(1) the neoclassical (and neoRicardian) economic notion of "equilibrium" as actually attained does have its use. The main use is to test the internal consistency of an economic model.
For example, back in the 1970s, Robert Lucas and his colleagues advanced Muth's notion of "rational" expectations (RatEx) as an important logic-check for macroeconomic models. For the uninitiated, RatEx does _not_ represent a case of rational calculation (by homo economicus), balancing the costs and benefits of collecting and using information, about what to expect about the future. Rather, it involves the assumption that "agents" (that's us) have an identical model of how the economy works -- eerily and inexplicably identical to the one held by the macroeconomist -- and then use that model in order to crank out what to expect. (Because everyone has exactly the same model of the economy, merge the entire population into a single "representative agent.") As an economist at Cal Tech (name forgotten) pointed out at a seminar I went to, this is the macro-equivalent of the Game Theorist's Nash equilibrium. In macro, these expectations are not assumed to be correct, however, since the fact of incomplete information is introduced by introducing random error around correct expectations. (This seems akin to the idea of "trembles" in Game Theory.) To sketch an example, initially ignore the random aspect. Because the macroeconomist uses a monetarist/classical model, the representative agent has one too -- and thus concludes that if the money supply is growing at 10% per year, the potential real GDP is increasing at 3% per year, and the velocity of money is constant, the inflation rate will be 7% per year. (The monetarist/classical model bizarrely assumes that in a non-random world, actual GDP always equals potential, under Say's "Law.") The agent then acts on this expectation, raising its price by 7%. The economist's model -- which predicted 7% -- is totally consistent with the agent's actions. Expectations are in equilibrium: what's real is rational and what's rational is real. On the other hand, if the agent simply adapts its expectations to experience with past inflation (following what's variously called "adaptive expectations," "partial adjustment," or "Bayesian learning"), it may easily expect an inflation rate that's completely inconsistent with the model: if the growth rate of the money supply, that of potential, and that of velocity, all stay constant, the agent's expected inflation rate will _never_ equal the actual inflation rate. The only exception occurs when the initial expectation equals 7%. Bringing in the random element, the representative agent using RatEx will not have correct or equilibrium expectations. But it will be correct on average. (2) So the idea of equilibrium -- here of expectations -- makes the model consistent, unlike the case of adaptive expectations, in which history (experience) plays the major role. -- Jim Devine / "The conventional view serves to protect us from the painful job of thinking." -- John Kenneth Galbraith; "Microeconomics is too important to leave to the microeconomists." -- yours truly.
