Schools of economic thought Lesson 5 of 9
Economics · Schools of economic thought
The Phillips curve: unemployment, inflation and expectations
From Phillips (1958) to Friedman and Phelps, the natural rate, the flattening debate and 2021–23, tested on US unemployment and core inflation by era since 1958.
On this page
The Phillips curve is the observed tendency for inflation to run higher when unemployment is low, and lower when unemployment is high. It is one of the most used and most argued-over relationships in economics, because it sits behind every central bank's judgment about whether a strong job market will push up prices.
The idea has been rebuilt three times. A. W. Phillips found a stable link between unemployment and wage growth in British data; American economists turned it into a menu of choices between inflation and jobs; Milton Friedman and Edmund Phelps argued that the menu disappears once people expect inflation; and after 2008 the curve looked so flat that some declared it dead, until 2021–23.
This lesson traces each version, then tests them on US data from 1958 to 2026: the unemployment rate from the Bureau of Labor Statistics, core CPI inflation, and the Congressional Budget Office's estimate of the natural rate of unemployment.
The original curve: Phillips, 1958
A. W. (Bill) Phillips, a New Zealand economist at the London School of Economics, plotted the yearly change in British money wages against unemployment for 1861 to 1957 and found a downward-sloping, curved relationship: when unemployment was low, wages rose fast; when it was high, wages rose slowly or fell. The curve was steep at low unemployment and flat at high unemployment, so a one-point fall in unemployment did far more to wages when unemployment was already low.
The logic is a labor market story. When jobs are plentiful, employers compete for workers and bid up pay; when they are scarce, workers accept smaller raises. Since wages are the largest cost for most businesses, faster wage growth tends to mean faster price growth.
The menu: Samuelson and Solow, 1960
Paul Samuelson and Robert Solow redrew the relationship for the US with price inflation in place of wage growth and presented it as a menu for policy: a country could choose a little more inflation to buy a little less unemployment. The 1960s seemed to confirm it.
The expectations version: Friedman and Phelps
Edmund Phelps (1967) and Milton Friedman, in his 1967 presidential address to the American Economic Association published in 1968, argued that the menu was an illusion. Workers and firms care about real wages. If inflation has been running at 5% and everyone expects it to continue, wages and prices will rise 5% a year with no change in unemployment. Only surprise inflation lowers unemployment, and only until expectations catch up.
That leaves one unemployment rate consistent with stable inflation, which Friedman called the natural rate and later economists the NAIRU (non-accelerating inflation rate of unemployment). Push unemployment below it and inflation does not settle at a higher level; it keeps accelerating.
The prediction was made just before the 1970s, when US inflation and unemployment rose together (Stagflation: high inflation, weak growth and the 1970s lesson). The old menu failed; the expectations version explained the failure. CBO publishes an estimate of the natural rate (FRED series NROU); it was about 4.4% in 2021–2026.
What the data show
Plot unemployment against inflation for every year since 1958 and the single curve disappears. The 1960s points trace one; the 1970s float far above it as expected inflation rose; 2008–2019 lie along an almost horizontal band, with inflation near 2% whether unemployment was 9.6% (2010) or 3.7% (2019). In levels, the correlation was −0.71 in 1958–1969 and +0.20 in 1970–1983, the wrong sign for the original curve.
The expectations version tests something different: whether the change in inflation falls when unemployment is above the natural rate. For each era we regressed the 12-month change in core CPI inflation on the unemployment gap (UNRATE minus CBO's NROU), month by month.
| Era | Slope | What it means |
|---|---|---|
| 1958–1969 | −0.51 | One point of slack lowers inflation 0.5 pt/yr |
| 1970–1983 | −0.86 | Strongest response: expectations moved fast |
| 1984–2007 | −0.25 | Weaker, as the Fed anchored expectations |
| 2008–2019 | −0.06 | Almost flat: slack barely moved inflation |
| 2020–2026 | −0.08 | Flat on average, despite the 2021–22 surge |
Every slope has the sign Friedman and Phelps predicted, which is a point in their favor. But the slope fell by more than 90% between the 1970s and the 2010s. That is the flattening.
Why the curve flattened
Economists offer several explanations, and they are not exclusive:
- Anchored expectations. Once people trust the Fed to keep inflation near 2%, a strong job market does not change what they expect, so inflation moves less. This is a success of policy, not a failure of the theory.
- The Fed hides the curve. If the Fed raises rates whenever it sees inflation coming, periods of low unemployment never last long enough to show inflation. Peter Hooper, Frederic Mishkin and Amir Sufi (2019) argued that national data understate the slope for this reason, and that state and city data still show a clear, nonlinear curve. Their title asked whether the curve was dead or "just hibernating".
