Rational expectations, the Lucas critique and rules over discretion
The New Classical program took an idea from John Muth (1961): expectations are essentially the same as the predictions of the relevant economic theory, so people do not waste information [1]. Robert Lucas carried it into macroeconomics. In his 1972 model, and in a 1973 study of annual data for eighteen countries over 1951-67, suppliers in scattered markets see their own prices but cannot be sure how much of a rise is general inflation, so only surprises in the price level move output away from its natural rate [2] [3].
In 1976 Lucas argued that policy evaluation based on estimated relationships is unreliable, because the relationships themselves shift when the policy rule changes: people's decisions depend on the rule they expect [4]. Sargent and Wallace (1975) showed that in a model with rational expectations and a Lucas supply curve, the distribution of output does not depend on which deterministic money-supply rule the central bank follows [5]. Kydland and Prescott (1977) added that discretion fails even when policymakers share the public's goals, because economic planning is a game against rational agents rather than against nature [6], and Barro and Gordon (1983) showed how discretion builds in higher inflation with no lasting gain, a bias that rules or reputation can remove [7].
Thomas Sargent's study of four hyperinflations after the First World War became the program's favorite historical evidence. In Austria, Hungary, Poland and Germany, inflation stopped abruptly rather than gradually once an independent central bank refused the government unsecured credit and fiscal policy changed at the same time [8]. The lesson drawn was that ending inflation requires a change in the policy regime that people believe, not a few temporary restrictive actions.
The record is mixed. The old Phillips relationship had already broken down in the United States before Lucas wrote: unemployment averaged 3.5 percent in 1969 and 8.5 percent in 1975, while consumer prices rose by between 3 and 12 percent a year [9] [10]. Credible institutions do seem to anchor expectations: far-ahead inflation compensation in UK bond markets stopped reacting to economic news after the Bank of England gained independence in 1997 [11]. But forecasts by professionals and others show pervasive information rigidities, a departure from full-information rational expectations [12], and Rudebusch (2005) finds that the shifts in US monetary policy were too small to change reduced-form models much [13].
What survives is a discipline. Modern models state the policy rule, the information people have and how they form expectations, and they test a policy against the response it will provoke. The stronger New Classical conclusion, that systematic stabilization policy is ineffective, has fared worse; this guide's limits section explains why.