Phillips Curve

The Phillips Curve is the inverse relationship between unemployment and inflation that A. W. Phillips identified in long-run UK data in 1958. Its postwar career as a contested empirical relationship — read variously as a stable policy menu, a short-run-only trade-off conditional on expectations, an artefact of monetary regime, and a discovery procedure rather than a causal mechanism — structures most of the substantive disagreement among twentieth-century macroeconomic traditions on the question of what causes inflation. No other single empirical regularity in macroeconomics has produced as much theoretical work, as much real-time policy consequence, or as much retrospective embarrassment for those who treated it as more stable than it turned out to be.

The original observation (Phillips 1958)

A. W. Phillips’s Economica paper “The Relation between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957” identified an inverse, nonlinear, statistically robust relationship between the unemployment rate and the rate of change of nominal wages across nearly a century of UK data. Phillips’s own framing was empirical and modest: he had identified a regularity, fitted a curve to it, and observed that the curve appeared stable across roughly homogeneous policy environments and shifted across regime breaks. He did not propose a causal mechanism; he did not propose a policy use; he certainly did not propose that the curve was structural in the sense the subsequent literature would treat it.

The neoclassical-synthesis reading (Samuelson-Solow 1960)

Paul Samuelson and Robert Solow, in a 1960 American Economic Review paper “Analytical Aspects of Anti-Inflation Policy,” reinterpreted Phillips’s wage-inflation relationship as a price-inflation relationship in US data and described it as a “menu” from which policymakers might choose. The accompanying mark-up assumption — that prices track wages with a stable margin — translated the wage-Phillips Curve into a price-Phillips Curve; the menu interpretation translated the empirical regularity into an exploitable policy trade-off. Samuelson and Solow themselves were cautious about the structural reading; the textbook tradition that followed them was less so. By the mid-1960s, the Phillips Curve as a stable, exploitable inflation-unemployment trade-off was the operating assumption of demand-management policy in the United States and the United Kingdom.

The natural-rate critique (Phelps 1967, Friedman 1968)

Edmund Phelps’s 1967 papers and Friedman’s 1968 AEA address argued — independently and in advance of the empirical breakdown that would substantiate the argument — that the Phillips Curve as the postwar mainstream had read it was a short-run relationship conditional on adaptive expectations of inflation. There exists, on this view, a unique unemployment rate (the natural rate) consistent with stable inflation, determined by structural labour-market characteristics. Attempts to hold unemployment below the natural rate via monetary stimulus would produce only accelerating inflation as expectations adjusted. The long-run Phillips Curve is vertical at the natural rate; there is no permanent inflation-unemployment trade-off.

The argument was pre-empirical: Friedman delivered it in December 1967 with US unemployment at 3.6 percent and the postwar Phillips trade-off apparently intact. Within five years the stagflation episode had visibly broken the postwar curve and given the natural-rate framework a substantial real-time vindication. The single best-corroborated prediction of postwar macroeconomics was, broadly, this one.

Expectations-augmented and rational-expectations versions

The natural-rate critique left the Phillips Curve formulation as a relationship between unanticipated inflation and unemployment: actual inflation minus expected inflation drives unemployment below or above the natural rate. Through the 1970s, this was operationalised with adaptive expectations (agents form expectations by extrapolating past inflation) producing accelerationist Phillips Curves. The New Classical reformulation in the mid-1970s replaced adaptive with rational expectations: agents anticipate the inflationary consequences of accommodative policy and front-run them, so even short-run trade-offs depend on the credibility of the policy regime. In its strong form, the rational-expectations Phillips Curve predicted that credibly-committed disinflations should be costless — a prediction not borne out by the Volcker disinflation and broadly rejected by the late 1980s.

The New Keynesian Phillips Curve

The synthesis that emerged in the New Keynesian program of the 1990s is a forward-looking Phillips Curve in which inflation depends on expected future inflation and the current output gap, with the slope determined by the underlying frequency of price adjustment (Calvo random-arrival pricing being the standard formulation). The framework retains the Phelps-Friedman natural-rate insight, replaces adaptive with model-consistent expectations, and adds explicit microfoundations for nominal price stickiness. Its empirical performance through the Great Moderation (1985–2007) was reasonable; its performance post-2008 has been substantially worse.

The post-2008 flatness debate

Through the 2010s, with US unemployment falling steadily from 10 percent in 2009 to under 4 percent by 2019, headline inflation remained persistently below the Federal Reserve’s 2 percent target. The Phillips Curve appeared, on this evidence, to have flattened — labour-market tightness no longer translated into proportional wage-and-price pressure. Several explanations were proposed: globalisation suppressing wage growth, anchored inflation expectations from credible inflation targeting, mismeasurement of slack, structural changes in labour-market institutions, or the curve genuinely having become approximately horizontal at low rates of inflation. None of these was decisively established; the working consensus by 2019 was that the curve had become substantially less informative as a policy guide than the previous twenty years of New Keynesian practice had assumed.

Post-2021 re-evaluation

The 2021–2023 inflation episode complicated the flatness reading. Inflation rose sharply (peaking near 9 percent CPI in mid-2022) and then fell substantially without a recession (down to 3 percent by mid-2023). The path fitted neither a simple New Keynesian Phillips Curve estimated over the prior decade nor any single competing framework cleanly. Supply-shock decompositions (oil, supply chains, labour-force participation), pent-up-demand stories, and conflict-inflation accounts each captured part of the picture; the New Keynesian framework’s coefficient estimates required substantial re-estimation. The most parsimonious read, by 2024, was that the Phillips Curve is real but considerably more dependent on the underlying composition of shocks (demand vs supply) than the prior consensus had assumed.

Each tradition’s reading

• Old Keynesian — Phillips Curve as a stable empirical menu (the original Samuelson-Solow reading); the framework’s most consequential mistake.
• Monetarist — Short-run only, conditional on adaptive expectations; vertical in the long run at the natural rate. The position vindicated by the 1970s.
• New Classical — Equilibrium artefact of the prevailing monetary regime; dissolves with regime change; the Lucas critique is the methodological version of this reading.
• New Keynesian — Forward-looking, microfounded via Calvo or menu-cost frictions, conditional on rational expectations and central-bank credibility; the operating framework of contemporary central banks.
• Post-Keynesian — A misleading framing; inflation is fundamentally distributional (conflict between wage- and profit-claimants), and the apparent unemployment-inflation correlation reflects unemployment’s effect on wage-bargaining power, not a structural macroeconomic relationship.
• Austrian — Largely uninterested in the Phillips Curve; treats inflation as a distortion of relative prices rather than as a level phenomenon traded off against employment.

Status today

The Phillips Curve survives in modern macroeconomics as a forward-looking, expectations-conditioned relationship whose slope and stability depend on the underlying composition of shocks and on the central-bank credibility regime. The flat naïve trade-off taught in 1960s textbooks is gone; the strong rational-expectations costless-disinflation claim is gone; the steeper New Keynesian curve estimated over 1990–2007 is widely treated as having been a regime-specific artefact. What remains is a working empirical scaffold that informs central-bank thinking without commanding the operational confidence of the prior generation. The single most-contested empirical relationship in macroeconomics remains contested.