From 1995 to 2022, U.S. GDP more than tripled, yet output did not even double

3.4× vs. 1.9× U.S. GDP in current dollars vs. U.S. output at fixed prices, 1995 to 2022

Both numbers are “GDP”. Which one tells us how much more the economy produces? Measuring the economy is the subject of today’s lecture.

Open with the puzzle. Nominal GDP was 7.6 trillion dollars in 1995 and 26.1 trillion in 2022 (the notebook’s national-accounts file): a factor of 3.41. Real GDP, in chained 2017 dollars, grew by a factor of 1.93. Ask: where did the rest go? Let students answer “prices” and then ask how anyone separates prices from quantities when an economy produces millions of different goods. We answer it in Act II.

A bit of history to set the stage: in the early 1930s, policymakers had no broad measure of total economic activity, so nobody knew how deep the Depression was. The national accounts, built by Simon Kuznets and colleagues at the U.S. Commerce Department, were the answer. Every chart in this course rests on them.

For the policy students: GDP growth targets, deficit-to-GDP ratios, debt limits and inflation targets are all defined with these measuring sticks.

Today: three questions

• The facts. Who buys U.S. output, who earns the income, and how fast have prices risen?
• The framework. How can one total be counted three ways, and how do we split its growth into prices and quantities?
• Uses and limits. How do we compare countries, and what does GDP leave out?

These are the three acts and the three branches of the mind map on the title slide. Act I shows five live charts from the national accounts. Act II derives the accounting identities behind them. Act III puts the tools to work: comparing countries, counting global poverty, the “beyond GDP” debate and the debate over labor’s falling share.

This chapter has no model of behavior. It builds the measuring sticks that every later model is tested against.

The facts

Act I

Fact 1: one kind of spending dominates U.S. GDP

What do Americans spend their GDP on? Which of consumption, investment, government purchases and net exports is the largest, and roughly how large?

Definitions first. GDP adds up, at market prices, everything a country produces in a year for final use (not the inputs used up along the way). Its spending side has four parts: consumption C (households), investment I (business structures, equipment, software and R&D, new housing, inventories), government purchases G (goods and services the government buys; not pensions or other transfers) and net exports NX, exports minus imports.

Ask the room to write down a share for each before you show the chart.

Consumption is about two-thirds of U.S. GDP; war and depression break the pattern

Data: BEA national accounts (NIPA Table 1.1.5) via FRED (GDPA, PCECA, GPDIA, GCEA, A019RC1A027NBEA); other countries: World Bank WDI. Chart inspired by Fig. 2.1 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the four shares of GDP, stacked, for every year since 1929. Net exports sit at the bottom; when they are negative (every year since 1976), the consumption band starts below zero, so the top of the stack is still 100 percent. Hover to read a value; click a component in the legend to hide it. The buttons switch the chart style and the menu picks another country (World Bank data for the G20 and OECD members).

Do live:

1. Point at the long, flat stretch after 1950: the shares hardly move.
2. Point at World War II (government purchases take almost half of GDP) and at 1932 (investment almost vanishes).
3. Choose China: in 2024 consumption is only 40 percent of GDP and investment also 40 percent. Choose Ireland: net exports are 40 percent of GDP in 2025, because multinational firms book large exports there.

The shares barely move, except in a war or a depression

• 1950–2025 averages: consumption 64%, government 20%, investment 17%, net exports −1.5%.
• World War II: government purchases peaked at 48% of GDP in 1944.
• Great Depression: investment fell to 3% of GDP in 1932, and the consumption share rose to 82%.
• The U.S. has run a trade deficit every year since 1976; the deepest was −5.7% in 2006.

Most of GDP is households buying goods and services; investment is the small part that swings the most.

The averages are the notebook’s key numbers for Fig. 1 (63.6, 20.5, 17.4 and −1.5 percent, 1950–2025). The last trade surplus was in 1975.

Why the consumption share rises in a depression: consumption falls less than GDP, because households smooth their spending, while firms stop investing almost completely. This is a first look at why investment is the most cyclical part of GDP, a theme of the short-run chapters.

Table 2.1 of the textbook breaks down U.S. GDP in 2022: 26.1 trillion dollars in today’s data, about 2 percent more than the textbook reports, because the BEA’s 2023 comprehensive revision raised the level of GDP. The shares barely changed: consumption 67.8 percent of GDP in 2022.

Fact 2: the spending mix has shifted since 1970

Since 1970, which shares of U.S. GDP rose and which fell, and by how many points?

Let the students recall the previous chart: consumption’s band got wider in recent decades. Ask: if consumption’s share rose, which other shares had to fall?

