But apart from extremists like Cotton, Washington has not actually abandoned the idea of win-win bargains: they just apply that logic to greater autarky, not trade. A recurring theme of Western criticism of China’s economic model is that it’s bad for Chinese people. That their wages, consumption, and general economic welfare would be better if China followed conventional economic advice, let its currency appreciate, and stimulated domestic demand.

Is it really plausible that Chinese people, on average, would be better off if China abandoned its spectacularly successful growth model? In the short term, perhaps, but over time? It takes great confidence to confront the greatest engine of growth and poverty reduction the world has ever seen and declare, “eh, could be better.”

Comparing similar GDP levels: Xi’s China vs LBJ’s America

Let’s turn back the clock to 1964.

Running through many Western critiques of China’s economic model is the suggestion that the Chinese population fares poorly relative to China’s GDP. Wages are low relative to GDP. Consumption is low as a share of GDP. Social services are meager relative to China’s (until recently) booming GDP.

This naturally invites the comparison to how other countries at a similar level of GDP performed. So let’s start with America, the last time its GDP looked like China’s.

All three widely-used sources for comparing real GDP growth over time across countries roughly agree on that date: the early- to mid-1960s. We’ll call it 1964 for simplicity. 1964 was an interesting inflection point for the U.S. in welfare terms. Lyndon Johnson was inaugurated just before Thanksgiving 1963. His 1964 State of the Union address laid out the ‘War on Poverty’. Jim Crow still ruled the South, and Johnson would sign the Civil Rights Act in July of ‘64. Medicare wouldn’t come into existence for another year.

Chinese people today are probably poorer than Americans were in 1964 – though it’s unclear how much

To track Johnson’s war on poverty, the Social Security Administration defined a poverty line at roughly $2.17 in daily pre-tax income per person – just under $20 today. At the time, just under 20% of Americans lived in poverty by that definition.

The closest comparison we can make for China relies on consumption, not income, a fudge which is common in global poverty measurement but probably understates China’s real performance somewhat. Nevertheless, the contrast is clear. Over 70% of Chinese people live on less than $19/day in PPP terms.

That’s a huge gap, but it may not tell the entire story.

My colleague Oliver Kim wrote a while back about how part of the mystery of China’s low consumption share of GDP – which underlies these poverty outcomes – is an artifact of measurement. There are signs that GDP is overestimated, and consumption is underestimated, a point Nick Lardy at PIIE has made before. Perhaps the best evidence for higher-than-reported Chinese consumption comes from a report by the China Finance 40 Forum, a Beijing think tank. The table below draws largely from that report, and juxtaposes it with comparable figures for the US circa 1964.

Setting aside messy issues of price deflators and purchasing power indices, Chinese people eat more and buy more consumer durables than Americans did in LBJ’s time – with the notable exception of cars. Housing is the main area where they are relatively deprived: China’s famous high-rise apartment blocks are not as spacious as suburban American houses.

Table: US vs Chinese consumption using physical metrics

Sources: US 1964 — FAO food balances via Our World in Data (calories, protein); Statistical Abstract of the United States 1966, Tables 819 and 1194 (factory sales/manufacturers’ shipments of cars and appliances, TV production; US population 191.9 million); Moura, Smith & Belzer (2015, PLOS ONE) (floor space, whole housing stock, ±10 sq ft). China — FAO via OWID (food, 2022) and Yu Fei & Guo Kai, “China’s Consumption Is Not Nearly as Low as It Appears,” CF40 Institute, July 2025 (durables ~2023, floor space 2020). The TV/smartphone row is not like-for-like across periods.

Chinese wages relative to GDP are very low by Western standards, but similar to other East Asian growth successes

As Biden economic adviser Brian Deese noted after leaving the White House, Chinese “workers are paid a smaller share of what they produce than in most maturing economies.” The word “maturing” is doing a lot of work here, but the answer seems clear regardless.

