In April 2020, about one in seven U.S. workers had no job

14.8% The U.S. unemployment rate in April 2020, up from 3.5% in February

Work is the main source of income for most people. What sets its price, why do some people who want a job not find one, and why are some paid so much more than others?

Open with the number. Both rates come from the notebook’s Fig. 2 key numbers: 3.5 percent in February 2020 and 14.8 percent in April 2020, the highest peak in the chart of this act.

Ask the room: how many of you, or of your families, were affected? Most people live mainly on what they earn from work, so the labor market is the market that matters most to them.

Link back: Chapters 3 to 6 explained income per person, which is mostly the wage. This chapter looks inside the labor market itself: how many people work, at what wage, and why some cannot find a job. It is the last long-run chapter before inflation, and the rigid wages here come back in the short-run chapters.

Today: three questions

• Who works, and at what wage? The facts on who has a job, and a supply-and-demand model of the labor market with taxes and demand shocks.
• Why are some people unemployed? Recessions, wages that do not fall, the flows into and out of jobs, and Europe.
• Why do some earn more? What a working life is worth, the rising college premium, manufacturing jobs and inequality.

These are the three acts of the deck and the three branches of the mind map on the title slide. Act I builds the supply-and-demand model and uses it for an income tax and a fall in labor demand. Act II adds the two ways the model produces unemployment: a wage that does not fall, and the constant churn of job losses and job findings. Act III uses the same model twice, in the markets for high school and college graduates, and ends with the data on inequality.

The headline result: wages and employment follow supply and demand; the natural rate of unemployment follows the flows of job losses and job findings; and the college premium rose because the demand for skills grew faster than their supply.

Who works, and at what wage

Act I

Every adult is employed, unemployed or out of the labor force

Ratio	Formula	What it measures
unemployment rate	\(u = U/LF\)	the share of the labor force that looks for work and has none
participation rate	\(LF/N\)	the share of adults who work or want to work
employment-population ratio	\(E/N\)	the share of all adults who have a job

The labor force is \(LF = E + U\). Students, retirees, carers and people who stopped searching are outside it.

The definitions follow the U.S. Bureau of Labor Statistics: N is the population aged 16 and over; E has a paid job; U has no job, has looked for one in the past four weeks and could start now. Everyone else is not in the labor force.

The key word is “searching”. A person who wants a job but has stopped looking is not counted as unemployed. These are the discouraged workers, and they matter for the next slide.

Ask: is a full-time student who works ten hours a week employed? (Yes. Any paid work counts.) Is a retiree who would take a job if offered one unemployed? (No, unless she is actively searching.)

Three statistics, one identity

\[\frac{E}{N} = (1-u)\times\frac{LF}{N}\]

100 adults	Before	One gives up searching
labor force \(LF\)	62	61
unemployed \(U\)	3	2
unemployment rate \(u\)	4.84%	3.28%
\(E/N\)	0.59	0.59

• \(E/N\) falls if more of the labor force is jobless or if fewer adults are in it.
• The unemployment rate sees only the first.
• One more point of \(u\) lowers \(E/N\) by about 1%.

The identity comes from writing E/N = (E/LF)(LF/N) and E = LF − U. Taking logs, ln(E/N) = ln(1 − u) + ln(LF/N), and ln(1 − u) is about −u when u is small: a one-point rise in unemployment lowers the employment-population ratio by about 1 percent.

The example is the notebook’s Check 1: 100 adults, a labor force of 62 and 3 unemployed. Both sides of the identity give 0.59. When one jobless person stops searching, U falls to 2 and LF to 61: the unemployment rate drops from 4.84 to 3.28 percent, and nobody has found a job.

This is why the chapter, and the news, should track both statistics.

A lower unemployment rate does not always mean more jobs

When the unemployment rate falls, more people must have found work.

Not necessarily. In the example, unemployment fell from 4.84% to 3.28% because one person gave up searching. Nobody found a job, and \(E/N\) stayed at 0.59.

The myth confuses a falling rate with rising employment. The unemployment rate falls when the unemployed find jobs, but also when they leave the labor force: they retire, return to school, or stop looking because they see no chance.

In deep recessions, part of the fall in unemployment during the recovery comes from people leaving the labor force. That is why economists also watch the employment-population ratio, Fig. 1, and the participation rate.

For the policy students: a program that pushes people out of the labor force, such as easier early retirement or disability benefits, can lower the unemployment rate while lowering employment. Judge it by E/N, not by u alone.

The share of adults with a job

Between 1960 and 2000, the share of U.S. adults with a job rose from 56% to almost 65%. Who were the new workers?

Let students guess before the reveal. Most will name women; some will mention the baby boom generation reaching working age.

The aggregate numbers, from the notebook’s key numbers: 56.1 percent on average in 1960 and a peak of 64.7 percent in April 2000.

Ask them also to predict what happened to men’s employment-population ratio over the same years.

Many more women and fewer men work than in the 1950s

Data: U.S. Bureau of Labor Statistics, Current Population Survey, employment-population ratios of men (LNS12300001) and women (LNS12300002), via FRED; NBER recessions. Inspired by Fig. 7.1 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the monthly employment-population ratios of men and women aged 16 and over, in percent, since 1948. The shaded bands are NBER recessions.

Do live:

1. Hover over 1960 and 2000 on each line: women rose from about 35 to about 58 percent, men fell from about 79 to about 72.
2. Point at the shaded bands: both lines fall in every recession, men’s more steeply, because recessions hit construction and manufacturing hardest.
3. Point at spring 2020: the deepest and shortest fall for both.

Data caveat: October 2025 is missing, because the survey was not carried out during that autumn’s federal government shutdown. The chart joins September and November.

The rise in employment came from women

Year	Men	Women
1960	78.9%	35.5%
2000	71.9%	57.5%
2019	66.6%	55.4%
2025	64.9%	54.9%

All adults	Ratio
1960 average	56.1%
peak, April 2000	64.7%
December 2006	63.4%
January 2010	58.5%
April 2020	51.2%
September 2026	59.2%

Women’s ratio peaked at 58.0% in April 2000. Since 2000 neither line has recovered its peak.

Yearly averages by sex; the right table is the monthly series of all adults. All numbers come from the notebook’s key-number cells.

In the supply-and-demand language of the next slides, the rise of women’s employment came from both sides. Supply: changing social norms, more education and the contraceptive pill (Goldin and Katz 2002, Journal of Political Economy; Goldin 2006, American Economic Review). Demand: less workplace discrimination and the growth of service jobs.

Men’s ratio has fallen for decades: longer schooling, earlier retirement, disability, and the decline of manufacturing (Act III).

The Great Recession cut the ratio by 4.9 points, from 63.4 percent in December 2006 to 58.5 percent in January 2010, and it took years to recover. One point of the ratio is about 2.7 million jobs.

Jobs are not a fixed pie

When women (or immigrants) enter the labor market, they take jobs that men would otherwise have held.

If the number of jobs were fixed, more women at work would mean fewer jobs in total. Instead the share of all adults with a job rose, from 56.1% in 1960 to a peak of 64.7% in 2000, while women’s ratio rose from 35.5% to 57.5%.

Economists call the myth the “lump of labor” fallacy: the belief that there is a fixed amount of work to share out. It shows up in debates on immigration, on retirement ages (“older workers should retire to make room for the young”) and on shorter working weeks.

In the model of the next slides, more people willing to work shift labor supply out. Employment rises, and the wage falls a little at first. But new workers also produce and spend, firms invest, and in the long run labor demand grows with productivity and capital.

Men’s ratio did fall over the same years, from 78.9 to 71.9 percent. Its causes (schooling, retirement, disability, the decline of manufacturing) are separate, and the fall began long before women’s employment surged.

Card (1990, ILR Review) studies the Mariel boatlift, which added many Cuban migrants to Miami’s labor force within months, and finds little effect on the wages or unemployment of other workers. The size of such effects is still debated.

