- The Phillips curve shows a short-run trade-off: lower unemployment tends to come with higher inflation, and higher unemployment with lower inflation.
- The modern, expectations-augmented version is \( \pi = \pi^e – \beta(u – u^*) \): inflation equals expected inflation, minus a term for how far unemployment is from its natural rate (the NAIRU).
- In the long run, expectations catch up, so the long-run Phillips curve is vertical at the NAIRU. Policy cannot buy permanently lower unemployment with permanently higher inflation.
- The curve looked “flat” for decades before 2020, then inflation surged in 2021–22. Research points to supply shocks, a very tight labour market and a curve that may be steeper when slack is low.
- Exam point: always say which curve you mean (short run or long run) and what is assumed about expectations.
What is the Phillips curve?
The Phillips curve is a relationship between the unemployment rate and the rate of inflation. In its simplest form it says that when unemployment is low, inflation tends to be high, and when unemployment is high, inflation tends to be low. You will usually see it drawn with unemployment on the horizontal axis and inflation on the vertical axis, as a curve that slopes downwards.
Why? When unemployment is low, firms struggle to find workers, so they bid up wages. Higher wages raise costs, and firms pass some of them on as higher prices. When unemployment is high, workers have little bargaining power, so wage growth and price rises are modest.
Phillips (1958): the original curve was about wages
The curve is named after the New Zealand-born economist A. W. Phillips. In 1958 he published a study in Economica using United Kingdom data from 1861 to 1957[1]. Notice the detail that students often miss: Phillips plotted the rate of change of money wages, not price inflation, against unemployment. He found a stable, curved relationship. His conclusion was that wage rates would be stable at an unemployment rate of about 5½ per cent, and that product prices would be stable at a little under 2½ per cent unemployment, assuming productivity grew by about 2 per cent a year[1].
Samuelson and Solow (1960): a “menu” for policy
Two years later, Paul Samuelson and Robert Solow applied the idea to the United States and used price inflation instead of wage growth[2]. Their article was widely read as offering policymakers a menu: accept a bit more inflation and you can have a bit less unemployment[3]. The economics profession took from it the idea that demand policy could buy a permanently lower unemployment rate at the cost of higher average inflation[7]. That “permanent” part is what later went wrong.
The expectations-augmented Phillips curve and the NAIRU
In the late 1960s, Milton Friedman and Edmund Phelps argued, separately, that the simple curve left out something vital: expected inflation. Friedman set out the argument in his 1968 presidential address to the American Economic Association[4]. Phelps developed it formally in a 1967 Economica paper[5]. Phelps pointed out that current inflation depends not only on unemployment but also on what people expect inflation to be. At a given unemployment rate, a one-point rise in expected inflation raises actual inflation by one point[6]. Phelps later received the 2006 Nobel memorial prize partly for this work[6].
The logic is simple. Workers care about real wages (wages adjusted for prices). If everyone expects 5% inflation, workers ask for at least 5% more pay just to stand still, and firms plan price rises of about 5%. Expected inflation is built in before the labour market has any effect.
Here is what each symbol means:
- \( \pi \) is the actual inflation rate this year.
- \( \pi^e \) is expected inflation: what workers and firms think inflation will be.
- \( u \) is the actual unemployment rate.
- \( u^* \) is the natural rate of unemployment, or NAIRU.
- \( \beta \) (beta) is a positive number measuring how strongly inflation responds to the unemployment gap. A larger \( \beta \) means a steeper curve.
Read the equation in three cases:
- If \( u = u^* \), the bracket is zero, so \( \pi = \pi^e \). Inflation is exactly what people expected. It neither rises nor falls.
- If \( u < u^* \) (a tight labour market), the bracket is negative, so \( \pi > \pi^e \). Inflation comes in above expectations.
- If \( u > u^* \) (a slack labour market), \( \pi < \pi^e \). Inflation comes in below expectations.
The NAIRU is not fixed by nature. It depends on how well job seekers and vacancies are matched, on benefits, on skills and on labour-market rules. In the United States, the Congressional Budget Office estimates the noncyclical rate of unemployment at about 4.4% in 2024–2025, down from about 4.7% in 2015[10]. Supply-side policies such as training and better job matching can lower it.
Worked example: why the trade-off disappears
The power of the expectations-augmented curve shows up when you run it over several years. The numbers below are illustrative, not real data.