There’s no doubt that Chinese workers today earn far less than their American counterparts, even 60 years ago. According to published NBS statistics from China, public sector workers fare best, earning about $30k a year on average in real (PPP) terms. That’s about 60% of the median male worker in 1964 America, and only slightly behind the median female worker. Outside the public sector, it’s more bleak. The average private-sector employee in China today is earning substantially less than US minimum wage circa 1964.

Venturing into more ambitious counterfactuals, Robert Lighthizer, Trump’s first-term trade representative, told the House Select Committee on the CCP in 2023 that

“China’s mercantilism comes at the expense of Chinese workers, who face lower wages and living standards than they would under a more balanced trade regime...”

Lighthizer’s assertion assumes that the spectacular wage gains that Chinese workers have experienced over decades would have occurred in the absence of that “mercantilist” system.

Is that true? Or has China’s mercantilist been responsible for its remarkable growth?

China’s low wages relative to GDP are pretty typical for fast-growing East Asian economies pursuing an export-led manufacturing model. The Occupational Wages around the World database put together by economists Richard Freeman and Remco Oostendorp shows wages in both Japan and South Korea at similarly low levels when they were at similar levels of GDP.

Arguably, comparing wages to GDP at a point in time is the wrong benchmark. The whole premise of this debate is that China produces “too much” GDP, in the form of investment and net exports, relative to its overall level of development. The more relevant question is whether from a given starting point, Chinese workers would see bigger gains under a different system.

Chinese people today enjoy more and better public services than Americans did in 1964

Another common complaint in the West, and among plenty of Chinese economists too, is that the Chinese welfare system is too stingy.

Jumping into the mind of the great leader himself, Paul Krugman summarized the standard view in 2024: “Xi views American-style consumption as wasteful, and fears that providing too much state support to households could encourage ‘welfarism.’” This line of concern was echoed recently by former Biden official Brad Setser in an interview with Ezra Klein, where he praised Lighthizer’s trade stance under Trump I, and noted that

“China actually has a rather thin system of social insurance... Personal income tax collections are like 1% of China’s GDP. It’s 8% here…. The basic retirement benefit that anyone in China gets, no matter what your residency status, is like tens of dollars a month. It’s really, really trivial.”

Contrary to this view, China spends more on social services today than America did at a similar per capita GDP level. Social welfare spending broadly defined is about 14 percent of GDP in China today, compared to just under 12 percent when LBJ took office.

Education expenditure was pretty comparable in 1960s America compared to contemporary China. But the system was more expensive on a per pupil basis, so reached fewer people. Tertiary enrollment in 1964 was 28.7 percent, compared to 60.8% in China per recent dat.

The boom in China’s college enrollment looks particularly dramatic juxtaposed with its fertility collapse. As a colleague at Peking University noted when showing me around the campus: there is no choice but for Chinese universities to shrink. As of 2022, the number of first year college students was larger than the number of babies born in all of China. Even if they all go to college, enrollment will decline.

Contrast that with 1964 America, which had college seats for only about a third of babies born that year. Even fast forwarding twenty-some years to when those kids actually reached college age, and there were only places for about three-fifths.

Healthcare is where China actually surpasses America’s spending patterns circa 1964. And that translates into better health as well.

Given global improvements in health technology – from new vaccines and drugs to MRIs and beyond – it’s no surprise that Chinese life expectancy today (78 years in 2024) exceeds Americans’ life expectancy at a similar GDP per capita (70.2 in 1964), and infant mortality in China is just 4.1 per 1000 live births, compared to nearly 25 deaths in Lyndon Johnson’s America. But again, the whole world has seen pretty rapid increases in life expectancy, driven largely by declines in infant mortality, so there’s a time effect here apart from any difference in health systems.

That said, it’s not just technological improvements that explains the China-US gap. China also boasts 3.6 physicians per 1,000 residents compared to America’s 1.5 in 1964, and has nearly twice as many hospital beds relative to its population. So the health system is simply bigger and more generous in some respects than America was at a similar income.

