I mean, the problem isn't the mathematics itself--it's not like they're making mistakes in solving equations or something like that--but the foundational assumptions behind the mathematical models. Virtually all economists will admit, when pressed, that their models only describe highly idealized systems that resemble the real world only if you take on board a set of very suspect assumptions that we actually have good reason to think are false (or at least true only in extremely limited edge cases). This wouldn't be useless exactly--in the same way that equations about the behavior of frictionless physical systems aren't useless--if it weren't for the fact that most of the time, most economists seem to conveniently forget about those simplifying assumptions and recommend policy as if their models are isomorphic to real world systems in all the relevant ways. There's a kind of pervasive amnesia in the discipline about the relationship between the map and the territory, and they end up making pronouncements (and, worse, policy recommendations) based on models that they know are deficient in pretty serious ways.

There's another pretty serious foundational problem involving value theory and system individuation concerns--the way the models are structured reflects some fairly hefty evaluative choices about what's important, what sorts of states should be thought of as relevantly different from one another, and so on--that's wrapped up with this one, but it's in the same family of concerns. At bottom, the big problem with the field is that they reify their own models in super suspect ways, and don't always seem cognizant of the difference between "here's a set of simplifying assumptions to make this problem tractable; we can learn some things from looking at this idealization, but have to remember it is different from the real world in significant ways" and "this is a good model that actually reflects the world."