Subscribe now

Why it matters

Low does not mean slow. China’s consumption, as previously argued, is a low share of the economy. But, also as previously argued, over the past quarter-century it grew faster than in any other major economy. Figure 1 shows it has continued to grow fast since the pandemic.

It is an easy-to-understand illustration, but should not be overweighted. Based on photos taken by audiences in China, it was one of my more popular pandemic-era charts. On the surface, it rebuts the weak consumption narrative by showing China winning this tale-of-three-economies horse race.

Fast does not mean fast as before. The biggest critique of Figure 1 is that consumption in China always grows faster. An alternative question is how much consumption slowed relative to normal. On that measure, a case could be made that China slowed the most. As discussed below, though, that question is harder to answer and, for China, hinges on how the normal growth rate is defined.

Fast also does not mean fast enough. The desirability of boosting consumption in China is a rare point of global agreement. China wants it, other countries want it, and international organizations (such as the IMF) also want it. Why China does is obvious. For the rest of the world, it is related to my previous posts on how faster consumption growth would help reduce global imbalances.

Details

Annualized average growth rates provide another way to compare the speed. Figure 2 shows that China’s consumption grew at a compound annual growth rate of 4.3 percent since the pandemic, compared with 2.2 percent in the US and an OECD median of 0.9 percent. So it grew almost twice as fast as in the US and around eight times faster than in the euro area.

Most OECD economies saw per person consumption grow much more slowly than China. Figure 3 is the same as Figure 1, but adds an OECD median line as well as a shaded area for the interquartile range (25th to 75th percentile). The euro area is at the slower end of the OECD, whereas both the US and China are much higher. Figure 4 shows all of the economies in the OECD with data. Set against them, China is the second highest, with Türkiye the one economy growing faster.

Where did consumption slow more than normal? That is the harder question, because it depends on what normal is. For the US and euro area, normal is safely the average of the 2017 to 2019 growth rates, since growth was not slowing before the pandemic. For China it is not: consumption, like GDP, had been slowing for years. Figure 5 measures each economy against its own 2017–19 average growth rate, assuming that is normal for the 2020 to present period. On this basis, China’s consumption was the most below normal of the three economies. That does not contradict Figure 1. Both are true: China grew the fastest and fell the furthest below its own normal during the pandemic (measuring normal by a pre-pandemic average).

But this finding is highly sensitive to how normal is defined. Figure 6 shows China’s index against two plausible trends. The first, labeled “Constant”, holds growth at 5.2 percent, the assumption behind Figure 5, and puts China 5.7 percent below trend in 2026Q1 and 6.0 percent by Q2 (Q2 is available for China but not the OECD). It fits the early data poorly: consumption was growing much faster than 5.2 percent a year in 2013 to 2017, which is why the trend runs well above actual through those years. The second, labeled “Decay”, lets growth decay towards a steady state (see the data note), fits the early data far better, and leaves China only 1.5 percent below trend in 2026Q2. The deeper problem is that nothing tells us which trend is right; that would take a much richer model than a line fitted through data. So I would read Figure 5 cautiously, and I have tended to show variants of it sparingly.

Data

Construction. China’s consumption comes from the NBS household survey, accessed through CEIC in August 2026. The data are year-to-date nominal, which I convert to real, deflating by the headline CPI, and seasonally adjust. The CPI is used because China publishes no deflator for household consumption. The CPI index itself is downloaded from the IMF’s Consumer Price Index dataset, accessed in August 2026, and, given publication lags, recent values are estimated using a weighted average of published year-on-year and month-on-month growth rates. These calculations closely match the NBS’s own published real year-to-date growth rates: across the 50 quarters both are available, the median absolute deviation is 0.04 percentage point and the largest is 0.16. The seasonal adjustment uses X-13ARIMA-SEATS and specifies three pandemic quarters as outliers: a permanent level shift in 2020Q1, and one-off breaks in 2022Q2 and 2022Q4. Data for the other economies come from the OECD Household Dashboard, downloaded in August 2026. These are published in real, seasonally adjusted, per capita terms already, so no adjustments are made. Euro area refers to Euro Area 21. The OECD data stop at 2026Q1 and China’s at 2026Q2.

Comparability. The OECD data are comparable by design. Even the US, which publishes personal consumption expenditure using a different definition, is included in the OECD dataset with a harmonized definition. China’s data, meanwhile, are from the household survey instead of the national accounts, which don’t have quarterly consumption. Moreover, the household survey and national account data do differ, which will be covered in a future post. But their growth rates are broadly similar and, since 2019, national accounts household consumption has grown a bit faster than the household survey.

The index. Every series is scaled so that its own 2019 full-year average equals 100. Averaging the four quarters of 2019 rather than anchoring on 2019Q4 avoids resting the whole comparison on a single quarter, which could have been unusually high or low. Because it is an index, the charts compare cumulative growth since 2019.

• Figure 1. Real consumption per person, seasonally adjusted, and indexed so that the 2019 average is 100.

• Figure 2. Compounded annualized growth rates version of Figure 1. The base is the 2019 average, not a single quarter, so the annualized growth rate is measured assuming the mid-point of each period. This means the base is 2-Jul-2019 and, for example, 14-Feb-2026 is the mid-point of 2026Q1, which puts the two 6.62 years apart and converts the 31.9 percent cumulative growth into 4.3 percent annualized.

• Figure 3. Repeats Figure 1 but adds the distribution of the other 36 OECD economies with data. The shaded band is the interquartile range and the grey line the median. At 2026Q1 the median is 105.8 and the band runs from 103.0 to 109.9.

• Figure 4. Similar to Figure 3, but shows each economy. Türkiye is the highest at 169.1 and Hungary the next at 128.7.

• Figure 5. Similar to Figure 1, except the line is the deviation relative to a counterfactual of constant growth at the pre-pandemic (2017–19) average growth rate. Specifically, China is 5.2 percent, the United States 2.1 percent, and the euro area 1.2 percent. China’s data are available since 2013, and regressions find that a time trend in the growth rate over this period is statistically significant for China but not the United States or the euro area.

• Figure 6. China compared to two counterfactuals. The first is the constant growth of Figure 5. The second allows for a trend decline in the growth rate as visible in the data. The trend assumes that consumption growth converges to a steady state at a constant proportional rate. The steady state is assumed to be 3.5 percent. The intercept, the initial growth rate, and the speed of convergence are estimated using non-linear least squares on the natural log of consumption over 2013–2019. The results do depend on the assumed steady state growth rate, with assumptions between 3 and 4 percent yielding shortfalls in 2026Q2 consumption between 0.7 and 2.5 percent.

The analysis, charts, and drafting in this post are mine. I use Claude as a copy editor, sounding board, and research assistant.

Subscribe now

Subscribe now

Why it matters

Domestic demand weakened considerably. The corollary to more reliance on foreign demand is less on domestic demand. As Figures 1 to 3 show, not only did the share of growth coming from domestic demand decline, but so too did the absolute amount as measured in percentage points. As argued in the last post, this is due in large part to the property market correction dragging down investment.

An increase in foreign demand helped avoid a recession. Real estate corrects, investment weakens, and either output falls or the trade surplus rises. A key reason the property correction did not cause a recession was this rotation in demand away from the domestic economy and toward foreigners. But foreigners were not going to buy houses.

Production rotated away from housing toward new economy goods that were sold to foreigners. This is the third leg of the story. The decline in the construction of housing was offset by an increase in production of other goods, such as EVs and batteries. These goods, in turn, found big markets abroad.

The combination of the above three factors explains the rise in the current account surplus since 2018. All three are needed. The investment decline, by itself, could have simply led to lower output and no rise in the surplus. Instead, foreign demand replaced the decline in domestic demand. But this required a rotation in production into goods that foreigners would want to buy. All three happened, which is in a nutshell why China’s current account surplus rebounded since 2018, and also how China avoided the property correction causing a big recession. This is a simplification. Higher investment in new economy sectors and resilient consumption also helped offset the real estate shock.

Details

The decline in domestic demand is easiest to see in percentage points. Figure 1 shows what percentage of growth came from net exports, with 33% in 2025 implying that 67% came from domestic demand. Figure 2 looks at domestic demand directly and switches to percentage points of growth. On average, domestic demand accounted for 9.2 percentage points of growth between 2002 and 2019, and never less than 5.4 percentage points (left-most columns in Figure 2). In 2020 and 2022 it was low, arguably reflecting the impact of the pandemic. Moreover, during the pandemic-affected period, 2020 to 2023, domestic demand contributed 4.3 percentage points on average. The most recent two years therefore stand out, as domestic demand contributed only around 3½ percentage points to growth in each year.

By construction, contributions to growth can be split into domestic and foreign. Figure 3 shows this decomposition, again in percentage points. The sum of the net export and domestic demand bars is GDP growth. This reinforces the messages from above. Pre-pandemic, domestic demand was the main contributor to growth, and the most recent two years stand out for the combination of a weak contribution from domestic demand being partly offset by a stronger contribution of net exports.

Net exports’ share of growth is at a post-WTO high, but its contribution in percentage points is not. Figure 4 shows the year-by-year contribution of net exports in percentage points. In 2006 and 2021 net exports had a larger contribution in percentage-point terms, with 2006’s share much lower because overall growth was so much higher (and this was the height of the WTO accession export boom), and 2021 reflecting pandemic-induced anomalies in global supply and demand. Thus 2024 and 2025 stand out for the large absolute contribution to growth in what could be considered normal times.

Is there evidence that the rise of the new economy offset the decline in real estate? Strong evidence comes from recent CF40 research by DU Yuchen (abstract free, full brief paywalled). Measuring each segment by the value added its final demand generates along the whole supply chain, the piece finds that since 2023 the new economy’s contribution to growth has been enough to offset the real estate drag. On the demand side, exports became the largest marginal source of the new economy’s growth in 2025, and this also shows up in the latest export data. Moreover, production of new economy goods depends more on external demand than real estate did, which leaves the economy more exposed to the external environment.

Further evidence comes from a year-by-year decomposition of the changes in the current account. Because the three shares sum to national income, their changes sum to zero: the current account rises by exactly as much as consumption plus investment falls. Starting in 2021, real estate investment began falling. In 2021 and 2022, this was largely offset by rising investment in other sectors, consistent with a rise in new economy investment. In 2023, with the reopening of the economy after the pandemic, consumption rebounded and offset declining real estate investment. The surplus starts to rise again in 2024 and 2025. Consumption is basically unchanged as a share of national income, and investment is declining, fueled by real estate. This shortfall in domestic demand was offset by a rising surplus, that is, by external demand. The accounting behind this decomposition, and the rest of what these posts have leaned on, is set out below.

Accounting primer

This section is a primer on the accounting used in the last several posts. The basic equation is Y = C + I + CA, where Y is a measure of national income. Y is usually thought of as GDP, which can be calculated equivalently from domestic production, domestic spending, or domestic income. GNDI is just GDP plus net income from abroad.

The intuition is simplest with production. Domestic production (Y) must be used for consumption (C), investment (I), or net exports. This assumes away foreign income flows for simplicity, so net exports (exports minus imports, NX = X − M) equals the current account (NX = CA) and GDP and GNDI coincide. With rearranging: Y − C − I = CA. This is the key accounting identity. Whatever is produced but not used at home is exported. And, similarly, if domestic demand (consumption plus investment) is greater than what is produced at home, the excess demand is imported. It can also be written as saving minus investment equals the current account, where saving is S = Y − C, which yields S − I = CA. This post focuses on Y = DD + CA, or GDP = domestic demand plus net exports, where domestic demand is consumption plus investment (DD = C + I).

How an investment decline affects the current account depends on what happens to the other variables. Mechanically, if investment falls, output (Y) could go down, consumption (C) could go up, or the current account (CA) could go up. And that is assuming just one variable adjusts. This is why last week’s post argued that the accounting identity does not imply causality. It must hold at the end of the day, but getting at causality requires tracing out how the investment decline affects the other variables.

A stylized example illustrates how this works. Table 2 is loosely calibrated on the 2018 to 2025 change in the current account surplus. The first column and top block are the starting point, in levels, for this simplified version of China’s economy. The second column shows an investment decline fully offset by less production. The current account surplus is unchanged in this case. Column 3 shows what happens if, instead, output does not change and the decline in real estate investment is offset by production of other goods that are then exported. Here output is unchanged yet the current account surplus goes up. This is the result that more closely resembles what happened in China. It also illustrates that the decline in investment did not necessarily mean that the current account surplus must rise. Indeed, for that to happen, the economy had to pivot out of building housing and into building tradable goods that could be sold abroad.

The idea is similar in income shares, shown in the lower block. Output is 100 percent of itself, which is why that row is 100 in every column. In column 2 the investment decline shows up as a lower ratio to income. Consumption, moreover, rises because the denominator falls, even though the level of consumption is unchanged. The current account surplus, near zero to begin with, stays there. In column 3, by contrast, output is unchanged and so is consumption, so that ratio is the same. The investment decline is fully offset by the rise in the surplus. Again, this result depends on output pivoting out of real estate and into goods that could be sold abroad. A decline in investment by itself is not enough to produce a rise in the surplus. Column 4 resembles what actually happened between 2018 and 2025, by adding a consumption rise to the investment decline.

Applying the accounting to China

My argument in this post is that the investment decline did not reduce growth. While this could be contested, the slowing is in line with trend, as argued in an earlier post. Thus, I argue that column 3 and column 4 are more relevant than column 2. The economic logic is that investment fell with the property market correction; the decline in the production of housing was offset by production of new economy goods like EVs; this prevented GDP from falling on the supply side; and on the demand side, these new goods were sold in large part to foreigners, causing the current account surplus to rise. Ex ante, the investment decline could have reduced output instead of raising the surplus. The accounting identity itself cannot answer that question. (Table 1 adds that investment in other sectors, which the CF40 research identifies as the new economy, also supported demand in 2021 and 2022.)

To close the loop: if China were manufacturing a lot of cars but could not sell them anywhere, the unsold cars would sit in a lot and be recorded as inventory investment. Without foreign demand the new production would have piled up that way, and investment would not have fallen as it did.

Related posts

On China’s growth

1. What to Make of China’s Q2 GDP? — growth was low by Chinese standards, but the level of GDP is broadly on trend.

2. China’s Real Estate Correction: How Big Is It? — big, and not yet over.

On the surplus behind it

3. Consumption is NOT to Blame for China’s Rising Surplus — consumption’s share of the economy has been rising since 2010, so it is not what is driving the surplus.

4. China’s Macro Race: Can Saving Fall Faster Than Investment? — from 2010 saving and investment fell in step, so the surplus went nowhere.

5. Accounting for China’s Post-2018 Surplus Rise: Investment, Not Consumption — the post-2018 jump is investment, tied to the property correction; consumption rose.

6. China’s Current Account Roller Coaster — the surplus up, down, and up again since 2000.

7. Global Imbalances: One Big Deficit, Many Big Surpluses — one large deficit economy, three surplus blocs.

Data

Construction. The data are taken from CEIC and accessed in August 2026. The National Bureau of Statistics (NBS) reports contributions to growth in both percentage points and share of total, from final consumption, gross capital formation, and net exports of goods and services. The tables use national accounts expenditure data and fixed asset investment data, also from CEIC. The figures are shares of GDP growth and the tables are shares of GNDI. An earlier post sets out how GNDI is constructed and why it is the better denominator for the ratios in the tables. The figures use GDP because NBS publishes contributions to GDP growth, not to GNDI.

• Figure 1. Net exports’ percent share of real GDP growth: the 2002–2019 average, then every one of those years as a dot, then each year from 2020. 2009 is excluded from the historical spread and the average because at −42% it distorts both.

• Figure 2. As Figure 1, but domestic demand instead of net exports and in percentage points instead of share.

• Figure 3. Stacked domestic demand and net exports contributions in percentage points. They sum to total GDP growth, which is shown as the black line.