Since 1970 consumption gained 7 points; government and net exports lost almost 10

Data: BEA national accounts (NIPA Table 1.1.5) via FRED (GDPA, PCECA, GPDIA, GCEA, A019RC1A027NBEA). Chart inspired by Fig. 2.2 in Jones (2024), Macroeconomics, 6th ed.

How to read it: each line is the change of one share since a base year, in percentage points; the slider picks the base year, and the title reports the four changes up to 2025.

Do live:

1. Start at 1970: consumption +7.4, investment +2.2, government −6.2, net exports −3.4.
2. Move to 2000: consumption gained only 1.6 points since then, so the big shift happened before 2000.
3. Move to 2010: investment rises and government falls by 3.6 points each, because 2010 came right after the Great Recession, when investment was low (13.3 percent of GDP in 2009).

Other countries (the textbook’s version of this chart, with levels): China’s consumption share fell from 53 percent in 1970 to 35 percent in 2010 as investment boomed, and was back at 40 percent by 2024; Japan’s rose from 45 to 53 percent.

The changes in the four shares always add up to zero

Share of GDP (%)	1970	2025	Change (points)
Consumption	60.3	67.6	+7.4
Investment	15.8	18.0	+2.2
Government purchases	23.5	17.4	−6.2
Net exports	0.4	−3.0	−3.4

\[1 = \frac{C}{Y} + \frac{I}{Y} + \frac{G}{Y} + \frac{NX}{Y} \quad\Longrightarrow\quad \Delta\frac{C}{Y} + \Delta\frac{I}{Y} + \Delta\frac{G}{Y} + \Delta\frac{NX}{Y} = 0\]

A rising share of one kind of spending must be paid for by falling shares of the others.

The changes in the table add up to zero (up to rounding: +7.4 + 2.2 − 6.2 − 3.4 = 0.0). That is not a fact about the U.S. economy but a consequence of the definitions: the four shares add up to one in every year, so their changes add up to zero.

Use it as a data check: if your four changes do not add up to zero, you made a mistake or mixed two sources.

For policy discussion: the rise in the consumption share was matched mostly by a lower government share and a trade deficit. Ask whether a rising consumption share is good or bad news. (It depends: more consumption now, but less saving for future investment unless foreigners lend to finance it; the trade deficit is the other side of that borrowing, a theme of the open-economy chapters.)

Fact 3: labor’s share of income

Has the share of U.S. income paid to workers risen, fallen or stayed flat since 1970?

Write down your prediction: “about two-thirds and steady”, “rising” or “falling”? If it changed, by how much?

GDP is also income: every dollar of output is paid to someone. Labor’s share is the fraction paid for work (wages, salaries and benefits, plus the labor part of the self-employed’s income).

A classic stylized fact of macroeconomics says labor earns about two-thirds of income, year after year. Ask the room whether they believe it, and collect predictions by show of hands before revealing the chart.

Labor’s share has drifted down since 1970

Data: BEA national accounts (NIPA Table 1.10) via FRED (A033RC1A027NBEA, GDIA, W254RC1A027NBEA, A041RC1A027NBEA); other countries: Penn World Table 11.0 (labsh). Chart inspired by Fig. 2.3 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the U.S. labor share for every year since 1929, with the textbook’s definition: compensation divided by income net of taxes on production and of proprietors’ income. The dashed line is a straight trend; the slider picks the year where it starts, and the title reports its slope in percentage points per decade. The menu adds countries from the Penn World Table, which treats self-employment income somewhat differently (its U.S. share is about 0.03 lower).

Do live:

1. Start the trend in 1970: −1.3 points per decade.
2. Move the start to 1990 and then 2000: the decline looks steeper.
3. Add the G7: Germany fell too, from 0.69 in 1970 to 0.64 in 2023, but Japan rose from 0.51 to 0.59. South Korea rose from 0.42 to 0.55.

Two-thirds is a rough rule, not a law of nature

• Average 1950–2025: 0.65, close to the classic two-thirds.
• Peak 0.69 in 1970; below 0.60 since 2022 for the first time, 0.586 in 2025.
• Trend since 1970: about −1.3 points per decade.
• Across countries the share differs a lot: Qatar 0.18, Kuwait 0.25, where most income is oil revenue.

The labor share is roughly stable over decades, but in the U.S. it has fallen since 1970.

Reveal the predictions. The averages: 0.668 in 1950–1969 and 0.609 in 2010–2025. The 0.586 for 2025 is the lowest in the series. Our values for the textbook’s last years are slightly lower than the book’s because of the BEA’s 2023 revision.

Why care? If productivity grows but the labor share falls, wages grow more slowly than output per worker. We return to the policy debate in Act III.