Labor demand slopes down; labor supply slopes up

Demand: firms

• Firms hire until the marginal product of labor, which falls as they hire, equals the wage.
• \(w = \bar{a}_d - \bar{b}_d L\)

Supply: people

• Leisure costs the wage it could earn, so a higher wage draws more people into work.
• \(w = \bar{a}_s + \bar{b}_s L\)

\[L^* = \frac{\bar{a}_d - \bar{a}_s}{\bar{b}_d + \bar{b}_s}, \qquad w^* = \frac{\bar{b}_s\bar{a}_d + \bar{b}_d\bar{a}_s}{\bar{b}_d + \bar{b}_s}\]

The demand curve comes from the production function of Chapter 4: with Y = ĀK̄^(1/3)L(2/3), the marginal product of labor is (2/3)ĀK̄^(1/3)L(−1/3), which falls as L rises. The textbook draws the curves without equations; the straight lines are the simplest curves with the right slopes, so we can compute with them.

The equilibrium comes from setting the two wages equal. The wage is a weighted average of the two intercepts, each weighted by the slope of the other curve.

The book’s baseline values: ā_d = 100, b̄_d = 1, ā_s = 10 and b̄_s = 1. So L* = 90/2 = 45 and w* = 110/2 = 55. These are illustrative numbers, not data: they make the diagrams easy to read.

When firms value workers more

Firms come to value every worker more: the demand intercept rises from 100 to 120. What happens to the wage, and to employment?

Ask students to sketch the shift before the reveal: the demand curve moves up, parallel to itself.

Most will get the direction of both changes right. Ask the harder question: how much of the shift shows up in the wage, and how much in employment? (With equal slopes, half and half.)

A demand shift moves the wage and employment the same way

Model figure: linear labor demand and supply with \(\bar{a}_d = 100\), \(\bar{b}_d = 1\), \(\bar{a}_s = 10\) and \(\bar{b}_s = 1\); the slider moves the demand intercept. Adapted from Fig. 7.3 in Jones (2024), Macroeconomics, 6th ed.

How to read it: employment L on the horizontal axis and the wage w on the vertical, as in the textbook. The dashed gray lines are the starting curves; the solid lines move with the slider. The title reports the new equilibrium.

Do live:

1. Start at 100: the market clears at L* = 45 and w* = 55.
2. Move to 120: w* = 65 and L* = 55. Each extra unit of the intercept adds half a unit to both.
3. Move to 80: both fall, to 45 and 35.

Ask: which real-world changes shift labor demand up? (Higher productivity, more capital per worker, a higher price for what the firm sells.)

Demand shocks move \(w\) and \(L\) together; supply shocks move them apart

\(\bar{a}_d\)	\(w^*\)	\(L^*\)
80	45	35
90	50	40
100	55	45
110	60	50
120	65	55

A 1% rise in …	\(L^*\)	\(w^*\)
\(\bar{a}_d\): demand up	+1.11%	+0.91%
\(\bar{a}_s\): supply up	−0.11%	+0.09%
\(\bar{b}_d\): steeper demand	−0.5%	−0.41%
\(\bar{b}_s\): steeper supply	−0.5%	+0.41%

Reading the data: when wages and employment rise together, labor demand has grown.

The left table is the slider of Fig. 3. The right table holds the elasticities of the equilibrium at the baseline values, from taking logs of the closed form; the notebook’s Check 2 reproduces them with a 1 percent change in each parameter (+1.111, +0.909; −0.111, +0.091; −0.498, ∓0.407).

The signs are what matter, and they hold for any positive values. A demand shift moves the wage and employment in the same direction. A supply shift moves them in opposite directions: when people need more pay to work, the wage rises and employment falls.

Over the long run in the United States, real wages and employment have both grown. Demand did most of the shifting: productivity growth, the subject of Chapters 4 to 6, raised what firms would pay for each worker.

Experiment: a tax on labor income

The government starts taxing labor income at 25%. What happens to employment? To the wage firms pay? To the wage workers keep?

Collect predictions on all three numbers separately. A common mistake is to think the whole tax falls on workers, so the wage firms pay stays at 55. Another is to forget that employment changes at all.

Hint for the sketch: workers decide how much to work from what they keep. Drawn against the wage firms pay, the supply curve shifts up and to the left.

A labor tax lowers employment and drives a wedge between two wages

Model figure: the baseline labor market with a proportional tax \(\tau\) on labor income; supply becomes \(w = (\bar{a}_s + \bar{b}_s L)/(1-\tau)\). Adapted from Fig. 7.4 in Jones (2024), Macroeconomics, 6th ed.

How to read it: A is the equilibrium without the tax, B with it. The taxed supply curve lies above the old one. The green bar at B is the tax on each unit of work: the gap between the wage firms pay and the wage workers keep.

Do live:

1. At τ = 25 percent: employment falls from 45 to 37.1. Firms pay 62.9, workers keep 47.1.
2. Move the slider to 50 percent: employment falls to 26.7, and the wedge grows.
3. Note the shape: the textbook draws the taxed curve parallel to the old one. A tax proportional to the wage makes it a little steeper, because it takes more when the wage is higher.

Firms pay more, workers keep less, and fewer people work

Tax rate \(\tau\)	Employment	Firms pay	Workers keep
0%	45.0	55.0	55.0
10%	42.1	57.9	52.1
25%	37.1	62.9	47.1
40%	31.2	68.8	41.2
50%	26.7	73.3	36.7

• At 25%, the tax wedge is 15.7 per unit of work.
• With equal slopes, the burden is shared: firms pay 7.9 more, workers keep 7.9 less.
• A steeper supply curve puts more of the tax on workers.
• \(E/N\) falls; the effect on \(u\) is unclear.

All rows come from the slider of Fig. 4; the wedge (15.7) and the split of the burden (7.9 each) come from the notebook’s Model overview.

The last bullet uses the identity of Act I. People who stop working because of the tax and stop searching leave the labor force: E/N falls, u does not. People laid off because labor got dearer are unemployed while they search; if they give up, u falls back. So a labor tax lowers employment for sure, but the unemployment rate may hardly move.

Who bears a tax does not depend on who writes the check: whether firms or workers send it to the government, the outcome is the same.

This mechanism is one candidate explanation of why Europeans work fewer hours than Americans, as we will see in Act II.

A fall in labor demand

New firing rules or dearer oil make firms want fewer workers at every wage: labor demand falls by 20. What happens to the wage and to employment?

This is the textbook’s second shock. Both examples raise the cost of hiring a worker, or lower what a worker adds to profits, so firms are willing to pay less at every level of employment.

Ask students to predict the direction of both changes and to guess how the fall is shared between them.

A fall in demand lowers both the wage and employment

Model figure: the baseline labor market; demand shifts down to \(w = (\bar{a}_d - \text{shift}) - \bar{b}_d L\). Adapted from Fig. 7.5 in Jones (2024), Macroeconomics, 6th ed.

How to read it: A is the old equilibrium and B the new one, after demand shifts down by the amount on the slider.

Do live:

1. At 20: the wage falls from 55 to 45 and employment from 45 to 35.
2. Move to 40: wage 35, employment 25.
3. With both slopes equal to 1, the fall is split equally between the wage and employment. A flatter supply curve would put more of it on employment.

Ask: where do the 10 workers go? Some are unemployed while they search; others leave the labor force. The employment-population ratio falls in either case.

The wage absorbs half of the shock

Fall in demand	Wage	Employment
0	55	45
10	50	40
20	45	35
30	40	30
40	35	25

\[\Delta L^* = -\frac{\Delta}{\bar{b}_d + \bar{b}_s}, \qquad \Delta w^* = -\frac{\bar{b}_s\,\Delta}{\bar{b}_d + \bar{b}_s}\]

A lower wage means some people choose to work less, so employment falls by less than the shift.

The formulas come from the closed form, which is linear in the intercepts: lowering ā_d by Δ lowers L* by Δ/(b̄_d + b̄_s) and w* by b̄_sΔ/(b̄_d + b̄_s). With both slopes equal to 1, each falls by half the shift.

The key word is “choose”. In this flexible-wage world the people who stop working do so because the wage they are offered is no longer worth it to them. Nobody is unemployed in the sense of wanting a job at the going wage and not getting one. Act II changes that by stopping the wage from falling.

Unemployment

Act II

Unemployment in recessions

Since 1950, when did U.S. unemployment reach its highest peaks, and which recovery was the fastest?