Suppose the NAIRU is \( u^* = 5\% \), \( \beta = 0.5 \), and inflation has been steady at 3%, so \( \pi^e = 3\% \). People form adaptive expectations: each year they expect last year’s inflation. The government stimulates demand and holds unemployment at 4% for three years. What happens to inflation? Then, how much unemployment would it take to bring inflation back to 3% in a single year?
Holding unemployment one point below the NAIRU did not buy a stable, slightly higher inflation rate. It made inflation accelerate by half a point every year. That is why the NAIRU is called “non-accelerating”. Undoing the damage is costly: cutting inflation by 1.5 points in one year needs unemployment three points above the NAIRU. Each point of inflation removed costs 1 / 0.5 = 2 “point-years” of extra unemployment. Economists call this cost the sacrifice ratio.
So holding unemployment below the NAIRU brings ever-rising inflation: Friedman’s point that monetary policy can lower unemployment only temporarily[3][4]. And the cost of disinflation depends on how fast expectations adjust. If, in Step 5, people believed a credible announcement and cut \( \pi^e \) straight to 3%, inflation could reach 3% with unemployment at 5%.
Short run vs long run Phillips curve: the diagram in words
Each SRPC is drawn for a given level of expected inflation. Changes in demand move the economy along it; changes in expected inflation, or supply shocks such as an oil price rise, shift it. The LRPC is vertical because, once expectations adjust, \( \pi = \pi^e \) and so \( u = u^* \) at any inflation rate.
Link to AD/AS
The Phillips curve is AD/AS seen from another angle. A rise in aggregate demand moves the economy up the short-run aggregate supply (SRAS) curve: output rises, unemployment falls and prices rise, which is a movement up along the SRPC. In the long run, wages and expectations adjust, SRAS shifts left, and output returns to the vertical long-run aggregate supply (LRAS) curve. The vertical LRAS and vertical LRPC say the same thing: in the long run, demand affects prices, not output or unemployment. A negative supply shock shifts SRAS left; on the Phillips diagram the SRPC shifts up, giving higher inflation and higher unemployment.
The 1970s, stagflation and rational expectations
The 1970s broke the menu idea. In the United States, consumer price inflation reached 12.3% in 1974 while unemployment averaged 5.6%; in 1975, unemployment rose to 8.5% and inflation was still 6.9%[8][9]. By 1979–1980, inflation was above 12% for two years running, with unemployment between 5.9% and 7.2%[8][9]. High inflation and high unemployment together is called stagflation, and the simple Phillips curve said it should not happen.
The expectations-augmented curve explains it. Years of rising inflation had pushed up \( \pi^e \), shifting the SRPC up. Oil price shocks then shifted it up again, from the supply side. The Bank for International Settlements (BIS) describes the 1970s oil shock as the main historical example of a self-reinforcing move from a low-inflation to a high-inflation regime[14]. Phelps’s theory made it possible to understand why inflation and unemployment rose together during that decade[6].
Rational expectations
In the 1970s, economists such as Robert Lucas, Thomas Sargent and Neil Wallace pushed the argument further[11][12]. Under rational expectations, people use all available information, including what they know about policy, when they form \( \pi^e \). They do not just look backwards. Two strong results follow:
- A predictable monetary expansion is built into expectations at once, so it raises inflation without lowering unemployment. Only surprises move unemployment. This is sometimes called the policy ineffectiveness proposition[12].
- A credible plan to cut inflation can lower expectations directly, so disinflation need not cost much unemployment.
The early 1980s offer a mixed test. US inflation fell from 13.3% in 1979 to 3.8% in 1982, but unemployment averaged 9.7% in 1982[8][9], so disinflation was far from free. Even so, later research finds that the sharp fall in inflation in the early 1980s owed more to changing beliefs about long-run monetary policy than to a steep Phillips curve[13].
Real-world data: US unemployment and inflation, 2015–2025
Inflation below is the December-to-December change in the US Consumer Price Index (CPI-U)[8]. Unemployment is the annual average of the monthly, seasonally adjusted unemployment rate[9]. The NAIRU column is the Congressional Budget Office estimate[10].