- Global competition and lower union membership reduce workers' bargaining power.
- Measurement. The natural rate is not observed. If CBO's estimate is wrong, the gap is wrong, and the estimated slope is biased toward zero.
2021–23: the curve wakes up?
In September 2022 core CPI inflation reached 6.6%, the highest since 1982, with unemployment at 3.5%. A straight line fitted to 2008–2019 cannot produce that. Two readings compete:
- Supply shocks. Ben Bernanke and Olivier Blanchard (2023) found that most of the surge that began in 2021 came from shocks to prices (energy, food, shortages in particular goods), while tight labor markets mattered less at first but had more lasting effects on wages.
- A nonlinear curve. Pierpaolo Benigno and Gauti Eggertsson (2023) argued the curve becomes steep when the labor market is very tight, measured by job openings per unemployed worker, and that the surge was mainly a labor shortage. Their model also implied that inflation could fall without a deep recession.
What happened next fits the nonlinear story in one respect: core inflation fell from 6.6% (September 2022) to 3.2% (December 2024) while unemployment rose only from 3.5% to 4.1%. The old flat-line rule would have required far more unemployment to remove that much inflation. In September 2026 unemployment was 4.2% and core inflation 2.8%.
The critics' case
- Rational expectations (Robert Lucas, Thomas Sargent) went further than Friedman: if people form expectations using all available information, even a temporary trade-off is available only by surprise, and a credible disinflation can be cheap.
- It is not a policy lever. The 1970s showed that trying to buy jobs with inflation raises inflation without lasting job gains.
- Unstable estimates. The slope and the natural rate shift over time, so any single number should be treated as a range.
- Supply shocks dominate some episodes, such as the 1973 and 1979 oil shocks and the 2021–22 shortages, and push inflation and unemployment the same way.
The versions compared
| Version | Claim | Fits best |
|---|---|---|
| Phillips (1958) | Low unemployment → fast wage growth | UK 1861–1957 |
| Samuelson–Solow (1960) | A stable menu of inflation vs jobs | US 1960s |
| Friedman–Phelps (1967–8) | Only surprises trade off; natural rate | US 1970s |
| New Keynesian (1990s–) | Expected future inflation + slack, flat | US 2008–19 |
| Nonlinear (2019–23) | Steep when the labor market is very tight | US 2021–23 |
How Finabase does it
The Economy page shows the two halves of the Phillips curve side by side: five-year monthly charts of CPI inflation (year over year) and the unemployment rate, next to the Fed funds rate, plus a calendar of CPI and jobs releases with actual, forecast and previous figures. It does not plot one against the other, show core CPI or the natural rate, or go back before the last five years; the scatter and slopes on this page are ours, from public BLS and CBO data. For how the Fed reads the labor market, Discover has a Fed topic; the Treasury page shows how bond yields priced the 2022 rate rises.
Frequently asked questions
Is the Phillips curve dead?
No, but it is much flatter than in the 1970s. In our test the change in core inflation still fell when unemployment was above CBO's natural rate in every era since 1958, but the slope was close to zero in 2008–2019, and 2021–23 suggested it steepens when the labor market is very tight.
What is the natural rate of unemployment?
It is the unemployment rate at which inflation neither speeds up nor slows down, once expectations have adjusted. It cannot be observed directly; CBO estimated it at about 4.4% in 2021–2026.
Why does the Fed care about the Phillips curve?
Because the Fed has two goals, maximum employment and stable prices, and the curve is its main guide to whether a strong job market threatens inflation. A flat curve lets the Fed run the job market hotter; a steep one forces a harder choice.
What is the difference between the Phillips curve and stagflation?
The Phillips curve says inflation and unemployment tend to move in opposite directions; stagflation is a period when both are high at once. Stagflation in the 1970s broke the original curve and led to the expectations-augmented version.
Summary
Phillips found in 1958 that British wages rose faster when unemployment was low; Samuelson and Solow turned that into a menu for US policy, which the 1960s seemed to confirm. Friedman and Phelps argued that only surprise inflation lowers unemployment, so holding unemployment below the natural rate makes inflation accelerate, and the 1970s proved them right. In US data since 1958 the expectations version holds in every era, but its slope fell from −0.86 in 1970–1983 to about −0.06 in 2008–2019, a flattening that most economists link to anchored expectations and Fed policy. The 2021–23 surge and fast disinflation revived the case for a curve that is flat most of the time and steep when labor is scarce.