Students will likely name 2020 and the Great Recession. Fewer remember the early 1980s, when the Federal Reserve raised interest rates sharply to stop inflation (Chapter 12).

Ask them to predict the shape of the recoveries too: does unemployment come down as fast as it goes up?

Unemployment jumps in every recession and falls slowly afterwards

Data: U.S. Bureau of Labor Statistics, Current Population Survey, Unemployment Rate (UNRATE), via FRED; NBER recessions. Inspired by Fig. 7.2 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the monthly unemployment rate, in percent of the labor force, with NBER recessions shaded.

Do live:

1. Hover over the three highest peaks: late 1982, October 2009 (10.0 percent) and April 2020 (14.8 percent).
2. Compare the recoveries: after 2009 the rate fell slowly for years; after 2020 it fell within months.
3. Point at the long stretches between recessions: unemployment drifts down gradually, often for many years.

The Covid jump was the largest and the shortest

Episode	Before	Peak	Back below 4%
Great Recession	4.4% (May 2007)	10.0% (Oct. 2009)	May 2018
Covid-19	3.5% (Feb. 2020)	14.8% (Apr. 2020)	Dec. 2021

After 2009 it took almost nine years to get back below 4%; after 2020, under two.

The notebook’s key numbers on peaks and recoveries. The textbook reports 14.7 percent for April 2020; today’s data say 14.8, because seasonal adjustment is revised every year.

Why was the Covid recovery so fast? Many layoffs were temporary: people expected to return to the same employer once the lockdowns ended. In the language of the bathtub model ahead, the job finding rate was very high.

About 100 million U.S. adults are outside the labor force

Millions of people	January 2023	September 2026
population 16+	265.96	275.55
labor force	165.94	170.26
employed	160.17	163.15
unemployed	5.77	7.11
not in the labor force	100.02	105.29
unemployment rate	3.48%	4.18%

One point of the employment-population ratio is about 2.7 million jobs.

The textbook’s table of the labor force, rebuilt from the monthly levels for the textbook’s month and for the latest month. January 2023 agrees with the textbook within rounding, except that today’s data count 5.77 million unemployed (the textbook: 5.7), so the rate is 3.48 rather than 3.4 percent.

The adults outside the labor force are students, retirees, carers, people with disabilities and discouraged workers. They far outnumber the unemployed, which is why the participation rate matters as much as the unemployment rate.

With 265.96 million people aged 16 and over in January 2023, one point of the employment-population ratio is 2.66 million people; in September 2026, 2.76 million.

In 2020, U.S. firms laid workers off; German and Japanese firms kept them

Unemployment rate	Feb. 2020	Mar. 2020	Apr. 2020	May 2020
United States	3.5%	4.4%	14.8%	13.2%
Canada	5.8%	8.5%	13.7%	14.2%
Germany	3.3%	3.4%	3.5%	3.7%
Japan	2.4%	2.5%	2.6%	2.8%

Germany and Japan subsidized firms that kept workers on shorter hours. The shock showed up in hours, not in unemployment.

The numbers are OECD harmonised unemployment rates, from the notebook’s “Beyond the United States” section of Fig. 2: every country counts the unemployed the same way.

Germany’s Kurzarbeit (short-time work) pays part of the wage for the hours a firm cuts, so the firm keeps the worker. Giupponi and Landais (2023, Review of Economic Studies) study Italy’s version and find that it saved jobs in firms hit by temporary shocks, but can also keep workers in firms that should shrink.

The comparison is also a warning about statistics: the same collapse in demand can show up as unemployment in one country and as fewer hours in another. Compare countries with more than one measure.

Same section, a different story: after the euro crisis, unemployment passed 26 percent in Spain (26.4 percent in February 2013) and Greece (28.3 percent in July 2013).

Unemployment has a normal part and a cyclical part

\[u = \underset{\text{natural rate } u^n}{\big(u_{\text{frictional}} + u_{\text{structural}}\big)} + u_{\text{cyclical}}\]

• Frictional: the time it takes to move between jobs, even in a well-working market.
• Structural: the rules of the market, such as firing costs, unemployment benefits and the minimum wage.
• Cyclical: \(u - u^n\), positive in recessions and negative in booms.

The natural rate is not fixed by nature: institutions and policies move it.

The natural rate is the unemployment rate of an economy that is neither booming nor in recession. The next two models explain where unemployment comes from: a wage that does not fall gives a cyclical story; the bathtub model gives the natural rate.

Ask: which kind of unemployment is a graduate who takes three months to find a first job? (Frictional.) A coal miner whose region has lost its mines? (Structural, if the skills or the location no longer match the jobs.) A waiter laid off in a recession? (Cyclical.)

Experiment: the wage does not fall

Labor demand falls by 20 again, but this time the wage stays at 55. Does employment fall by more or by less than with a flexible wage? Who is left out?

Collect predictions. Most students expect a larger fall; ask them by how much. Some will see that employment is now set by the demand curve alone.

The second question is the important one: at a wage of 55, how many people want to work? (Still 45, from the supply curve.)

A rigid wage turns a fall in demand into unemployment

Model figure: the baseline labor market; after demand falls, the wage stays at \(\bar{w} = w^* = 55\). Adapted from Fig. 7.6 in Jones (2024), Macroeconomics, 6th ed.

How to read it: A is the old equilibrium. With a flexible wage the economy would move to B; with the wage stuck at w̄ = 55, firms hire only what they want at that wage, at C on the new demand curve.

Do live:

1. At a fall of 20: L’’ = 25 with the rigid wage, against L’ = 35 with a flexible one. The jobless gap is 20.
2. Move to 40: L’’ = 5 against 25, a gap of 40.
3. Point at the horizontal line from C to the supply curve: these people want to work at the going wage and cannot find a job. That is unemployment in the everyday sense.

With a rigid wage, the whole shock falls on jobs

Fall in demand	Flexible \(L'\)	Rigid \(L''\)	Jobless
0	45	45	0
10	40	35	10
20	35	25	20
30	30	15	30
40	25	5	40

\[L'' = L^* - \frac{\Delta}{\bar{b}_d}\]

• The fall is larger by the factor \((\bar{b}_d + \bar{b}_s)/\bar{b}_d\): twice, with our slopes.
• Keynes’s explanation of the mass unemployment of the 1930s.

The table is the slider of Fig. 6. The formula: at the rigid wage, firms hire what the new demand curve says, and since ā_d − w* = b̄_d L, that is L − Δ/b̄_d. The number of people who want a job at w̄ but have none is the gap L* − L’’ = Δ/b̄_d.

Compare with the flexible case of Act I: there employment fell by Δ/(b̄_d + b̄_s), here by Δ/b̄_d. With equal slopes, twice as much.

Rigid wages and prices are the core of the short-run model in Chapters 9 to 13. This slide is the first appearance of the idea.

Why would the wage not fall?

Rules and contracts

• A minimum wage above the market wage.
• Union contracts that fix wages for several years.
• Pay cuts can be illegal or need workers’ consent.

The firm’s own choice

• Efficiency wages: paying above the market rate can raise effort, cut quitting and attract better applicants.
• Morale: managers fear that pay cuts anger the workers who stay.

Whatever the reason, a wage above the market-clearing level leaves people who want work at that wage without a job.

The textbook’s case study is Henry Ford’s “five-dollar day”, which sharply raised pay at his car plants when he could have hired at the old wage. Ford said the higher pay cut costs: workers stayed, worked harder and turned up. Raff and Summers (1987, Journal of Labor Economics) study the episode and find that the higher wage did lower quitting and absenteeism.

Bewley (1999, Why Wages Don’t Fall during a Recession) interviewed many managers. Most said they avoided pay cuts because cuts hurt morale and productivity, and preferred to lay off a few workers instead.

Efficiency wages also explain why some unemployment persists in good times: if every firm pays above the market rate, the market never fully clears.

The bathtub: unemployment settles where job losses equal job findings

\[\Delta U_{t+1} = \underset{\text{jobs lost}}{\bar{s}E_t} - \underset{\text{jobs found}}{\bar{f}U_t} \qquad\Longrightarrow\qquad u^* = \frac{\bar{s}}{\bar{f}+\bar{s}}\]

• With \(\bar{s} = 0.01\) and \(\bar{f} = 0.20\) a month: \(u^* = 0.01/0.21 \approx\) 4.8%.
• Each month about 0.95% of the labor force loses a job and the same share finds one: large flows, a small stock.
• The gap to \(u^*\) closes by 21% a month: from 10%, unemployment is 7.34% after 3 months and 5.07% after a year.
• An average spell lasts \(1/\bar{f} = 5\) months.