• Figure 4. Net exports’ contribution to GDP growth in percentage points with a 5-year moving average. The axis is truncated below −1.5 for readability, which affects the 2009 bar where the actual value is shown.

• Table 1. Changes (first difference) in each component’s share of GNDI. Current account, consumption, and investment come from the national accounts. Real estate is estimated from FAI construction and installation, and other investment is calculated as a residual. See last week’s post for how this series was constructed and caveats on mixing with national accounts investment data, as done in this table.

• Table 2. Illustrative and made up numbers. They are loosely calibrated to match China’s 2018 shares of GNDI in column 1 and its 2025 shares in column 4. In the lower block, note that a component whose level does not change still shows a larger share when output falls, because the denominator is smaller. The surplus starts near zero here so this is not visible, but it is why consumption’s share rises in column 2 on an unchanged level.

  Subscribe now

The analysis, charts, and drafting in this post are mine. I use Claude as a copy editor, sounding board, and research assistant.

Why it matters

Subscribe now

Declining investment accounts for the rebound in the current account surplus. To be specific, the surplus increased by 3.5 percentage points of national income between 2018 and 2025. Investment declined 4.5 points, which by itself would have raised the surplus by the same amount. Consumption rose by 1 point, which is why the surplus rose 3.5 points rather than 4.5. Given the enormous real estate correction under way, it is not surprising that investment is falling as a share of the economy. The details below show how real estate is weighing on total investment.

What is surprising is the resilience of consumption. The real estate shock should have lowered consumption as well, given that housing is the bulk of household wealth, as I discussed two weeks ago. Instead consumption rose by a percentage point of the economy.

The focus so far has been on accounting, which is not the same as causality. It is tempting to conclude from Figure 1 that the decline in investment caused the surplus to rise. Causality, however, requires moving beyond accounting identities and into economics. In this case, I believe declining investment was a key cause of the rising surplus. But by itself it is not sufficient. To give away next week’s punchline, the pivot toward producing new-economy goods such as EVs, and selling them abroad, is also a big part of the story.

Details

This post focuses on the re-emergence of the current account surplus since 2018. As Figure 2 shows, and as I discussed in an earlier post, the current account surplus as a share of the economy bottomed out just above zero in 2018. Since then it has been on an upward trend of, using a simple linear fit, nearly 0.4 percentage points of national income a year. At first the rise could be discounted as a pandemic side-effect, with supply disruptions and demand shocks both favoring Chinese exports. That reading no longer works. At 3.7 percent of national income in 2025, the surplus is well above anything it reached during the pandemic years.

Since 2018, consumption is up and investment is down. Figure 3 is much the same chart as the first two in last week’s post, with consumption in place of saving. Consumption in 2025 was a percentage point of national income higher than it was in 2018, 57.2% against 56.2%. Investment, meanwhile, was 4.5 percentage points lower, 39.1% against 43.6%. The accounting identity, with all variables scaled by national income, is 1 − c − i = ca, where c is consumption, i is investment, and ca is the current account. Finally, where Δ gives the change between two periods, Δca = −Δ(c + i) = −Δc − Δi. This is why the rise in consumption reduces the surplus, while the fall in investment raises it.

Investment in real estate slowed sharply with the housing market correction. Figure 4 shows fixed asset investment (FAI) in real estate construction and installation, which excludes land purchase among other items, as a share of national income. It fell by 3.4 percentage points between 2018 and 2025, about three-quarters of the 4.5-point fall in total investment over the same period. An important caveat: FAI and the national accounts’ gross capital formation are not the same concept. Focusing on construction and installation helps, but does not remedy the differences. A fuller discussion is parked for another day. See the data section for how I reconstruct the FAI data from growth rates.

Related posts

On consumption

1. The Paradox of the Chinese Consumer — consumption has grown faster in China than anywhere else, yet its share of the economy fell.

2. How Low Is China’s Household Consumption Share? — very low, and the ranking holds whichever year, denominator, or sample you pick.

3. Why Do Consumers Still Feel So Bad? — weak consumer confidence is not a Chinese phenomenon, it is a global one.

4. What Is Worrying China’s Consumers? — the labor market, mostly.

On global imbalances

5. Consumption is NOT to Blame for China’s Rising Surplus — consumption’s share of the economy has been rising since 2010, so it is not what is driving the surplus.

6. China’s Macro Race: Can Saving Fall Faster Than Investment? — from 2010 the two fell in step, so the surplus went nowhere.

7. China’s Current Account Roller Coaster — the surplus up, down, and up again since 2000.

8. Global Imbalances: One Big Deficit, Many Big Surpluses — one large deficit economy, three surplus blocs.

On real estate

9. China’s Real Estate Correction: How Big Is It? — big, and not yet over.

Data

Construction. Unless noted otherwise, the national accounts and balance of payments series are taken from CEIC, accessed in July 2026. Investment is gross capital formation, the published expenditure component, which covers fixed investment plus the change in inventories. Consumption is the published total, both household and government. The current account is net exports of goods and services plus net primary and secondary income. The income components, which enter both national income and the current account, come from two places. From 2010 they are the renminbi balance of payments in CEIC. Before 2010 that renminbi series does not exist, so they come from the IMF’s Balance of Payments database in US dollars, converted to renminbi at the implied exchange rate from the IMF World Economic Outlook. Where the two overlap they agree closely, so the join is not visible in any of the charts.

GNDI. As I argued two weeks ago, GNDI is a better denominator for these charts. Gross national disposable income is expenditure GDP plus net income from abroad, meaning net primary income plus net secondary income. It is the income actually available to residents to spend or save. The longer version of that argument is in the first post of this series.

FAI. Figure 4 uses fixed asset investment, which the National Bureau of Statistics includes in its monthly data releases. Those series were downloaded from CEIC in August 2026.

• Figure 1. The left panel shows the current account balance as a percent of GNDI in 2018 and 2025. The right panel’s first bar shows the change in the surplus over that period, a rise of 3.5 percentage points of GNDI. The next two bars show how much consumption and investment contributed to this change. What may be confusing is that the bars are scaled so that a positive value increases the surplus and a negative value decreases it. Consumption is −1.0 points, which is the size of the rise in consumption’s share of GNDI between 2018 and 2025 and, equivalently, how much that rise contributed to decreasing the surplus. Likewise, investment is +4.5 because that is the fall in investment over the period and how much it contributed to increasing the surplus.

• Figure 2. The current account balance as a percent of GNDI, 2000 to 2025, with ordinary least squares trends fitted separately from 2010 to 2018 and from 2018 to 2025. The fitted slopes are not the same as the change between the end years, because the surplus does not move in a straight line between them.

• Figure 3. Consumption and gross capital formation as a percent of GNDI, 2000 to 2025, with 2018 and 2025 marked. No trends are fitted.

• Figure 4. Real estate development investment as a percent of GNDI, the construction and installation component only, which excludes equipment purchase and other expenses, the largest of which is land. I show this component rather than the total because it is closest conceptually to the national accounts’ gross capital formation, the investment measure in the figures above. The levels I use are not the published ones. NBS’s published growth rates are not always consistent with those calculated from changes in the published levels, and the growth rates are generally preferred because they adjust for changes in the sample. I therefore rebuild the levels by chaining the published year-on-year growth back from 2025. It matters: on the published levels the decline in construction and installation is 4.8 points, more than the entire decline in investment. But as noted above, FAI and national accounts gross capital formation are not conceptually the same, and comparisons such as the one in this chart should be made with some caution.

The analysis, charts, and drafting in this post are mine. I use Claude as a copy editor, sounding board, and research assistant.

Subscribe now

Subscribe now

Why it matters

The race between saving and investment determines the current account surplus. By accounting identity, in terms of share of the economy, saving minus investment equals the current account surplus (see Figure 2). So for the surplus to decline, saving has to fall faster than investment.

Lowering the surplus and China’s high investment at the same time is hard. In my days at the IMF, we spent a lot of time on this. Lower investment was desirable, so too was lowering the current account surplus. The cure for both required higher consumption. So consumption’s share had to rise by more than enough to lower the current account surplus; it also had to cover the decline in investment.

Consumption did rise considerably since 2010, but this only offset the decline in investment. This reconciles last week’s post: consumption is up 7 percentage points of national income since 2010, with nothing left over for the surplus (see Figure 3). Thus, the current account surplus is right back to where it was in 2010. This does not mean that China did not adjust, but rather it did not adjust enough; the required move in consumption is a bigger challenge than it would appear by just considering the current account. The choice of 2010 as the starting point is post-GFC, but admittedly a bit arbitrary. Next week, I’ll show what has happened just since the current account trough in 2018.

Details

External imbalances are mirrored by domestic imbalances. I’ve argued this in many posts; Figure 2 shows it visually. The current account surplus is the difference between saving and investment. For the surplus to move, the gap between saving and investment must move. This is why any discussion of the current account is also a discussion of saving and investment. The figures scale everything by gross national disposable income (GNDI), which as explained last week is better suited to these charts than GDP, though the two give much the same picture.

Since 2010, the rise in consumption almost exactly offset the fall in investment. This was argued above, and Figure 3 shows it graphically. The current account surplus went from 3.9 percent of GNDI in 2010 to 3.7 percent in 2025. With rounding, the right part of Figure 3 shows this as a decline of 0.1 percentage points of GNDI. The rise in consumption by itself would have lowered the surplus by 7.4 percentage points. The fall in investment would have increased the surplus by 7.3 percentage points. The net change is a decline of 0.1 percentage points. So the surplus is unchanged, but this masks what are actually large changes in investment and consumption. For example, if investment had stayed at its 2010 share, the increase in consumption’s share would have reduced the current account from a surplus of 3.9 percent of GNDI in 2010 to a deficit of 3.6 percent in 2025. The 2018-25 version of this chart will be next week’s lead. It tells a different story: the rise since 2018 is clearly led by falling investment.

Saving is simply income that is not consumed. In terms of share of the economy, saving equals one minus consumption (s = 1 - c). The rise in consumption’s share of the economy, therefore, is exactly the same as the fall in saving’s share. This is shown in Figure 4, which plots each as a share of GNDI, with trends of 0.43 percentage points a year in opposite directions.

Related posts

On consumption

1. Consumption is NOT to Blame for China’s Rising Surplus — consumption’s share of the economy has been rising since 2010, so it is not what is driving the surplus.

2. The Paradox of the Chinese Consumer — consumption has grown faster in China than anywhere else, yet its share of the economy fell.

3. How Low Is China’s Household Consumption Share? — very low, and the ranking holds whichever year, denominator, or sample you pick.

4. Why Do Consumers Still Feel So Bad? — weak consumer confidence is not a Chinese phenomenon, it is a global one.

5. What Is Worrying China’s Consumers? — the labor market, mostly.

On global imbalances

6. China’s Current Account Roller Coaster — the surplus up, down, and up again since 2000.

7. Global Imbalances: One Big Deficit, Many Big Surpluses — one large deficit economy, three surplus blocs.

Data

Construction. Unless noted otherwise, the national accounts and balance of payments series are taken from CEIC, accessed in July 2026. Investment is gross capital formation, the published expenditure component, which covers fixed investment plus the change in inventories. Saving is gross national saving, measured as national income less total consumption, both household and government. The income components used to build national income come from two places. From 2010 they are the renminbi balance of payments in CEIC. Before 2010 that renminbi series does not exist, so they come from the IMF’s Balance of Payments database in US dollars, converted to renminbi at the implied exchange rate from the IMF World Economic Outlook. Where the two overlap they agree closely, so the join is not visible in any of the charts.

GNDI. As I argued last week, GNDI is a better denominator for these charts. Gross national disposable income is expenditure GDP plus net income from abroad, meaning net primary income plus net secondary income. It is the income actually available to residents to spend or save.

• Figure 1. Gross national saving and gross capital formation as a percent of GNDI, 2000 to 2025, with ordinary least squares trends fitted from 2010.

• Figure 2. The same two series, with the area between them shaded.

• Figure 3. The left panel shows the current account balance as a percent of GNDI in 2010 and 2025. The right panel’s first bar shows the change in the surplus over that period, which rounds to a decline of 0.1 percent of GNDI. The next two bars show how much consumption and investment contributed to this change. What may be confusing is that the bars are scaled so that a positive value increases the surplus and a negative value decreases it. Consumption is −7.4 points, which is the size of the rise in consumption’s share of GNDI between 2010 and 2025 and, equivalently, how much that rise contributed to decreasing the surplus. Likewise, investment is +7.3 because that is the fall in investment over the period and how much it contributed to increasing the surplus.

• Figure 4. Consumption and saving as a percent of GNDI, fitted the same way as Figure 1. The two lines cross in 2008, at almost exactly 50% each.

The analysis, charts, and drafting in this post are mine. I use Claude as a copy editor, sounding board, and research assistant.

Subscribe now

Subscribe now

Why it matters

While low, consumption’s share of the economy is trending up. True, consumption today is a much lower share than it was 25 years ago: it is some 7 percentage points of GNDI below where it was in 2000. But the decline was virtually over by 2008 and coincided with the rapid rise in China’s current account surplus. Since then, it has been on a clear upward trend. It remains low relative to comparators and could certainly rise faster. But given the pandemic and the large real estate correction, consumption is proving remarkably resilient.

Consumption, therefore, is not behind the recent rebound in the current account surplus. Recall that the current account surplus rose from 0.2 percent of GNDI in 2018 to 3.7 percent in 2025. By accounting identity, this rise has to be mirrored by an equivalent drop in the sum of consumption and investment. Only, consumption didn’t fall. Thus, as I will show next week, investment declining as a share of the economy more than covers the full increase in the current account surplus since 2018.

Household, not government, consumption is the key factor behind the upward trend. This is shown in Figure 2 below: roughly three-quarters of the trend rise in consumption since 2010 is explained by rising household consumption. Looking just at 2025 versus 2018, government consumption didn’t change, and rising household consumption fully explains the rise in consumption.

Details

Household consumption is the key driver of changes in total consumption. Figure 2 shows the total, household and government shares, each with its trend. Government consumption has moved within a narrow band, between 14.6 and 17.6 percent of GNDI over 2000 to 2025, and has sat close to 17 percent for most of the past decade. Household consumption, meanwhile, fell from 2000 to 2010 and has been on a gradual rise since. It has even risen 1 percentage point since 2018.

Recent research highlights how much the real estate correction weighs on consumption. Rogoff and Yang estimate that the decline in housing prices, on the order of 20 to 40 percent, implies a consumption loss of 2 to 4 percent of GDP (Rogoff and Yang 2026; Stanford Center on China’s Economy and Institutions 2026). The intuition is that real estate is the main asset Chinese households own, accounting for nearly 70 percent of household wealth, so changes in house prices have a significant impact on consumption through wealth effects. Separate work from staff at the Bank for International Settlements provides further evidence of the link between housing prices and consumption. Using household transaction data, they find that a 10 percent rise in house prices raises consumption by about 1.6 percent across 33 Tier 1 and Tier 2 cities, though there is no such effect in smaller cities (Mojon et al. 2025). Thus, it is striking that consumption’s share of the economy rose, rather than fell, during the real estate correction.

China’s current account surplus resembles a roller coaster. Figure 3 shows that it rose sharply following WTO accession and peaked at nearly 10 percent of GNDI in 2007. It subsequently fell just as sharply, to nearly zero in 2018, and has been rising again since. I covered this in two earlier posts, on global imbalances and on China’s own current account.

Using GDP instead of GNDI does not change the story. This is shown in Figure 4. As I have argued before, GNDI is the better denominator for these charts. It captures all of the income available to be used for consumption and investment. In most economies, including China, GDP and GNDI track each other closely, so the choice does not matter much. But since these posts ultimately link developments to the current account, which itself includes net income flows from abroad, it makes sense to scale consumption and investment by GNDI too.