| Year | Unemployment rate (%) | CPI inflation, Dec to Dec (%) | CBO NAIRU estimate (%) |
|---|---|---|---|
| 2015 | 5.3 | 0.7 | 4.7 |
| 2016 | 4.9 | 2.1 | 4.6 |
| 2017 | 4.4 | 2.1 | 4.6 |
| 2018 | 3.9 | 1.9 | 4.5 |
| 2019 | 3.7 | 2.3 | 4.5 |
| 2020 | 8.1 | 1.4 | 4.5 |
| 2021 | 5.3 | 7.0 | 4.4 |
| 2022 | 3.6 | 6.5 | 4.4 |
| 2023 | 3.6 | 3.4 | 4.4 |
| 2024 | 4.0 | 2.9 | 4.4 |
| 2025 | 4.3* | 2.7 | 4.4 |
Sources: US Bureau of Labor Statistics via FRED, series CPIAUCNS and UNRATE[8][9]; Congressional Budget Office via FRED, series NROU[10]. *2025 unemployment is the average of the 11 monthly readings published; no October 2025 observation is available.
- 2017–2019: the “missing inflation” puzzle. Unemployment fell below the CBO’s NAIRU estimate, to 3.7% in 2019, yet inflation stayed close to 2%. On a steep Phillips curve, inflation should have risen. This is the period that started the “flattening” debate.
- 2021–2022: the surge. Inflation hit 7.0% in 2021 and 6.5% in 2022. On a monthly basis, the 12-month CPI rate peaked at about 9.1% in June 2022[8]. Unemployment, meanwhile, fell back to 3.6%.
- 2023–2025: disinflation without a big rise in unemployment. Inflation fell to 2.7% by 2025 while unemployment rose only to about 4.3%. The old sacrifice-ratio arithmetic would have predicted a much costlier disinflation.
Is the Phillips curve flat, dead or “baaack”? The evidence
From the 1990s to 2020, inflation in many advanced economies barely responded to swings in unemployment. The BIS calls the “flattening” of the Phillips curve “a well documented stylised fact”[14]. Three main explanations compete.
1. Anchored expectations
Hazell, Herreño, Nakamura and Steinsson used price data for individual US states, going back to 1978, to estimate the slope[13]. They found the slope is small, was already small in the early 1980s, and has declined only modestly since. Their conclusion is that the greater stability of inflation from 1990 to 2020 came mainly from long-run inflation expectations becoming more firmly anchored[13]. In the equation, \( \pi^e \) stopped moving, so inflation barely moved either.
2. Structural change
The BIS argues that structural forces, such as globalisation and stronger competition, have reduced how sensitive inflation is to domestic slack. It adds that low-inflation regimes are self-stabilising, while high-inflation regimes are not[14].
3. A non-linear curve
The 2021–23 surge brought a new idea: the curve may be flat when the labour market is slack but steep when it is very tight. Benigno and Eggertsson measure tightness as the ratio of job vacancies to unemployed workers. They find significant non-linearity and argue the surge was mostly driven by an exceptionally tight labour market[17]. Economists at the Federal Reserve Bank of San Francisco estimated that the US curve becomes much steeper when the unemployment-to-vacancy ratio drops below 1.0. That implied a relatively small increase in slack could bring inflation down, provided expectations stayed anchored[18]. This fits the 2023–2025 data: inflation fell a long way while unemployment rose only a little.
What caused the 2021–23 surge?
Most research gives a two-part answer:
- Supply shocks shifted the curve. Bernanke and Blanchard find that most of the 2021 inflation surge came from price shocks, such as commodity price spikes and sector-specific price rises, rather than from overheated labour markets[16]. Ball, Leigh and Mishra, in an IMF working paper, link swings in headline inflation mainly to energy prices and supply-chain backlogs[15].
- A tight labour market moved the economy along it. Ball, Leigh and Mishra link the rise in core inflation to a tighter labour market and to pass-through from earlier headline shocks[15]. Bernanke and Blanchard find that labour-market overheating has more persistent effects on wages and inflation[16].
A quick illustrative check: suppose \( \beta = 0.5 \) and expectations stayed anchored at 2%. Using 2022 data, \( u = 3.6 \) and \( u^* = 4.4 \):
Actual inflation in 2022 was 6.5%[8]. A linear curve with anchored expectations explains only a small part of the surge; the rest must come from supply shocks, a steeper curve at low unemployment, or both.
Evaluation: how useful is the Phillips curve?
Top answers weigh points against each other rather than listing them.
Strengths
- It gives a clear, testable link between the labour market and inflation.
- The expectations-augmented version correctly predicted that persistent stimulus would cause accelerating inflation, and it explains 1970s stagflation[6].
- It shows why credibility matters. If expectations stay anchored, as they largely did in 2022–2025, disinflation can be cheaper.