Before Medicare and Medicaid launched in 1965, about 78% of Americans had private hospital insurance of some kind, per the Health Insurance Association of America (which was higher than survey estimates). By contrast, 94% of Chinese people today are covered by basic public health insurance – though that coverage is indeed pretty basic.

Chinese growth has been good for Chinese households – perhaps more so than in America

Any defense of the Chinese economic model from a worker perspective hinges on wage increases over time, not their meager level, and the extent to which the fruits of world history’s greatest economic growth miracle have trickled down to the masses.

The FT’s Martin Wolf is a good example of many people’s skepticism: .

“Is this good valuable GDP, by which I mean GDP that produces now or in future welfare for the Chinese people, or the GDP which consists of creating stuff which will actually never be useable and that’s really difficult to tell. Obviously, a lot of it has been the latter and there’s no doubt that China has got much better off - there’s no doubt. But quite a bit of it has also been the former and that is always the question.”

As Wolf notes, Chinese leadership sometimes voices similar concerns. But the data doesn’t really bear them out.

In America, much has been made of the disconnect between economic growth and household earnings. Before Reagan’s arrival in the 1980s, median household income tracked GDP growth roughly one-for-one. Post 1980, that relationship fell apart.

In China, the story is roughly the opposite: the link from growth to wages and household consumption has been strong and getting stronger over time.

Formal wages in China have basically kept pace with economic growth for the last 30 years, or even surpassed GDP growth. The elasticity of median household consumption with respect to GDP in China started off low in the 1990s and early 2000s, about 0.7. Since then it’s picked up to almost 1.2 – implying consumption for the median Chinese person is growing faster than the economy as a whole. That’s also true for formal wages. The one group which has notably not kept up is migrant workers, whose wages show a growth elasticity (relying again on official NBS figures) of about 0.7.

In short, household gains from growth in China today look similar to America’s post-war glory days, and are much clearer than in America today. There is no denying that Chinese workers get a smaller share of the pie. But their income and consumption gains track growth remarkably well.

And again, regardless of how growth is shared, there has just been a lot more growth to go around in China. From the 1960s to today, Americans saw median incomes roughly double in real terms. China achieved that same feat (using median consumption as the metric) in less than a decade, and more than once.

A win-win retreat from globalization?

In sum, China’s population has a lot to lose. Its per capita GDP is roughly where America’s was when LBJ launched the war on poverty. Yes, Chinese people remain poorer and earn lower wages, but have access to more and better social services, and have enjoyed the gains from growth in a way Americans (perhaps rightly) feel they no longer do. The real material gains China’s model has produced for its population are unrivaled, and any suggestion that a different model would do better bears an enormous burden of proof.

Of course, Chinese welfare isn’t the only consideration. As a development economist, I worry a lot about whether the Chinese export juggernaut is currently making it more difficult for other, poorer developing countries to climb the traditional ladder of development (see the recent piece by my friends Shoumitro Chatterjee and Arvind Subramanian, or this by Wolfgang Krieger of Germany’s state-owned development bank). So there are a lot of normative issues unanswered here.

I don’t for a second think that a deeper appreciation of China’s social indicators is going to quiet American or European complaints about Chinese currency manipulation or lead Washington or Brussels to rollover and accept the death of whole industries. Nor should it, necessarily.

From a positive perspective, though, one can understand why Chinese policymakers might look at these numbers and wonder about the wisdom of deviating from the status quo.

In an era of national security rivalry and massive and persistent global trade imbalances, Katherine Tai and others have good grounds to question the econ 101 promise that free trade is always win-win. But there is also no economic theorem guaranteeing that a move toward autarky will be win-win either. Rectifying global imbalances may produce losers, many of them Chinese people who remain poorer than Americans today.

A few weeks ago, my family and I relocated from China to South Africa. So far we love it, but there are tradeoffs. Moving from Beijing’s technological marvels, low crime, and sedate politics, to roughly the opposite here outside Johannesburg, it’s tempting to get a bit cynical about the relationship between democracy and development.

