In April 2020, the United States lost 13.6% of its jobs in a single month

−13.6% Change in U.S. payroll employment, April 2020

From February 2020 to the low point, 21.9 million jobs disappeared. The long-run model of Chapters 3 to 8 has nothing to say about this. Why does the economy swing, and what do the swings do to inflation and jobs?

Open with the number. It comes from the notebook’s Fig. 4: the monthly percent change of nonfarm payroll employment was −13.57 percent in April 2020. The jobs count of the key numbers: 152.29 million at the February 2020 peak, 130.43 million at the low point, so 21.87 million jobs lost.

The contrast: the four recessions from 1980 to 2001 cost between 1.3 and 2.8 million jobs each, and the Great Recession 8.7 million.

The long-run model explains where the economy is heading on average: capital, ideas, the natural rate of unemployment, money growth and long-run inflation. Keynes’s famous complaint was that a theory of the calm sea after the storm is no help during the storm. This chapter starts the short run: the storms.

Ask the room: was April 2020 a recession like the others, or something different? Keep the question open; Chapter 10 returns to it.

Today: three questions

• How do we measure the cycle? Potential output, short-run output, and four U.S. facts: GDP, the gap, jobs and inflation.
• What does the cycle do to inflation? The Phillips curve, its fit in U.S. data, a boom that lasts, and the cost of bringing inflation down.
• What does it do to jobs, and what can policy do? Okun’s law, the chain from interest rates to output, and steering by a gap nobody observes.

These are the three acts of the deck and the three branches of the mind map on the title slide.

The headline result: output swings around potential. A boom raises inflation year after year and pushes unemployment down, so bringing inflation down takes a slump, and the slope of the Phillips curve fixes how big that slump must be in total.

This chapter is a framework, not yet a full model. It lets policymakers choose short-run output directly. Chapters 11 to 13 explain how monetary and fiscal policy actually move output.

Measuring the cycle

Act I

The long-run model sets the trend; the short-run model explains the swings

Model	What it determines	Where it comes from
long-run model	potential output \(\bar{Y}_t\) and long-run inflation \(\bar{\pi}\)	Chapters 3 to 8
short-run model	actual output \(Y_t\) and inflation \(\pi_t\), year by year	Chapters 9 to 15

• Potential output is what the economy produces when capital, labor and ideas are used at rates it can sustain. It is not a ceiling: for a while output can run above it.
• Shocks push output away from potential: oil prices, financial panics, tax changes, new technologies, disasters, booms abroad.
• The short run is how long such deviations last, typically about two years.

In the short-run model, \(\bar{Y}_t\) and \(\bar{\pi}\) are given from outside: that is what the bars mean.

The bars are the book’s notation for exogenous variables. The short-run model takes the long-run model’s answers as given and explains the deviations from them. The endogenous variables of the short run are actual output and actual inflation.

Long-run inflation has no time subscript: in the end it is a stable target chosen by policymakers (Chapter 13).

Firms can produce more than potential for a while: overtime shifts, deferred maintenance, extra temporary workers. They cannot do it for ever, and that is exactly why a boom raises inflation (Act II).

Ask: name a shock from the last five years. (The pandemic; the energy price surge of 2021–22; supply-chain disruptions; interest-rate hikes.)

Short-run output is the percent gap between actual and potential output

\[\widetilde{Y}\!_t \equiv \frac{Y_t - \bar{Y}_t}{\bar{Y}_t} \quad\Longleftrightarrow\quad Y_t = \bar{Y}_t\,(1 + \widetilde{Y}\!_t)\]

\[\widetilde{Y}\!_t \approx \ln Y_t - \ln\bar{Y}_t\]

• Boom: \(\widetilde{Y} > 0\). Slump: \(\widetilde{Y} < 0\). At potential: \(\widetilde{Y} = 0\).
• Year 2 of our baseline: potential 104.04, actual 106.12, so \(\widetilde{Y} = +2\%\).
• A recession starts when \(\widetilde{Y}\) turns negative and is dated over at its low point, while output is still below potential.

Measured in percent, a gap means the same thing in a small economy and a large one.

Left top: the definition, and the same equation solved for actual output. Actual output is the trend times a cycle factor slightly above or below one.

Left bottom: take logs of Y = Ȳ(1 + Ỹ) and use ln(1 + x) ≈ x for small x. The approximation is very good for usual gaps: the notebook’s Check 2 gives an error of 0.02 points at ±2 percent, 0.13 at −5 percent and 0.54 at −10 percent. Agencies often report gaps in logs.

The baseline of every model figure: potential output starts at 100 and grows 2 percent a year; the gap follows a wave between +2 and −2 percent that repeats every 8 years. In year 6, the bottom of the slump: potential 112.62, actual 110.36, gap −2 percent.

Other names: the output gap, detrended output. In the United States, a committee of the National Bureau of Economic Research dates recessions from many indicators, not only GDP.

Actual growth is trend growth plus the change in the gap

\[g_{Y,t} \approx \bar{g} + \Delta\widetilde{Y}\!_t\]

• Year 1 → 2: the gap rises from +1.41 to +2, so output grows by about \(2 + 0.59 = 2.59\%\).
• Year 4 → 5: the gap falls from 0 to −1.41, so output grows only \(0.59\%\): still growing, yet sliding into a slump.

Baseline year	\(\widetilde{Y}\)	growth of \(Y\), exact
1	+1.41%	3.44%
2	+2.00%	2.59%
4	0.00%	0.58%
5	−1.41%	0.56%
6	−2.00%	1.39%
8	0.00%	3.46%

Growth of 5% fits fast potential growth with a steady gap, or slow potential growth with a widening gap. Growth alone cannot tell them apart.

Write the log equation for year t and year t − 1 and subtract. The change in a log is approximately a growth rate, so actual growth is potential growth plus the change in the gap. The notebook’s Check 3 compares the exact growth rates with the approximation: they agree within 0.05 points.

Quarterly data: GDP growth is reported at annualized rates. A quarter with 1 percent growth is reported as 1.01⁴ − 1 = 4.06 percent.

The takeaway is the policymaker’s problem. We observe actual growth; we do not observe potential growth or the change in the gap. Potential output must be estimated: with a smooth trend, an average of surrounding quarters, or, like the Congressional Budget Office, a production function with labor and capital at sustainable levels. We return to this in Act III.

Experiment: potential output grows faster

Our economy’s potential output grows 2% a year, and actual output swings 2% above and below it. Now raise potential growth to 4% a year. What happens to the slump in a chart of actual output? And to short-run output \(\widetilde{Y}\)?

Collect guesses before moving the slider. A common answer: faster growth makes the slump smaller, or makes it go away. Ask students to say which panel they mean: the level of output, or the gap.