Weaknesses
- The NAIRU cannot be observed. It must be estimated, and estimates move over time. The CBO’s own figure fell from about 4.7% to about 4.4% between 2015 and 2025[10]. A policymaker using the wrong NAIRU will misread the economy.
- The slope is unstable. It looked very flat before 2020 and possibly much steeper in 2021–22[13][17].
- Supply shocks dominate at times. Energy and supply-chain shocks shift the short-run curve, so unemployment alone predicts inflation poorly in those periods[15][16].
- Expectations are hard to measure. The BIS notes that household and firm expectations tend to be backward-looking, while market-based measures can be distorted by risk premia and liquidity[14].
- Unemployment is not the only measure of slack. The vacancy-to-unemployment ratio seems to explain recent inflation better than the unemployment rate on its own[17][18].
Judgement. The short-run trade-off is real but unreliable in size; the long-run trade-off does not exist. The expectations-augmented curve is most useful when combined with measures of supply shocks and expectations. How far policy can rely on it depends on credibility: the more firmly expectations are anchored, the more stable the short-run curve.
Practice questions
EasyQ1. What does the short-run Phillips curve show, and what is on each axis? Show worked answer
EasyQ2. Why is the long-run Phillips curve vertical? Show worked answer
MediumQ3. Suppose \( \pi^e = 2\% \), \( \beta = 0.4 \) and \( u^* = 5\% \). Calculate inflation when unemployment is (a) 3% and (b) 6.5%. Show worked answer
(b) \( \pi = 2 – 0.4(6.5 – 5) = 2 – 0.4(1.5) = 2 – 0.6 = 1.4\% \).
Unemployment below the NAIRU pushes inflation above expectations; unemployment above it pulls inflation below.
MediumQ4. Inflation is 6%, expected inflation is 6%, \( u^* = 4\% \) and \( \beta = 0.5 \). The central bank wants inflation of 4% next year, and expectations are adaptive (\( \pi^e \) stays at 6%). What unemployment rate is needed? What is the sacrifice ratio? Show worked answer
Check: \( 6 – 0.5(8 – 4) = 6 – 2 = 4\% \).
Inflation falls by 2 points at a cost of 4 point-years of unemployment above the NAIRU, so the sacrifice ratio is 4 / 2 = 2. If the central bank were fully credible and expectations fell straight to 4%, the cost could be close to zero.
ExamQ5. Evaluate the view that the Phillips curve is no longer a useful guide for monetary policy. [15 marks] Show worked answer
Knowledge and understanding (up to 4): definition of the short-run Phillips curve; expectations-augmented equation \( \pi = \pi^e – \beta(u – u^*) \); vertical LRPC at the NAIRU; correctly labelled diagram showing an SRPC shift when expectations rise.
Application (up to 4): real data, for example US inflation near 2% in 2019 with unemployment at 3.7%, below the CBO NAIRU estimate (flat curve); inflation of 7.0% in 2021 and 6.5% in 2022; disinflation to 2.7% in 2025 with unemployment only rising to about 4.3%.
Analysis (up to 4): anchored expectations flatten the observed curve; supply shocks shift the SRPC so unemployment alone predicts inflation poorly; a non-linear curve is steep when the labour market is very tight, which explains both the surge and the cheap disinflation.
Evaluation (up to 3): the NAIRU is unobservable and estimates change; the slope is unstable; but the expectations-augmented framework still explains stagflation and why credibility matters. A supported judgement (for example, “useful as a framework, not as a precise forecasting tool”) earns the top band.
Frequently asked questions
What is the difference between the short-run and long-run Phillips curve?
The short-run curve slopes downwards and holds expected inflation fixed, so lower unemployment comes with higher inflation. The long-run curve is vertical at the NAIRU, because once expectations adjust, unemployment returns to its natural rate at any inflation rate.
What is the NAIRU in simple terms?
It is the lowest unemployment rate the economy can sustain without inflation rising. Below it, inflation accelerates; above it, inflation slows. The CBO puts the US figure at about 4.4%.
Does the Phillips curve still exist?
Yes, in the short run, but its slope is small and unstable. It looked almost flat from the 1990s to 2020, and research suggests it may steepen sharply when the labour market is very tight, as in 2021–22.
How does stagflation contradict the Phillips curve?
It contradicts only the original curve, which ruled out high unemployment and high inflation together. In the expectations-augmented version, rising expected inflation and supply shocks shift the short-run curve up, so both can rise at once.
