At the system level, economic growth (and ultimately, wealth) is truly just a mindset.
This is not surprising: so are freedom and justice, for example. You can put people in the exact same material conditions, with the same tools and infrastructures and technologies and everything, and convince them to act as if they are free or not, as if laws exist or not, and as if the economy is growing or not.
The crucial intuition I wish to convey is that the science of economics does not study a general property of multi-agent organizations. It studies how to get people into, maintain and track the objective consequences of one very particular possible belief/mindset about social life.
It’s like an entire science devoted to how to convince people to believe in Baʿal Ḥammon (or express parental love, or experience the qualia of greenness), and what ensues in the real world when they do.
What it demonstrates is that, if enough of your society shares a certain subjective phenomenon, you can linearize the coupled multi-agent dynamics to make them look as if agents are interacting with an objective phenomenon.
This is something that economics does far better than law or politics, where the individual (ideological, subjective, consensual) level is much more obvious and prominent, whereas tools to deal with the collective level are much less developed and quantitative. We have a diffuse sense of superorganisms in legal or political contexts, of large emergent collective forces like law or war, but (with few exceptions, like voting systems) we do not quite behave towards them with the technicality and quasi-objectivity that we grant to wealth and growth.
But many complexities of economics arise from the moments where this abstraction breaks down. Moments where economists and governments switch desperately between Keynesian and Monetarist and Marxist and every other frame of analysis to make sense of what is happening, because those are all different linearizations, around different equilibria or in different directions, and what you are seeing is truly nonlinear dynamics.
The economy (as tracked by all sorts of constructions, e.g. inflation or recession) is not a single-valued objective function of how much capital, how large a bourgeoisie, how many jobs or resources or new technologies there are. That is not going to improve by just "adding more factors". It's not just complexity in the sense of high-dimensional physics.
It is, deep down, a collective reaction of people, sharing a certain set of beliefs, to what they perceive of their world. That includes many small signals that we can group into working abstractions such as “signals of capital/bourgeoisie” or “signals of jobs/productivity” when society is close to a certain dynamical state, but also entirely other sets of signals (e.g. fears, pandemics, power shifts, etc.) that arise in transient situations and have nothing to do with any existing economic framework. And all these signals are, in a deep deep sense, arbitrary. There is no "real best framework" independently of what people currently believe about they own society.
Even though the whole field is purely about signalling (creating a society-wide mindset, a coordinated equilibrium if you wish), this shows that signalling is in no way as simple as naive game theory makes it out to be. Signalling is how you program a society, and programming has complex data structures, nested subroutines, bugs, patches, conflicting levels of abstraction.
Micro economics remains at the trivial Algorithmics 101 level of how to signal and manage one or a few bits of information about individual agents (e.g. am I cooperative, trustworthy, high or low status, etc.).
Macro deals with heuristics and infrastuctures to capture and signal, at the very least, O(log N) bits of information about collective states and modes of a society-wide dynamics. It manages to do so by creating abstractions that are relevant only close to an extremely specific setting of society, one where everyone believes in Baʿal Ḥammon.
Its only deep failure is that, in trying to emulate sciences that deal with truly objective phenomena, it has failed to develop a proper epistemology for patching together various local objectivizations (linearizations) and understanding how they fail as you move away from very particular states of society.