The model: a fixed labor force L̄ = E + U. Each month a share s̄ of the employed lose their job (the job separation rate) and a share f̄ of the unemployed find one (the job finding rate). Job losses are the tap, job findings the drain, and the unemployed are the water in the tub.

In the steady state the inflow equals the outflow: s̄(L̄ − U) = f̄U, which gives u* = s̄/(f̄ + s̄). The size of the labor force does not matter, only the two rates.

The path: u_t − u* = (1 − s̄ − f̄)^t (u_0 − u*). The factor is 0.79, so the gap halves in about 2.9 months. The notebook simulates the model month by month and matches the formula: 7.3445 percent after 3 months, 5.0714 after 12, 4.7619 after 60.

The textbook stresses the gross flows: each month the U.S. economy creates and destroys millions of jobs, and the net change is small.

Experiment: making firing costly

A new law makes firing costly. Layoffs fall: the job separation rate drops from 1.0% to 0.8% a month. Firms also hire more cautiously. Does the natural rate of unemployment fall or rise?

Collect predictions. Most students say it falls, because fewer people lose their jobs. Push them: what does a firm that cannot easily fire do when it is unsure about the future? (It hires less.)

The answer depends on how much each rate moves. The chart lets you change the job finding rate; the next slide gives both cases.

Faster job finding means lower unemployment

Model figure: our own simulation of the bathtub model with \(\bar{s} = 0.01\) a month, starting from 10% unemployment; the slider sets the job finding rate \(\bar{f}\). Adapted from the bathtub model (Section 7.4) in Jones (2024), Macroeconomics, 6th ed.

How to read it: the unemployment rate month by month, starting from 10 percent. The dashed line is the natural rate for the chosen rates. The textbook draws no figure for this model; this is the notebook’s own chart.

Do live:

1. At f̄ = 0.20: unemployment falls from 10 percent and settles at 4.8 percent within about a year.
2. Lower f̄ to 0.15: the natural rate rises to 6.2 percent, and the path settles more slowly.
3. Lower it to 0.10: 9.1 percent. A slower drain means more water in the tub.
4. Raise it to 0.50: 2.0 percent, and the path settles within months.

Firing costs can lower or raise the natural rate

Job finding \(\bar{f}\)	\(u^*\)
0.10	9.1%
0.15	6.2%
0.20	4.8%
0.30	3.2%
0.50	2.0%

Case	\(\bar{s}\)	\(\bar{f}\)	\(u^*\)
before the law	0.01	0.20	4.76%
fewer layoffs only	0.008	0.20	3.85%
fewer layoffs, slower hiring	0.008	0.15	5.06%

A 1% rise in \(\bar{s}\) raises \(u^*\) by about 0.95%; a 1% rise in \(\bar{f}\) lowers it by about as much.

Left: the slider of the bathtub chart, with s̄ = 0.01. Right: the notebook’s firing-cost example. If the law only cuts layoffs, the natural rate falls to 3.85 percent. If it also slows hiring to 0.15, it rises to 5.06 percent, above where it started.

The elasticities come from logs: ln u* = ln s̄ − ln(f̄ + s̄), so the elasticity with respect to s̄ is f̄/(f̄ + s̄) = 1 − u*, about 0.95, and with respect to f̄ it is about −0.95 (the notebook’s 1 percent change gives −0.943).

Only the two flow rates move the natural rate. A policy aimed at unemployment has to work through how often people lose jobs or how fast the jobless find new ones.

Protecting jobs does not always lower unemployment

Making it harder to fire workers protects jobs, so it must lower unemployment.

A firing cost slows the tap but can also narrow the drain. If hiring slows as well, the natural rate rises, from 4.76% to 5.06% in our example, and the jobless wait longer for work.

The myth looks only at the inflow. Strict employment protection does cut layoffs, but firms that cannot easily adjust also hire less, so the unemployed find jobs more slowly.

Blanchard and Portugal (2001, American Economic Review) compare Portugal, with strict employment protection, and the United States. The two had similar unemployment rates, but Portugal’s flows into and out of unemployment were far smaller and its spells far longer: a stagnant pool rather than a fast-flowing stream. The same unemployment rate can mean very different lives for the unemployed.

The evidence across countries on the net effect of employment protection on the unemployment rate is mixed; its effect on flows and durations is clearer.

How generous should unemployment benefits be?

Benefits protect people who lose their jobs, but they can also slow the search for a new one. How should they be designed?

Insurance

• Benefits let families keep spending while they search.
• Workers short of cash may take the first job offered; a cushion allows a better match.
• Without insurance, a job loss can hit the poorest hardest.

Incentives

• Benefits lower the cost of staying unemployed, so the job finding rate \(\bar{f}\) falls.
• Many people find a job just when their benefits run out.
• Design matters: a lump sum at job loss keeps the incentive to search.

Left column. Gruber (1997, American Economic Review) finds that unemployment insurance protects household consumption after a job loss. Ganong and Noel (2019, American Economic Review) show with bank-account data that spending drops sharply when benefits run out. Chetty (2008, Journal of Political Economy) shows that much of the longer search under higher benefits comes from people with little cash, who use benefits to search for a better job rather than to take time off: a liquidity effect, which is a benefit, not a cost.

Right column. Card, Chetty and Weber (2007, American Economic Review Papers and Proceedings) document a spike in exits from the unemployment rolls at benefit exhaustion in Austria. Schmieder, von Wachter and Bender (2012, Quarterly Journal of Economics) find that longer benefits in Germany lengthen spells of non-employment, by less in recessions. Baily (1978, Journal of Public Economics) set out the trade-off that this literature measures: insurance value against longer spells.

In the bathtub model, benefits work through f̄. The textbook’s exercise compares two reforms: replacing 100 percent of income for 26 weeks, or paying a lump sum of ten weeks’ income at the start of the spell. Ask the policy students which they would choose, and why. (The lump sum gives insurance without paying people for staying unemployed, but it gives nothing to the long-term unemployed.)

Unemployment around the world

Up to 1975, unemployment in Europe was about half the U.S. rate. What do you expect for the decades since 1985?

The numbers behind the question, from the notebook’s key numbers: from 1960 to 1975, Europe averaged 2.6 percent and the United States 5.21 percent.

Ask students to guess, and to name one reason for their guess. Keep the reasons on the board; the policy slide comes back to them.

Europe’s unemployment rose above the U.S. rate after the 1970s and stayed there

Data: U.S. Bureau of Labor Statistics (UNRATE), annual average, and OECD harmonised rate for Japan (LRHUTTTTJPA156S), via FRED; Europe: euro area of 12 countries, European Commission, AMECO (ZUTN). Inspired by Fig. 7.7 in Jones (2024), Macroeconomics, 6th ed.

How to read it: annual unemployment rates since 1955 for the United States, Europe (the euro area of 12 countries) and Japan. The slider marks a year; the title reports the three rates.

Do live:

1. 1960: U.S. 5.5, Europe 2.4, Japan 1.6 percent.
2. 1985: Europe 9.4, above the U.S. 7.2. Europe first rose above the United States in 1979.
3. 2013: Europe’s peak, 12.1 percent, after the euro crisis.
4. 2025: U.S. 4.3, Europe 6.4, Japan 2.5.

Data caveat: the textbook uses an OECD series for Europe, which runs up to about 1 point higher in the mid-1980s; the notebook uses the European Commission’s.

Europe went from low to high unemployment

Average	1960–1975	1985–2025
United States	5.21%	5.74%
Europe	2.6%	9.08%
Japan	1.33%	3.45%

Average since 1985	Unemployment
Japan, South Korea	3.5%
Switzerland	3.6%
Norway	4.0%
Spain	16.6%

There is no single “European” rate: Spain’s unemployment went from 4.5% in 1975 to 24.1% in 1994, while Switzerland’s and Norway’s stayed close to Japan’s.