For decades, the Chinese political system rewarded provincial leaders for economic growth. In African politics, by contrast, “neopatrimonialism” is still a pretty dominant explanation for how things work: a veneer of formal legal institutions is combined with executives who wield immense discretionary power. They dole out state resources to reward political allies and punish political enemies, with no incentive to invest in public goods.

Following the news from here in Southern Africa over the last couple years, I’m more optimistic.

Voters reward growth, especially in poor countries

My optimism got a boost with this week’s election news from Zambia, where a strong economic track record appears to have propelled Hakainde Hichilema to reelection. While there are plenty of concerns about the harassment of the opposition during the campaign, tentative results give Hichilema 60% of the vote. HH has become something of a darling in the west for opening up FDI in mining and restoring macroeconomic stability after Zambia’s 2020 sovereign debt default. That recipe was enough to deliver steady economic growth around 4.5 percent since the pandemic slump.

Zambia is less of an aberration than you might think.

Contrary to the view that poor countries’ electorates fail to reward economic growth, the cross-country empirical literature actually finds exactly the opposite. Across democracies in the developing world, Brender & Drazen (AER 2008) found that 1 percentage point faster economic growth over a leader’s term is associated with a 7 to 9 percentage point higher probability of reelection. Strikingly, that effect is totally absent in developed economies.

Growth failures also have consequences

For leaders in southern Africa who lack Hichilema’s strong economic track record, things have not gone so well. As Sishuwa Sishuwa put it for the African Arguments blog, “while Zambians may be poor at choosing good leaders, they are good at removing bad ones.”

It seems Zambia’s neighbors also know how to punish slow economic growth. The graph below plots election margins against economic growth rates over the 5 years in the run-up to elections. A regression line through 8 data points is a bit silly, but if you squint, it does seem Southern Africa fits the global pattern of a positive electoral payoff to growth.

Take Malawi, for instance. Prior to elections last year, Malawi had suffered multiple years of growth that barely scratched 2 percent, punctuated by a forex crisis, queues for fuel, and an El Niño famine in the south of the country. In September, incumbent Lazarus Chakwera suffered defeat at the polls to his longtime rival Peter Mutharika.

In Botswana, elections in 2024 saw the end of the BDP’s 58-year reign, with incumbent Mokgweetsi Masisi losing to Duma Boko as the country’s diamond-driven economy falls to synthetic rivals and falling Chinese demand for luxury goods. As political scientist Bontle Tumediso wrote in African Affairs, “Contrary to the common adage, diamonds [aren’t] forever,” and amid 27% unemployment, voters blamed the BDP for failing to diversify the economy.

The region’s biggest economy, South Africa, has seen particularly dismal economic growth for over a decade, hovering around 1 percent per annum with the exception of the Covid slump and brief recovery. In 2024, Cyril Ramaphosa returned to the presidency, but the ANC lost its majority and was forced into a coalition with the DA.

Angola is closest to a clear exception to this pattern in southern Africa. João Lourenço has some of the hallmarks of Hichilema, but has been less successful in spurring growth. “While foreign investors hail Lourenço as a ‘courageous’ great reformer, hunger, poverty, and popular dissatisfaction are increasing”, noted Jon Schubert and Gilson Lazaro before the 2022 elections. In the end, Lourenço was victorious despite his patchy growth performance. It was, however, his MPLA’s weakest showing in the post-war period, losing 26 parliamentary seats.

This pattern doesn’t always hold under conditions of complete electoral probity

Even autocrats are accountable to somebody, sometimes. Extending the cross-country analysis mentioned above, Burke (2012) confirms a link between growth and regime survival in a sample spanning both democracies and autocracies in the developing world, with no real difference between the two.

In that vein, nobody was shocked that Zimbabwe’s Emmerson Mnangagwa was reelected in 2023, officially with 52.6 percent of the vote. But for its faults, even the World Bank credited the ZANU-PF government for generating “one of the fastest-growing economies in the Southern African Development Community” in 2021, 2022, and 2023.