Data lesson: the Penn World Table reports Venezuela at exactly 0.0844 in every year from 2019 to 2023, a value carried forward, not measured. Look at the extremes before you trust a cross-country file.

Labor’s share is more than wages

Labor’s share is just wages and salaries divided by GDP.

It also counts benefits and the labor part of the self-employed’s income, and leaves out taxes on production. In 2022, wages were 43% of GDP and all compensation 52%, but labor’s share of income was 0.60.

\[\text{labor share} = \frac{\text{compensation}}{\text{GDI} - \text{taxes on production} - \text{proprietors' income}}\]

Numbers from the notebook’s Table 2.2 for 2022: wages and salaries 42.8 percent of GDP, benefits 8.9 percent, compensation 51.6 percent, and the labor share of Fig. 3 0.598.

Why the formula works: proprietors’ income pays the self-employed both for their work and for their equipment. If we assume labor gets the same fraction θ of it as of all other income, then labor’s share of income net of taxes is exactly θ, which is the formula on the slide. Leaving proprietors out of both the top and the bottom gives the same answer as splitting their income.

The simple ratio compensation/GDP is lower because its denominator still holds the taxes and all of proprietors’ income. The notebook’s toy example: 0.643 with the formula, 0.540 with the simple ratio.

Fact 4: prices rise, but not at a steady pace

How many times higher is the U.S. price level today than in 1947?

The GDP deflator is the price level of everything counted in GDP: nominal GDP divided by real GDP, scaled to equal 100 in a base year (here 2012). Ask for guesses: twice? Five times? Twenty times?

The U.S. price level is about 12 times its 1947 level

Data: BEA GDP implicit price deflator (GDPDEF), 2012 = 100, and NBER recession dates, via FRED; other countries: World Bank WDI (NY.GDP.DEFL.ZS). Chart inspired by Fig. 2.4 in Jones (2024), Macroeconomics, 6th ed.

How to read it: the quarterly U.S. GDP deflator since 1947 (2012 = 100); shaded bands are NBER recessions. The buttons switch between a linear and a ratio scale; the menu adds groups of countries (yearly World Bank data, rebased to 2012 = 100).

Do live:

1. On the linear scale, the line looks about as steep in the 2000s as in the 1970s.
2. Switch to the ratio scale: now the 1970s clearly stand out, because on a ratio scale the slope is the percentage rate of change.
3. Add the G7 and point at Japan: its deflator fell from 119 in 1994 to 100 in 2013, two decades of deflation.
4. Add Turkey on the ratio scale: its price level in 2025 was more than 30,000 times its 1990 level.

On a ratio scale, the slope of the price level is the inflation rate

Period	1947–1965	1965–1983	1983–2019	2019–2025
Average inflation per year	2.1%	6.0%	2.2%	3.6%

\[\bar{g} = \left(\frac{P_T}{P_0}\right)^{1/T} - 1\]

The price level only tells us how far prices have come; its growth rate, inflation, tells us how fast.

The average annual growth rate ḡ is the constant rate that turns the first value into the last: P_T = P_0 (1 + ḡ)^T. The four periods are the notebook’s key numbers for Fig. 4.

Why the linear scale misleads: a 5 percent rise adds 5 index points when the index is 100, but 7 points when it is 140. The same percentage change looks bigger later. On a ratio (log) scale, ln P rises by about π each year, so constant inflation is a straight line.

Rebasing an index (2017 = 100 instead of 2012 = 100) moves its level but leaves every growth rate unchanged: the constant cancels in every ratio.

Fact 5: inflation comes in waves

When was U.S. inflation highest since 1947, and did prices ever fall?

Inflation is the growth rate of the price level. FRED’s standard measure compares each quarter with the same quarter a year earlier: π_t = 100 × (P_t / P_{t−4} − 1). Ask students to name the decade with the highest inflation, and whether they think prices ever fell after World War II.

Inflation came in two great waves: the 1970s and 2021–22

Data: BEA GDP implicit price deflator (GDPDEF) and NBER recession dates, via FRED; other countries: World Bank WDI (NY.GDP.DEFL.KD.ZG). Chart inspired by Fig. 2.5 in Jones (2024), Macroeconomics, 6th ed.

How to read it: quarterly inflation of the GDP deflator, in percent; shaded bands are recessions. The slider switches between the change from a year ago and the one-quarter change at an annual rate; the menu adds countries (yearly data).

Do live:

1. Hover over the peaks of 1975, 1981 and 2022.
2. Move the slider to the quarterly measure: it is much noisier, because one unusual quarter moves it a lot. That is why FRED’s default compares with a year ago.
3. Add Japan: its deflator fell in 18 of the 20 years from 1994 to 2013. Add China: prices fell in 2023, 2024 and 2025.