Hint if needed: write the gap for the baseline. Y − Ȳ is Ȳ times 0.02 sin(2πt/8), and we divide by Ȳ.

Faster trend growth hides the slump, but the gap does not change

Model figure: potential output \(\bar{Y}_0 = 100\) growing at \(\bar{g}\), short-run output \(\widetilde{Y}\!_t = 2\sin(2\pi t/8)\) percent. Adapted from Fig. 9.1 in Jones (2024), Macroeconomics, 6th ed.

How to read it: panel (a) draws actual output Y around potential output Ȳ over one 8-year cycle; the gray bars are the gaps Y − Ȳ at the top of the boom and the bottom of the slump. Panel (b) draws short-run output Ỹ in percent.

Do live:

1. Start at the baseline, 2 percent. Panel (b) looks like panel (a) without the trend: detrended output.
2. Move the slider to 4 percent. In panel (a) the slump nearly disappears: even at the bottom of the cycle, output keeps growing by more than 2 percent a year. Panel (b) does not move.
3. Move it to 0. Potential output is a constant 100, so the two panels show the same wave.

The message carries straight to Fig. 2: with real GDP growing about 3 percent a year, the U.S. cycle is hard to see in a chart of levels.

The trend drops out of the gap

Potential growth \(\bar{g}\)	\(\widetilde{Y}\) in year 6
0% a year	−2.00%
2% a year	−2.00%
4% a year	−2.00%

• Dividing by \(\bar{Y}_t\) cancels the growth factor \((1+\bar{g})^t\).
• Faster growth changes what we see in levels, not the cycle.
• In a slump, output can keep growing, only more slowly than potential.

A recession is a statement about the gap, not about whether output is falling.

The numbers are the notebook’s Check 1: short-run output in year 6 is −2.00 percent whatever the growth rate of potential output.

Comparative statics of the framework, from the notebook’s table: faster potential growth steepens the trend and raises actual growth one for one, but leaves Ỹ unchanged; bigger swings of the cycle change Ỹ but not the trend; a larger economy scales everything up and changes no percentage.

Bridge to the misconception on the next slide.

A recession does not require output to fall

A recession is when GDP shrinks.

In our baseline, output never shrinks: its slowest growth is 0.56% a year. Yet the economy spends 4 years of every 8 below potential. A slump is output below potential, whether or not it is falling.

The popular rule “two quarters of falling GDP” is a news shorthand, not the definition economists use. The NBER dating committee looks at employment, income, sales and industrial production, and the book’s definition is about the gap.

Why it matters for policy: in an economy whose potential grows 2 percent a year, growth of 1 percent means unemployment is rising (Okun’s law, Act III). An economy can be in a slump while its GDP sets a record every quarter.

Ask: if potential output grew 0.5 percent a year, as in Japan since 1995, what growth rate would be “normal”? (Half a percent. The same 1 percent that is weak in the United States would be a boom there.)

U.S. real GDP since 1929

Real GDP and the CBO’s potential output, since 1929. Can you see the recessions?

Let students guess what the chart will show before the reveal. Ask them to name the biggest swings of the last century: the Great Depression and World War II are the usual answers.

Ask also: on a ratio scale, what does a straight line mean? (Constant growth: Chapter 3.)

Growth dwarfs the cycle, except in the Depression and the war

Data: BEA real GDP (GDPCA) and CBO potential GDP (GDPPOT), via FRED, in 2022 dollars; 1929–1948 potential is our own smooth trend; other countries: OECD Economic Outlook No. 119. Inspired by Fig. 9.2 in Jones (2024), Macroeconomics, 6th ed.

How to read it: actual and potential real GDP, trillions of 2022 dollars, ratio scale. The buttons switch between the ratio and the linear scale; the menu picks another country (OECD data, at PPP, starting later).

Do live:

1. On the ratio scale: since about 1950 the two lines are almost on top of each other. The illusion of scale that Fig. 3 removes.
2. Hover over 1933 (output far below potential) and 1944 (far above).
3. Click “linear scale”: everything before 1960 is squeezed near the bottom.
4. Choose Greece: after 2008 output fell far below potential, to 10.6 percent below in 2012, and the OECD now estimates that potential itself shrank. Choose Japan: potential output has grown by less than 1 percent a year since 1995.

Data caveat: nobody published potential output for 1929–1948, so the early gaps are our own smooth trend. The 1929 value, 25 percent above potential, is an artefact of where the trend starts, not a real boom.

The Depression and the war were swings of a third of output

United States	Gap or growth
1933, the Depression	−23.3%
average of the 1930s	−11.4%
1944, World War II	+35.1%
largest gap since 1950 (1982)	−6.2%
potential growth 1990–2007	3.0% a year
potential growth 2007–2019	1.8% a year

• Since 1950, no slump has come close to the 1930s.
• Potential growth slowed after the Great Recession: weak investment meant less capital in the CBO’s production function.
• Real GDP in 2022 was 26.06 trillion dollars.

The trend is an estimate too, and at the edge of a sample it is least reliable. Always ask where a number comes from.

All numbers are from the notebook’s Fig. 2 key numbers.

The Depression: output 23 percent below potential in 1933, on average 11 percent below over the decade. The war: output 35 percent above potential in 1944, as the economy ran overtime to produce armaments.

The slower potential growth after 2007 matters for Act III: if potential grows more slowly than policymakers think, they will see a slump that is not there.

Ask the policy students: why would a recession lower potential output itself? (Less investment, so less capital; workers out of work for years lose skills; fewer new firms and ideas.)

U.S. economic fluctuations

Take the trend out: plot short-run output, quarter by quarter, with the recessions shaded. How deep do U.S. recessions go, and how often is the gap larger than 4%?

Ask for a number: in a typical U.S. recession, how far does output fall below potential? Students who remember 2008 or 2020 will give large numbers; the typical recession is milder.

Short-run output falls sharply in every recession

Data: BEA real GDP (GDPC1) and CBO potential GDP (GDPPOT), via FRED, today’s data and the 2023 vintage from ALFRED; NBER recession dates; other countries: OECD output gap. Inspired by Fig. 9.3 in Jones (2024), Macroeconomics, 6th ed.

How to read it: short-run output Ỹ in percent, quarterly since 1970, with the NBER recessions shaded. Two lines: today’s data and the data as published in 2023, when the textbook was written.

Do live:

1. Every recession begins near a peak of Ỹ and ends close to its trough. Hover over the troughs of 1982, 2009 and the second quarter of 2020.
2. Compare the two vintages in 2020: −8.8 percent today, −10.4 percent in the 2023 data. Click a line in the legend to hide it.
3. Click Latvia and Estonia in the legend: 18 and 14 percent above potential in 2007, then 12 and 11 percent below in 2009–10. Click Greece: a slump that lasted a decade.