Left: the notebook’s key numbers. Right: from the 43-economy chart in the same section of the book chapter (AMECO data).

Spain’s path: 4.5 percent in 1975, 17.8 in 1985, 24.1 in 1994, 26.1 in 2013 and 10.5 in 2025. Its average since 1985, 16.6 percent, is the highest of the large economies in the file.

The textbook’s explanation combines two forces. Adverse shocks in the 1970s (the oil price shocks and the productivity slowdown of Chapter 6) raised unemployment everywhere. Labor market institutions (generous benefits, firing costs, high labor taxes) then kept it high in parts of Europe. In the bathtub model, both work through a lower job finding rate.

Data caveat: some values in the wider file are exactly 0 (Luxembourg 1960–1974, Switzerland 1970–1973), which usually means unemployment was barely measured, not that nobody was looking.

Europeans now work fewer hours per person than Americans

Hours worked per person (U.S. in 2019 = 100)	1960	1990	2019
United States	88	103	100
France	111	79	75
Germany	131	93	88
Italy	107	87	86
United Kingdom	114	96	96
Japan	130	125	110
South Korea	89	134	122

In 1960 all four European countries worked more hours per person than Americans; by 2019 all four worked fewer.

Hours per person combine three things: how many adults work, how many weeks a year, and how many hours a week. The table uses Penn World Table 10.01, which reproduces every cell of the textbook’s table.

The latest version, 11.0, revises several countries down (the U.K. by about 9 points in 2019, Japan by 7), but the ranking is the same. The notebook’s lesson: when a data source is revised, say which version you used.

Two explanations: Europeans choose more leisure (longer vacations, earlier retirement), or taxes and institutions discourage work. The next slide weighs them.

Did Europe’s institutions cause its unemployment?

Generous benefits, firing costs and high labor taxes are often blamed for Europe’s high unemployment and short hours. Should Europe change them, and would that work?

Shocks meet institutions

• Europe had these institutions when unemployment was low, so they cannot explain the rise alone.
• The consensus: the shocks of the 1970s raised unemployment, and the institutions kept it high.
• Taxes on labor income lower hours worked, as in our tax experiment.

Reforms and their limits

• Several countries cut benefit durations and eased hiring rules; unemployment later fell in Spain, Ireland and the Netherlands.
• Reforms shift risk onto workers.
• Culture and unions may matter as much as taxes.

Left column. Blanchard and Wolfers (2000, Economic Journal) find that shocks alone and institutions alone each explain little of the rise in European unemployment, while the interaction of the two explains much of it. Prescott (2004, Federal Reserve Bank of Minneapolis Quarterly Review) argues that differences in labor taxes account for most of the gap in hours between Europe and the United States. This is the tax experiment of Act I, at the scale of a country.

Right column. Alesina, Glaeser and Sacerdote (2005, NBER Macroeconomics Annual) argue that union power and labor regulations, not taxes, explain most of the gap in hours, and that Europeans may value coordinated leisure. Germany’s Hartz reforms cut benefits for the long-term unemployed and reorganized job placement; Krause and Uhlig (2012, Journal of Monetary Economics) attribute a substantial part of the later fall in German unemployment to them. The textbook notes that reforms in Spain, Ireland and the Netherlands were followed by lower unemployment than in the 1980s.

For the policy students: a reform that raises the job finding rate by cutting benefits trades insurance for incentives, the same trade-off as on the unemployment insurance slide. Which countries would you study to decide?

Wages, skills and inequality

Act III

A dollar in the future is worth less than a dollar today

\[pdv = \frac{\text{future value}}{(1+R)^T}, \qquad pdv = \text{payment}\times\frac{1-\left(\frac{1}{1+R}\right)^n}{1-\frac{1}{1+R}}\]

A payment of …	Interest rate	Worth today
$10,000 in 10 years	3%	$7,441
$1,000 in 5 years	10%	$621
$100 a year for 20 years	10%	$936
$1 a year for 45 years	3%	$25.25
$1 a year forever	3%	$34.33

The first formula: a deposit of pdv today grows into pdv(1 + R)^T in T years, so the value today of a payment in T years is the payment divided by (1 + R)^T. Dividing by (1 + R)^T is called discounting.

The second formula adds up n equal payments, the first one today. With a = 1/(1 + R), the sum is 1 + a + a² + … + a^(n−1) = (1 − a^n)/(1 − a), the geometric-series trick: multiply the sum by a, subtract, and every middle term cancels.

The textbook’s choice: $1,000 in five years or $100 a year for 20 years, at 10 percent? The stream is worth $936 today and the single payment $621, so the stream wins.

The last two rows: 45 years of $1 are worth only $25.25 today, not $45, and even a payment that lasts forever is worth $34.33. The 45 years already give 74 percent of forever.

The value of a working life

You expect to earn $100,000 a year for 45 years. With an interest rate of 3%, what is that whole stream worth today?

Let students guess. Common answers are $4.5 million (no discounting) and something under $1 million (too much discounting).

The value of this stream of future earnings is the value of a person’s human capital, the subject of the chart.

A typical new worker is a millionaire

Our own calculation with the present-value formula and illustrative numbers: a wage of $100,000 a year for 45 years at \(R\) = 3%; no data. Inspired by Section 7.6 of Jones (2024), Macroeconomics, 6th ed.

How to read it: each bar is one year’s $100,000 wage valued today, year 0 to year 44. The title reports the total: $2,525,427.

Do live:

1. Hover over year 0 and year 44: the first wage counts in full; the last counts for much less.
2. The total is the area of all the bars, about $2.53 million: 25.25 dollars today for each dollar of yearly pay.

Ask: what happens to the total at a 1 percent interest rate? At 5 percent? (Higher rates shrink every bar, the distant ones most. One more point of interest lowers the value of a payment ten years away by about 9.2 percent.)

A college degree is worth several hundred thousand dollars

\[NPV = \underset{\approx\ 673{,}000}{(w_{\text{college}} - w_{\text{high school}})\times S_{45}\times\frac{1}{(1+R)^4}} \;-\; \underset{\approx\ 287{,}000}{(w_{\text{high school}} + \text{tuition})\times S_4}\]

• Four years of a lost $60,000 wage plus $15,000 tuition; then $90,000 instead of $60,000 for 45 years.
• At \(R\) = 3%: \(NPV \approx\) $386,000.
• The interest rate at which the degree stops paying is its rate of return: about 8.6% a year.

S_n is the value today of one dollar a year for n years: S_4 ≈ 3.83 and S_45 ≈ 25.25. The gain starts in year 4, so it is discounted four more years (1/1.03^4 ≈ 0.888).

The numbers are illustrative: the textbook’s wages of $60,000 and $90,000, and the notebook’s tuition of $15,000 a year. The notebook computes NPV = $385,997 and finds the rate of return by raising R in small steps until the NPV turns negative: 8.59 percent.

Two cautions. First, the extra $30,000 is not all caused by the degree: people who go to college may have earned more anyway. Studies that address this still find large returns. Second, the return is an average; it varies a lot by field and by college.

More college graduates, and their wage premium

Between 1979 and 2020, the share of U.S. jobs held by college graduates more than doubled. What happened to how much more they earn than high school graduates?

Ask students to predict using supply and demand. A pure supply story says: more graduates, so a lower premium. Let the room vote: up, down or flat?

The numbers in the question come from the notebook’s key numbers: 18.3 percent of jobs in 1979, 40.9 percent in 2020.

College graduates became more numerous and better paid at the same time

Data: Economic Policy Institute, State of Working America Data Library, from the Current Population Survey: regression-adjusted college wage premium and share of employment with a bachelor’s degree or more. Inspired by Fig. 7.8 in Jones (2024), Macroeconomics, 6th ed.

How to read it: two lines on the same scale. The college wage premium, in log points, is how much more college graduates earn per hour than high school graduates with similar gender, race, age and region. The share of employment is the percent of workers with a bachelor’s degree or more.

Do live:

1. The premium rose steeply in the 1980s, more slowly in the 1990s, and has moved between about 44 and 49 log points since 2000.
2. The share rose the whole time.

Data caveats: the textbook uses private data from Acemoglu and Molina, which start in 1963 and show a dip in the 1970s; the public data start in 1979. The textbook’s share is of hours worked, about twice the share of employment.