Tanzania’s elections late last year were marred by violence, and opposition parties were mostly banned. So, yes, take the Tanzania vote margins in the graph with a big grain of salt. Nevertheless, looking at Tanzania’s macroeconomic performance, it should come as no surprise that Samia Suluhu Hassan and CCM – which has ruled since independence – cruised to reelection. Samia achieved that feat on the back of 5 percent growth, putting most regional neighbors to shame. CCM elites seem to be hoping for a Chinese-style political bargain, where civil liberties are held in abeyance as long as they can deliver economic progress.

Madagascar is a messier story, admittedly: incumbent Andry Rajoelina was reelected in November 2023, on the back of stable 4 percent growth (again, with a Covid exception). But less than two years later he was ousted amid Gen Z protests about the long-run economic travails Madagascar has faced – falling gradually behind its regional peers over decades – culminating in a military coup in late 2025.

Punishing failure is not the same as producing success

To avoid overstating the case here, it’s worth noting that none of these growth rates are that impressive. Zambia’s 4.5% growth with 2.8% population growth means a doubling-time of 42 years for per capita incomes. China achieved that in 7 years in the mid-2000s.

Repeatedly punishing failure is also not the same as producing success. Malawi has seen an amazing run of party turnover in the president’s mansion, reflecting in part voter dissatisfaction with the country’s long-term economic slump. But those switches have failed to produce much real change. It remains to be seen whether getting its nose bloodied in the last elections will lead the ANC to make any radical changes, and similarly Duma Boko’s government in Botswana still faces the harsh reality of the declining diamond market.

But there is some small consolation in the signs that voters are paying attention. Electoral politics and pro-growth reforms do not seem fundamentally at odds, even in some of the world’s poorest countries.

“Is there some action a government of Nigeria could take that would lead the Nigerian economy to grow more like an AI-powered United States’ or China’s? If so, what, exactly? If not, what is it about the ‘nature of Nigeria’ that makes it so? The consequences for human welfare involved in questions like these are simply staggering: Once one starts to think about them, it is hard to think about anything else.” — the spirit of Bob Lucas (1988)1

In 1800, Mali’s per capita GDP was roughly 5x smaller than Britain’s. Today it is roughly 25x smaller.2 The Industrial Revolution accelerated growth in the UK and the benefits of the new tech eventually reached the whole world. But the process has taken so long that, two centuries later, the gap is still 5x larger than it was before the new tech arrived.

More and more economists are taking seriously the possibility that AI will cause a rapid acceleration of growth in frontier economies.3 Investors are betting trillions on AI. For their bets to be right, AI companies need to generate even more trillions in revenue, which would imply a big growth acceleration in frontier economies. The economic consequences are serious enough that the Fed just launched a high-profile task force on AI and the economy. So it doesn’t seem crazy to think that AI can really accelerate frontier growth. What happens to Mali this time?

This post argues that Mali, and all non-frontier economies, are likely to be left behind, at least in the initial phase when frontier economies are rapidly becoming more automated.4 The broad argument depends on two simple and, in my opinion, fairly reasonable assumptions:

1. AI accelerates frontier growth. At least in the period immediately after the arrival of Transformative AI (TAI), the growth rate of frontier economies increases. That’s the period I focus on here, and I call it the “detachment window.”5

2. Poor countries will also see a boost to growth — but there will be a lag. For reasons I will discuss below, poor countries lag behind in AI adoption in productive activities, so the productivity increase due to AI arrives in non-frontier economies with a lag.6

Surprisingly (to me at least), these two ingredients are sufficient to conclude that the GDP gap between frontier and non-frontier economies increases in the detachment window. You can see the math in the technical note. The intuition is best presented via an image:

Figure 1. On the flat, the gap is constant; when the leader hits the downhill, the same 30 seconds becomes a growing gap. The frontier is the leader; the slope is accelerating growth; the head start is the tech adoption.