Data caveat: the other countries’ points are annual OECD estimates drawn at mid-year; the shaded bands are U.S. recessions only.

The same recessions look different in two vintages of the data

Recession starts	Lowest \(\widetilde{Y}\), data of 2026	Lowest \(\widetilde{Y}\), data of 2023
1973-11	−4.72%	−4.75%
1981-07	−7.29%	−7.88%
1990-07	−2.24%	−3.60%
2001-03	−2.17%	−2.56%
2007-12	−4.98%	−5.48%
2020-02	−8.83%	−10.42%

• 92.9% of the quarters since 1970 lie between −4% and +4%.
• Today’s data put every trough higher than the data of 2023.
• Nothing about 2020 changed: we only know more now.

When you quote an output gap, say which vintage of the data it comes from.

Both columns are the notebook’s own calculation: the lowest Ỹ from the start of each recession to one year after its end. The 2023 vintage reproduces the textbook’s numbers: nearly 8 percent in 1982, more than 5 percent in 2009, more than 10 percent in 2020.

Why the numbers move: the Bureau of Economic Analysis revises real GDP as more complete data arrive, and the CBO re-estimates potential output. Potential output is never observed, so every new estimate moves the gap.

This sets up the misconception and the policy slide of Act III.

A deep slump costs a fifth of a year’s output

Slump	Years below potential	Lost output, % of a year’s GDP
1974–77	3.0	7.8
1980–87	7.7	17.2
1990–94	3.5	4.6
2001–04	3.2	4.1
2008–18	9.5	21.3
2020–21	1.2	3.4

• Add up the shortfall, quarter by quarter, until output is back at potential.
• The textbook’s typical slump, 6% of 2022 GDP, is $1.56 trillion: $4,738 per person, $18,953 for a family of four.

Long slumps cost far more than deep ones: the early 1980s and the decade after 2008.

The notebook walks through the quarters: each time Ỹ turns negative it starts adding −Ỹ/4 percent of a year’s GDP, until output is back at potential. The milder slumps around 1970, 1990–91, 2001 and 2020 cost between about 1.5 and 5 percent of a year’s GDP; the 1974–75 slump about 8 percent.

The dollar numbers use the revised 2022 GDP and the textbook’s population of 330 million. The textbook rounds to 1.5 trillion, 4,500 and 18,000 dollars; ours are a little higher because GDP was revised up.

Note the 2020 row: the deepest gap of the period, but short, so a small total cost. The 2008 slump was less deep but lasted almost ten years.

Ending the business cycle would not save the cost of a recession

If policy could abolish booms and busts, we would gain the output lost in every recession.

In our baseline, the slump years cost −4.83% of a year’s potential output, but the boom years add +4.83%: over a whole cycle the gaps sum to zero. Abolishing the cycle would also abolish the booms.

The numbers are the notebook’s Check 4: the gaps of years 5 to 7 add up to −4.83 percent; the whole cycle (years 1 to 8) sums to 0.00.

The real gain from stability is smaller than the cost of a recession, but not zero. In the data, booms and slumps are not symmetric: long slumps like 1980–87 and 2008–18 cost far more than the booms that preceded them. And the losses fall on the people who lose their jobs, while the gains of a boom are spread thinly.

For the policy students: the case for stabilization policy rests on that asymmetry and on who bears the losses, not on recovering the whole output lost in recessions.

Changes in U.S. employment

Each bar will be one month since 1980: the percent change in the number of U.S. jobs. How many jobs does a typical recession destroy, and how does 2020 compare?

Give the textbook’s reasoning as a hint if needed: a typical recession raises unemployment by 1 or 2 points; with a labor force of about 160 million, that is 1.6 to 3.2 million jobs.

Most months add jobs; 2008–09 and April 2020 stand out

Data: BLS nonfarm payroll employment (PAYEMS), via FRED, monthly percent change; other countries: OECD Economic Outlook No. 119, total employment, quarterly. Inspired by Fig. 9.4 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the monthly percent change in U.S. payroll employment since 1980. The buttons switch between the full range and a zoom on normal months; the menu picks another country (quarterly household-survey data, so not directly comparable with the monthly U.S. bars).

Do live:

1. Full range: April 2020 dwarfs everything, −13.6 percent in one month. Then the rebound: +2.0 and +3.5 percent in May and June.
2. Click “zoom on normal months”: now the small negative bars of the early 1980s, 1990–91 and 2001 appear, and the long run of negative months in 2008–09.
3. Choose Germany: in the second quarter of 2020 its employment fell by only 1.2 percent, against 12.7 percent in the United States (household survey) and 11.5 percent in Canada.

Germany’s case: firms kept their workers on short-time work, Kurzarbeit, with the government paying part of the wages. Output fell; jobs did not.

Recessions cost millions of jobs, and 2020 cost more than any other

Recession starts	Jobs lost (millions)
1980-01	1.28
1981-07	2.83
1990-07	1.56
2001-03	2.60
2007-12	8.69
2020-02	21.87

• The rule of thumb, 1 to 2 points of a 160-million labor force, gives 1.6 to 3.2 million jobs.
• It fits the four recessions from 1980 to 2001.
• The last two recessions broke it, in different ways: 2008 was long, 2020 was sudden.

Behind every point of short-run output are millions of people out of work.

The table is the notebook’s count from the peak month to the lowest employment up to two years later.

The Great Recession’s losses built up month after month for almost two years; the Covid-19 losses came in two months and were half recovered within a year.

Since 2023 monthly job gains have been small, about 0.13 percent a month in 2023 and close to zero in 2025, with a few negative months in 2025–26. Ask: does that mean a recession? (Not by itself. Slower labor-force growth lowers the normal pace of job gains: the growth of potential matters, as on the growth slide.)

Inflation and recessions

Now U.S. inflation since 1960, with the recessions shaded. During a recession, does inflation rise or fall? Where do you expect its peaks to sit relative to the shaded bands?

Collect predictions. Some students reason that a recession is a bad time, so prices must be rising; others that weak demand slows price increases. Ask them to commit before the reveal.

Inflation peaks near the start of a recession and then falls

Data: BLS CPI for All Urban Consumers (CPIAUCSL), via FRED, annual and 12-month inflation; NBER recession dates; other countries: World Bank WDI and Global Inflation Database. Inspired by Fig. 9.5 in Jones (2024), Macroeconomics, 6th ed.

How to read it: annual CPI inflation (points) and the monthly 12-month rate (thin line), with U.S. recessions shaded. The menu picks a group of countries; a click on a name in the legend adds it.

