Non-Equilibrium Economics: What Changes When Nothing Settles
The phrase sounds technical and slightly forbidding. The underlying idea is not: an economy is a system that never finishes adjusting, and models that assume it has finished will miss exactly the behaviour you care about.
Equilibrium entered economics as a borrowed metaphor from nineteenth-century mechanics, and the borrowing was explicit. A system of prices and quantities was imagined as a system of forces that balance. It was a productive metaphor. It made economics mathematical, and it made a large class of questions answerable.
The difficulty is that the metaphor also imported an assumption that physics itself abandoned: that the interesting state of a system is the resting state. In most physical systems that matter - weather, turbulent flow, ecosystems, the climate - the resting state is either never reached or is reached only in a laboratory. Economies look far more like the second class of system than the first.
What "non-equilibrium" actually asserts
It is a narrower claim than it sounds. It does not say prices are arbitrary or that markets never clear. It says three things:
- Adjustment takes time, and the time matters. If the adjustment period is long relative to the interval between disturbances, the system spends its entire life in transit. The equilibrium is a point it is always heading towards and never occupies.
- The path affects the destination. Firms that go bankrupt during an adjustment do not come back when conditions improve. Skills that atrophy during unemployment do not return on demand. This is path dependence, and it means the end state is not independent of the route.
- Aggregates can be stable while the mechanism is not. A stable unemployment rate can conceal enormous churn. Averaging over a turbulent system produces a number, and that number can be calm while nothing underneath it is.
Heterogeneity is not a technicality
The representative-agent device replaces a population with a single optimising household. It is defensible when the population is roughly homogeneous or when aggregation happens to be well behaved. Neither condition holds in the cases that matter most. Households at different points in the wealth distribution respond to the same policy in opposite directions. Firms with different leverage respond to the same credit shock with different survival probabilities.
Once heterogeneity is taken seriously, the analytical toolkit changes. You are no longer solving for a fixed point; you are propagating a distribution forward. That is a simulation problem, which is how agent-based modelling became the standard method of the field rather than a curiosity.
Interaction structure
Standard models let agents interact only through prices. Everyone sees the same signal and responds independently. Real economic agents interact through supply relationships, credit exposures, employment and imitation, and those relationships form a network with a very specific shape. A shock to a highly connected node behaves differently from a shock to a peripheral one, which is why the structure of the network is part of the economics and not a separate subject.
The honest trade-off
Dropping equilibrium costs you something real. Equilibrium models are disciplined: they have few free parameters and their predictions are sharp enough to be wrong in an informative way. Non-equilibrium models are flexible, and flexibility without discipline is how you fit noise. Anyone selling complexity economics as a pure improvement is overselling it. The correct claim is that it answers different questions - structural ones about fragility, distribution and mechanism - and that it needs its own discipline, which is the subject of validation and calibration.
Where it has paid off
The clearest returns have been in financial stability, where the equilibrium framework had almost nothing to say about how a solvent system becomes insolvent in a fortnight. Network models of interbank exposure produced the concept of a systemically important institution, defined by position rather than size, and that concept made it into regulation. See financial instability and crises. The second area is inequality, where a distributional question cannot even be posed inside a representative-agent model.