still failed to deliver. Ironically, current fundamental physics is in a similar predicament for the opposite reason. The Standard Model is so good that any theory subsuming it will be extremely hard to falsify.

If you agree that in many areas we can't meet the preconditions for falsification's effectiveness why do you recommend it as preferred methodology, and suggest that we subjugate theory development to its convenience? Isn't it more reasonable to look for discriminatory alternatives that fit the context?

And aren't you concerned that transplanting success standards without having the (perhaps unmeetable) preconditions met might only be counterproductive? As would seem to be the case in modern fundamental physics, economics or psychology. If there is a "something" in methodology it is that it should be justifiable by a rational analysis of conditions in the field, including observational/experimental capabilities, and theory development phase.