Do live:

1. The peaks of 1970, 1974, 1980 and 1990 sit at or just before a shaded band, and inflation falls during and after it.
2. Hover: the highest annual rate is 13.5 percent in 1980, the lowest −0.3 percent in 2009.
3. The 2020 recession lasted two months, so the annual points barely notice it, but the monthly line dips in spring 2020. Then the surge of 2021–22.
4. Add Japan, Germany and the United Kingdom: Japan peaked at 23.2 percent in 1974, the United Kingdom at 24.2 percent in 1975, Germany at only 7.0 percent in 1973.

Data caveat: October 2025 has no CPI, because prices could not be collected during the federal government shutdown, so the monthly line has a small gap there.

Inflation falls during recessions and usually stays lower after them

Recession starts	Inflation that year	Highest in the recession	Year after it ends
1969	5.44%	5.88%	4.23%
1973	6.26%	11.01%	5.78%
1980	13.50%	13.50%	10.38%
1981	10.38%	10.38%	3.16%
1990	5.42%	5.42%	3.04%
2007	2.87%	3.82%	1.64%
2020	1.25%	1.25%	4.68%

2020 is the exception in annual data only: from February to April 2020, 12-month inflation fell from 2.34% to 0.31%.

A slump slows inflation. That fact is the heart of the short-run model.

The table is the notebook’s Fig. 5 key numbers (1960 and 2001 are left out for space: 1.49 to 1.18 and 2.82 to 1.60).

In every recession before 2020, inflation in the year after the recession ended was lower than at its start. The Covid-19 recession lasted only two months, February to April 2020, and the monthly data show the same fall; the 2021–22 surge came after it.

The 1981 row is the Volcker disinflation, the textbook’s main example: inflation of 11 to 13 percent in 1979–80 fell to about 3 percent by 1983. Act II explains it.

Bridge: why would a slump slow inflation? That is the Phillips curve.

The Phillips curve

Act II

The short-run model rests on three premises

1. Shocks never stop. Something is always pushing output away from potential, or inflation away from \(\bar{\pi}\).
2. Policy can move output in the short run. Monetary and fiscal policy raise or lower spending for a while. In this chapter, policymakers simply choose \(\widetilde{Y}\!_t\).
3. Output and inflation trade off over time. A hot economy speeds up inflation; a cold one slows it down.

If policy can raise output, why not keep it high? The third premise is the answer.

These are the three premises of the short-run model in our words.

Premise 2 is a shortcut for this chapter only. Chapter 11 shows how the real interest rate moves spending (the IS curve), and Chapter 12 how the central bank sets the real interest rate (the MP curve).

Premise 3 is the Phillips curve, the subject of this act. The question in the takeaway is the one the textbook poses: a government that can raise output would want to keep it as high as possible, unless doing so has a cost. The cost is rising inflation.

The Phillips curve: a hot economy speeds up inflation

\[\Delta\pi_t \equiv \pi_t - \pi_{t-1} = \bar{\nu}\,\widetilde{Y}\!_t, \qquad \bar{\nu} > 0\]

• \(\bar{\nu}\) (nu) is the slope: the points by which inflation rises in a year for each percent of output above potential. For the United States, about 1/3.
• Why: when demand runs hot, firms pay overtime, run machines without a break, and goods are scarce. Each firm raises prices faster than the inflation everyone expected, so inflation itself goes up.
• In a slump, firms with idle capacity hold back their price increases to win customers.

The curve is about the change in inflation, not its level.

The textbook’s image is the hot holiday toy: its maker runs overtime and the toy is scarce, so its price rises faster than usual. In a boom, it is as if every firm made the hot toy and every month were December.

The level of inflation that persists when output is at potential can be anything: 2 percent, 5 percent, 10 percent. That level comes from the long run and from policy.

A full version in Chapter 12 adds a term for price shocks, such as oil. Ask: what would a jump in oil prices do to this line? (Shift it up: inflation rises even at potential.)

Experiment: policymakers push output above potential

Inflation is steady at 2%, and the slope is \(\bar{\nu} = 0.33\). Policymakers push output 2% above potential this year. What happens to inflation? And if instead they push it 3% below potential?

Ask for numbers, not only directions. Students should get +0.66 points and −0.99 points from the equation.

A frequent wrong answer: “inflation rises to 2.66 percent and stays there”. Ask what happens next year if output is still 2 percent above potential. (It rises again.)

Above potential inflation rises; below it falls; only at potential is it steady

Model figure: the Phillips curve \(\Delta\pi = \bar{\nu}\widetilde{Y}\) with \(\bar{\nu} = 0.33\). Adapted from Fig. 9.6 in Jones (2024), Macroeconomics, 6th ed.

How to read it: short-run output Ỹ across, the change in inflation Δπ up. The black dot is the economy at the short-run output on the slider. The title computes the change.

Do live:

1. At Ỹ = 0 the dot sits at the origin: inflation is steady.
2. Move to +2: Δπ = +0.66 points, inflation rises.
3. Move to −3: Δπ = −0.99 points, inflation falls.
4. Ask: to lower inflation by 2 points in one year, how far below potential must output be? (−2/0.33 = −6.06 percent, beyond the slider’s range.)

The line always goes through the origin: only output at potential keeps inflation steady.

The sign of the gap decides the direction of inflation

Short-run output	Change in inflation	Inflation
boom, \(\widetilde{Y} > 0\)	\(\bar{\nu}\widetilde{Y} > 0\)	rising
at potential, \(\widetilde{Y} = 0\)	0	steady, at any level
slump, \(\widetilde{Y} < 0\)	\(\bar{\nu}\widetilde{Y} < 0\)	falling

• A steeper curve (higher \(\bar{\nu}\)) rotates the line around the origin: booms raise inflation faster, and slumps lower it faster.
• A price shock shifts the whole line up.
• A different long-run inflation \(\bar{\pi}\) does not move the line.

\(\widetilde{Y} = +2\%\) adds 0.66 points to inflation; \(\widetilde{Y} = -3\%\) takes 0.99 points off.

The table is the notebook’s “What the model predicts”. The slope is the effect of one more percent of short-run output on the change in inflation; both are in percent, so it plays the role of an elasticity.

Comparative statics are the habit of the book: which line moves, and in which direction. Here: the slope parameter rotates the line, a price shock shifts it, and long-run inflation leaves it in place because the curve is about changes.

Solved forward, inflation is a running total of past gaps

\[\pi_t = \pi_0 + \bar{\nu}\sum_{s=1}^{t}\widetilde{Y}\!_s\]

• A boom that is over still leaves inflation higher, until a slump of the same total size brings it back.
• Inflation peaks where the gap crosses zero from above: at the start of a recession.

Baseline year	\(\widetilde{Y}\)	Inflation
1	+1.41%	2.47%
2	+2.00%	3.13%
3	+1.41%	3.59%
4	0.00%	3.59%
5	−1.41%	3.13%
7	−1.41%	2.00%

This is the pattern of the inflation chart: peaks at the start of each recession, falls during it.