Have you noticed in the Tour de France that when riders reach a downhill stretch, the distance between them tends to open up? That’s because the slope accelerates the bikes, which turns a fixed gap in time into a growing gap in space, even if they are pedaling equally hard. Here the leader is the frontier economy, TAI is the downward slope, the fixed gap is in years for adoption of AI in productive activities, and the gap is in GDP.

The rest of this post presents a simple economic model that helps me justify these assumptions and quantify how much the gap opens up.

The stakes are enormous. In the central scenario, delayed adoption leaves average incomes in the developing world at the end of the window about a fifth of what they would have been with no lag. Each year of faster productive AI adoption in lagging economies is worth about $42.6 trillion in present-value income (more than a third of the world’s GDP), or the equivalent of saving 115 million lives if we adopt Coefficient Giving’s standard methodology.

If you are skeptical of the assumptions, you can change all the parameters in this calculator. You can also download the source code here.

Subscribe now

A simple model of takeoff in frontier economies

Think of the economy as a series of tasks: drive the truck, weld a joint, file taxes, make Doritos Screamin’ Sriracha, sell Doritos Screamin’ Sriracha at Dollar General, etc. Each task can be done by either a person (L, for labor) or a machine (K, for capital), and all tasks are combined using a Cobb-Douglas aggregator (a staple of economics) to get us total output (Y). A bit of math gets us to:

To see the math, check the accompanying technical note.

This is the tasks model, the workhorse of the automation literature.7 Here f is the fraction of tasks in the economy allocated to capital. Since f tasks are allocated to capital, (1 − f) tasks are left to labor. Capital has to perform f tasks, so the total amount of capital available to produce each task is K/f.

I think of an AI-induced growth takeoff in frontier economies as increasing f. This is not something I invented. In fact, I just copied the idea from one of the most cited papers on the economics of AI, Aghion, Jones, and Jones. As AI capabilities grow, machines become more capable of performing economic activities, so more tasks are allocated to machines, and the growth rate in frontier economies increases. In fact, in this model, the higher f is, the higher the growth rate.8

To calibrate the speed of growth we first need to assume a trajectory for f. I assume f follows an S-shaped curve because that’s basically how all technologies spread (Geroski 2000). In the beginning, AI doesn’t exist and doesn’t drive automation. Then it’s invented. Then, as it gets better and better, tasks are automated at a growing rate. As we get closer to 100% of tasks automated, only the harder, more expensive tasks to automate are left. We’ve reached diminished marginal returns and f tapers off.

Maybe f never reaches 100%, as Alex Imas believes. That is, a number of tasks can be permanently left to humans. These can be heavily regulated tasks, like the final steps of getting a green card, or tasks where people have an intrinsic preference for humans, like yoga teachers or sports.

Once we have a shape and a cap, we can calibrate how long it takes for frontier economies to go from the current level of automation to the final level. Call that T.9 My central guess is 15 years, which is the mid-point between optimistic and pessimistic predictions related to the speed of automation of the US economy.10

Finally, to determine the full path of frontier growth, we need to set the speed of growth under full automation. I base that on Korinek and Suh 2024, where a fully automated economy eventually settles into growth of approximately 18% per year (more on this parameter in the last section of the post). The growth rate at the end of the detachment window then depends both on the maximum level of automation and the growth rate under full automation.

Step 1. Falling machine costs push the automation share f up an S-curve, and the model converts that share into a growth rate, side by side below. The key fact: the relationship is log-linear, so equal steps in f multiply growth by equal factors. Drag T:

Every parameter behind these panels is exposed in the full calculator.

A rising tide lifts some boats faster than others

Today, the latest AI models reach poor countries quickly. My friend in Benin uses ChatGPT all the time (though getting advice on crypto trading might not be the most productive AI use-case). So shouldn’t f just grow at the same rate in lagging and frontier countries? Isn’t there a chance that poor countries adopt AI even faster than rich ones? Africa has more young people than any other continent, and they did overtake the US in the adoption of digital payments.