Related posts

On consumption

1. The Paradox of the Chinese Consumer — consumption has grown faster in China than anywhere else, yet its share of the economy fell.

2. How Low Is China’s Household Consumption Share? — very low, and the ranking holds whichever year, denominator, or sample you pick.

3. Why Do Consumers Still Feel So Bad? — weak consumer confidence is not a Chinese phenomenon, it is a global one.

4. What Is Worrying China’s Consumers? — the labor market, mostly.

On global imbalances

5. China’s Current Account Roller Coaster — the surplus up, down, and up again since 2000.

6. Global Imbalances: One Big Deficit, Many Big Surpluses — one large deficit economy, three surplus blocs.

Data

Construction. Unless noted otherwise, the national accounts and balance of payments series are taken from CEIC, accessed in July 2026. Total consumption is published along with its split into household and government. The income components used below come from two places. From 2010 they are the renminbi balance of payments in CEIC. Before 2010 that renminbi series does not exist, so they come from the IMF’s Balance of Payments database in US dollars, converted to renminbi at the implied exchange rate from the IMF World Economic Outlook. Where the two overlap they agree closely, so the join is not visible in any of the charts.

Which GDP, and where the statistical discrepancy goes. GNDI is constructed using expenditure GDP. China publishes two nominal GDP measures, one built up from production and one from expenditure, and they differ by a statistical discrepancy. Either can be used to build GNDI, and there is a case for each. Production GDP is the headline measure, is published earlier than the expenditure levels, and is the one usually treated as better measured, so production GDP plus net income gives the more timely and arguably more accurate estimate of national income. That is the version to reach for when all you need is a denominator. But this post rests on an adding-up argument, that consumption plus investment plus the current account exhausts national income, and that only closes exactly on the expenditure measure. The cost of that choice is small. The gap between the two GDP measures has averaged about a quarter of a percent of GDP since 2000 and exceeded half a percent in only three years, 2010, 2011 and 2020. Since 2023 it has been essentially zero.

GNDI. Gross national disposable income is that expenditure GDP plus net income from abroad, meaning net primary income (mainly investment income and compensation earned abroad, less the same paid to non-residents) plus net secondary income (transfers, such as remittances). It is the income actually available to residents to spend or save, which is why it belongs in the denominator for consumption. Those same two income items are also what separate the current account from the trade balance, so scaling consumption and the current account by GNDI keeps the two on a common basis. For China the choice does not matter that much, because net income from abroad is small: in 2025 total consumption was 57.2% of GNDI and 56.9% of GDP, which is what Figure 4 shows.

• Figure 1. Total consumption as a percent of GNDI, 2000 to 2025, with an ordinary least squares trend fitted from 2010. The share was 49.8% in 2010 and 57.2% in 2025, a gain of about 7.4 percentage points, or roughly 0.43 points a year on the fitted trend. The 2000 value (64.4%) and the 2025 value are marked, along with 2018 (56.2%), the year the current account surplus bottomed.

• Figure 2. The same total split into its two components, household and government consumption, each with its own trend fitted from 2010. Household consumption rose from 34.8% of GNDI in 2010 to 40.2% in 2025, a trend of about 0.31 points a year; government consumption rose from 14.9% to 17.0%, about 0.12 points a year. Households therefore account for roughly three-quarters of the increase in the total. The two are the published components of the total, so they sum to it exactly. Values are marked for 2000, 2018 and 2025.

• Figure 3. The current account balance as a percent of GNDI, 2000 to 2025. The balance is net exports plus net primary and secondary income, the same items that turn GDP into GNDI, so the ratio sits on exactly the denominator used for consumption in the other figures. The surplus peaked at 9.7% of GNDI in 2007, fell to 0.2% in 2018, and has risen since to 3.7% in 2025. Only those three years are labeled, to keep the chart readable.

• Figure 4. Total consumption measured against GNDI and against GDP, on the same axes. The two lines are close enough to be hard to tell apart: the largest gap between them over 2000 to 2025 is 0.8 percentage points, and in 2025 the shares are 57.2% and 56.9%. This is the check behind the point made above, that the choice of denominator does not carry the result.

References

Mojon, Benoit, Han Qiu, Fang Wang, and Michael Weber. 2025. Housing Wealth Effects in China. December 2025. https://www.bis.org/publ/work1319.htm.

Rogoff, Kenneth S., and Yuanchen Yang. 2026. A Tale of Two Countries: The Real Estate Crises in 1990s Japan and Contemporary China. Working Paper No. 35054. Cambridge, MA: National Bureau of Economic Research. https://doi.org/10.3386/w35054.

Stanford Center on China’s Economy and Institutions. 2026. China’s Property Slump Can Curb Growth for Years, Even Without a Banking Crisis. SCCEI China Briefs. https://sccei.fsi.stanford.edu/china-briefs/chinas-property-slump-can-curb-growth-years-even-without-banking-crisis.

The analysis, charts, and drafting in this post are mine. I use Claude as a copy editor, sounding board, and research assistant.

Subscribe now

Subscribe now

Why it matters

Outbound migrants are a sizable part of China’s workforce. They are equivalent to roughly a quarter of all employment in China, and, as Figure 2 below shows, to more than a fifth of the working-age population.

The data arrive more often than other employment indicators. Total employment, urban employment, and the other categories of migrant worker are published only once a year. The outbound series is quarterly and thus provides a more timely indicator of developments.

Their numbers suggest that the lower-wage end of the labor market remains soft. Outbound migrants are generally lower-skill, lower-paid workers. Thus, the series is most informative about that part of the market.

Details

What are outbound migrant workers? They are workers holding rural household registration (hukou) who are employed outside their home township for six months or more of the year. The test is geographic rather than urban, so they could be working in either urban or rural areas. Migrants who do non-farm work within their home township are counted instead as “local” migrant workers. Both the total and the local counts are published only annually, whereas the outbound series has the advantage of being quarterly. For perspective, outbound migrants made up about 60% of all migrant workers in 2019, and essentially the same share, just under 60%, at the end of 2025.

The shortfall is sensitive to how the trend is drawn. The linear trend in Figure 1 passes an eyeball test and has the appeal of simplicity: it assumes that outbound migrants increase by a constant amount each year: the trend implies about 2 million. But the population they are drawn from, China’s working-age adults, has been shrinking by about 7 million a year recently. An alternative is to scale by the working-age population, much as Part 3 did for urban employment. Since that population is shrinking, this counterfactual allows the number of new outbound migrants it implies each year to decline, rather than holding at the constant amount the linear trend assumes. Figure 2 does this, and is analogous to Figure 1 in the missing urban jobs post. The bottom line is that the shortfall of outbound migrants on this counterfactual is much smaller: 2.5 million workers, instead of the 6 million using the methodology in Figure 1.

Figure 3 compares the two methodologies. The results of Figure 2 are converted from a share of the working-age population into the implied number of outbound migrants, which allows direct comparison with Figure 1. The linear methodology of Figure 1 suggests a counterfactual of 194 million outbound migrants, whereas the share-of-working-age methodology of Figure 2 suggests 191 million. The actual was 188 million.

There is no obviously right counterfactual. Both the linear trend and the share-of-working-age trend fit the historical data well and have some intuitive appeal. Both also lack a theoretical basis. The idea, though, is not to write a model of outbound migrants, but rather to see how the market is doing relative to a reasonable counterfactual. The linear trend is the simplest. And the two counterfactuals overlap for much of the pre-pandemic period in any event.

The implicit assumption is that labor demand is the determinant of outbound migrants. As long as there is an available pool of outbound migrant workers, the equilibrium quantity should be largely driven by changes in demand. Wages offer some evidence for this: if weak demand is behind the shortfall, then that should map into lower wage growth.

Figure 4 suggests that outbound migrant income growth has indeed been sluggish. It compares household wage income with outbound migrant wage income, both available quarterly, against a common 5% annual benchmark. Income is the wage rate times hours worked, which for these purposes is an advantage: weak demand shows up in fewer hours as well as in lower pay. Outbound migrant income has been growing much slower than other wage income, quite consistent with sluggish demand for outbound migrant labor. One caveat on the benchmark: the 5% line is an assumed growth rate, anchored at the 2019 average, not a trend fitted to the data the way those in Figures 1 through 3 are.

Labor market series

This is the sixth post in my series on China’s labor market.

1. China’s Stable Unemployment Rate — the surveyed unemployment rate is too stable to say much about the labor market.

2. China’s Population Is Aging and Shrinking — the demographic backdrop: fewer workers supporting more retirees.

3. China’s Missing Urban Jobs — urban employment is running some 37 million below its pre-pandemic trend.

4. China’s Urban Employment Is Weak — But What About the Total? — whether adding rural employment makes the picture look any better.

5. How High Is China’s Youth Unemployment? — much higher than the overall rate, but no global outlier.

6. This post — outbound migrant workers, a quarterly indicator of employment conditions.

Data

Construction. The migrant series is quarterly outbound rural migrant workers (外出农民工) from China’s National Bureau of Statistics (NBS), accessed via CEIC on 16 July 2026 (who is counted is explained in the Details). The series is not published seasonally adjusted, so I adjust it myself with the X-13 procedure; the first quarter of 2020 was not surveyed and is shown as a gap. Figure 1 plots the level in millions; Figure 2 expresses it as a percent of the working-age population; Figure 3 puts the same counterfactual back in millions. In each, the pre-pandemic trend is a linear fit on 2016Q1–2019Q4, extended to the latest quarter. The working-age denominator comes from the UN World Population Prospects 2024, downloaded directly from the UN, and uses the 16–59 band, consistent with China’s own working-age definition and with the urban-employment post (Part 3). Throughout, “below trend” means outbound migration relative to the extended pre-pandemic trend, not a fall in the level, since outbound migration rose over the period.

Wages (Figure 4). Migrant pay is the NBS quarterly series on outbound migrants’ average monthly income, from the household survey, deflated by the consumer price index and seasonally adjusted. One wrinkle matters: the NBS quarterly release reports this series for the first three quarters only, so I interpolate each fourth quarter from the quarters either side of it, within the same survey. The economy-wide comparison is the household survey’s per-capita wage income, deflated and seasonally adjusted the same way. Both are indexed to the 2019 average (specifically the first three quarters of 2019) rather than to the fourth quarter alone. I avoid a 2019Q4 base for three reasons: the migrant fourth quarter is interpolated rather than surveyed, it sits next to the un-surveyed first quarter of 2020, and in seasonally adjusted terms it is a local low for the migrant series, so anchoring there would flatter the whole post-2019 path. Both series are then compared against an assumed 5% annual real growth path anchored so that it, too, averages 100 over the first three quarters of 2019. That 5% line is a benchmark I use across this series, not a trend fitted to the data, unlike the trends in Figures 1 through 3, which are estimated on 2016Q1–2019Q4. The comparison between the two series does not depend on the level of the benchmark.

• Figure 1. Seasonally adjusted outbound migrants in millions, against a linear trend fit on 2016Q1–2019Q4 and extended to the latest quarter. At 2026Q2 the level is 188 million against a trend of 194 million, a gap of roughly 6 million workers, or 3.1% below the trend.

• Figure 2. Outbound migrants as a percent of the working-age (16–59) population, seasonally adjusted, against the extended pre-pandemic trend. At 2026Q2 the figure is 21.7% against a trend of 21.9%, a shortfall of 2.5 million workers, or 1.3% below trend.

• Figure 3. Seasonally adjusted outbound migrants in millions, with the naive linear trend and the working-age counterfactual (the pre-pandemic rate re-imposed on the realized 16–59 population). At 2026Q2 the naive trend implies 194 million and the working-age counterfactual 191 million, against an actual 188 million; the residual against the working-age line is the same 2.5 million.

• Figure 4. Real, seasonally adjusted wage indices (2019 first-three-quarters average = 100) against an assumed 5% annual path on the same base. From that 2019 base to 2026Q2, economy-wide real wage income is up a cumulative 40%, an annualized 4.9% that essentially matches the 5% benchmark. Outbound migrant pay is up just 22%, about 3% a year.

The analysis, charts, and drafting in this post are mine. I use Claude as a copy editor, sounding board, and research assistant.

Subscribe now

Subscribe now

Why it matters

GDP is the most watched summary of economic performance, but it can be hard to interpret. Year-on-year growth gets the most attention in China, but it is influenced both by the base quarter — a weak year-earlier quarter flatters this year’s growth, a strong one drags it down — and by developments over the intervening year. The seasonally adjusted annualized rate (SAAR) gives a better picture of current momentum, how fast the economy is growing right now, though it can be noisy and subject to revision. That is why my preferred indicator for China is the level of GDP, shown in Figure 1 — my go-to chart.

Q2 growth, excluding the pandemic quarters, is one of the lowest in decades. In the post-2000 sample, at 4.3 percent, the year-on-year print is the lowest non-pandemic growth rate. The SAAR of 3.6 percent tells a similar story: it is only the third time since these data were first published that the rate has dipped below 4 percent, again setting the pandemic aside.

Overall, however, GDP is broadly in line with the 5 percent trend. That the level stays close to trend is, in my view, the key point, and why Figure 1 is my go-to chart: it looks through the quarterly noise and short-term fluctuations. Moreover, as shown below, the gap to the trend line is growing but not out of line with earlier quarters (Figure 2).

Details

A 5 percent trend is a reasonable, if not ambitious, anchor for the eye. China’s official growth target was around 5 percent in 2023, 2024, and 2025; for 2026 it was set as a range of 4.5 to 5.0 percent, while striving for better in practice. A trend matters more for levels than for growth rates, because it supplies an intuitive counterfactual — a path against which to judge where output actually is. The ambition lies in the assumptions: 5 percent presumes no lasting scarring from the pandemic, and has the economy simply growing at 5 percent a year from the end of 2019 — even as China’s trend rate of growth has been slowing, a tension I return to below.

Still, the economy recovered quite quickly to the 5 percent trend. That is especially notable given that China was contending not only with the pandemic but also with a large real estate correction, which I discussed in China’s Real Estate Correction. Figure 2 plots the gap, in percent, between the level of GDP and the 5 percent trend. It shows how quickly output recovered and, for a time, even ran above trend. The gap then held near half a percent below trend for a couple of years. It has widened this year, reaching about 0.8 percent below trend in the second quarter — the widest in over three years, though still no larger than the gaps China ran in 2022, and far shallower than the pandemic plunge.

The SAAR data are bouncy, underscoring the value of not over-interpreting any single print. Figure 3 shows quarterly SAAR growth going back to when these data were first published. One quirk is worth keeping in mind: the NBS reports the underlying quarter-on-quarter growth to a single decimal place, so the annualized values come in discrete steps. A 0.9 percent quarterly gain annualizes to 3.6 percent and a 1.0 percent gain to 4.1 percent, with no readings possible in between. Excluding the pandemic quarters, the only other times growth dipped below 4 percent were 2019Q4 and 2023Q4. The chart is also truncated; showing the pandemic outliers in full would flatten everything else.

Figure 3 also makes clear that the trend rate of growth has slowed. For simplicity I proxy that trend with a five-year rolling median, which is robust to the pandemic outliers; a rolling average, by contrast, would lurch up and down as those extreme quarters moved into and out of the rolling window. On this measure, the trend has slipped from about 7.8 percent in the mid-2010s to 4.7 percent now — a fall of roughly 3 percentage points.

The year-on-year growth rate tells a similar, and even cleaner, story. Figure 4 plots year-on-year growth against its five-year rolling median, back to 2000. Two things stand out. First, as noted above, the Q2 reading is the lowest non-pandemic print in the sample. Second, the median makes the trend slowdown in growth unmistakable — it has more than halved, from about 12 percent around 2010 to 4.8 percent now. Note, too, that this chart, unlike Figure 3, is not truncated: the pandemic swings — a drop of nearly 7 percent and a rebound of almost 19 percent — stretch the vertical axis, which if anything makes the trend slowdown look milder than it is.