Add π_{t−1} to both sides of the Phillips curve: π_t = π_{t−1} + ν̄Ỹ_t, a difference equation. Apply it year after year from π_0 and the gaps pile up in a sum. The notebook’s Check 4 computes inflation both year by year and with the running-total formula: they agree to three decimals.

Inflation is steady when, and only when, Ỹ = 0. That is why the Phillips curve does not pin down the level of long-run inflation.

In the baseline cycle, inflation rises in years 1 to 3, stays at 3.59 percent in year 4 where the gap is zero, and falls back to 2 percent by year 7, because the gaps of a full cycle add up to zero.

Lowering inflation takes a slump whose total size the slope fixes

\[\sum_{s=1}^{T}\widetilde{Y}\!_s = \frac{\pi_T - \pi_0}{\bar{\nu}}\]

• With \(\bar{\nu} = 0.33\), cutting inflation by 1 point takes −3.03 percent-years of output.
• \(1/\bar{\nu}\) is the sacrifice ratio.

Path to cut inflation from 2% to 1%	Ends at
one year at \(\widetilde{Y} = -3.03\%\)	1.00%
three years at \(\widetilde{Y} = -1.01\%\)	1.00%

By Okun’s law, the same disinflation costs about 1.52 point-years of extra unemployment.

Policy chooses how to spread the pain over time, not how much pain there is.

Set t = T in the running-total equation, subtract π_0 and divide by ν̄. The left side is the cumulative gap, in percent-years.

The notebook’s Check 5 runs both paths: one sharp year or three mild years both end at exactly 1.00 percent. A cold turkey disinflation and a gradual one cost the same total output in this model.

The Okun number previews Act III: half a point of unemployment per percent of output, so ½ × 3.03 ≈ 1.52 point-years.

A steeper curve makes disinflation cheaper; a flat curve makes it very expensive. That is why the slope matters so much for central banks, and why its measurement on the next slides matters.

Measuring the Phillips curve

Each point will be one U.S. year since 1961: short-run output across, the change in inflation up. Guess the slope of the best line through the cloud. Will the points sit close to it?

Ask students to write a number. The textbook’s answer is “about 1/3”. Ask also how tight they expect the cloud to be, and which years they expect far from the line. (Good guesses: the oil shocks of the 1970s and 2021–22.)

U.S. years slope up, with a slope close to 1/3

Data: BEA real GDP (GDPC1), CBO potential GDP (GDPPOT) and BLS CPI (CPIAUCSL), via FRED; other countries: OECD output gap and World Bank CPI inflation. Inspired by Fig. 9.7 in Jones (2024), Macroeconomics, 6th ed.

How to read it: one point per year since 1961, annual average Ỹ across and the change in CPI inflation up. The line is the model Δπ = ν̄Ỹ with the slope on the slider; the title adds up the squared misses, the squared vertical distances from the points to the line. The 2023–2025 points have their own color.

Do live:

1. Move the slider from 0 to 1. The squared misses are smallest near 0.33 to 0.40; the OLS slope is 0.35.
2. Hover over the points far from the line: 1974 and 1979 (oil shocks), 2021–22 (the inflation surge). Shocks move inflation too.
3. Look at 2023–2025: inflation fell while output was above potential, the opposite of the curve.
4. Click countries in the legend: Germany 0.42, Japan 0.20, France 0.16; Mexico is negative (−0.90), because in 1995 a currency crisis pushed output 6 percent below potential and raised inflation by 28 points.

The slope is about 0.35, and shocks scatter the points around it

Model slope \(\bar{\nu}\)	Squared misses
0	199.7
0.25	166.4
0.33	163.6
0.40	164.3
0.50	170.4
1.00	290.7

• The OLS line is \(\Delta\pi = 0.10 + 0.35\,\widetilde{Y}\) (1961–2025).
• Even the best line misses a lot: output is not the only thing moving inflation.
• Revisions move points too. In the 2023 data, 2020 sat at \(\widetilde{Y} = -4.78\%\); today it sits at −3.28%.

The Phillips curve gets the direction right, but price shocks push many years far off the line.

The squared misses come from the slider of the live figure; the OLS fit from the notebook’s np.polyfit.

The drop in squared misses from ν̄ = 0 to the best slope is modest, from 199.7 to 163.6. Most of the variation in the change of inflation is not explained by the output gap alone. Chapter 12 adds price shocks to the curve, which is where most of the remaining variation lives.

With today’s data, 2021–22 moved from slightly below potential (−0.86, −0.67 in the 2023 data) to about 1 percent above it (0.81, 0.92). Data revisions can move a point across the vertical axis.

Hold the 2023–25 puzzle for the Devil’s advocate at the end.

Should a central bank cause a recession to end high inflation?

Inflation was 11.25% in 1979. Volcker’s Fed raised interest rates sharply; output fell 6.22% below potential in 1982, and inflation was 3.16% by 1983. Was the slump worth it?

What the model says

• Lowering inflation requires a slump.
• Right direction in 1981–83; inflation fell even faster than predicted.
• The early-1980s slump cost 17.2% of a year’s GDP.

What is debated

• Credibility: if people believe the bank, the slump can be smaller.
• Speed: cold turkey or gradual? Here the total is the same.
• Who pays: the jobless, not the savers.

Numbers: the notebook’s Volcker check. Predicted changes in inflation with ν̄ = 0.33 against the data: 1981, Ỹ −1.82, predicted −0.60, actual −3.12; 1982, Ỹ −6.22, predicted −2.05, actual −4.22; 1983, Ỹ −4.80, predicted −1.58, actual −3.00. Inflation 11.25 percent in 1979, 13.50 in 1980, 3.16 in 1983. Lost output of the slump from 1980 to 1987: 17.2 percent of a year’s GDP.

Left column: the model’s prediction is a smaller fall than happened, a sign that something else was moving inflation down too, such as falling oil prices and expectations.

Right column, the research. Sargent (1982, in Hall, ed., Inflation: Causes and Effects) argues that the great hyperinflations ended quickly and cheaply because the policy change was credible. Ball (1994, in Mankiw, ed., Monetary Policy, University of Chicago Press) measures sacrifice ratios across disinflations in OECD countries and finds that fast disinflations had smaller ratios. Goodfriend and King (2005, Journal of Monetary Economics) call the Volcker episode an “incredible” disinflation: the Fed was not believed at first, which made the slump deep. These are contested findings; the simple Phillips curve of this chapter has no role for credibility, which Chapter 12 adds through expected inflation.

Ask the policy students: if you could promise a smaller slump by being believed, how would you make a central bank believable?