More supply and a higher price: demand must have grown faster

Year	Premium (log points)	Share of jobs
1979	25.7	18.3%
1990	40.7	23.8%
2000	45.9	27.6%
2020	48.1	40.9%
2025	44.0	41.8%

• The premium peaked at 49.2 log points in 2016.
• 44 log points means college graduates earn about 55% more.
• Across 38 countries, more graduates go with a lower premium (correlation −0.58), mostly because of four countries where graduates are scarce.

Log points are 100 times the log of one plus the gap: for small gaps the two units agree, for large ones the log version is smaller. In 2025 the premium is 43.95 log points, or 55.19 percent.

The textbook’s summary speaks of a premium rising from 50 to 100 percent, which does not fit its own figure; the notebook’s 48.1 log points in 2020 matches the figure’s “nearly 50 percent”.

Across countries (OECD data, a simpler measure not adjusted for age or gender): the U.S. premium, 61 log points in 2024, is among the highest; Norway, Denmark and Sweden have the lowest, 16 to 26. The correlation of −0.58 across 38 countries comes mostly from South Africa, Brazil, Chile and Costa Rica, where graduates are scarce and earn 2.2 to 3.1 times as much; among the other 34 it is only −0.20. Demand for skills differs across countries as much as supply.

Experiment: more graduates every year

Every year more young people finish college: supply in the college labor market shifts out by 10. Must the college wage fall? What would it take for it to rise?

Collect predictions. Push students to the condition: a supply shift alone lowers the wage, so demand has to shift too. By how much?

The college market in the model: demand w = 90 − 0.4L, supply w = 30 + 0.6L, so the college wage starts at 66 against a high school wage of 55, a premium of 20 percent.

The college wage rises when demand shifts out by more than supply

Model figure: high school market as in Act I; college market with demand \(w = 90 - 0.4L\) and supply \(w = 30 + 0.6L\); supply shifts out by 10. Adapted from Fig. 7.9 in Jones (2024), Macroeconomics, 6th ed.

How to read it: panel (a) is the market for high school graduates, which stays put; panel (b) is the market for college graduates. Dashed gray curves are before the shifts, solid ones after. The single star marks the old equilibrium and the double star the new one.

Do live:

1. At a demand shift of 15: college employment rises from 60 to 85 and the college wage from 66 to 71. The premium rises from 20 to 29 percent.
2. Set the demand shift to 0: the extra supply alone lowers the college wage to 62.
3. Ask: what is the smallest demand shift that keeps the wage from falling?

The race between the demand for skills and their supply

Demand shift	College wage	Premium
0	62	13%
5	65	18%
10	68	24%
15	71	29%
20	74	35%

\[\Delta w^* = \frac{\bar{b}_s\,x_d - \bar{b}_d\,x_s}{\bar{b}_d + \bar{b}_s} = 0.6\,x_d - 0.4\,x_s\]

• The wage rises only if \(x_d/x_s > \bar{b}_d/\bar{b}_s = 2/3\).
• Candidate demand shifters: skill-biased technical change and globalization.

The table is the slider of Fig. 9, with the supply shift fixed at 10. The formula follows from the closed form, which is linear in the intercepts: demand rising by x_d and supply shifting out by x_s change the wage by 0.6x_d − 0.4x_s with these slopes. The defaults give 9 − 4 = 5.

Skill-biased technical change: computers and software raise the productivity of workers who can use them, more than that of workers who do routine tasks. Globalization: trade with countries where skilled workers are scarce raises the value of skills here, and trade with countries rich in unskilled labor lowers the wages of the unskilled.

Katz and Murphy (1992, Quarterly Journal of Economics) estimate this supply-and-demand framework for the U.S. wage structure. Goldin and Katz (2008, The Race between Education and Technology) tell the long history: when education grew faster than the demand for skills, inequality fell; when it slowed, inequality rose.

The 1970s: the college premium fell

In the 1970s, the baby boom and the Vietnam War sent many more young people to college, while demand for graduates kept growing. What happened to the college wage?

This is the same model used in reverse. Ask students which shift was larger in the 1970s, and what that implies for the wage.

The draft could be deferred by going to college, which drew many young men in; the baby boom generation reached college age at the same time.

When supply wins the race, the college wage falls

Model figure: the college market of the previous figure; demand shifts out by 10 and the slider sets the supply shift. Adapted from Fig. 7.13 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the college market only. Demand shifts out by 10; the slider sets how far supply shifts out.

Do live:

1. At a supply shift of 25 (the default): college employment rises from 60 to 95, but the wage falls from 66 to 62.
2. Move to 15: the wage stays at 66, because 0.6 × 10 = 0.4 × 15.
3. Move to 0: the wage rises to 72.

The point: the 1970s and the decades after use the same model. Whether the premium rises or falls depends only on which shift wins.

Same model, two decades: which shift wins decides the premium

Supply shift	Employment	College wage
0	70	72
10	80	68
15	85	66
25	95	62
30	100	60

• With demand shifting out by 10, a supply shift of 15 leaves the wage unchanged: \(0.6\times 10 = 0.4\times 15\).
• In the 1970s, supply won: employment 60 → 95, wage 66 → 62.
• Since 1980, demand won: premium 20% → 29% in Fig. 9’s example.

The table is the slider of Fig. 13, with the demand shift fixed at 10.

This is the most useful lesson of the chapter for reading data: a price and a quantity that move in the same direction point to a demand shift; in opposite directions, to a supply shift. College employment rose in both periods. The wage fell in the 1970s (supply dominated) and rose afterwards (demand dominated).

More graduates do not have to mean a lower premium

If many more people earn college degrees, their wage advantage must shrink.

Only if demand for their skills stands still. From 1979 to 2020 the share of jobs held by graduates rose from 18.3% to 40.9%, and their premium rose from 25.7 to 48.1 log points. In the 1970s, supply did outrun demand, and the premium fell.

The myth applies supply-and-demand reasoning with only one curve. More supply alone does lower the price, which is what happened in the 1970s. Over the following four decades, demand for skilled workers grew faster still.

For policy, the lesson cuts both ways. Expanding college access does not have to destroy the return to a degree, as long as demand for skills keeps growing. And if the premium is high because supply lags demand, more education is one of the few policies that can lower wage inequality while raising incomes (Goldin and Katz 2008).

Note the recent slip: the premium fell from 49.2 log points in 2016 to 44.0 in 2025. Whether this is the start of a new 1970s is an open question.

Manufacturing jobs

U.S. manufacturing employment peaked in 1979. When did it fall the most, and by how much?

Most students will guess the recessions of the 1980s or the 2008 crisis. The answer is the decade after 2000, which includes the 2001 recession, the Great Recession and China’s entry into the World Trade Organization.

Manufacturing jobs peaked in 1979 and fell sharply after 2000

Data: U.S. Bureau of Labor Statistics, Current Employment Statistics, All Employees, Manufacturing (MANEMP), via FRED; NBER recessions. Inspired by Fig. 7.10 in Jones (2024), Macroeconomics, 6th ed.

How to read it: monthly employment in manufacturing, in millions of workers, from the payroll survey. Recessions are shaded.

Do live:

1. Hover over the 1979 peak: 19.55 million in June 1979.
2. Hover over January 2000 (17.28 million) and February 2010 (11.44 million).
3. Point at the shape: falls in every recession, a flat 1990s, a collapse after 2000, and no recovery since.

A third of manufacturing jobs disappeared in a decade

U.S. manufacturing	Millions
peak, June 1979	19.55
January 2000	17.28
February 2010	11.44
September 2026	12.65

2000–2010: −5.84 million, −34%

Share of jobs in manufacturing	1991	2025
United States	16.0%	9.5%
Germany	27.9%	18.0%
Japan	23.6%	15.6%
United Kingdom	21.8%	8.0%
China	18.4%	21.0%
Vietnam	5.6%	24.0%

Left: the notebook’s key numbers. Research cited in the textbook finds that rising imports from China explain about a quarter of the decline after 2000: about 1.46 million jobs. The rest came from automation and other forces.