This long-run slowdown is natural and expected. As I argued in a 2013 IMF blog post, the world should welcome a slowdown in China’s growth to a more sustainable pace. It does, however, raise a question about whether a constant 5 percent line is too ambitious an anchor going forward — trend growth has already slipped below it, and is still easing. But that is a topic for a future post on the sources of China’s growth — including the role of slowing total factor productivity.

Data

Construction. All figures use China’s quarterly real GDP as reported by the National Bureau of Statistics, retrieved through CEIC (data as of 2026-07-17). Two official series underlie them: the headline year-on-year growth rate, and the NBS seasonally adjusted quarter-on-quarter growth rate. NBS publishes the seasonally adjusted quarterly rate to one decimal place and only from 2010Q4 onward; I annualize it as (1 + g)⁴ − 1, where g is the quarterly growth, to state it at an annual rate (SAAR) — NBS itself publishes the quarterly figure, not the annualized one. The seasonally adjusted level index chains that quarterly series and is normalized to 100 in 2019Q4. The 5% trend is an assumed benchmark — a path growing 5% a year from the same 2019Q4 anchor, not a statistically fitted trend. The gap from trend is the percentage difference between the index and that path, 100 × (index / trend − 1). The 5-year trailing median in Figures 3 and 4 is the median of the preceding twenty quarters; unlike a trailing average it steps over the pandemic spikes rather than being distorted by them.

• Figure 1. Seasonally adjusted real GDP index (2019Q4 = 100) against the 5% benchmark path; the inset table shows the seasonally adjusted annualized growth rate of the last four quarters.

• Figure 2. Percentage gap between the index and the 5% path. The vertical axis is truncated to ±1.5%; the two off-panel quarters (2020Q1, −11.6%; 2022Q2, −2.2%) are labeled, with a torn-edge mark on the truncated bars.

• Figure 3. Seasonally adjusted annualized quarterly growth from 2011, with its 5-year trailing median. The axis is truncated to −5 to 15, so the two pandemic quarters (2020Q1, −35.8%; 2020Q2, +61.9%) run off-panel; their values are given in the caption. The shaded band marks the pandemic quarters (2020Q1–2023Q1).

• Figure 4. Year-on-year real GDP growth since 2000, with its 5-year trailing median. The shaded band marks the pandemic quarters (2020Q1–2023Q1).

  Subscribe now

For the last several years, youth unemployment has often run more than 10 percentage points above the overall urban rate — which, as flagged in Part 1, has itself stayed low and steady, near 5 percent. Figure 2 puts China’s youth unemployment in global perspective, showing it to be far from an outlier.

Subscribe now

Why it matters

Youth unemployment is watched closely, and for good reason. It is one of the few monthly readings we get on the labor market, and it has risen well above its pre-pandemic level. The original series reached 21.3 percent in June 2023, its last value before suspension — about 11 percentage points higher than the 10 percent it registered in June 2018. It points to real stress in an important part of the workforce. The number is also harder to read than it looks, due to strong seasonality and a mid-series break, both taken up below.

The demographics, moreover, are such that the youth cohort is growing in the near term. It is rising in absolute numbers and — because the overall working-age population is shrinking — climbing as a share of that population too. High unemployment, in other words, has settled on precisely the cohort set to expand over the next several years (Figure 3, in the details below).

Youth unemployment is an issue in many economies around the world. On the harmonized definitions in Figure 2, several economies have higher youth unemployment than China. Moreover, China’s roughly 10-percentage-point gap between overall and youth unemployment is not a large outlier either (Figure 4, in the details below). None of this makes it less painful for those out of work, but it is important context.

Details

This is the fifth post in my series on China’s labor market. Part 1 argued the unemployment rate is too stable to be informative; Part 2 set out the demographic backdrop; Part 3 showed urban employment well below its pre-pandemic trend; and Part 4 asked whether the total employment picture looked any better. This post turns to the youth — a soft corner of the labor market.

The youth unemployment numbers are hard to read for two reasons. The first is seasonality: youth unemployment jumps every summer, as a fresh cohort of graduates reaches the market before being absorbed over the following months, so any single reading depends heavily on the time of year. Moreover, the series are too short to seasonally adjust reliably. Related, and second, is the 2023 break in the series. The old series counted job-seeking students among the unemployed; the new one does not. The two are therefore not comparable, and the series cannot be read straight across the break — the lower readings since 2024 partly reflect the change in definition, not a genuine improvement.

The youth cohort is rising in the near term. The population aged 15–24 bottomed out around 2022, after falling through the 2010s, and is set to rise by some 15 million before peaking in the early 2030s (Figure 3, top). Because the working-age population is contracting at the same time, the youth share of it climbs even faster, cresting around 2033 (Figure 3, bottom). It is a near-term bulge within the longer demographic decline traced in Part 2 — a larger group of young people arriving at a weak entry-level market.

The cross-country figures come from the International Labour Organization (ILO), which harmonizes national data so economies can be compared on a consistent basis. The ILO’s China figure is a modeled estimate built on the country’s urban survey data and averaged over the year. Thus, it does not exactly match the NBS data, though the two are fairly close: for 2024, the annual average of the NBS figure is 15.8 percent versus 15.0 percent for the ILO’s model-based one. On the basis of the ILO data, China’s youth unemployment sits well above its overall rate — but the gap is far from the widest in the world. Among comparable economies, China’s gap ranks around seventh, behind several European and emerging markets (Figure 4). And across the sixty largest economies, youth unemployment runs above the overall rate almost everywhere; China lands near the middle of the field on both (Figure 5).

There is a more standard way to sidestep the student problem altogether: NEET — the share of young people not in employment, education, or training. Because it rests on a population denominator rather than a labor force, it does not turn on whether job-seeking students are counted — the very issue behind China’s 2023 redefinition. China, however, neither publishes the data needed to construct one nor appears in the ILO’s annual modeled series for it. A few scattered estimates exist — from occasional surveys, and on broader age bands — but there is no regular, comparable figure.

Data

Construction. Figure 1 uses China’s surveyed urban unemployment rate — overall and for 16–24 year-olds — from the National Bureau of Statistics, accessed through CEIC. The youth series breaks in 2023: the Bureau suspended it after the June 2023 reading of 21.3 percent and resumed it in January 2024 on a revised basis that measures unemployment only among 16–24 year-olds who are not enrolled students — students are removed from both the unemployed count and the labor-force base, so the rate reflects the non-student youth labor market rather than the whole cohort. The two segments are built differently and are not comparable, and are drawn as separate lines rather than spliced. Figures 2, 4, and 5 use the World Bank’s World Development Indicators, which republish the International Labour Organization’s modeled estimates of youth (15–24) and total unemployment for 2024 — figures harmonized so that economies are measured on a consistent basis. The ILO figure for China is a modeled estimate built on the country’s urban survey data. It sits well below the 21 percent that made news in 2023, which came from the older, student-inclusive definition at its monthly peak. Figure 3 draws on the United Nations World Population Prospects 2024.

• Figure 1. China’s surveyed urban unemployment rate: overall (all ages) and youth (16–24), with the pre-2023 and post-2024 (student-excluding) youth segments shown separately.

• Figure 2. Youth unemployment rate for selected economies (ILO-modeled), China highlighted; South Africa (youth ~60 percent) omitted as an off-scale outlier. The selected set is the 19 G20 members plus Malaysia, the Philippines, Thailand, Singapore, Vietnam, Hong Kong SAR, and Spain, Greece, and the Netherlands.

• Figure 3. China’s population aged 15–24 (level, in millions) and its share of the 16–59 working-age population, 2010–2040 (2025 onward are UN projections).

• Figure 4. The youth-minus-overall unemployment gap for the same economies, sorted by the size of the gap.

• Figure 5. Youth against overall unemployment for the 60 largest economies by GDP (ILO-modeled), with a 45-degree line for reference.

  Subscribe now

Subscribe now

At market exchange rates (MER), though, advanced economies are still bigger. The emerging world has still gained ground — its share has risen from about a quarter of world output in 1980 to 43% by 2031 — but it remains the smaller half, with the advanced economies still near 58% (Figure 2). The gap between the two measures is itself the story: PPP and MER differ considerably on how much the emerging world has risen.

Why It Matters

• Because size matters. Economic size and economic importance go together: a larger economy, all else equal, exerts more influence on the global macro numbers. The bulk of global growth now comes from the emerging world — and from China and India above all (the subject of an earlier post on the two engines of global growth). That shift also raises a question of global governance: has the voice and representation of emerging economies in international organizations kept pace with their rising economic weight?

• But measurement matters too. Ideally the relative size of the two groups wouldn’t hinge on which exchange rate you use — but increasingly it does. In 1980 it hardly mattered: PPP put the advanced economies at 65% of world output and market rates at 76%, a gap of about 11 percentage points, and either way the story was the same — the advanced economies were much bigger. By 2031 the choice is decisive: the two measures differ by more than 20 points and no longer even agree on which group is larger. PPP makes the emerging world the bigger half; market rates keep the advanced economies ahead.

• Which measure is right depends on the question. PPP is the better gauge of real output and living standards — of how much an economy actually produces and consumes. Market exchange rates are the better gauge of financial and commercial weight — the size of an economy in globally traded dollars, what matters for trade balances (exporters, for example, care about size in US$) and currency markets. Neither is “right.” They answer different questions.

Details

Why do the two pictures differ? Price levels. Goods and especially services that are not traded across borders — housing, haircuts, restaurant meals — tend to be cheaper in poorer economies, so a unit of output there converts into fewer dollars at market exchange rates than it is actually worth in real terms. PPP tries to correct for this; market exchange rates do not. Of course, distilling all of an economy’s relative prices into a single number is, in essence, an intractable problem — PPP is a reasonable attempt at it.

The MER-versus-PPP gap matters more for some economies than others. Figure 3 shows the difference for the largest economies — a twist on the point made in an earlier post. For every large emerging economy — China, India, Russia, Brazil, Indonesia — the PPP share (P) sits above the market-rate share ($): their output is worth more in PPP terms than the dollar figures suggest. For the advanced economies — the United States, the European Union, Japan, Canada, Korea — the reverse holds. In fact the connector color, which simply flags which measure is larger, recovers the advanced-versus-emerging divide on its own: the rich economies all run one way, the emerging economies all the other. The two widest gaps, moreover, belong to the United States — far bigger at market rates than in PPP — and India, far bigger in PPP.

The upshot is that “who is bigger” has no single answer. However, no matter how it is measured, the share of EMDEs has been rising. This reflects the simple fact that they grow faster, but more on this in a future post.

This week is a detour from my labor-market series, which resumes next week with youth unemployment. My motivation for this week’s post was a podcast I did with Interactive Brokers, which included a discussion of PPP versus MER.

Data

Construction. The figures show the advanced-economy and emerging-market-and-developing-economy (EMDE) groups’ shares of world GDP, using the IMF’s analytical country groupings, from the IMF World Economic Outlook (International Monetary Fund. Research Dept. 2026) (April 2026 vintage, accessed 15 April 2026 at data.imf.org). The two groups together sum to 100% of world output by construction. PPP shares are each group’s share of world GDP valued at purchasing-power parity. Market-exchange-rate shares are each group’s US-dollar GDP divided by the world total; following the IMF’s convention, the dollar figures are smoothed using a three-year trailing average to damp year-to-year exchange-rate swings (1980–1982 use the current year, where the lags are unavailable). Values from 2026 onward are IMF projections.

• Figure 1. Advanced-economy and EMDE shares of world GDP, PPP basis, 1980–2031.

• Figure 2. Advanced-economy and EMDE shares of world GDP, market-exchange-rate basis (three-year trailing average), 1980–2031.

• Figure 3. The largest economies in 2025 — those in the top ten by either PPP share or market-rate share, a set of twelve — each shown as a PPP share (P) and a market-rate share ($) joined by a connector. Taking the union of the two rankings avoids privileging one measure. EU member states are folded into the European Union aggregate to avoid double-counting; the United Kingdom, outside the EU, is shown separately. Market-rate shares again use the three-year trailing average. Economies are ordered by PPP share.

References

International Monetary Fund. Research Dept. 2026. World Economic Outlook, April 2026: Global Economy in the Shadow of War. USA: International Monetary Fund. https://doi.org/10.5089/9798229042758.081.

Subscribe now

Why it matters

First, total employment is the broadest measure of the job market. On the surface, a case could be made that it is the better indicator. But as argued above, it does not tell a clean story of what is actually happening — there is no trend to measure against, and a large and growing student body clouds the picture.

Second, urban employment is arguably the better indicator in China’s case. As I argued in the last post, a worker who loses an urban job and cannot find another may return to the countryside — where they still count as employed. Rural employment thus serves as de facto unemployment insurance, cushioning the total even when urban jobs are scarce. But it is small comfort: urban wages are far higher, so a worker who falls back on rural work usually takes a steep pay cut (Figure 2).

Third, the headcount says nothing about underemployment. By the international standard China follows, a person counts as employed if they worked as little as one hour in the survey’s reference week. Someone working far fewer hours than they want is therefore still counted as employed. This matters most in the countryside, where part-time and seasonal work — especially farming — is the norm: a returning migrant who picks up a few hours on the family plot is, rightly, counted among the employed, even if working and earning far less than they did in the city.

Details

This is the fourth post in my series on China’s labor market. Part 1 argued the unemployment rate is too stable to be informative; Part 2 set out the demographic backdrop; Part 3 showed urban employment well below its pre-pandemic trend. This post explains why I prefer urban employment to the total in reading the labor market.

Returning to the countryside likely means a steep drop in income. Rural per-capita wage income is less than a third of urban — about 30% in 2025, though up from 16% in 2000 (Figure 2). Counting all income, not just wages, the rural share is higher, around 43%, because rural households lean more on farming and small business to fill the gap. Either way, a worker who loses a city job and falls back on the village almost certainly takes a large pay cut. The income decline, moreover, would reflect both lower wages and fewer hours worked — the income data cannot separate the two.

Urbanization has been a big driver of growth in China — a subject I will take up in its own post. Figure 3 shows the trend clearly: a declining share of rural workers as people leave the countryside for higher-paying jobs in the cities, mirrored by a rising urban share. Figure 4 adds undergraduate enrollment to the picture, a small but growing group; it is not clear, however, how much of that growth reflects the difficult job market, which can encourage more young people to enroll in college in the hope of a better one later.

Data

Construction. Employment data are annual urban, rural, and total employment (year-end, millions) from China’s National Bureau of Statistics (NBS), accessed via CEIC on 10 June 2026. Higher-education enrollment is from the Ministry of Education, also via CEIC. The working-age denominator is the population aged 16–59 from the UN World Population Prospects 2024, downloaded directly from the UN — China’s own working-age definition, reflecting statutory retirement at 60 for men and 50–55 for women. I use the UN series rather than the NBS annual age structure because the latter carries census-rebasing breaks in 2000, 2010, and 2020. “Employed or in college” (Figure 1) adds enrolled undergraduate and specialized-college students to employment; I leave out postgraduate students, since some are on the job and already counted as employed. This undergraduate-only count is slightly narrower than the one in Part 3, which deliberately added all enrolled students — postgraduates included — as a generous robustness check, to show that urban employment was below trend even if every student were counted as working. Here the college adjustment is part of the baseline rather than a stress test, so I exclude postgraduates to avoid double-counting those already employed and to keep the trend consistent. The difference is about half a percentage point and does not change the picture. Figure 2 uses per-capita income from the NBS household survey: rural income as a percentage of urban, both for wage income alone and for total disposable income (the sum of wage, business, property, and transfer income). Figure 4 stacks urban employment, rural employment, and undergraduate (plus specialized-college) enrollment, each as a share of the 16–59 population. No trend line is fitted anywhere in this post — as the headline showed, the total has no stable trend to extrapolate.