Four peaks and two troughs, and deflation is not only history

• Peaks: 8.6% in 1951 (Korean War), 10.9% in 1975, 10.2% in 1981, 7.8% in 2022.
• Troughs: −2.0% in late 1949, when prices fell, and 0.1% in mid-2009.
• Latest: 4.0% in the second quarter of 2026.

High inflation and falling prices are both live risks; why they happen is a question for the short-run chapters.

The peaks and troughs are the notebook’s key numbers for Fig. 5 (dates show the first month of each quarter). They match the textbook’s figure, except that the 2022 peak is a little higher after the data revisions.

Ask the policy students: why would a central bank worry about deflation as much as about high inflation? (Debts become heavier in real terms, and interest rates cannot fall much below zero.) We come back to this in the monetary-policy chapters.

Transition: Act I used five charts built from the national accounts. Act II asks how those accounts are put together.

The accounting framework

Act II

Value added counts each stage of production exactly once

Producer	Sells	Revenue	Buys inputs	Value added
Farm	grain to the mill	$40	$0	$40
Mill	flour to the bakery	$70	$40	$30
Bakery	bread to households	$120	$70	$50
Total		$230		$120 = final sales

\[VA_j \equiv R_j - M_j \qquad\Longrightarrow\qquad \sum_j VA_j = \text{value of final goods and services}\]

Adding up all revenues ($230) would count the grain three times; value added counts it once.

This is the notebook’s Check 1, a made-up chain of three producers. Value added is revenue R minus purchased intermediate inputs M. When we add value added along the chain, every intermediate sale appears once as a seller’s revenue and once as a buyer’s cost, so the sum telescopes to the final sale.

This is the production approach to GDP, and it explains why GDP counts only final goods. A final good is bought by its last user; an intermediate good is used up in making another good in the same period. Ask: is a used car sold this year part of GDP? (No, it was counted when it was new; but the dealer’s margin is, because it pays for a service produced this year.)

Production = expenditure = income

Production

Add the value added of every producer.

Expenditure

\(Y = C + I + G + NX\), with \(NX = X - IM\).

Income

Compensation + taxes on production + operating surplus + depreciation.

In 2022: compensation 51.6% of GDP, net operating surplus 25.3%, depreciation 16.5%, taxes 6.6%. Income (GDI) and spending (GDP) differ by only 0.03%.

Every dollar of output is bought by someone and paid to someone: three counts, one total.

Production = income: value added is paid out as wages and benefits, taxes on production, depreciation (the wear and tear of capital) and the operating surplus. The operating surplus is a residual, so the income split always adds up.

Production = expenditure: final goods are bought by households (C), firms (I), the government (G) and foreigners (X). Goods made but not sold this year count as inventory investment, as if the firm bought its own output. Imports are subtracted because C, I and G include goods made abroad.

In the data, income is measured from other sources (tax records, payroll data) than spending, so the two totals differ by a “statistical discrepancy”. The notebook finds GDI of 26,068 billion and GDP of 26,060 billion dollars in 2022.

“Gross” means depreciation is not subtracted; net domestic product is GDP minus depreciation.

The identity tells us which entries move

Event, worth $p	\(C\)	\(I\)	\(G\)	\(NX\)	GDP \(Y\)
A household buys a new car made at home	\(+p\)	0	0	0	\(+p\)
A household buys a new imported car	\(+p\)	0	0	\(-p\)	0
A firm makes goods it cannot sell this year	0	\(+p\)	0	0	\(+p\)
The government hires a teacher	0	0	\(+p\)	0	\(+p\)
The government pays a pension	0	0	0	0	0

Spending counts in GDP only when it pays for new production at home.

Go row by row and ask the room to fill in each cell before you reveal the slide (or cover it). The tricky rows:

• the imported car: it enters consumption with a plus sign and net exports with a minus sign, so GDP does not change;
• unsold goods: they count as inventory investment, so production still has a “buyer” in the accounts;
• the pension: a transfer moves income from taxpayers to retirees but buys nothing, so it is not part of G. When the retiree spends the pension, that spending shows up in C.

For policy students: this is why “government spending” in the budget is much larger than “government purchases” in GDP. Most of the federal budget is transfers.

Imports do not lower GDP

The minus sign in \(Y = C + I + G + X - IM\) means that buying imports reduces GDP.

Imports enter \(C\), \(I\) or \(G\) with a plus sign and leave through \(-IM\), so they cancel. The minus sign only removes goods made abroad that were already counted in domestic spending.