Right: ILO estimates, from the “Beyond the United States” section. Manufacturing’s share of jobs fell in every rich country since 1991, so the decline is not only an American story. China’s factories employ about 155 million people, almost ten times as many as American ones.

Data caveat: the ILO counts 16.2 million U.S. manufacturing workers in 2025, about 3.5 million more than the payroll survey, because household surveys also count the self-employed. Compare shares across countries, and levels only within one source.

Ask: in the college-market model, which curve does the loss of manufacturing jobs shift? (It lowers the demand for workers without a degree, in the high school market.)

Who should help the workers hurt by trade and automation?

Trade and new technology raise national income, but the losses are concentrated in particular places and workers. What should policy do for them?

What the evidence shows

• Places more exposed to Chinese imports lost more manufacturing jobs.
• Wages there fell and disability and other benefit claims rose.
• Workers did not move away as much as models assumed.

Policy options

• Retraining and job-search help: results vary by program.
• Wage insurance and income support for displaced workers.
• Place-based aid, or help to move to where jobs are.

Left column. Autor, Dorn and Hanson (2013, American Economic Review) compare U.S. local labor markets with different exposure to Chinese imports and find larger falls in manufacturing employment, lower wages and higher transfer payments in the more exposed ones. Acemoglu, Autor, Dorn, Hanson and Price (2016, Journal of Labor Economics) estimate the national employment effects. Pierce and Schott (2016, American Economic Review) link the timing to the end of uncertainty about U.S. tariffs on Chinese goods. Autor, Dorn and Hanson (2016, Annual Review of Economics) review the slow adjustment of local labor markets. Acemoglu and Restrepo (2020, Journal of Political Economy) find similar local effects of industrial robots.

Right column. Card, Kluve and Weber (2018, Journal of the European Economic Association) review hundreds of evaluations of active labor market programs: training and job-search help do little in the short run and more in later years, with large differences across programs and groups. The U.S. Trade Adjustment Assistance program combines training with income support; its evaluations are mixed.

The trade-off for the policy students: help that targets trade victims only is hard to justify to workers who lost their jobs to other causes, such as automation. Help for all displaced workers is fairer but more costly.

A century of top incomes

From 1913 to 1929, the richest 0.1% of Americans received close to 9% of all income. What happened to that share over the next hundred years?

The top 0.1 percent is the richest one adult in a thousand. The number in the question is the 1913–1929 average from the notebook: 8.7 percent in the United States, 9.4 percent in France.

Ask students to sketch the path they expect for the United States, and then for France.

Top incomes fell after 1929 in both countries, but rose again only in the U.S.

Data: World Inequality Database, share of pre-tax national income of the top 0.1%, via Our World in Data; it extends the tax-record series of Piketty and Saez (2003). Inspired by Fig. 7.11 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the share of national income, before taxes and benefits, received by the top 0.1 percent of adults, in percent, for the United States and France since 1913. The data come from tax records, which show the very top of the distribution well.

Do live:

1. The 1910s and 1920s: high in both countries, above 10 percent in peak years.
2. The Depression and World War II cut both shares sharply, and they stayed low until about 1980.
3. After 1980 the U.S. share rose back to almost 10 percent, while France’s stayed much lower.

Data caveat: the U.S. value is exactly 9.93 in 2022, 2023 and 2024. The database seems to repeat its last estimate until new tax data arrive; treat the last two years with care.

Inequality at the very top: a U-shape in the U.S., an L-shape in France

Average share	U.S.	France
1913–1929	8.7%	9.4%
1950–1959	5.0%	3.7%
1960–1979	3.9%	3.3%
2015–2022	8.8%	4.3%

Top 0.1% share	1980	2023
United Kingdom	1.8%	5.0%
Japan	2.3%	1.9%
China	2.4%	7.9%
India	2.1%	9.8%
Russia	0.6%	9.8%

Early in the century top incomes came mostly from capital; since 1980 they come largely from salaries and business income.

Left: the notebook’s key numbers. The textbook’s “around 4 percent” in the postwar decades fits the 1960s and 1970s; the 1950s average is 5.0 percent in today’s data, with a spike in 1950. The French share after 2015 is higher in today’s data than in the textbook’s figure: the database is revised as new tax records arrive.

Right: from the 58-country chart in the same section (WID). The United Kingdom follows the U.S. U-shape, from 14.1 percent in 1913 down to 1.8 in 1980 and up to 5.0 in 2023. Japan follows France. China, India and Russia rose fastest after 1980, as their markets opened.

The takeaway is the textbook’s reading of Piketty and Saez: the top of the distribution used to be people living on capital; today it is largely highly paid executives, professionals and entrepreneurs.

Growth across the income distribution

Since 1980, U.S. income per person has kept growing. Did incomes grow at the same pace for the poor, the middle class and the rich?

So far the book has studied growth of average income per person. This figure asks how that growth was shared.

Ask students to sketch growth rate (vertical) against income percentile (horizontal) for 1946–1980 and for 1980–2014.

Since 1980 the richest have seen the fastest income growth

Data: Piketty, T., E. Saez and G. Zucman (2018), Distributional National Accounts, Quarterly Journal of Economics 133(2); income after taxes, transfers and noncash benefits. Inspired by Fig. 7.12 in Jones (2024), Macroeconomics, 6th ed.

How to read it: each point is the average annual growth of real income at one percentile of the distribution over 34 years, 1946–1980 and 1980–2014. Above the 99th percentile the horizontal axis zooms in to the top 0.1, 0.01 and 0.001 percent. The diamonds are group averages from the authors’ 2022 update, which also covers 1985–2019.

Do live:

1. The 1946–1980 curve slopes down: the poorer half grew fastest, above 2.5 percent a year at the bottom.
2. The 1980–2014 curve slopes up: under 1 percent a year up to the 57th percentile, about 6 percent at the very top.
3. Click the legend to show or hide the 1985–2019 group averages: still favoring the top, less steeply.

Growth used to lift the bottom fastest; since 1980 it has lifted the top

• Median income grew 1.93% a year in 1946–1980 and only 0.91% in 1980–2014.
• Only the bottom five percentiles lost income after 1980.
• At the very top: up to 5.96% a year.

Pre-tax growth, 1980–2014	10th	50th	99th	99.99th
U.S.	−1.25%	0.58%	2.10%	4.08%
France	0.72%	0.88%	1.46%	2.19%
China	3.29%	3.87%	6.30%	7.83%
India	1.69%	1.71%	5.06%	7.34%

Left: the notebook’s key numbers for the U.S. after taxes and transfers. The 1980–2014 curve beats the old one only from the 96th percentile; at the top, growth was 3.24 percent a year at the 99.9th percentile, 4.04 at the 99.99th and 5.96 at the 99.999th. The textbook gives average growth of about 2 percent before 1980 and 1.4 percent after; the file has no overall average, but the median tells the same story.

Right: WID’s pre-tax incomes at selected percentiles, from the “Beyond the United States” section. Before taxes and transfers, the U.S. 10th percentile fell by 1.25 percent a year; after them it rose by 0.26 percent. Taxes and transfers lift the bottom. France’s curve is almost flat: the steep U.S. curve is not the only possible pattern. China grew fast at every percentile, and fastest at the top: fast growth can lift all incomes and raise inequality at the same time.

Data caveat: WID’s numbers for countries with poor tax data are partly model estimates.

Should top incomes be taxed more?

Top incomes have grown much faster than the rest since 1980 in the United States. Should top tax rates rise, and what would that do to work, innovation and revenue?

For higher top rates

• Part of top pay may come from bargaining power, not extra output; taxing it costs little output.
• Higher top rates in the past coincided with lower top shares.
• Revenue could fund transfers that lift the bottom.

Against, or with limits

• Top earners respond to taxes, partly by moving income into other forms.
• Inventors and other high earners move across borders.
• If top incomes reward innovation, taxing them may slow growth.

Left column. Piketty, Saez and Stantcheva (2014, American Economic Journal: Economic Policy) argue that when top earners can bargain for a larger share of the pie, high top rates reduce bargaining rather than work, which favors higher optimal rates. Diamond and Saez (2011, Journal of Economic Perspectives) lay out the case for progressive taxation from optimal-tax theory.