Overstimulating the economy

The economy starts at potential with inflation at 2%. Policymakers then hold output 3% above potential every year, with a slope of \(\bar{\nu} = 1\). Where is inflation after three years? Does it settle at a new, higher rate?

This is the textbook’s exercise “Overstimulating the economy”. Ask for the three-year number and for a sketch of the path. Many students expect inflation to jump to a higher level and stay there, because they read the Phillips curve as a curve in the level of inflation.

A boom held in place raises inflation every year, without end

Model figure: \(\pi_t = \pi_{t-1} + \bar{\nu}\widetilde{Y}\) with \(\widetilde{Y} = +3\%\) and \(\pi_0 = \bar{\pi} = 2\%\). Adapted from Fig. 9.9 in Jones (2024), Macroeconomics, 6th ed., the worked answer to Exercise 3; the simulated path is our own addition.

How to read it: panel (a) is the Phillips curve with the economy at potential and in the boom (two black dots). Panel (b) is the inflation path the boom produces, year by year, against long-run inflation π̄ (dashed).

Do live:

1. At ν̄ = 1: inflation goes 2 → 5 → 8 → 11 percent in three years. The title shows the path.
2. Move the slope to 0.33, the U.S. value: 2 → 2.99 → 3.98 → 4.97 percent. Slower, but the line in panel (b) still climbs without end.
3. Move to 1.5: 2 → 6.5 → 11 → 15.5 percent.

The dot in panel (a) does not move once the boom is in place: the same Δπ every year, so inflation climbs in a straight line.

A permanent boom buys ever-rising inflation

Year	Inflation, \(\bar{\nu} = 1\)	Inflation, \(\bar{\nu} = 0.33\)
0	2.00%	2.00%
1	5.00%	2.99%
2	8.00%	3.98%
3	11.00%	4.97%
5	17.00%

• A gap held at \(\widetilde{Y}\) adds \(\bar{\nu}\widetilde{Y}\) points every year: \(\pi_t = \pi_0 + \bar{\nu}\widetilde{Y}\,t\).
• A flatter curve slows the climb; it does not stop it.
• Run it backwards: a gap held below zero brings inflation down, which is how a recession fights inflation.

There is no lasting trade-off: output above potential cannot be bought with a one-time rise in inflation.

The ν̄ = 1 column is the notebook’s simulation: [2, 5, 8, 11, 14, 17] for years 0 to 5. The ν̄ = 0.33 column is the slider’s title at that slope (the U.S. slope of about 1/3 gives about 1 point a year).

The algebra is the running total with a constant gap: t copies of the same Ỹ.

Exercise for the room, from the notebook’s “Your turn”: inflation starts at 6 percent and ν̄ = 0.5. What constant gap brings it to 3 percent in exactly three years? (−1 point a year, so Ỹ = −2 percent.)

Running the economy hot cannot last

If policymakers keep the economy a little above potential, they get more output and jobs for good, at the price of slightly higher inflation.

A gap held at \(+3\%\) with \(\bar{\nu} = 1\) takes inflation from 2% to 11% in three years and 17% in five. The price is not a higher inflation rate but an inflation rate that rises every year.

The myth is the old reading of the Phillips curve as a menu of inflation levels and unemployment rates: choose a little more inflation and get a little less unemployment, for ever. The curve of this chapter is in the change of inflation, so there is no such menu.

Running hot for a while can make sense: after a deep slump, or when policymakers are unsure where potential output is. Running hot for ever cannot.

Bridge to the policy slide on the 1970s.

Can policy buy lasting low unemployment with a little inflation?

One reading of the 1970s: policymakers held output above potential year after year, thinking inflation cheap or the trade-off permanent. What should a government that wants more jobs learn from that?

The lesson of the model

• A lasting boom raises inflation every year.
• It then takes a slump to undo, as in 1981–83.
• The boom is borrowed, not bought.

Why it was hard to see

• Potential was overestimated in real time, so stimulus looked right.
• Oil shocks raised inflation too.
• Once people expect inflation, the trade-off is gone.

The model lesson is Fig. 9.

Right column, the research. Friedman (1968, American Economic Review) and Phelps (1967, Economica) argued, before the 1970s, that no permanent trade-off between inflation and unemployment exists once expectations adjust: the natural-rate hypothesis. Orphanides (2003, Journal of Monetary Economics) uses the Fed’s real-time data and argues that 1970s policy looked reasonable given the output gaps believed at the time, which later turned out to be far too negative. Romer and Romer (2002, Federal Reserve Bank of Kansas City, Jackson Hole symposium) argue instead that policymakers’ beliefs about the economy, including an optimistic view of the natural rate of unemployment, drove the policy. The weight of each explanation is debated.

The slide adds no numbers beyond the book’s: the oil shocks show up in Figs. 5 and 7.

Ask: today’s version of the question is “running the economy hot” after a deep slump to pull people back into work. When is that a good idea? (When output is clearly below potential; the risk is that potential is lower than thought.)

Okun’s law and policy

Act III

Okun’s law: each point of output gap is half a point of unemployment

\[u_t - \bar{u} = -\tfrac{1}{2}\,\widetilde{Y}\!_t\]

Short-run output \(\widetilde{Y}\)	+2%	0	−2%	−4%
Cyclical unemployment \(u - \bar{u}\) (points)	−1	0	+1	+2

• \(\bar{u}\) is the natural rate of unemployment (Chapter 7), the labor-market cousin of potential output.
• The slope is less than 1 in size: in a slump firms also cut hours and keep some workers, and some job losers stop looking for work.

Okun’s law lets us work with output alone and translate to unemployment whenever we need to.

The table is the notebook’s “What the law predicts”. Okun’s law is a rule of thumb fitted to U.S. data, named after Arthur Okun.

Going backwards: multiply by −2. Output 2 percent below potential means unemployment 1 point above the natural rate; unemployment 1 point below the natural rate signals output 2 percent above potential.

Why less than one-for-one: labor hoarding (firms keep workers they will need later), shorter hours, and discouraged workers who leave the labor force and so are not counted as unemployed. Japan’s very small slope, on the next slides, is the extreme case.

Unemployment rises whenever output grows more slowly than potential

\[\Delta u_t \approx -\tfrac{1}{2}\left(g_{Y,t} - \bar{g}\right)\]

• Baseline, year 4 to 5: output grows 0.56%, so unemployment rises by about +0.7 points.
• Steady growth at the trend rate only holds unemployment where it is.

\[\Delta\pi_t = -2\bar{\nu}\,(u_t - \bar{u})\]

• The Phillips curve in terms of unemployment.
• With \(\bar{\nu} = 0.33\): each point of unemployment above the natural rate lowers inflation by 0.66 points a year.

Growth below potential growth, not negative growth, is what raises unemployment.