• Figure 1. Total employment, and total employment plus higher-education (undergraduate and specialized) enrollment, each as a share of the UN working-age (16–59) population, 2000–2025. Employment fell from 84.1% in 2019 to 82.8% in 2025, against a 2010–2019 average of 84.0%; counting students, the share was 87.4% in 2019 and 87.3% in 2025, a touch above its 2010–2019 average of 86.9%.

• Figure 2. Rural per-capita income as a share of urban, 2000–2025, for wage income (about 30% in 2025) and for total disposable income (about 43%). The wage gap is wider because rural households draw more of their income from farming and small business. The measure is per capita across all residents, so it reflects both lower rural wages and the smaller share of rural residents in wage work — it is not a like-for-like comparison of pay for the same job. On a median rather than mean basis the rural-to-urban income ratio is a little lower — about 40% in 2025 against 43% on the mean — but the trend is identical; the median series is only available from 2014.

• Figure 3. Total, urban, and rural employment as shares of the UN working-age (16–59) population, 2000–2025.

• Figure 4. Urban employment, rural employment, and undergraduate (plus specialized-college) enrollment, each as a share of the UN working-age (16–59) population, stacked, 2000–2025.

  Subscribe now

Subscribe now

Why it matters

First, the labor market is weaker than it looks. Whereas unemployment data look benign, this urban job shortfall tells a different story. Urban employment, moreover, is the right metric — better than total employment (which I will cover in a future post). The logic: urban jobs pay much more, and rural work, especially farming, often serves as the fallback.

Second, it explains weak consumer sentiment. As highlighted in a previous post on China’s consumer sentiment, the main cause of low sentiment is concern over the labor market. This indicator justifies why consumers may be concerned: the urban job market is indeed weak.

Third, generating urban jobs is a key economic challenge. Despite healthy GDP growth, employment has fallen short of its pre-pandemic path. How much of the shortfall reflects automation is hard to say: factory automation has long displaced workers, and newer technology may be adding to the pressure. Baidu’s Apollo Go robotaxis, for instance, charge a fraction of what a human cabbie does; in Wuhan, drivers losing business to them petitioned the city in 2024 to rein the service in. And China is hardly alone: across the world, economies are grappling with how to generate enough jobs against a backdrop of rapid technological progress — a challenge AI only sharpens.

Details

This is the third post in my series on China’s labor market. Part 1 argued the unemployment rate is too stable to be informative; Part 2 set out the demographic context — a population that is aging and shrinking. This post turns to my preferred indicator of labor-market health: urban employment relative to its pre-pandemic trend. The rest of this section shows the same shortfall in the raw hiring numbers, explains how the trend is drawn and why it is reasonable, and checks that it holds under other ways of counting the working-age population.

Since the pandemic, the economy has been adding urban jobs at a much slower pace. Urban employment grew by ~12 million jobs a year in 2010–19. Since then: +5 mn (2021), –8 mn (2022), +11 mn (2023), +3 mn (2024), and +2 mn (2025) (see Figure 2). Thus, the finding of missing urban jobs is not being driven by the specification of the trend or the denominator — we see it in the raw employment numbers too.

The trend is not cherry-picked. The dashed line simply extends the 2015–2019 path. The longer view in Figure 3 shows urban employment climbed steadily for two decades on this measure, so projecting the pre-pandemic pace forward is reasonable — and the smooth, two-decade rise is itself a sign that scaling by the working-age population makes sense. The break after 2019 is the anomaly, not the trend.

The shortfall is not an artifact of how we count the working-age population. Figure 1 uses the 16–59 band, consistent with China’s own definition and its low retirement ages — the conservative choice. Figure 4 repeats the analysis on the standard international 15–64 measure: the gap is larger, 44.4 million, or 8.5% below trend. Using China’s own published (NBS) population instead of the UN’s gives 37.5 million, 7.3%. Every reasonable denominator puts the shortfall between roughly 35 and 44 million jobs.

Adjusting for rising college enrollment barely dents the shortfall. Enrollment in undergraduate and graduate programs has climbed from about 6 million in 2000 to roughly 44 million today. Figure 5 repeats the analysis treating every college student as employed — adding enrollment to urban employment. The gap is still 32.5 million, or 3.7% of the working-age population. That assumption is conservative on more than one front: it counts every student as an urban worker, though many are from rural areas, and it double-counts any graduate students who are already employed.

One caveat: this indicator is only available annually. Urban employment is reported just once a year, so it is slow to flag turning points. Timelier, higher-frequency measures — such as the quarterly migrant-worker data I will turn to in a later post — can help fill that gap.

Data

Construction. The employment series is annual urban employment (year-end, millions) from China’s National Bureau of Statistics (NBS), accessed via CEIC on 10 June 2026. The working-age denominator is from the UN World Population Prospects 2024, downloaded directly from the UN. The headline (Figure 1) uses the 16–59 band — China’s own working-age definition, reflecting statutory retirement at 60 for men and 50–55 for women — because urban employment effectively ends at the retirement age, so the denominator covers the same ages as the numerator. Figure 4 uses the standard international 15–64 band for comparison. Figure 5 adds higher-education enrollment (undergraduate, specialized, and postgraduate students) from the Ministry of Education via CEIC to the numerator. I use the UN series rather than the NBS annual age structure because the NBS series carries census-rebasing breaks in 2000, 2010, and 2020 — the apparent 27-million drop in the NBS working-age population between 2019 and 2020 is mostly a statistical splice landing on the COVID year. Reassuringly, the two series differ by no more than 2.3% over 2000–2025; I use the UN simply because it is smooth, whereas the NBS series steps at each census. The pre-pandemic trend in Figures 1, 3, 4, and 5 is a linear fit on 2015–2019, extended through 2025; the missing-jobs figure converts the gap between the extended trend and the actual share back to millions of people using that year’s denominator. Throughout, “missing jobs” means employment relative to the pre-pandemic trend, not a fall in the level (urban employment rose over the period). The two percentages on each chart are “X% below trend” = (trend − actual) ÷ trend, and “X% of the working-age population” = the gap in people ÷ that year’s working-age population.

• Figure 1. Urban employment as a share of the UN working-age (16–59) population, 2015–2025; trend fit 2015–2019, extended through 2025. Over the fit window the share rose about 1.3 percentage points a year — roughly 11 million urban jobs annually; the extended trend reaches 58.5% of the working-age population in 2025 against an actual 54.3%. Gap in 2025: 36.8 million people, 4.2% of the working-age population, 7.2% below trend.

• Figure 2. Year-on-year change in urban employment, millions, 2010–2025.

• Figure 3. The same series and trend as Figure 1, shown from 2000 to place the 2015–2019 trend in long-run context.

• Figure 4. As Figure 1 but on the standard international 15–64 band; gap in 2025: 44.4 million people, 4.5% of the working-age population, 8.5% below trend. For further robustness: on the NBS 15–64 series the gap is 37.5 million (7.3% below trend); against total population, 29.2 million (5.8%); in raw levels, 43.3 million (8.4%).

• Figure 5. Urban employment plus all higher-education enrolled (treating every student as employed), as a share of the UN working-age (16–59) population; trend fit 2015–2019. Gap in 2025: 32.5 million people, 3.7% of the working-age population, 5.9% below trend. Using undergraduate-only enrollment (excluding postgraduates, some of whom are on-the-job and already counted as employed) gives a near-identical 32.3 million.

  Subscribe now

Subscribe now

China’s working-age population is also shrinking. Figure 2 shows it peaked in 2015 at around 1 billion. It holds near that level for a few more years before declining steadily — to roughly 745 million by 2050.

Why it matters

First, the shrinking working-age population is a headwind to growth. In simple terms, fewer workers means less output. In more technical terms, GDP growth is decomposed into the contributions from labor, capital, and total factor productivity — something I’ll take up in a future post. For now the simple story is enough: a declining working-age population means, all else equal, labor contributes less and growth is lower.

Second, aging will strain the social protection system. Spending on pensions and healthcare will rise considerably as China ages. And because China essentially runs a pay-as-you-go system — current benefits paid from taxes on current workers — a rising number of retirees per worker spells financial pressure on an already stretched system (in present-value terms, future benefit payments far exceed future revenue). This is one reason the government has gradually raised retirement ages, and why it is cautious about social security benefit increases or tax cuts.

Third, aging should eventually lower saving and narrow the current account surplus. An aging society would typically be expected to save less, as retirees draw down their savings once work income stops. Less saving means more consumption and, all else equal, a smaller current account surplus (the trade surplus plus net income from abroad). That tendency should hold in the long run, but it is not yet visible in China’s case. The consumption side is also where policymakers see an opportunity: the push to develop the “silver economy” is, in effect, an effort to raise consumption by developing the market for goods and services aimed at older consumers.

Details

Part 1 of this series argued the unemployment rate tells us little about how China’s labor market is doing. Demographics provides the overall context for understanding the labor market. The population’s size and age structure, and especially the working-age cohort, identify the potential supply of workers. This, in turn, is important for interpreting changes in employment numbers, which will be taken up in the next post in this series. The charts below put China’s aging in international context, then turn to the two forces behind it: fewer births and longer lives.

China is aging rapidly and forecast to have one of the highest old-age dependency ratios by 2050. Figure 3 shows that among the G20 and select Asian economies, China will have the fifth highest old-age dependency ratio by 2050 — behind only Korea, Japan, Italy, and Germany. China also has one of the larger increases in the ratio. Korea, Japan, China, Thailand, and Singapore are aging fast and at the upper end of the chart. India, in contrast, is aging much more slowly and forecast to have a much lower dependency ratio in 2050 — less than half of China’s.

China is getting old before it gets rich. Figure 4 shows the old-age dependency ratio (vertical axis) against income (purchasing power parity, on a natural log scale). China is ‘older’ than most other economies at a similar income level: in fact, only two economies with lower income have a higher old-age dependency ratio (Thailand and war-hit Ukraine). Generally, it is the advanced economies that have high old-age dependency ratios.

China’s total population has begun to shrink too. Figure 5 revisits the workforce line and adds total population. Total population is forecast to decline somewhat more slowly in absolute numbers. The decline in both lines is due to the decline in fertility (Figure 6), which has plummeted to around one birth per woman — well below the 2.1 needed to stabilize the population. This decline reflects the one-child policy (1980–2015) and the normal decline in fertility as economies become richer. Moreover, declining fertility is a worldwide phenomenon, in rich and poor economies alike — so income is only part of the story. Population, however, is declining more slowly because China has also made considerable gains in life expectancy. Figure 7 shows the gains over time, with further improvements forecast. Finally, Figure 8 shows that China’s life expectancy is relatively high for its income level.

Data

Construction. Population figures are from the United Nations, World Population Prospects 2024 (medium-variant projections from 2024 onward), downloaded 5 June 2026 from population.un.org/wpp. Age cohorts are summed into the working-age (15–64) and 65-and-over populations; the old-age dependency ratio is the number aged 65 and over per 100 of working age. Life expectancy at birth and the total fertility rate are taken directly from the WPP indicators. Income is GDP per capita at purchasing-power parity, 2025, from the IMF World Economic Outlook, April 2026 vintage, accessed 15 April 2026 at https://data.imf.org/en/datasets/IMF.RES:WEO, matched to each economy. The scatter samples cover all economies with a population above 5 million. Shaded bands mark UN projections (after 2023).

• Figure 1. Old-age dependency ratio — people aged 65+ per 100 aged 15–64 — for China, Japan, Korea, and the United States, 1990–2050.

• Figure 2. China’s working-age (15–64) population, in millions, 1990–2050.

• Figure 3. Old-age dependency ratio in 2025 and 2050, ordered by the 2050 level, for the 19 G20 member countries plus Malaysia, the Philippines, Singapore, Thailand, and Vietnam. China highlighted.

• Figure 4. Old-age dependency ratio (people aged 65+ per 100 aged 15–64) against GDP per capita (PPP), 2025; one dot per economy with a population above 5 million, income on a natural log scale. China highlighted; reference economies labelled.

• Figure 5. China’s total population and working-age (15–64) population, in millions, 1970–2050.

• Figure 6. China’s total fertility rate (births per woman), 1990–2050.

• Figure 7. China’s life expectancy at birth (years), 1990–2050.

• Figure 8. Life expectancy at birth against GDP per capita (PPP), 2025; economies with a population above 5 million, China highlighted.

  Subscribe now

Source: OECD; National Bureau of Statistics of China. See data note.

Subscribe now

Why it matters

First, unemployment is a key indicator of an economy’s health. It usually gets equal play to GDP. For good reason: when people who want to work cannot find jobs, the economy is sick. Too low a rate is a concern too: it could signal future inflation. So unemployment matters both for people and for reading the business cycle. No statistic is perfect, but few are watched as closely. For China, however, we have to look elsewhere: the series does not move enough to be informative.

Second, more data — and a wider range of it — would help us better read the labor market. China offers relatively little here, and not enough to identify stress or turning points in the cycle. Labor force participation — the share of the working-age population in the labor force (workers plus those looking for a job) — would be especially useful, since people entering and leaving the labor force can be as telling a data point as unemployment. Moreover, the US and others publish alternative measures of unemployment that also capture underemployment (people working less than they want to) and discouraged workers (people who left the labor force because they couldn’t find a job).

Third, the stability of China’s unemployment has a structural explanation. It is tempting to argue that China is simply fudging the numbers to paint a rosy picture. But there is a more benign — and more telling — explanation: many workers who lose their jobs go home to the countryside. Thus, technically, they do not count as urban unemployed. They are no longer in the urban area — and they may well be working anyway, just back on the farm.

Details

The US and euro area are not cherry-picked to make China look too stable. Most OECD economies show more volatility in unemployment rates. Figure 2 plots two measures of volatility over the period since 2017. The horizontal axis is the standard deviation and the vertical axis is the range (high minus low in percentage points). China is near the bottom by both measures. Japan and Germany are the other dots near China (see below).

Source: OECD; NBS. See data note.

It is not just a pandemic story — China was already the least volatile before COVID. The sample is small as we only have consistent data for China from 2017. But in this period, China had the smallest variation by both measures (standard deviation and range). Still, Germany and Japan are quite close to China.

Source: OECD; NBS. See data note.

China, Germany, and Japan are stable for different reasons. In Japan, firms tend to ride out slumps by cutting hours and bonuses rather than staff — a feature of its lifetime-employment system — so the unemployment rate barely moves. The government reinforced this during the pandemic through the Employment Adjustment Subsidy and a related job-security subsidy, short-time-work schemes that Japan’s labor ministry credits with holding the jobless rate down by around 2.6 percentage points (Japan Institute for Labour Policy and Training 2022). Germany’s rate stays steady largely because the state subsidizes shorter hours through Kurzarbeit, letting firms hold onto workers through downturns rather than lay them off. Aiyar and Dao (2021) estimate that, absent its pandemic expansion, German unemployment would have risen by an additional 3 percentage points at the trough of the recession. China’s stability, as argued above, reflects that migrant workers return home, so they are no longer in the urban labor force to be counted as unemployed.

Source: OECD; NBS via CEIC. See data note.

China subscribes to the IMF’s Special Data Dissemination Standard (SDDS), the benchmark for timely, comprehensive official statistics. For more information, see the IMF SDDS page and China’s country page. The standard lets each subscriber take two “flexibility options” to report a category less often or with a longer lag — and China has spent both on the labor market, publishing employment and wages only annually rather than quarterly (wages also with a roughly five-month lag). Given the importance of the labor market, more data — and more timely data — would help policymakers and analysts alike.

Data

Construction. OECD economies — monthly harmonised unemployment rate, seasonally adjusted, downloaded 4 June 2026 from the OECD Data Explorer (“Monthly unemployment rates”). China — the surveyed urban unemployment rate (城镇调查失业率), from NBS via CEIC. The published series is unadjusted, so to line it up with the seasonally adjusted OECD data I average January and February (to net out the shifting Lunar New Year), then apply an X-13 adjustment. China’s measure is not identical to the OECD’s: the chief difference is coverage — China’s rate is urban-only, the OECD’s national — which is part of why migrants who return to the countryside drop out of China’s count.