Check your understanding
4 questions on this lesson. Each answer comes with the reason, and nothing is scored.
Inflation has run at 4% for several years and everyone expects 4% to continue. The central bank wants unemployment one point below the natural rate and keeps it there. What does the Friedman–Phelps view predict?
- Inflation settles at a new, higher level, say 5%, and stays there
- Inflation keeps rising year after year as expectations catch up
- Inflation falls, because more people are working
Show the answer
B. Inflation keeps rising year after year as expectations catch up Only surprise inflation lowers unemployment. Each time expectations catch up, the bank has to create a bigger surprise to keep unemployment below the natural rate, so inflation accelerates. That is why the natural rate is also called the NAIRU.
From 2008 to 2019 US unemployment ranged from 3.7% to 9.6% while core inflation stayed near 2%. Which explanation is consistent with the Phillips curve still existing?
- Inflation expectations were anchored near 2% and the Fed leaned against inflation, which flattens the measured curve
- Unemployment has no effect on wages
- Core inflation leaves out the prices that respond to unemployment
Show the answer
A. Inflation expectations were anchored near 2% and the Fed leaned against inflation, which flattens the measured curve If people expect 2% whatever happens and the Fed offsets demand pressure, national data show little link even when one exists underneath. Hooper, Mishkin and Sufi found a clearer curve in state and city data, where the Fed cannot offset local booms.
In 1974 an oil embargo quadruples crude prices. Unemployment and inflation both rise. How does the expectations-augmented Phillips curve account for this?
- It cannot: the curve says they always move in opposite directions
- A supply shock shifts the whole curve up, so inflation rises at any level of unemployment
- The natural rate fell
Show the answer
B. A supply shock shifts the whole curve up, so inflation rises at any level of unemployment The modern curve has a supply-shock term. A jump in energy costs raises inflation for a given amount of slack, while the same shock cuts output and jobs. That is the stagflation case.
Core inflation fell from 6.6% in September 2022 to 3.2% in December 2024 while unemployment rose only from 3.5% to 4.1%. Under a straight-line curve with the 2008–2019 slope of about −0.06, what would that disinflation have required?
- About what happened: a small rise in unemployment
- A very large and long rise in unemployment above the natural rate
- A fall in unemployment
Show the answer
B. A very large and long rise in unemployment above the natural rate With a slope of −0.06, each point-year of unemployment above the natural rate lowers inflation by only 0.06 points, so removing 3.4 points would take dozens of point-years of slack. The actual path is more consistent with a curve that was steep while the labor market was very tight, plus fading supply shocks.
Exercise
One task to do in Finabase. Nothing is graded or saved.
Open the Economy page and put the CPI inflation and unemployment charts side by side. Find the months in 2022 when inflation peaked and read the unemployment rate then. Then move to late 2024 and read both again. The lesson says core inflation fell from 6.6% to 3.2% while unemployment rose only from 3.5% to 4.1%. Does the headline chart show the same shape: a big fall in inflation for a small rise in unemployment?
Open the Economy pageWhat you should find
Inflation falls a long way from its 2022 peak while unemployment drifts up by well under a point. A flat curve like 2008–2019 would have needed far more unemployment to remove that much inflation, which is why the episode supports a curve that is steep when labor is scarce. The page shows headline, not core, CPI, so the numbers differ.
Sources and further reading
- Phillips, The relation between unemployment and the rate of change of money wage rates in the United Kingdom, 1861–1957, Economica 25(100), 1958
- Samuelson and Solow, Analytical aspects of anti-inflation policy, American Economic Review 50(2), 1960
- Phelps, Phillips curves, expectations of inflation and optimal unemployment over time, Economica 34(135), 1967
- Friedman, The role of monetary policy, American Economic Review 58(1), 1968
- Hooper, Mishkin and Sufi, Prospects for inflation in a high pressure economy: is the Phillips curve dead or is it just hibernating?, NBER Working Paper 25792 (2019), retrieved 2026-11-18
- Benigno and Eggertsson, It’s baaack: the surge in inflation in the 2020s and the return of the non-linear Phillips curve, NBER Working Paper 31197 (2023), retrieved 2026-11-18
- Bernanke and Blanchard, What caused the US pandemic-era inflation?, NBER Working Paper 31417 (2023), retrieved 2026-11-18
- BLS — Unemployment rate (FRED series UNRATE), downloaded 2026-10-30
- BLS — CPI-U less food and energy (FRED series CPILFESL), downloaded 2026-10-30
- CBO — Noncyclical rate of unemployment (FRED series NROU), downloaded 2026-10-30