Left: write Okun’s law for two years and subtract; the natural rate cancels. Then replace ΔỸ by g_Y − ḡ from the growth slide of Act I. The notebook’s Check 6 gets +0.71 points from the levels and +0.72 from the growth form; the difference is the approximation.

Right: substitute Ỹ = −2(u − ū) into the Phillips curve. This is the form in which the trade-off was first found in the data, with unemployment on the horizontal axis: the history is in Chapter 12.

Ask: potential output grows 2 percent a year and GDP grows 1 percent. What happens to unemployment? (It rises by about half a point.)

Okun’s law in U.S. data

Each point will be one U.S. year since 1960: short-run output across, unemployment minus the CBO’s natural rate up. How tight will this cloud be, compared with the Phillips curve? Which year will sit furthest from the line?

Ask students to compare with the loose Phillips cloud they just saw, and to name a year that might break the rule. (2020 is the usual guess, rightly.)

Output and unemployment move together tightly

Data: BEA real GDP (GDPC1), CBO potential GDP (GDPPOT) and natural rate (NROUST), BLS unemployment rate (UNRATE), via FRED; other countries: OECD output gap and unemployment, our own trend. Inspired by Fig. 9.8 in Jones (2024), Macroeconomics, 6th ed.

How to read it: one point per year since 1960, short-run output across and cyclical unemployment (the unemployment rate minus the CBO’s natural rate) up. The line is the model u − ū = slope × Ỹ, with the slope on the slider; the title adds up the squared misses.

Do live:

1. At the textbook’s −0.5: squared misses 33.34. Move to −0.6: 28.59, the best value on the slider. The OLS slope is −0.62.
2. Hover over 2020: far above the line. Unemployment jumped much more than the fall in output alone predicts, as businesses closed and laid off workers.
3. Click Spain: slope −0.59, cyclical unemployment swinging by up to 6 points either way. Click Japan: only −0.13, Japanese firms keep their workers through slumps.

Data caveat: for the other countries the natural rate is a smooth trend through unemployment, which gives smaller slopes: for the United States, −0.38 with the trend against −0.62 with the CBO’s natural rate.

The Okun line fits far better than the Phillips line, but 2009 and 2020 break it

Okun slope	Squared misses
−1.0	69.62
−0.6	28.59
−0.5	33.34
0	147.05

Year	\(\widetilde{Y}\)	Okun predicts \(u - \bar{u}\)	Actual
2009	−4.66%	+2.33	+3.94
2020	−3.28%	+1.64	+3.60

• The right slope cuts the squared misses by four fifths; for the Phillips curve, by less than a fifth.
• In both recent recessions unemployment rose far more than the rule predicts.

Output and unemployment tell the same story most years, so the short-run chapters can focus on output.

Numbers: the slider titles of the live figure and the notebook’s Okun key numbers.

Four fifths: from 147.05 with a zero slope to 28.59 at −0.6. For the Phillips curve: from 199.7 to 163.6, less than a fifth.

Part of the misses in 2009 and 2020 comes from the data revisions of the Phillips-curve slide, which moved both years to the right. The textbook’s own example, output about 6 percent below potential with unemployment just over 3 points above normal, fits the rule exactly.

Ask: why might unemployment rise more than usual in a recession that hits services, such as 2020? (Restaurants and shops cannot hoard labor when they are closed.)

The full short-run model: from the interest rate to output, inflation and jobs

\[\underbrace{R_t}_{\text{real interest rate}} \;\xrightarrow{\ \text{IS curve}\ }\; \widetilde{Y}\!_t \;\xrightarrow{\ \text{Phillips curve}\ }\; \Delta\pi_t, \qquad \widetilde{Y}\!_t \;\xrightarrow{\ \text{Okun's law}\ }\; u_t - \bar{u}\]

• IS curve (Chapter 11): a higher real interest rate lowers investment and other spending, and so output.
• MP curve (Chapter 12): the central bank sets a short-term nominal rate, and with it the real rate.
• Signs: a higher \(R_t\) → lower \(\widetilde{Y}\!_t\) → falling inflation and higher unemployment.

This chapter skipped the first link by letting policymakers pick \(\widetilde{Y}\!_t\) directly.

The road map of the next chapters. Chapter 10 applies the framework of this chapter to the Great Recession and the Covid-19 pandemic. Chapters 11 to 13 build the full model: the IS curve, the MP curve and the Phillips curve together. Chapter 14 returns to the financial crisis, and Chapter 15 to modern DSGE models.

What the full model predicts, in four lines: policy can lean against shocks and keep the gap near zero, which keeps inflation steady; it cannot keep output above potential without inflation climbing; high inflation comes down only through a slump; and the job is harder than in the model because potential output is not observed.

The output gap is an estimate, not a measurement

The output gap is a statistic, like GDP or unemployment.

Potential output is never observed. The 2020 trough moved from −10.42% to −8.83% between two vintages of the data. The CBO and the OECD agree on every big swing, but their U.S. gaps correlate only 0.80, and in 2025 they disagree on the sign: −0.06% for the OECD, +1.40% for the CBO.

Numbers: the notebook’s Fig. 3 key numbers (vintages) and “two agencies, two estimates” (correlation 0.80 for 1985–2025, largest difference 2.0 points in 1986).

GDP is measured, with errors and revisions. Potential output is the output of a model: a smooth trend, an average of nearby quarters, or a production function. Different models give different gaps, and the same model gives a new gap every time the data are revised or a new year is added at the end of the sample.

For other countries every gap in the book is the OECD’s estimate, from its own model.

Should policy steer by an output gap it cannot see?

Central banks set interest rates partly on the size of the output gap. The gap is estimated, revised for years, and differs across agencies. How much weight should policy put on it?

Lean on the gap

• Inflation responds to the gap: without an estimate, policy flies blind.
• The direction of big swings is clear in every vintage.
• Unemployment, which is observed, cross-checks it through Okun’s law.

Lean less on it

• Real-time gaps have been revised by as much as the gaps themselves.
• An overestimate of potential calls for too much stimulus, as in the 1970s.
• Respond more to observed inflation and unemployment.

The research. Orphanides and van Norden (2002, Review of Economics and Statistics) show that real-time estimates of the U.S. output gap are unreliable, and that most of the problem is not data revisions but the difficulty of knowing the trend at the end of the sample. Orphanides (2003, Journal of Monetary Economics) connects that to the 1970s. Staiger, Stock and Watson (1997, Journal of Economic Perspectives) show that the natural rate of unemployment is also measured with wide uncertainty, so the Okun cross-check is not free of the same problem.

The left column’s claims are the book’s: the vintages agree on direction (the trough table of Act I); Okun’s law links the gap to unemployment.

The right column’s first claim is the research above; the slide adds no outside numbers.