References
- Phillips, A. W. (1958). “The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957”. Economica, 25(100), 283–299. https://doi.org/10.1111/j.1468-0335.1958.tb00003.x
- Samuelson, P. A. and Solow, R. M. (1960). “Analytical Aspects of Anti-Inflation Policy”. American Economic Review, 50(2), 177–194. https://www.jstor.org/stable/1815021
- Lacker, J. M. (2010). “Unemployment and Monetary Policy: Lessons from Half a Century Ago”. Speech, Federal Reserve Bank of Richmond, 14 November 2010. https://www.richmondfed.org/press_room/speeches/jeffrey_m_lacker/2010/lacker_speech_20101114
- Friedman, M. (1968). “The Role of Monetary Policy”. American Economic Review, 58(1), 1–17. https://www.jstor.org/stable/1831652
- Phelps, E. S. (1967). “Phillips Curves, Expectations of Inflation and Optimal Unemployment over Time”. Economica, 34(135), 254–281. https://doi.org/10.2307/2552025
- Royal Swedish Academy of Sciences (2006). “Edmund Phelps’s Contributions to Macroeconomics”. Advanced information, Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel. https://www.nobelprize.org/uploads/2018/06/advanced-economicsciences2006.pdf
- Nelson, E. (2005). “Paul Samuelson and Monetary Analysis”. Economic Synopses, No. 8. Federal Reserve Bank of St. Louis. https://fraser.stlouisfed.org/files/docs/publications/frbsl_econosynops/economicsynopses_stls_20050401.pdf
- US Bureau of Labor Statistics (2026). “Consumer Price Index for All Urban Consumers: All Items in U.S. City Average, not seasonally adjusted (CPIAUCNS)”. Retrieved from FRED, Federal Reserve Bank of St. Louis. https://fred.stlouisfed.org/series/CPIAUCNS
- US Bureau of Labor Statistics (2026). “Unemployment Rate, seasonally adjusted (UNRATE)”. Retrieved from FRED, Federal Reserve Bank of St. Louis. https://fred.stlouisfed.org/series/UNRATE
- US Congressional Budget Office (2026). “Noncyclical Rate of Unemployment (NROU)”. Retrieved from FRED, Federal Reserve Bank of St. Louis. https://fred.stlouisfed.org/series/NROU
- Lucas, R. E. (1972). “Expectations and the Neutrality of Money”. Journal of Economic Theory, 4(2), 103–124. https://doi.org/10.1016/0022-0531(72)90142-1
- Sargent, T. J. and Wallace, N. (1975). “‘Rational’ Expectations, the Optimal Monetary Instrument, and the Optimal Money Supply Rule”. Journal of Political Economy, 83(2), 241–254. https://doi.org/10.1086/260321
- Hazell, J., Herreño, J., Nakamura, E. and Steinsson, J. (2022). “The Slope of the Phillips Curve: Evidence from U.S. States”. Quarterly Journal of Economics, 137(3), 1299–1344. https://doi.org/10.1093/qje/qjac010
- Bank for International Settlements (2022). “Inflation: a look under the hood”. Annual Economic Report 2022, Chapter II. BIS. https://www.bis.org/publ/arpdf/ar2022e2.htm
- Ball, L. M., Leigh, D. and Mishra, P. (2022). “Understanding U.S. Inflation During the COVID Era”. IMF Working Paper, 28 October 2022. International Monetary Fund. https://www.imf.org/en/Publications/WP/Issues/2022/10/28/Understanding-U-S-525200
- Bernanke, B. S. and Blanchard, O. J. (2023). “What Caused the US Pandemic-Era Inflation?”. NBER Working Paper 31417. National Bureau of Economic Research. https://www.nber.org/papers/w31417
- Benigno, P. and Eggertsson, G. B. (2023). “It’s Baaack: The Surge in Inflation in the 2020s and the Return of the Non-Linear Phillips Curve”. NBER Working Paper 31197. National Bureau of Economic Research. https://www.nber.org/papers/w31197
- Crust, E. E., Lansing, K. J. and Petrosky-Nadeau, N. (2023). “Reducing Inflation along a Nonlinear Phillips Curve”. FRBSF Economic Letter 2023-17. Federal Reserve Bank of San Francisco. https://www.frbsf.org/research-and-insights/publications/economic-letter/2023/07/reducing-inflation-along-nonlinear-phillips-curve/