• Figure 1. China vs the US and euro area, all seasonally adjusted; monthly, Jan-2017–Mar-2026 (trimmed to the common last month).

• Figure 2. Volatility scatter since 2017: standard deviation (x) vs the high–low range (y), in percentage points. Each grey dot is an OECD member economy (the OECD, EU, and G7 aggregates are excluded); China, the US, and the euro area are highlighted. China 0.26 / 1.2 pp — second-lowest of 41, just behind Japan (0.20); the US is the far outlier (≈1.8 / 11.4 pp).

• Figure 3. The same scatter for the pre-pandemic window 2017–2019 (the years China’s data cleanly cover). China 0.12 / 0.4 pp — the lowest of 41; the US 0.32 / 1.2.

• Figure 4. China vs Japan and Germany, monthly unemployment rate, all seasonally adjusted; Jan-2017–Mar-2026.

References

Aiyar, Shekhar, and Mai Chi Dao. 2021. “The Effectiveness of Job-Retention Schemes: COVID-19 Evidence From the German States.” IMF Working Papers 2021 (242): 1. https://doi.org/10.5089/9781513596174.001.

Japan Institute for Labour Policy and Training. 2022. “The Impact of COVID-19 on Employment and Labor: MHLW’s White Paper on the Labor Economy 2021.” Japan Labor Issues 6 (37). https://www.jil.go.jp/english/jli/documents/2022/037-01.pdf.

Subscribe now

Source: National Bureau of Statistics of China via CEIC; author’s calculations. Construction. Both series measure newly built residential floor space in square meters — “starts” is floor space on which construction began; “sales” is newly built floor space sold — re-based to 2020 = 100, so each line reads as the share of its 2020 level retained in a given year. See the data note at the end for series definitions.

Subscribe now

Why it matters

First, China’s economy has held up remarkably well. A correction of this size would very likely have triggered a full-blown economic crisis almost anywhere else. The adjustment also coincided with the pandemic — two big shocks in 2020–2022. IMF staff estimate, in Box 5 of the 2024 Article IV report, that the property contraction has been a multi-point drag on growth — around 1.3 percentage points a year over 2026–2030, and more in the immediate downturn, relative to the sector’s boom-era pace (International Monetary Fund. Asia and Pacific Dept 2024). Around 90 percent of urban households own their home, so the downturn hits household wealth directly, not just construction activity. That China has so far avoided a systemic crisis is the genuinely remarkable part.

Second, the bust was policy-induced — and overdue. It reflects a policy choice to rightsize a market that had run hot for years, beginning with the “Three Red Lines” curbs on developer leverage in August 2020. The need was long apparent: the 2015 IMF staff report — full disclosure, I ran the IMF’s China division at the time — was already calling for a multi-year correction (see Figure 3 below). To my mind the government does not get enough credit for finally undertaking an adjustment that was always going to be as painful as it was necessary. Many will quibble with how — myself included: I would have preferred a faster correction, with the state absorbing losses upfront. Whether the losses surface at the banks, local governments, or elsewhere, only the central government’s balance sheet is healthy enough to bear them in the end — so better to recognize them quickly and clear the headwind to growth. But gradualism is more China’s style; it may prove more costly, yet it is also less risky. Either way, when the correction ends — and it will — the economy will be on much sounder footing.

Third, real estate is important for understanding China’s trade surplus. My last two posts looked at trade imbalances — technically, current account imbalances — in the world and in China. Recall that, by accounting identity, China’s current account surplus must equal domestic saving minus investment. So today’s post connects directly to those: the sharp correction in real estate is having a big effect on investment, and perhaps on consumption too. That is a teaser for an upcoming post on how saving (the flip side of consumption) and investment explain China’s rising surplus — and the collapse in housing starts is already a strong hint that investment will be a big factor.

Details

What about prices? Here the picture is more nuanced. Figure 2 shows two measures. New-build prices — the average price of newly sold homes — have been surprisingly stable, down only about 4 percent from their 2020 level (and around 8 percent from their 2023 peak). Resale prices have fallen further: the official second-hand index is down about 15 percent from 2020, and more from its 2021 peak — and it is the better gauge of what an existing home is actually worth (see the data note). Even so, the broad point holds: the adjustment has been far larger in construction volumes than in prices.

Source: National Bureau of Statistics of China via CEIC; author’s calculations. Construction. Both series are re-based to 2020 = 100. New-build is the nationwide average selling price of newly built homes (total residential sales value ÷ floor space sold) — an average realized price that shifts with the mix of what sells. Resale is the NBS 70-city second-hand price index, a quality-controlled measure that holds the type and quality of unit roughly constant rather than averaging over a shifting mix; shown here as the CEIC composite of the city-level series, averaged to an annual figure. See the data note below.

As argued above, IMF staff were already calling for a multi-year real estate adjustment in 2015. Figure 3 reproduces that report’s chart, which showed that a protracted period of below-equilibrium construction would be needed to work through the inventory overhang. By 2020, when the adjustment actually began, the scale of the problem was even bigger.

Figure 3. The IMF’s 2015 housing-adjustment scenario

Source: Reproduced from IMF (2015), People’s Republic of China: 2015 Article IV Consultation — Staff Report, IMF Country Report No. 15/234 (chart accompanying ¶34, “Real Estate: A Multi-Year Adjustment”). (International Monetary Fund 2015)

For further reading, real estate features prominently in every recent IMF staff report, and I’m a big fan of Rogoff and Yang’s work on China — especially Peak China Housing and A Tale of Tier 3 Cities (Rogoff and Yang 2020, 2022).

A note on the data

All three series are annual, cover residential property nationwide, and are published by China’s National Bureau of Statistics (NBS), accessed here through CEIC. The CEIC series, by their published descriptors, are:

• Sales — Floor Space Sold: Residential (商品房销售面积：住宅): the floor area, in square meters, of newly built (“commodity”) residential units sold during the year.

• Starts — Floor Space Started: Residential (房屋新开工面积：住宅): the floor area of residential buildings on which construction was newly begun during the year.

• New-build price — Average Selling Price: Residential (商品住宅平均销售价格), in RMB per square meter.

• Resale price — NBS 70-city second-hand residential price index. NBS publishes the 70 cities individually; the national figure used here is CEIC’s arithmetic-mean aggregate of those city-level series. Monthly, here averaged to an annual figure.

The two volume series (Figure 1) are reported in the same physical unit (in the raw CEIC data, thousands of square meters). In level terms the 2020 figures were roughly 1.40 billion m² sold and 1.64 billion m² started; by 2025 these had fallen to about 733 million m² and 430 million m², respectively.

The two price series in Figure 2 measure different things. The new-build series is a unit value: total residential sales value divided by floor space sold, not a constant-quality index. Because it averages over every transaction, it moves with the geographic and unit-size mix of what sells as well as with underlying prices — if sales shift toward lower-tier cities or smaller units, the average can soften even when like-for-like prices are flat (and vice versa). In level terms it rose from about RMB 10,400/m² in 2020 to a peak near RMB 10,900/m² in 2023, then eased to about RMB 10,000/m² in 2025 — only a few percent below its 2020 level. The resale series is CEIC’s national aggregate of the NBS 70-city second-hand index, a constant-quality measure that better captures what an existing home is worth; it falls more (to about 85, or ~15% below 2020, by 2025).

Subscribe now

References

International Monetary Fund. 2015. “People’s Republic of China: Staff Report for the 2015 Article IV Consultation.” IMF Staff Country Reports 15 (234): 1. https://doi.org/10.5089/9781513512907.002.

International Monetary Fund. Asia and Pacific Dept. 2024. “People’s Republic of China: 2024 Article IV Consultation-Press Release; Staff Report; and Statement by the Executive Director for the People’s Republic of China.” IMF Staff Country Reports 2024 (258). https://doi.org/10.5089/9798400284281.002.

Rogoff, Kenneth, and Yuanchen Yang. 2020. Peak China Housing. No. w27697. Cambridge, MA: National Bureau of Economic Research. https://doi.org/10.3386/w27697.

———. 2022. “A Tale of Tier 3 Cities.” IMF Working Papers 2022 (196): 1. https://doi.org/10.5089/9798400222320.001.

Source: SAFE BOP via CEIC (from 2010), IMF Balance of Payments (2000–2009, USD converted via WEO). Author’s calculations. Construction. Trend lines are linear OLS fits on each window; “pp / yr” is the average annual change in the surplus (as a share of GNDI), in percentage points — so “−0.3 pp / yr” means the surplus fell by 0.3 percentage points of GNDI each year over the window. The dotted vertical line at 2018.5 marks the regime break between the post-GFC decline and the post-2018 rebound.

Subscribe now

China is much bigger now, so as a share of world GDP the surplus is back to its 2007 peak. Relative to China’s own economy the surplus is far smaller than it was in 2007. But because China is now a much bigger share of world GDP, the surplus is roughly the same share of world GDP as it was at the 2007 peak. Figure 2 shows the surplus measured both ways — as a share of China’s own GDP and of world GDP — along with the IMF’s five-year-ahead forecasts, which point to the surplus staying elevated.

Source: IMF World Economic Outlook (Spring 2026). Construction. China’s current account balance (WEO BCA) scaled by China’s own GDP and by world GDP (both WEO NGDPD). Each series is labelled at its peak — 2007 for the own-GDP share, 2008 for the world-GDP share — plus 2025 (latest actual) and 2031 (end of forecast); the 2018 trough is marked but not labelled. The shaded band marks WEO forecast years. As a share of China’s own economy the surplus is far below its 2007 peak (3.7% vs 9.8%); as a share of world GDP it is essentially back to 2007 (0.6% in both years) and just shy of the 2008 peak (0.65%). The world-share peak comes a year later than the own-GDP peak because in 2008 the rise in China’s share of world GDP more than offset the fall in its CA/GDP ratio — so the surplus loomed larger globally even as it began shrinking relative to China’s own economy.

Why This Matters

First, this explains why global imbalances are back in the news. As I argued last week, China’s rising current account surplus coincides with a rising US deficit and, relatedly, with renewed global attention to trade imbalances. As I highlighted, though, the US is the sole large deficit economy, whereas there are three surplus blocs (see also Figure 3 below).

Second, it reflects rising domestic imbalances in China. By accounting identity, a rising current account surplus must be matched by rising domestic imbalance: specifically, a growing gap between domestic saving and investment. Or, put more simply, China is producing more than it is consuming and investing—and the excess is shipped abroad. A future post will dig into whether the culprit is a falling investment share or a rising saving share—the necessary data for 2025 should be published next month. But, as a teaser, the falling investment share, which was both welcome and necessary, plays a much bigger role than many may think.

Third, reducing global imbalances requires action by many global players. Rising consumption in China would reduce saving and lower the trade surplus. Smaller fiscal deficits in the US would increase savings and reduce the US current account deficit. Likewise, policies are also needed in the other surplus economies to reduce the domestic imbalances that underpin the global ones. See, for example, the recent CEPR Paris Report chapter by Weder di Mauro and Zettelmeyer (Weder di Mauro and Zettelmeyer 2026), the G7 Economists Memo (Bai et al. 2026), and the IMF’s recent policy paper Understanding Global Imbalances (International Monetary Fund 2026).

The Details

Figure 3 is an updated version of the global imbalances chart from last week. European surplus is now defined as the EU27 plus Switzerland. What is striking, based on these IMF forecasts, is that by 2031 each of the three surplus blocs will have virtually identical contributions to global surpluses.

Source: IMF World Economic Outlook (Spring 2026). Author’s calculations. Construction. Each bloc’s current account balance as a share of world GDP at 2007, 2019, 2025, and 2031; the US deficit is plotted as positive so all four series sit on the same side of zero. Asia Surplus aggregates seven East Asian economies (Japan, Hong Kong, Korea, Malaysia, Singapore, Taiwan, Thailand). The shaded band marks WEO forecast years.

Trade in goods is the driver of China’s rising current account surplus. Figure 4 breaks the current account into its three components: the goods balance (teal), the services balance (gold), and the income balance (purple = primary plus secondary). The story is straightforward. Goods drove the 2007 peak; its collapse drove the 2008–2018 decline; and its rebound since 2020 is driving the recent climb.

Source: SAFE BOP via CEIC (from 2010), IMF BOP (2000–2009, converted). Construction. Bars stack the three CA components (goods, services, income) as % of production GDP; the black line is the total CA balance (equal to the sum of the bars by construction). “Income” combines primary income (mostly investment earnings) and secondary income (mostly transfers).

The quarterly picture confirms the rebound is recent and persistent. Figure 5 is the same decomposition at quarterly frequency (rolling 4-quarter sums). It zooms in on the post-2010 era and shows the quarterly dynamics: the goods surplus has expanded steadily since 2020, services have stayed roughly flat as a drag, and income has been slowly shrinking in deficit. The CA balance line traces a clean upward path from the 2018 trough.

Source: CEIC (SAFE BPM6 quarterly BOP). Construction. Same decomposition as Figure 4 but at quarterly frequency, using rolling 4-quarter sums to smooth seasonality. Each component (goods, services, income) is shown as a share of rolling 4-quarter GDP. The dataset starts in 2010Q4 because SAFE’s BPM6 quarterly BOP series begin in 2010.

A small technical aside: GDP or GNDI in the denominator? Some readers may prefer GDP as the denominator; for China, the choice barely matters. Gross National Disposable Income (GNDI) is equal to GDP plus net income (primary and secondary) from abroad. As you can see in Figure 4, net income from abroad in China is relatively small, usually around 1 percent of GDP—so GNDI will be only 1 percent smaller than GDP. So the current account surplus looks the same visually whether GDP or GNDI is in the denominator. Since the current account surplus includes income flows, it makes sense to use an income measure (GNDI) that includes them too.

The Data

Annual data, 2000–2025. The current account balance is taken from SAFE’s BPM6 Balance of Payments from 2010 onward (via CEIC), and from the IMF Balance of Payments database for 2000–2009 (USD converted to RMB using the WEO NGDP / NGDPD implied exchange rate). Where the two sources overlap (2010+), they agree on every component of the current account — goods, services, primary income, secondary income — to within 0.01 percentage points of GDP in every year. The splice at 2009/2010 is therefore visually invisible.

Denominator. GDP is NBS production-side nominal GDP, the official headline aggregate used by WEO and the World Bank. GNDI = production GDP + net primary and secondary income from abroad. The series used is constructed by adding the BOP income components to NBS production GDP.

Annual vs quarterly publications. Figure 4 (annual decomposition) draws on SAFE’s annual BOP series; Figure 5 (rolling 4-quarter) draws on SAFE’s quarterly BOP series. These two SAFE publications agree exactly for 2010–2018, but diverge in 2019–2024 by as much as 0.3 percentage points of GDP. The mismatch is entirely on the CA side — NBS’s annual GDP equals the sum of its quarterly GDP to the last decimal in every year, and the comparison is RMB-to-RMB throughout (no exchange-rate conversion is involved). The biggest gaps line up with the pandemic reporting period (2020–2022 peak at +0.3 pp in 2021); 2025 is preliminary and the two sources again match.

References

Bai, Chong-En, Gita Gopinath, Hélène Rey, and Axel Weber. 2026. G7 Economists Memo on Global Imbalances. https://www.g7.utoronto.ca/finance/260330-imbalances.html.

International Monetary Fund. 2026. Understanding Global Imbalances. IMF Policy Paper 2026/006. Washington, DC: International Monetary Fund. https://www.imf.org/-/media/files/publications/pp/2026/english/ppea2026006.pdf.