A very common error in the press and in political debate: “imports subtract from GDP growth”. In the accounts, imports have no direct effect on GDP at all. They could affect GDP indirectly (if imports replace goods that would otherwise have been made at home), but that is a question about behavior, which needs a model, not an identity.

Check with the table on the previous slide: the imported car leaves GDP unchanged.

Real GDP holds prices fixed; the price level links it to nominal GDP

\[\text{nominal GDP}_t = \sum_i p_{i,t}\,q_{i,t} \qquad\quad \text{real GDP}_t = \sum_i p_{i,b}\,q_{i,t}\]

\[\text{nominal GDP}_t = P_t \times \text{real GDP}_t\]

Nominal GDP can rise because goods cost more or because there are more goods. Real GDP changes only when quantities change; the price level \(P_t\), the GDP deflator, picks up the rest.

Notation: good i has price p_{i,t} and quantity q_{i,t} in year t. Nominal GDP values each year’s goods at that year’s prices. Real GDP values them at the prices of a base year b, so it is measured in “year-b dollars”. The price level, the GDP deflator, is defined as the ratio of the two, so the identity holds by construction. In the base year, real equals nominal and P = 1 (or 100 as an index).

With one good, P is its price and real GDP is its quantity. The difficulty is that a real economy has millions of goods whose relative prices change. Which year’s prices should we use? The next slide shows that the answer matters.

Apples and computers: two price bases give two growth rates

	2026	2027
Apples: price	$2	$3
Apples: quantity	500	550
Computers: price	$1,000	$1,000
Computers: quantity	5	6
Nominal GDP	$6,000	$7,650

• Nominal growth: 27.5%.
• At 2026 prices (Laspeyres): 18.3%.
• At 2027 prices (Paasche): 17.7%.
• Chain-weighted, the average: 18.0%.

Computers became relatively cheaper and their quantity grew faster, so the price base changes the answer.

This is the textbook’s example economy (Table 2.4), computed in the notebook’s Check 3. Apples grew 10 percent, computers 20 percent. The price of apples rose by half while computers kept their price.

Laspeyres: value both years at 2026 prices: 7,100 / 6,000 − 1 = 18.3 percent. Paasche: value both years at 2027 prices: 7,650 / 6,500 − 1 = 17.7 percent. At 2026 prices computers carry more weight (they were relatively expensive), and computers grew fastest, so the Laspeyres rate is the higher one.

Ask: which number is “right”? Neither: each answers a slightly different question. The statistical agencies take an average and link the years into a chain (next slide).

Both indexes average the goods’ growth rates, with different weights

\[g^L = \sum_i s_{i,t}\,g_i \qquad\quad \frac{1}{1+g^P} = \sum_i \frac{s_{i,t+1}}{1+g_i} \qquad\quad g^{\text{chain}} = \tfrac{1}{2}\left(g^L + g^P\right)\]

From year \(t\) to \(t+1\) …	Laspeyres \(g^L\)	Paasche \(g^P\)	Price level \(P\)
every price \(\times\lambda\), same quantities	0	0	\(\times\lambda\)
every quantity \(\times\mu\), same prices	\(\mu-1\)	\(\mu-1\)	unchanged
quantities grow fastest where relative prices fall	highest	lowest

The two indexes differ only when relative prices change, and the gap grows with the length of the period.

s_{i,t} is the expenditure share of good i in year t, and g_i its quantity growth rate. Laspeyres growth is a weighted average of the goods’ growth rates with the first year’s spending shares; Paasche is a weighted harmonic average with the last year’s shares. In the apples-and-computers economy, computers’ share is 0.833 in 2026 and 0.784 in 2027, which reproduces 18.3 and 17.7 percent.

If all quantities grow at the same rate, both give that rate. If all prices change by the same factor, it cancels. So they disagree only when relative prices move, which over long spans makes a fixed base year go out of date: Laspeyres overstates growth, Paasche understates it.

Chain weighting applies the average to each pair of neighboring years and links the growth rates into a chain of levels (“chained 2017 dollars”). The U.S. switched to chain weighting in the 1990s, when computer prices were falling fast. Agencies use the Fisher index, the geometric mean of the two growth factors; here it gives 18.0124 percent against 18.0128 for the simple average.

One caution for data work: chained components do not add up to chained GDP. Compute shares from nominal values, and real growth from chained values.

