An Adaptive Macro-Network Model: A Co-evolving Recovered Dependence Operator with Bounded-Rational Learning

Economists usually model the economy as a fixed network of links between sectors. This paper lets that network become a moving part: participants choose how much exposure to take on, guided by a rule that reflects their limited capacity to process information, and the web of dependence they build then shapes their own rewards. As the market's price for shared risk falls, learning drives the system through a gradual collapse of diversity, with a single hub coming to dominate; the change is smooth, with no sudden jump and no point of no return. One discipline runs throughout: the paper acts only on what daily stock-market data can actually recover, never on the finer question of which sector drives which, which those data cannot pin down. A pre-registered test on four real stock-market datasets finds that how concentrated the network has become marks crises as they happen but does not forecast them even one step ahead. Recovering the structure of the system is not the same as predicting returns.