Weder di Mauro, Beatrice, and Jeromin Zettelmeyer. 2026. “The New Global Imbalances: Why Care, Why Now, and What Can Be Done.” In Paris Report 4: The New Global Imbalances, edited by Hélène Rey, Beatrice Weder di Mauro, and Jeromin Zettelmeyer, 1–34. London: CEPR Press. https://cepr.org/publications/books-and-reports/paris-report-4-new-global-imbalances.

Subscribe now

Source: IMF WEO and author’s calculations. Construction. Current account balances scaled by world GDP, at four time points: 2007 (pre-Global Financial Crisis, peak of the previous imbalance era), 2019 (pre-Covid, after a decade of compression), 2025 (latest actual data), and 2031 (end of WEO forecast horizon). The US line shows the magnitude of its deficit as a positive number, so the comparison with surpluses is visually direct. Asia Surplus aggregates Japan plus six persistent Asian export-manufacturing economies (Hong Kong SAR, Korea, Malaysia, Singapore, Taiwan, and Thailand). EU Surplus aggregates the seven European economies in surplus in essentially every year since 2000 (Germany, Netherlands, Switzerland, Sweden, Denmark, Luxembourg, and Austria). The shaded band marks WEO forecast years.

Subscribe now

The US deficit is in a class of its own. At 0.94% of world GDP in 2025 — and 0.88% in the 2031 forecast — the US deficit is nearly nine times the next-largest single deficit economy, and is forecast to be ten times larger by 2031. The pattern is durable: since the late 1990s, the US has been by far the world’s largest single deficit, typically five to ten times the next-largest. No other economy has come close in either size or persistence.

Surpluses, by contrast, are split across several comparably-sized players. China at 0.49% in the 2031 forecast, the Asia Surplus bloc at 0.48%, the EU Surplus bloc at 0.40% — the latter two together almost exactly mirror the US deficit. China is the most-talked-about. Many may be surprised that in 2019 China’s surplus was actually way smaller than either of the other blocs.

The asymmetry sharpens when each bloc is measured against its own GDP. Table 1 lays out the basic geography. Comparing China — a single economy — to multi-economy blocs is more than fair: China’s GDP and population both exceed those of either bloc. As shares of their own economies, the US deficit (–3.6%) and China’s surplus (+3.7%) are nearly mirror images. But the Asia Surplus bloc runs +7.3% — nearly twice as imbalanced as either. The EU bloc runs +5.7%. By the natural “how imbalanced is this economy” measure, China is the least imbalanced of the structural surplus blocs; it dominates the absolute numbers only because it is so large. The US, by contrast, is both the largest single deficit economy in absolute terms (–0.94% of world GDP in 2025) and the largest economy at market prices (26% of world GDP). The annual IMF External Sector Report (International Monetary Fund. Research Dept. 2025) provides a rigorous assessment of imbalances that looks beyond size alone to their underlying domestic-policy roots.

Why This Matters

First, China’s role in global current account surpluses is overstated. It is one among many surplus economies. In the years before the pandemic it had relatively small surpluses — and surpluses that were much smaller than the Asian Surplus bloc and the EU Surplus bloc. Ironically, China’s big surpluses only re-emerged after the trade war escalated in 2018 (but that’s a topic for another day). Even in 2031, the Asian Surplus bloc as a whole is forecast to have basically the same total surplus as China. The EU bloc ran persistently higher surpluses than China through 2010-2019, though it has been undergoing a trend decline.

Second, trade imbalances mirror domestic imbalances. It is an accounting identity that a trade surplus means an economy produces more than it uses at home; a trade deficit means an economy uses more than it produces. Or, equivalently, the current account surplus is the gap between domestic saving and investment. For the US trade deficit to fall, US saving minus investment has to rise — most likely through higher saving (for example, smaller government deficits). The reverse is true for the surplus economies: more consumption in China would translate into lower saving and a smaller surplus. But — tying back to the first point — China alone is not enough; adjustment would have to take place across all the big surplus economies. Finding durable cures for trade imbalances requires looking at their domestic counterparts in each economy — the spirit of the IMF External Sector Report.

Third, trade fragmentation would be costly — and the global economy is too integrated for clean decoupling. Trade tensions raise barriers and uncertainty, reduce efficiency, and shrink global growth. Trade integration has been one of the great growth engines of the post-World War II era, primarily through higher total factor productivity. With a trend decline in growth, as I wrote at HKU Business School, the world economy needs more trade, not less. And the integration is structurally deep: my last post showed how intertwined global balance sheets are, and trade flows and supply chains are also deeply entangled. As I wrote in my earlier post on the trade war, the IMF has highlighted that fragmentation in an extreme scenario could cost around 7% of global GDP, with technology decoupling raising the hit to 8–12% for some economies (Aiyar et al. 2023). The endgame is slower growth, lost economies of scale, and economies stuck investing in redundant supply chains.

The Details

The four-year snapshot in Figure 1 isn’t cherry-picked. Figure 2 plots the same four series year by year from 2000 to 2031. The asymmetry is persistent: the US deficit (green) sits above the surplus blocs in essentially every year, and the three surplus blocs cluster together near 0.5% of world GDP. China is the most volatile of the surplus players — swinging from 0.6% in 2007 to near zero in 2018 and back to 0.6% in 2025.

Source: IMF WEO and author’s calculations. Construction. Annual current account balances scaled by world GDP, 2000-2031. The US deficit is shown as a positive number for direct comparison with the surplus blocs (defined in Figure 1). The shaded band marks WEO forecast years.

The asymmetric pattern shows up just as cleanly at the economy level. Figure 3 plots the six largest surplus and six largest deficit economies in 2025. The contrast between the two sides is stark. The US deficit (0.94% of world GDP) is more than three times the next five deficit economies combined — the UK, Brazil, Australia, Saudi Arabia, and India together add up to just 0.27%. On the surplus side, the picture is reversed: the next five economies after China actually sum to more than China itself — Germany, Japan, Taiwan, Korea, and the Netherlands together total 0.71%, against China’s 0.62%. China is large in absolute terms, but it does not dominate the surplus side the way the US dominates the deficit side.

Source: IMF WEO and author’s calculations. Construction. Top six surplus and top six deficit economies in 2025 by current account balance, scaled by world GDP. Economy labels include the value (the chart axis is unsigned magnitude; deficit economies’ bars represent negative CA balances). Aggregates (Euro Area, etc.) excluded — individual economies only.

A conventional way to view global imbalances is to sum all surpluses and all deficits each year. Figure 4 shows this view — consistent with the standard presentation in the IMF External Sector Report (International Monetary Fund. Research Dept. 2025, fig. 1.1.2) and the CEPR Geneva 28 report (Chari et al. 2025, fig. 8b). Aggregate imbalances peaked near 3% of world GDP in 2007, compressed through the 2010s, and have rebuilt since 2020. The 2031 forecast puts the total close to its long-run median — a return to historical norms, not a new extreme.

Source: IMF WEO and author’s calculations. Construction. Each year’s bar stacks economies by group. Above zero: China and Germany broken out individually; “Other Asian Surplus” (Japan, Korea, Taiwan, Singapore, Hong Kong SAR, etc.); “Commodity Surplus” (oil and gas exporters); “Other Surplus” (residual — mostly other AEs in surplus such as the Netherlands, Switzerland, Ireland). Below zero: United States broken out; “Other AE Deficit” (UK and other advanced economies in deficit); “RoW Deficit” (rest of world in deficit). The black “Discrepancy” bar closes the books each year. Group composition is reassigned each year by the sign of each economy’s current account. Dashed line: median total surplus 1980-2025 (1.72%); dotted line: median 2000-2025 (1.94%).

Figure 5 redraws the same picture using the fixed blocs from this post. Members of the four surplus blocs are assigned to their bloc regardless of whether they happened to be in surplus or deficit in a particular year; everyone else is netted into “Rest of World (net).” The medians are slightly lower than in Figure 4 (1.46% / 1.73% vs 1.72% / 1.94%) because the fixed blocs net out some flickering surplus from non-bloc economies.

Source: IMF WEO and author’s calculations. Construction. Stacked bars by fixed bloc, 1980-2031. Above zero: China, Asia Surplus and EU Surplus (as defined in Figure 1), and the Commodity bloc — oil and gas exporters across the Gulf, other MENA, sub-Saharan Africa, the former Soviet Union, Norway, Brunei, and Latin America. Below zero: United States, Rest of World (net = all economies not in a defined bloc), and Discrepancy (closes the books each year). Dashed line: median total surplus 1980-2025 (1.46%); dotted line: median 2000-2025 (1.73%).

Finally, a focused 25-year view of surplus blocs vs. the US deficit. Figure 6 stacks the four surplus blocs from 2000 to 2031 and overlays the US deficit (shown as positive). The surplus stack typically runs roughly twice the US deficit — the rest goes to absorbing European-side deficits, EM deficits, and the measurement discrepancy.

Source: IMF WEO and author’s calculations. Construction. Stacked bars sum the four surplus blocs (Commodity on top, then EU Surplus, Asia Surplus, China at bottom), as defined in Figures 1 and 5. Green line is the US current account deficit shown as a positive number. All scaled by world GDP. The shaded band marks WEO forecast years.

Data

All charts and tables in this post draw on the IMF World Economic Outlook database, April 2026 vintage, accessed 15 April 2026 at https://data.imf.org/en/datasets/IMF.RES:WEO. Bloc aggregates are author’s calculations using economy-level current account (BCA) and nominal GDP (NGDPD) series. Population uses series LP; PPP shares use PPPSH.

References

Aiyar, Shekhar, Anna Ilyina, Jiaqian Chen, Alvar Kangur, Juan Trevino, Christian Ebeke, Tryggvi Gudmundsson, et al. 2023. “Geo-Economic Fragmentation and the Future of Multilateralism.” Staff Discussion Notes 2023 (001): 1. https://doi.org/10.5089/9798400229046.006.

Chari, Anusha, Nathan Converse, Arnaud Mehl, Gian Maria Milesi-Ferretti, and Isabel Vansteenkiste. 2025. Geneva 28: Geopolitical Tensions and International Fragmentation: Evidence and Implications. October 2025. https://cepr.org/publications/books-and-reports/geneva-28-geopolitical-tensions-and-international-fragmentation.

International Monetary Fund. Research Dept. 2025. External Sector Report 2025: Global Imbalances in a Shifting World. Washington, D.C.: International Monetary Fund. https://doi.org/10.5089/9798229017329.085.

Subscribe now

Construction. The numerator is the sum of foreign assets and foreign liabilities (gross IIP) for each economy at end-2024, from the IMF’s International Investment Position dataset. The horizontal axis scales that sum by the economy’s own GDP — a measure of how globalized the economy is — and is plotted on a natural-log scale, because the ratio spans orders of magnitude (financial centers like Luxembourg sit far above ordinary economies). The vertical axis scales the same numerator by world GDP — a measure of how much the economy weighs in the global system — and is plotted on a linear scale, which is why the US towers over everyone else. Sample: all IIP-reporting economies.

Subscribe now

Foreign assets and liabilities exceed GDP in almost every economy. 85 percent of economies clear 100 percent of GDP; the median sits around 200 percent. Because assets and liabilities are stocks rather than flows, this captures how much of an economy’s balance sheet is wired into the rest of the world. Financial centers — Luxembourg, Hong Kong SAR — anchor the far right. The US stands out on the vertical axis: no other economy is close in global weight.

Why This Matters

First, the world is massively integrated. Talk of fragmentation, de-risking, and friend-shoring dominates the press, but the data tell a different story. The median economy has foreign assets and liabilities above 200 percent of its own GDP — balance-sheet linkages on a scale that simply cannot be unwound. We often measure integration through trade flows — the income statement of integration — but Figure 1 is the balance sheet, and it dwarfs the flow. (My earlier post on the trade war happened to show the trade-flow analog.)

Second, dis-integration would be costly. With foreign assets and liabilities above 100 percent of GDP in nearly every economy, the world is simply too interconnected to unwind. And even mapping ownership — which assets are held by allies, which by rivals — is enormously difficult, which makes it hard to even assess what fragmentation would look like in practice. Chapter 1 of Chari et al. (2025) provides an outstanding summary of the state of the field on this question.

Third, the US is the outlier — and that has global consequences. US assets and liabilities dominate Figure 1; whatever happens in US markets reverberates everywhere. As Figure 2 shows, the US is also the world’s largest net debtor by a wide margin — partly because of persistent trade deficits, but more because foreign holders of US assets (equities especially) have done very well.

The Details

Today’s post draws on comments I gave as a discussant on Geneva Report 28 (Chari et al. 2025) at a recent Hong Kong Monetary Authority event. The report is essential reading for anyone interested in how rising geopolitical tensions are reshaping global finance — careful on the evidence, thoughtful on the implications.

Largest creditors and debtors

The US is both the world’s most integrated economy and its largest net debtor. Figure 2 ranks the top net creditors and net debtors as a share of world GDP. The US is in a class of its own — the largest net position of any economy by a wide margin, mirroring its outsized gross balance sheet from Figure 1.

How net positions have evolved

The US net debt has grown sharply over the past two decades. Figure 3 traces the net positions of the US, Germany, China, and Japan as a share of world GDP. The US line is the dominant feature: the world’s biggest debtor, getting bigger. The three largest creditors — Germany, China, Japan — are not close in absolute size, and have moved much less.

Why the US is so deeply negative

More than half of the US net debt buildup is not US borrowing — it’s foreigners earning strong returns on their US holdings. Figure 4 splits the cumulative change in US net IIP into two pieces: cumulated current-account deficits (US borrowing from the rest of the world) and valuation effects (the gap between how much US-held foreign assets appreciated and how much foreign-held US assets appreciated).

BOP accounting. In any year, the change in an economy’s net foreign position equals the current-account balance plus valuation changes. The US has run persistent current-account deficits, but rising US equity prices have made foreign-held US assets appreciate much faster than US-held foreign assets — a valuation drag on net IIP that owes more to market performance than to borrowing.

Finally, for those who prefer more details, Table 1 summarizes the US balance sheet. The US has a net foreign debt equivalent to 91 percent of US GDP. More staggering, though, is that the US has liabilities to foreigners in excess of 200 percent of GDP — over half of that is foreign holdings of US equities and bonds (portfolio investment). Thus, as argued above, this reflects the attractiveness of US financial markets to foreigners as much as US borrowing to finance its trade deficit.

Net debt this large would be alarming in any other economy — but the US is special. Most US external liabilities are denominated in dollars (its own currency), so there is no currency-mismatch risk. And US markets are simply too big and too liquid for any plausible mass exit: investors pulling funds out of the US en masse would have nowhere comparable to redeploy them. Those are stories for another day; today’s point is simply how integrated the global economy is, and how central the US sits within it.

Data

• Cross-border positions: IMF International Investment Position (IIP) dataset, end-2024 vintage, downloaded 23 March 2026 from https://data.imf.org/en/datasets/IMF.STA:IIP.

• GDP for scaling: IMF World Economic Outlook (April 2026), accessed 15 April 2026 at https://data.imf.org/en/datasets/IMF.RES:WEO.

References

Chari, Anusha, Nathan Converse, Arnaud Mehl, Gian Maria Milesi-Ferretti, and Isabel Vansteenkiste. 2025. Geneva 28: Geopolitical Tensions and International Fragmentation: Evidence and Implications. October 2025. https://cepr.org/publications/books-and-reports/geneva-28-geopolitical-tensions-and-international-fragmentation.

Subscribe now

Subscribe now

Figure 1 also shows inflation rising. Higher commodity prices from the Iran war are, ironically, helping in this case. Q1 only partly captures the effect; Q2 will likely show a stronger impact, possibly pushing all three indicators into positive territory. But such cost-push inflation — driven by higher commodity prices rather than stronger demand — is hardly the way to durably escape deflation.

Where inflation settles is less clear. Without a pickup in demand, a protracted stretch of lowflation — positive inflation, but below the government’s target of around 2 percent — is quite possible. And whether the print lands a touch above or below zero is largely noise: what matters is that, absent a durable pickup in demand, inflation will likely still be running below target once the oil price shock washes through the system.