Nominal growth is about inflation plus real growth

\[1 + g_{\text{nominal}} = (1+\pi)(1+g_{\text{real}}) \qquad\Longrightarrow\qquad g_{\text{nominal}} \approx \pi + g_{\text{real}}\]

• 2022: nominal GDP +9.7% ≈ prices +7.1% + real GDP +2.4%.
• 1995 to 2022: 3.41 = 1.76 × 1.93. Prices rose 76%; output almost doubled.
• The rule fails when rates are large: in the apples-and-computers economy, 27.5% − 18.0% = 9.5%, but prices rose only 8.0%.

That answers the opening puzzle: most of the 3.4-fold rise in nominal GDP was higher prices.

Derivation: write nominal GDP = P × real GDP in two neighboring years and divide. The growth factor of a product is the product of the growth factors. Expanding, g_nominal = π + g_real + π·g_real; the cross term is the product of two small numbers. With 2 percent inflation and 2 percent growth it is only 0.04 points. In the apples-and-computers economy it is about 1.5 points: 8.04 + 18.01 + 1.45 = 27.50 percent.

With natural logs the split is exact: Δln(nominal GDP) = Δln P + Δln(real GDP).

The 1995–2022 factors are the notebook’s (nominal 3.41, real 1.93, prices 1.76). The textbook’s 7.4 trillion dollars for 1995 is now 7.6 trillion after revisions.

Rising nominal GDP does not mean the economy grew

Nominal GDP rose 11% in 2021, so the economy grew 11%.

Real GDP grew 6.3% in 2021; prices rose 4.5%. In 2020, nominal GDP fell only 0.8% while real GDP fell 2.1%. Always ask whether a number is nominal or real.

The yearly numbers are the notebook’s (2021: nominal 11.09, real 6.27, prices 4.54 percent; 2020: nominal −0.76, real −2.08, prices +1.34 percent).

Where this matters for policy: budget numbers, wages, tax brackets and debt ratios are usually reported in nominal terms. A “record” budget, a “record” minimum wage or “record” GDP may simply reflect inflation. Ratios to nominal GDP (deficit/GDP, debt/GDP) put the price level in both the top and the bottom, so they are free of this problem.

At common prices, China was almost as large as the U.S. in 2019

\[\text{real GDP}^{\text{U.S. prices}}_{\text{China}} = \frac{\text{price level}_{\text{U.S.}}}{\text{price level}_{\text{China}}} \times \text{nominal GDP}^{\text{dollars}}_{\text{China}} = \frac{Y^{\text{yuan}}}{\text{PPP}}\]

China, 2019	At the exchange rate	At common prices (PPP)
Yuan per dollar	6.91	4.73
GDP	$13.8 trillion	$20.1 trillion
Share of U.S. GDP	67%	98%

Where goods are cheap, the same dollars stand for more goods.

The textbook’s 2019 numbers, checked in the notebook (Check 7): U.S. GDP of 20.6 trillion dollars; China’s GDP of 95.1 trillion yuan; 6.91 yuan per dollar; Chinese prices at 68.4 percent of U.S. prices.

Steps: converting at the exchange rate gives 95.1 / 6.91 = 13.8 trillion dollars. But goods in China cost less, so each of those dollars buys more. Dividing by China’s relative price level, 13.8 / 0.684 = 20.1 trillion dollars at U.S. prices. The PPP conversion factor, e × (price level China / price level U.S.) = 4.73 yuan per dollar, does both steps at once.

In logs: the gap at common prices (−0.02) equals the gap at the exchange rate (−0.40) minus the log of the relative price level (−0.38).

This is the cross-country version of the Laspeyres index (China’s goods at U.S. prices). The World Bank’s International Comparison Program and the Penn World Table use average world prices, “international dollars”.

A currency swing changes dollar GDP, not GDP at common prices

A 1% rise in …	GDP at the exchange rate	GDP at common prices
the exchange rate \(e\) (a weaker yuan)	−1%	0
every yuan price in China	+1%	0
China’s output of every good	+1%	+1%

Only real production moves GDP at common prices; currency markets and local inflation do not.

The notebook’s check: a 1 percent weaker yuan makes China’s GDP 0.99 percent smaller in dollars, while GDP at common prices does not move at all, because China’s goods become cheaper in dollars by the same amount.

This is why comparisons of living standards use PPP, and why “China overtook / fell behind the U.S.” headlines based on market exchange rates can flip from one year to the next without any change in production.

Uses and limits

Act III

Market exchange rates exaggerate gaps in living standards

Converting GDP at market exchange rates is the right way to compare living standards across countries.

Services that cannot be traded, like haircuts and restaurant meals, use a lot of labor, so they are cheap where wages are low. Poor countries have low price levels, and market rates make them look poorer than they are.

The China example on the previous slides is one case: 67 percent of U.S. GDP at the exchange rate, 98 percent at common prices. The same pattern holds generally: the poorer the country, the lower its price level, and the bigger the correction.