Right column. Saez, Slemrod and Giertz (2012, Journal of Economic Literature) review how taxable income responds to tax rates: much of the response is avoidance and timing, which depends on how broad the tax base is. Kleven, Landais and Saez (2013, American Economic Review) find that top football players move in response to tax rates across European countries; Akcigit, Baslandze and Stantcheva (2016, American Economic Review) find the same for star inventors. Jones and Kim (2018, Journal of Political Economy) model top incomes as the reward for entrepreneurial innovation, in which case taxing them can slow growth, the Chapter 6 link.

The textbook’s own position: whether rising inequality is good or bad is hard to say, and the profession has no consensus. Ask the policy students which evidence would change their minds.

“Robots and artificial intelligence will soon make most workers unnecessary”

A skeptic says: “Machines already do much of what factory workers did, and artificial intelligence will do what office workers do. Labor demand will collapse, and most people will have no job.”

Reply: technology shifts labor demand between tasks and skills rather than abolishing it. In the college market it raised the demand for some workers, and manufacturing lost jobs, but the share of U.S. adults with a job was 59.2% in September 2026, above the 56.1% of 1960, after decades of rapid productivity growth. The real risk is who gains and who loses.

Take the objection seriously: the losses are real. Manufacturing lost 5.84 million jobs between 2000 and 2010, and Acemoglu and Restrepo (2020, Journal of Political Economy) find that each additional industrial robot in a local labor market reduced employment and wages there.

The reply has three parts. First, the fear is old: Keynes (1930, “Economic Possibilities for our Grandchildren”) predicted “technological unemployment”, and employment kept growing. Second, technology creates new tasks as well as replacing old ones: Acemoglu and Restrepo (2019, Journal of Economic Perspectives) show that new tasks have offset automation for most of the past century, and Autor (2015, Journal of Economic Perspectives) explains why machines that replace some tasks raise the value of the tasks they complement. Third, in the model of this chapter, a technology that raises productivity shifts labor demand up for the workers it complements: that is the skill-biased technical change behind the rising college premium.

What the reply does not claim: that every worker will be fine. The transition can take decades, and the gains can go to a few. That is why the policy slides on displaced workers and on top incomes matter.

Ask: which tasks in your future job do you expect AI to do, and which new tasks might it create?

What to remember from this chapter

• Read the employment-population ratio and the unemployment rate together: \(E/N = (1-u)\,LF/N\). Women drove the rise in employment.
• In supply and demand, a demand shock moves the wage and employment together; a tax lowers employment; a rigid wage turns a fall in demand into unemployment.
• The natural rate is set by job flows, \(u^* = \bar{s}/(\bar{f}+\bar{s})\), and policies that change \(\bar{s}\) can also change \(\bar{f}\).
• Present values compare payments over time: a working life is worth about $2.53 million, a degree about $386,000.
• The college premium rose with more graduates because demand for skills grew faster than supply; since 1980 the top has grown fastest.

Ask each student to write one sentence answering the opening question: who works, why can’t everyone find a job, and why do some earn more? A good answer: employment and wages follow supply and demand; unemployment comes from wages that do not fall and from the time it takes to find a job; and workers whose skills are in growing demand earn more.

The next chapter turns to the other long-run topic of the book: inflation. Then the short-run chapters bring back the rigid wages and prices of this one.

Walk the three branches of the mind map in about a minute each: who works and at what wage (cyan), unemployment (lavender), wages, skills and inequality (green). Click the image to zoom in.

Wages follow supply and demand, unemployment follows job flows, and the premium follows the race between skills and technology

Next: Chapter 8, Inflation

This chapter used one tool, supply and demand, in three markets: all labor, high school graduates and college graduates. It added two ideas the simple model lacks: wages that do not fall, and the flows of job losses and job findings.

Chapter 8 completes the long run with inflation: why prices rise, and why rapid money growth makes them rise fast.

Sources: data and models

• Figs. 1 and 2, U.S. employment and unemployment. U.S. Bureau of Labor Statistics, Current Population Survey: LNS12300001, LNS12300002, EMRATIO, UNRATE, via FRED; NBER business cycle dates (USREC); OECD harmonised unemployment rates (LRHUTTTT{CC}M156S), via FRED. Inspired by Figs. 7.1 and 7.2 and Table 7.1 in Jones (2024).
• Figs. 3 to 6, 9 and 13, and the bathtub model. Model figures adapted from Figs. 7.3 to 7.6, 7.9 and 7.13 and Section 7.4 in Jones (2024); the bathtub chart and the present-value chart are our own.
• Fig. 7 and hours worked. U.S. BLS (UNRATE); OECD (LRHUTTTTJPA156S); European Commission, AMECO (ZUTN); Penn World Table 10.01 and 11.0. Inspired by Fig. 7.7 and Table 7.2 in Jones (2024).
• Fig. 8, the college premium. Economic Policy Institute, State of Working America Data Library; OECD, Education at a Glance. Inspired by Fig. 7.8 in Jones (2024).
• Figs. 10 to 12, manufacturing and inequality. U.S. BLS, MANEMP, via FRED; ILOSTAT; World Inequality Database, via Our World in Data; Piketty, T. and E. Saez (2003), Quarterly Journal of Economics 118(1); Piketty, T., E. Saez and G. Zucman (2018), Quarterly Journal of Economics 133(2), doi:10.1093/qje/qjx043. Inspired by Figs. 7.10 to 7.12 in Jones (2024).
• 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 7.

Sources: research cited

• Who works, rigid wages and job flows. Goldin, C. and L. Katz (2002), Journal of Political Economy 110(4); Goldin, C. (2006), American Economic Review 96(2); Card, D. (1990), ILR Review 43(2); Raff, D. and L. Summers (1987), Journal of Labor Economics 5(4); Bewley, T. (1999), Why Wages Don’t Fall during a Recession, Harvard University Press; Blanchard, O. and P. Portugal (2001), American Economic Review 91(1); Giupponi, G. and C. Landais (2023), Review of Economic Studies 90(4).
• Unemployment insurance. Baily, M. (1978), Journal of Public Economics 10(3); Gruber, J. (1997), American Economic Review 87(1); Chetty, R. (2008), Journal of Political Economy 116(2); Card, D., R. Chetty and A. Weber (2007), American Economic Review 97(2); Schmieder, J., T. von Wachter and S. Bender (2012), Quarterly Journal of Economics 127(2); Ganong, P. and P. Noel (2019), American Economic Review 109(7).
• Europe. Blanchard, O. and J. Wolfers (2000), Economic Journal 110(462); Prescott, E. (2004), Federal Reserve Bank of Minneapolis Quarterly Review 28(1); Alesina, A., E. Glaeser and B. Sacerdote (2005), NBER Macroeconomics Annual 20; Krause, M. and H. Uhlig (2012), Journal of Monetary Economics 59(1).
• Skills, trade and automation. Katz, L. and K. Murphy (1992), Quarterly Journal of Economics 107(1); Goldin, C. and L. Katz (2008), The Race between Education and Technology, Harvard University Press; Autor, D., D. Dorn and G. Hanson (2013), American Economic Review 103(6), and (2016), Annual Review of Economics 8; Acemoglu, D., D. Autor, D. Dorn, G. Hanson and B. Price (2016), Journal of Labor Economics 34(S1); Pierce, J. and P. Schott (2016), American Economic Review 106(7); Card, D., J. Kluve and A. Weber (2018), Journal of the European Economic Association 16(3); Acemoglu, D. and P. Restrepo (2019), Journal of Economic Perspectives 33(2), and (2020), Journal of Political Economy 128(6); Autor, D. (2015), Journal of Economic Perspectives 29(3); Keynes, J. M. (1930), “Economic Possibilities for our Grandchildren”.
• Top incomes and taxes. Piketty, T., E. Saez and S. Stantcheva (2014), American Economic Journal: Economic Policy 6(1); Diamond, P. and E. Saez (2011), Journal of Economic Perspectives 25(4); Saez, E., J. Slemrod and S. Giertz (2012), Journal of Economic Literature 50(1); Kleven, H., C. Landais and E. Saez (2013), American Economic Review 103(5); Akcigit, U., S. Baslandze and S. Stantcheva (2016), American Economic Review 106(10); Jones, C. I. and J. Kim (2018), Journal of Political Economy 126(5).