Ask the policy students: if you had to set interest rates tomorrow, which three numbers would you look at? (Inflation, unemployment, and some estimate of the gap; and you would ask how often each has been revised.)

“The Phillips curve is dead”

A skeptic says: “Look at the cloud. 1974, 1979 and 2021–22 sit far above the line. In 2023–2025 inflation fell while output was above potential. Even the best line explains less than a fifth of the variation. A curve that misses this often cannot guide policy.”

Reply: over 65 years the cloud still slopes up, about 0.35, and the big misses are price shocks that the full model adds as a shift of the curve. Every large disinflation in the data came with a slump, 1981–83 above all. A flat, noisy curve is a reason to expect costly disinflations, not a reason to ignore the gap.

Take the objection seriously. The simple curve of this chapter has no price shocks and no expected inflation, and a flat curve with large shocks is hard to see in a scatter plot.

The research. Blanchard (2016, American Economic Review Papers and Proceedings) finds that the U.S. Phillips curve has become flatter but is still there, with inflation now anchored by expectations. Hooper, Mishkin and Sufi (2020, Research in Economics) ask whether the curve is dead or “hibernating” and find a stronger relation in city- and state-level data and in periods of high labor-market pressure. Hazell, Herreño, Nakamura and Steinsson (2022, Quarterly Journal of Economics) estimate the slope from U.S. states and find it small but positive, and argue that the fall of inflation in the 1980s owed more to changing expectations than to the slope. These are contested readings of a hard identification problem.

The book’s own numbers for the reply: OLS slope 0.35 for 1961–2025; squared misses fall from 199.7 to 163.6; the Volcker years all moved in the model’s direction.

Ask: if the curve is flat, is that good or bad news for a central bank fighting inflation? (Bad: the sacrifice ratio 1/ν̄ is large. Good: a boom raises inflation only slowly.)

What to remember from this chapter

• Short-run output is the percent gap from potential, \(\widetilde{Y} = (Y - \bar{Y})/\bar{Y}\). Actual growth is potential growth plus the change in the gap.
• U.S. recessions push output a few percent below potential, cost millions of jobs and up to a fifth of a year’s GDP. The measured gap changes when the data are revised.
• Inflation usually falls in a recession. The Phillips curve, \(\Delta\pi = \bar{\nu}\widetilde{Y}\) with a U.S. slope near 1/3, makes inflation a running total of past gaps.
• A lasting boom raises inflation every year; cutting inflation by a point costs \(1/\bar{\nu}\) percent-years of output.
• Okun’s law, \(u - \bar{u} = -\tfrac{1}{2}\widetilde{Y}\): unemployment rises whenever growth falls short of potential growth.

Ask each student to write one sentence answering the opening question: why does output swing around its trend, and what do booms and slumps do to inflation and jobs? A good answer: shocks push output away from potential; a boom pushes inflation up year after year and unemployment down; a slump does the opposite, and bringing inflation down requires one.

For the policy students, three habits: ask whether output is growing faster or slower than potential, not only whether it is growing; ask which vintage and which agency a gap comes from; and treat any promise of a lasting boom as a promise of rising inflation.

Walk the three branches of the mind map in about a minute each: measuring the cycle (cyan), the Phillips curve (lavender), Okun’s law and policy (green). Click the image to zoom in.

Output swings around potential, and inflation keeps a running score of the swings

Next: Chapter 10, The Great Recession and the Covid-19 Pandemic: A First Look

This chapter set up the short-run frameworks: the gap, the Phillips curve and Okun’s law. Chapter 10 uses them on the two biggest shocks of recent decades, the global financial crisis and the pandemic. Chapters 11 to 13 then explain how policy moves output.

Sources: data and models

• Figs. 2 and 3, output and the gap. BEA, real GDP (GDPCA, GDPC1) and GDP deflator (A191RD3A086NBEA); CBO, potential GDP (GDPPOT), via FRED; 2023 vintage from ALFRED; NBER business-cycle dates (USREC; NBER dating page); OECD, Economic Outlook No. 119 (GDPVD, GDPVTR, GAP), OECD Data Explorer. Inspired by Figs. 9.2 and 9.3 in Jones (2024).
• Fig. 4, employment. BLS, nonfarm payroll employment (PAYEMS), via FRED; OECD, Economic Outlook No. 119, total employment (ET). Inspired by Fig. 9.4 in Jones (2024).
• Figs. 5 and 7, inflation and the Phillips curve. BLS, CPI for All Urban Consumers (CPIAUCSL), via FRED; World Bank, World Development Indicators (FP.CPI.TOTL.ZG) and the Global Inflation Database. Inspired by Figs. 9.5 and 9.7 in Jones (2024).
• Fig. 8, Okun’s law. BLS, unemployment rate (UNRATE); CBO, noncyclical rate of unemployment (NROUST), via FRED; OECD unemployment rate (UNR), with our own Hodrick–Prescott trend. Inspired by Fig. 9.8 in Jones (2024).
• Models, Figs. 1, 6 and 9. Short-run output, the Phillips curve and Okun’s law follow Sections 9.2 to 9.4 of Jones (2024); the illustrative baseline, the running-total and disinflation-cost derivations and the simulated inflation path are the book chapter’s own.
• Textbook. Jones, C. I. (2024). Macroeconomics, 6th ed. W. W. Norton.
• Book chapter, with the data, code and every figure. Data Science and Macroeconomic Models, Chapter 9.

Sources: research cited

• Disinflation and credibility. Sargent, T. (1982), “The Ends of Four Big Inflations”, in R. Hall (ed.), Inflation: Causes and Effects, University of Chicago Press; Ball, L. (1994), “What Determines the Sacrifice Ratio?”, in N. G. Mankiw (ed.), Monetary Policy, University of Chicago Press; Goodfriend, M. and R. King (2005), Journal of Monetary Economics 52(5).
• The natural rate and the 1970s. Phelps, E. (1967), Economica 34(135); Friedman, M. (1968), American Economic Review 58(1); Orphanides, A. (2003), Journal of Monetary Economics 50(3); Romer, C. and D. Romer (2002), “The Evolution of Economic Understanding and Postwar Stabilization Policy”, Federal Reserve Bank of Kansas City, Rethinking Stabilization Policy.
• Measuring the gap in real time. Orphanides, A. and S. van Norden (2002), Review of Economics and Statistics 84(4); Staiger, D., J. Stock and M. Watson (1997), Journal of Economic Perspectives 11(1).
• Is the Phillips curve dead? Blanchard, O. (2016), American Economic Review 106(5); Hooper, P., F. Mishkin and A. Sufi (2020), Research in Economics 74(1); Hazell, J., J. Herreño, E. Nakamura and J. Steinsson (2022), Quarterly Journal of Economics 137(3).