When are market rates the right tool? For questions about buying power on world markets: paying for imports, servicing foreign-currency debt, the size of a country’s economy for trade negotiations. For living standards, productivity and poverty, use PPP.

Global poverty is counted in PPP dollars

How should the World Bank decide who is “extremely poor” in countries with very different prices?

Why PPP

• A poverty line should buy the same basket of goods everywhere.
• At market rates, a dollar a day buys more in a poor country, so poverty would be overstated.
• PPP factors come from price surveys in many countries (the International Comparison Program).

Why it is debated

• New price surveys revise PPPs, and the count of the poor can move a lot.
• The basket behind PPP reflects average spending, not what poor households buy.
• The line itself is a judgment, updated as new prices arrive.

The World Bank sets an international poverty line in PPP dollars and counts how many people live below it. In June 2025 it updated the line to 3.00 dollars a day at 2021 PPPs (from 2.15 dollars at 2017 PPPs), using the latest International Comparison Program prices (World Bank, Poverty and Inequality Platform, June 2025 update).

Deaton (2010, American Economic Review) shows how a single revision of the PPPs, from the 2005 price survey, changed the estimated number of the world’s poor substantially, and argues that the price indexes behind PPP are not designed for the poor. Ferreira et al. (2016, Journal of Economic Inequality) explain how the line is updated when new PPPs arrive. Present the method as the best available, and its uncertainty as real.

Discussion question: if a new price survey says prices in poor countries are higher than we thought, does the number of poor people rise or fall? (It rises: the same local income buys fewer PPP dollars.)

Beyond GDP: health, leisure and inequality also matter for welfare

Should governments add, or even replace, GDP with a broader measure of well-being?

What GDP leaves out

• Work at home: cooking, cleaning, caring for children.
• Longer and healthier lives, and leisure.
• Depletion of natural resources and damage to the environment.
• How income is shared.

What a broader measure finds

• Western Europe is closer to the U.S. than its GDP suggests: longer lives, more leisure, less inequality.
• Most poor countries are further behind.
• Across 100+ countries, welfare and GDP are still strongly correlated.

The right column summarizes Jones and Klenow (2016, American Economic Review), the study the textbook’s “Beyond GDP” case study reports. They start from consumption per person and add life expectancy, leisure and consumption inequality, using standard utility functions. The book chapter explains why we do not rebuild their measure: it needs household-survey data on consumption inequality that are not openly available.

The policy side: the Stiglitz–Sen–Fitoussi commission (2009), set up by the French government, recommended a dashboard of indicators (income and consumption by household, health, education, environment, inequality) rather than a single number. Several countries now publish well-being indicators next to GDP.

Ask: which omitted item would change rankings the most for your own country?

Who gains from growth? The labor-share debate

If labor’s share keeps falling, should policy respond, and how?

Proposed explanations

• Cheaper machines and software replace workers.
• “Superstar” firms with high profits and few workers win larger market shares.
• Globalization and weaker unions lower workers’ bargaining power.

Why it is contested

• Part of the fall comes from how the accounts treat software and R&D.
• Self-employment income and depreciation are hard to split.
• Trends differ across countries: down in the U.S. and Germany, up in Japan and Korea.

Fig. 3 showed the U.S. decline since 1970. Influential studies: Karabarbounis and Neiman (2014, Quarterly Journal of Economics) document a global decline and link it to the falling price of investment goods; Autor, Dorn, Katz, Patterson and Van Reenen (2020, Quarterly Journal of Economics) link it to the rise of “superstar” firms; Elsby, Hobijn and Şahin (2013, Brookings Papers) show that how self-employment income is treated changes the size of the decline; Koh, Santaeulàlia-Llopis and Zheng (2020, Econometrica) argue that most of it reflects the capitalization of intellectual property products in the national accounts. The book chapter notes that this measurement question is still debated.

Policy angles: the explanation matters for the response. Market power points to competition policy; automation to training and taxes on capital versus labor; bargaining power to labor-market institutions. A measurement artifact would need no response at all. Present the debate as open.

Same spending, different box: check the definitions before comparing

U.S., share of GDP (%)	1970	2000	2024
Investment, BEA	15.8	19.9	18.1
Investment, World Bank	21.4	23.7	21.5
Government purchases, BEA	23.5	17.8	17.5
Government purchases, World Bank	18.0	14.0	13.6

Two sources, one economy, a 4-point gap: the World Bank counts public investment as investment. Data are also revised: U.S. GDP for 2022 is now 2.2% higher than in the textbook.