Note on construction. Both Core CPI and PPI are published monthly. The quarterly values shown here are the simple average of the three monthly year-on-year inflation rates within each quarter. The GDP deflator is quarterly by construction, computed as the ratio of nominal to real GDP YoY growth.

Why This Matters

First, deflation is costly. Falling prices may sound appealing, but they aren’t. They reflect an imbalance between supply and demand — people aren’t buying enough, so prices fall. Wages tend to follow prices down, undoing the apparent benefit. And since nominal wage cuts are hard to push through, firms tend to cut hours or workers instead — weakening demand further. For an economy carrying a lot of debt (like China), nominal GDP growth also helps the economy grow out of its debt — deflation pulls in the opposite direction.

Second, deflation episodes are hard to end. Just ask Japan, which spent roughly two decades with falling or flat prices. Once prices and wages have fallen for a protracted stretch, people expect more of the same. Those expectations become self-reinforcing — workers accept slower nominal wage increases, firms hesitate to raise prices, and the cycle entrenches itself.

Third, the sign of the print isn’t the substantive issue. The forces keeping inflation below target — such as a depressed property sector and weak labor market sentiment (as I argued recently) — haven’t gone away. The commodity spike from the Iran war is doing the immediate lifting, but it doesn’t strengthen demand: higher commodity prices erode households’ purchasing power rather than expanding their spending. And when oil normalizes — likely within a year — base effects could pull the print back below zero. Whether 2027 lands at +0.4 or −0.4, the substantive issue is the same: inflation will likely keep running below target unless the underlying demand-supply imbalance is addressed.

The Details

Today’s post takes on a controversial issue: was China in deflation, and where does it go next? You may agree or disagree with my read. Either way, the charts below are my actual workbook — the go-to ones I use to track inflation.

A few rules of thumb for reading China’s inflation indicators:

• Headline CPI is pushed around by food prices, so look at core.

• Producer prices often decline, but rarely pass through to consumer prices.

• Producer prices themselves are largely driven by global commodity prices.

Consumer Price Index (CPI)

In China, headline inflation is driven largely by food prices. Figure 2 plots monthly year-on-year inflation for headline CPI, core, and food. Large swings in food are clearly the main driver of headline swings. Dig deeper, and within food it’s often pork.

Core inflation is the better indicator for monetary policy. This is true generally — not just in China — but it matters more here because food carries such a heavy weight in the total CPI. Food and energy prices are pushed around by factors outside monetary policy’s control (weather, wars); core strips those out, isolating prices that respond to policy. The shortcoming is that China’s core CPI is relatively smooth, with few cyclical signals to read.

Headline inflation would have been negative for years if housing were measured differently. China calculates imputed rent on owner-occupied housing — a fancy way of saying how much rent an owner would pay to live in their own home — using a depreciated construction-cost (“acquisition”) approach. Most other economies instead anchor housing on observed market rents (the US, UK, Japan), and a handful (Canada, Sweden, Iceland) tie it to home market values via the risk-free return (“opportunity cost”). Sheng, Zhongming and He, Zhu (2024) apply this last method to China.

Two things change: the housing weight in the basket rises from roughly 23% to 32%, and the housing price series follows falling home prices rather than the flat construction-cost series. The revalued headline CPI turns negative in late 2022 and bottoms near −3% in late 2023 / early 2024, even as the official CPI stays roughly flat. Core CPI, though the authors didn’t compute it, would feel an even larger drag from the housing reweighting than headline CPI does — housing weights more heavily in core after food and energy are excluded.

Producer Price Index (PPI)

Since 2006, China’s producer prices have spent much more time in negative territory than the CPI. The pass-through from producer to consumer prices in China looks fairly weak. Figure 3 shows that headline CPI has rarely been firmly negative — except for a short stretch around the 2008–09 global financial crisis. The PPI, meanwhile, has gone through four episodes of “deflation” — including a multi-year stretch in 2013–16 — without translating into falling consumer prices.

The PPI, in turn, is heavily influenced by commodity prices. Just as food drives the CPI, commodities drive the PPI. Figure 4 plots global commodity prices (in RMB), the PPI for Raw Materials, and headline PPI. Global commodity swings clearly pass through to Raw Materials prices, and the link from Raw Materials to headline PPI is even tighter. This suggests headline PPI should turn firmly positive in the coming months as the energy price spike feeds through.

Finally, and importantly, the consumer goods component of PPI has not shown signs of recovery. Figure 5 shows PPI consumer goods prices in their longest stretch of declines since 2006, with no clear sign of having turned the corner. As with headline PPI, eyeballing the data suggests little pass-through into core CPI. Even so, continued declines in the producer prices of consumer goods are a cautionary signal on persistent demand-supply imbalances.

References

Sheng, Zhongming, and He, Zhu. 2024. When Inflation Meets Housing Prices. 24-060. China Finance 40 Forum (CF40).

Subscribe now

This post returns to that paradox. Despite the rapid growth, consumption’s share of the economy remains low — the economy has simply grown even faster than consumption. Figure 1 shows the result: at 40 percent of Gross National Disposable Income (GNDI), China ranks second from the bottom among OECD economies.

Subscribe now

China’s low share is robust to measurement issues. As shown in “The Details,” 2024 is not special. The same chart for 2019 — just before the pandemic — yields the same result. Likewise, the 5-year average (2015-19) shows the same. Nor does it matter whether I use GDP or GNDI (though I argue GNDI is the correct denominator). Broadening the sample to include non-OECD economies yields qualitatively the same result: China sits near the bottom.

Why this matters

First, it shows China has plenty of room for consumption growth. Yes, this is putting a positive spin on the data. With China already posting the fastest consumption growth, it is hard to argue it should have been faster still. But the low ratio means that consumption can grow faster than the economy for many years. This is important. The economy is undergoing a natural trend slowing as it catches up (or ‘converges’) to advanced economy status. And, more important, consumption growth need not slow as much as GDP growth. Since consumption is a better measure of living standards than GDP, the convergence-driven slowdown need not translate into a comparable slowdown in living standards.

Second, it underscores why the Government is emphasizing boosting consumption. Coming out of the annual policy meetings in China — the “Two Sessions” — was an explicit call to raise consumption’s share of the economy.

Third, the consumption ratio matters for the rest of the world. More consumption in China should translate to a smaller trade surplus. But, this is a topic for a future post.

The details

The cross-country bar chart in Figure 1 tells a clean story but raises the question of whether it is fair to compare China to advanced economies. The scatter plot in Figure 2 answers that directly. Using Penn World Tables (Feenstra, Inklaar, and Timmer (2015)) data for 2023 (the last available year), it plots household consumption as a share of GDP against GDP per capita in current US dollars for the 75 largest economies. The trend line suggests that richer economies on average have lower consumption shares. China stands out near the bottom of the chart. All of its peers at the bottom are either energy-producing economies or have large net primary income outflows (meaning GNDI is smaller than GDP or, put differently, their ratios would rise if scaled by GNDI). In fact, in 2023, the economies with the three largest net primary income outflows were Luxembourg (31 percent of GDP), Ireland (23 percent of GDP), and Singapore (18 percent of GDP), the only non-energy exporters with lower consumption to GDP ratios than China.

One might also worry that 2024 is an unusual year — with China still recovering from pandemic disruptions and consumer confidence still soft. Figure 3 addresses this by going back to 2019, the last pre-pandemic year, and adding a dot for the five-year historical average (2015-19). The picture is nearly identical: China near the bottom, well below the OECD median, and the five-year average confirms the 2019 reading is not an outlier.

Finally, Figure 4 replaces the five-year average dot with household consumption as a share of GDP rather than GNDI. The two denominators give very similar rankings. China’s net income from abroad is small relative to GDP, so GNDI and GDP are close in value. The China comparison is essentially unchanged whichever denominator you use. Some economies, like Luxembourg and Ireland (as noted above), have large income payments abroad — meaning that GNDI is smaller than GDP. This illustrates one reason why it is conceptually correct to use GNDI — which is the income available for consumption — instead of GDP, which includes production where the income accrues to foreigners and so is not available for domestic consumption. But this also shows that, in practice, the choice of denominator rarely matters — except for the few economies where such income payments are large.

The four charts taken together make a consistent case. China’s low household consumption share is not an artifact of the choice of year, the choice of denominator, or the income level of the country — however you slice it, China is near the bottom.

Data

Figures 1, 3, and 4 — OECD and CEIC. These charts use OECD national accounts data: household final consumption expenditure (transaction code P3, sector S1M — households and non-profit institutions serving households) from the OECD’s DF_TABLE1_EXPENDITURE dataset, and Gross National Disposable Income (B6G) from DF_TABLE14_EXP. The denominator is GNDI — the income actually available to residents — rather than GDP, because GDP includes income that accrues to non-residents and is not available for domestic consumption. For most countries the difference is small; for a handful of financial centres and large energy exporters it matters more (see the discussion of Figure 4).

China is not an OECD member and does not report sectoral accounts to the OECD. The China data points use NBS expenditure-side household consumption (CEIC series “Final Consumption Expenditure: Household”) divided by GNDI constructed as GDP plus net primary and secondary income from the IMF Balance of Payments Statistics. In practice China’s net primary income from abroad is small relative to GDP, so the GNDI and GDP denominators yield nearly identical results for China.

Figure 2 — Penn World Tables. The scatter uses Penn World Tables version 11.0 (Feenstra, Inklaar, and Timmer (2015)), the most recent release with broad country coverage, for 2023. The consumption share is household consumption divided by GDP at current prices (v_c / v_gdp). The income measure on the horizontal axis is GDP per capita in current US dollars at market exchange rates (v_gdp / xr / pop), plotted on a natural log scale — the standard approach in cross-country income comparisons, since what matters is the proportional gap between countries rather than the dollar gap. The sample is the 75 largest economies by 2023 USD GDP, together about 97 percent of world output.

A simple regression of the consumption share on log income per capita across the 75 economies yields a slope of roughly −6.7 percentage points per unit of log income (p < 0.001, R² ≈ 0.39) — richer countries do, on average, consume a smaller share of GDP. Removing financial centres (Luxembourg, Ireland, Singapore, Hong Kong) and major energy exporters (Saudi Arabia, Kuwait, Qatar, UAE, Iraq, Algeria, Oman) leaves 64 economies; the slope barely changes (≈ −5.8 pp, R² ≈ 0.43), confirming the relationship is not driven by structural outliers. China in 2023 sits roughly 18–20 percentage points below the fitted line for its income level, meaning its low consumption share cannot be fully explained by the income–consumption relationship common to other countries.

Subscribe now

References

Feenstra, Robert C., Robert Inklaar, and Marcel P. Timmer. 2015. “The Next Generation of the Penn World Table.” American Economic Review 105 (10): 3150–82. https://www.ggdc.net/pwt.

More broadly, the IMF data show that emerging markets are the main drivers of global growth — and this has been true for decades, not just last year. Summing contributions over the full 2001–2025 period, emerging market and developing economies account for about three-quarters of cumulative world growth, with China alone responsible for roughly 30 percent. The result in part reflects that the IMF uses purchasing power parity (PPP) weights to aggregate global growth. As I documented in my first post, by this metric China is already the world’s largest economy. Moreover, China and India not only are large: they grow much faster than the other economies. Thus, it follows naturally that together they account for nearly half of global growth.

Subscribe now

Why It Matters

• China and India drive GDP growth, at least mechanically. Thus, what happens in these economies will have an outsize influence on the aggregate numbers. So even small changes in their forecasts will move the global numbers. The advanced economies, by contrast, play a much smaller role.

• The economics is more complicated than the mechanics. Growth may be in China, but the underlying demand may be coming from elsewhere. For example, roughly one-third of China’s growth in 2025 came from net exports — foreign demand, in other words, was a key factor in China’s growth. Thus, the mechanics of global growth (how much China contributes) and the underlying economics (where does the demand originate) can be different.

• The Asian engine is actually a trio. Add the ASEAN-5 to China and India and the three together account for well over half of global growth. The center of gravity of the global economy has shifted decisively to Asia — and the data suggest it will stay there through the forecast horizon.

• Methodology matters. PPP weights are but one way to aggregate growth. The World Bank and OECD, for example, use market exchange rates. As I showed in my first post, the weighting scheme matters — the US is the largest economy using market exchange rates. PPP-based weights give more weight to emerging markets, which explains why the World Bank and OECD have lower global GDP forecasts (this would happen, solely due to weighting, even if the forecast for every economy was the same across organizations). But does the weighting scheme matter for who drives growth? Surprisingly, not so much for China — but a lot for India and the US. More in a future post.

Details

Figure 2 shows a stacked bar variant of Figure 1. Rather than the percent of world growth, the bars are the percentage point contributions of each economy. It is visually clear that China has been the number one contributor to global growth from 2000 to the present. India’s contribution is rising. Meanwhile, the share of advanced economies like the US and European Union has been shrinking.

Figure 3 shows that 2025 is not unusual: the picture looks essentially the same for 2026, the first forecast year. Figure 3 repeats Figure 1 but uses 2026 data (and adds some additional information about each economy). The broad story is the same. China is the biggest contributor to global growth. India is second. China, India, and the ASEAN-5 account for over half of global growth. This broad story is forecast to continue. Figure 4 repeats Figure 3, but uses 2031, which is the last year in the data. With the forecast of continuing slowing in China, the contribution from China and India are basically equal by 2031. India is still half the size, but is growing twice as fast.

Table 1 puts numbers on this. It shows each economy’s share of cumulative world growth across three periods: 2001–2025 (the full post-2000 window), 2016–2025 (the recent decade), and 2026–2031 (the pure forecast). Using cumulative contributions rather than annual averages avoids distortion from negative-growth years.

Finally, what about market exchange rates? The weighting scheme matters — a lot. Figure 5 compares global growth measured using PPP weights versus market exchange rates. Market exchange rates (MER) put much more weight on advanced economies that generally grow more slowly, so MER-weighted global growth is both lower and more sensitive to advanced economy forecasts. A deeper dive into what changes when you switch weighting schemes — and what it means for reading global growth headlines — is a topic for another post.

Data

How is global growth measured? The IMF calculates global growth as a weighted average of individual economy growth rates: . For the headline (PPP-weighted) number, the weight is each economy’s share of world GDP at purchasing-power-parity prices. The IMF also publishes a market-exchange-rate-weighted aggregate, where the weight is each economy’s share of world GDP in current US dollars; to reduce volatility from exchange rate swings, this uses a three-year trailing average of nominal GDP shares.

All data are from the International Monetary Fund’s World Economic Outlook (WEO), April 2026 release, downloaded from https://data.imf.org/en/datasets/IMF.RES:WEO on 15 April 2026. Growth contributions are computed as , where NGDP_RPCH is real GDP growth and PPPSH is each economy’s share of world GDP at purchasing-power-parity weights. Rest of World is the residual: world growth minus the sum of the six featured economy contributions. The European Union (WEO code 998) is used rather than the Euro Area, as the EU aggregate carries a published PPP share. Years 2026–2031 are IMF projections.

ASEAN-5 refers to Indonesia, Malaysia, the Philippines, Singapore, and Thailand. The IMF publishes a composite aggregate for this group (SDMX code G510; WEO country codes: Indonesia 536, Malaysia 548, Philippines 566, Singapore 576, Thailand 578). Adding Vietnam (growing at roughly 6–7 percent per year) to create an ASEAN-6 would raise the group’s recent growth rate by roughly 0.3–0.5 percentage points per year.

For market-exchange-rate-weighted world growth (Figure 5), the series shown is the IMF’s published MER-weighted world growth aggregate (WEO series NGDP_RPCHMK), which uses a three-year trailing average of nominal GDP shares to reduce volatility from exchange rate movements.

Subscribe now