The numbers are the notebook’s comparison of the BEA and World Bank series for the U.S. The World Bank follows the UN System of National Accounts, in which government spending on roads, buildings and equipment counts as investment and government purchases cover only government consumption. The BEA puts government investment in G. Consumption and net exports agree.

Revisions: the BEA’s 2023 comprehensive update raised 2022 GDP to 26.06 trillion dollars, against 25.5 trillion in the textbook. Shares changed much less than levels.

For policy analysts this is the most practical lesson of the chapter: before you compare two countries, or two reports, check that the numbers use the same definitions and the same vintage of data.

“GDP is the wrong target for policy”

A skeptic says: “GDP ignores leisure, health, inequality and the environment. Governments should stop chasing GDP growth and target well-being instead.”

Reply: GDP was built to measure market production, and it does that consistently across years and countries. Broader measures are strongly correlated with it, and they are built from the same accounts. Use GDP for what it measures, and add other indicators for what it leaves out, rather than replacing it.

Take the objection seriously: a policy that raises GDP by polluting more or working longer hours may lower welfare. That is a reason to look at more than one indicator.

But replacement has costs. Well-being measures need weights (how much is a year of life worth relative to consumption?), and those choices are contestable. GDP’s identities make it checkable: production, spending and income must agree. And across countries, the Jones and Klenow welfare measure and GDP per person are highly correlated, so GDP is an informative, if imperfect, guide to living standards.

Ask the policy students: which single extra indicator would you put next to GDP in a budget speech, and why?

What to remember from this chapter

• GDP counts final goods once. Production, expenditure and income give the same total, and \(Y = C + I + G + NX\) splits it by buyer.
• In the U.S., consumption is about two-thirds of GDP; labor’s share is near two-thirds but has fallen since 1970.
• Real GDP holds prices fixed. Laspeyres and Paasche average growth with different weights; chain weighting averages the two.
• Nominal growth ≈ inflation + real growth. To compare countries, use common prices (PPP).

Ask each student to write one number from today they will remember, and one question about measurement they still have. Good questions to seed: how are prices of new goods, like a new phone model, compared with old ones? How is the output of a hospital or a school measured?

Walk the three branches of the mind map in about a minute each: the facts (cyan), the accounting framework (lavender), uses and limits (green). Click the image to zoom in.

Measure before you model: one total counted three ways, split into prices and quantities, compared at common prices

Next: Chapter 3, An Overview of Long-Run Economic Growth

Chapter 3 uses real GDP per person at common prices, exactly the measure built today, to study why some countries are so much richer than others.

Sources

• Figs. 1 and 2, the expenditure shares. BEA national accounts (NIPA Table 1.1.5) via FRED: GDPA, PCECA, GPDIA, GCEA, A019RC1A027NBEA; World Bank WDI shares of GDP. Inspired by Figs. 2.1 and 2.2 in Jones (2024).
• Fig. 3, labor’s share. BEA national accounts (NIPA Table 1.10) via FRED: A033RC1A027NBEA, GDIA, W254RC1A027NBEA, A041RC1A027NBEA; Penn World Table 11.0 (Feenstra, Inklaar and Timmer 2015, doi:10.34894/FABVLR). Inspired by Fig. 2.3 in Jones (2024).
• Figs. 4 and 5, the GDP deflator and inflation. BEA GDPDEF and NBER business cycle dates, via FRED; World Bank WDI (NY.GDP.DEFL.ZS, NY.GDP.DEFL.KD.ZG). Inspired by Figs. 2.4 and 2.5 in Jones (2024).
• Worked examples. Apples and computers: Table 2.4 in Jones (2024); China and the U.S., 2019: Section 2.4 in Jones (2024); both recomputed in the book chapter.
• Global poverty and PPP. World Bank (2025), Poverty and Inequality Platform, June 2025 update; Deaton, A. (2010), American Economic Review 100(1); Ferreira, F. et al. (2016), Journal of Economic Inequality 14(2).
• Beyond GDP. Jones, C. I. and P. J. Klenow (2016), American Economic Review 106(9); Stiglitz, J., A. Sen and J.-P. Fitoussi (2009), Report by the Commission on the Measurement of Economic Performance and Social Progress.
• The labor share. Karabarbounis, L. and B. Neiman (2014), Quarterly Journal of Economics 129(1); Autor, D., D. Dorn, L. Katz, C. Patterson and J. Van Reenen (2020), Quarterly Journal of Economics 135(2); Elsby, M., B. Hobijn and A. Şahin (2013), Brookings Papers on Economic Activity, Fall; Koh, D., R. Santaeulàlia-Llopis and Y. Zheng (2020), Econometrica 88(6).
• 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 2.