Unfortunately, my friend is just an exception that proves the rule. Using data on AI-related activity on Windows devices, this recent IMF paper finds that LLM use lags behind in poor countries.

And I think that amidst a wave of AI-driven growth in frontier economies, the AI gap between frontier and non-frontier economies will not decrease. That’s especially true of AI adoption in productive activities – that is, in activities that actually generate money, move the automated fraction of the economy, and cause GDP growth. Here are the main reasons:

1. GDP of frontier economies relies more on cognitive tasks, and those will probably be automated first by AI. Increasing f means automating production, which involves both cognitive and physical tasks. AI is currently much better at cognitive than physical tasks. A high fraction of US GDP is in sectors with a high cognitive component, like banking, consulting, law, software, and biotech. Mali has a lot of GDP in agriculture and mining, where physical tasks are relatively more important. So the first wave of automation has far more surface area to work on in the US than in Mali.11

2. Automation costs money. The hard part of AI-driven automation is in physical tasks. Drones that spray fields, humanoid robots, and self-driving trucks are all expensive machines. Poor countries have little money to pay for them. They also face higher interest rates, so any investment project gets more expensive.

   

   Source: Panel A: World Bank World Development Indicators, gross savings (% of GDP) and GDP per capita, 2024 (NY.GNS.ICTR.ZS, NY.GDP.PCAP.CD), retrieved 31 July 2026. Panel B: Aswath Damodaran, Country Default Spreads and Risk Premiums, updated 5 January 2026. Cost of equity in US dollars = risk-free rate in US dollars plus the country equity risk premium, at a beta of 1. Risk-free rate in US dollars = 4.67% US 10-year Treasury on 30 July 2026 less Damodaran’s 0.23% US default spread = 4.44%. Damodaran builds the country premium by scaling a sovereign default spread by the ratio of emerging-market equity to emerging-market government-bond volatility, a factor of 1.52. The spread used here is the traded one, not the one implied by the credit rating: Damodaran’s sovereign CDS premium for the United States, China, Brazil and Kenya, and for Mali, which has no CDS, its own 3-year CFA-franc Treasury bond at 8.17% (UMOA-Titres auction, 18 March 2026) less the 3-year German Bund at 2.56%, the CFA being pegged to the euro. Ratings would put Kenya at 18.4% and Mali at 20.3% instead; they agree with the market to within 0.2 percentage points for the other three. This is a hurdle rate for equity: a firm’s dollar borrowing cost sits below it and the sovereign’s below that.

   In the period of AI-induced frontier growth acceleration, capital will likely flow uphill (from poorer to richer countries), raising interest rates in poor countries relative to rich countries and making it even harder to pay for these machines.12 (Lack of money is less of a factor in AI adoption for cognitive tasks, but it can be an issue. While people in poor countries can download a frontier LLM app for free, the price of tokens starts to bite at scale. In the US, AI-use bills are increasingly too big even for rich companies.)

3. Automation makes more economic sense when wages are high. Suppose a robot that replaces one worker costs $50,000 a year to own and run (after capital costs, maintenance, and power). In Seattle, the worker it replaces costs $65,000 a year, so automation saves the employer $15,000. In Lilongwe, a worker costs $2,000, so automation loses $48,000 a year. This is a stylized example, but the point is just that the higher the labor costs, the higher the economic gain from automation.13

4. Complementary inputs. Automation is more attractive if you have reliable and cheap electricity, internet connectivity, sensors producing data, skilled managers who can reorganize firms, etc. All these complementary inputs are more abundant in frontier economies. China already has a lot of robots making cars, so the next AI-powered robot can slot right in. If you are Zambia and have no car factories, making the AI-powered robot produce cars will require a lot of complementary investment.14

In the model, these reasons can be summarized by a single lag (D).15 That is, the poor country’s automation curve f is the frontier’s curve shifted right by D years. It’s not hard to see why. A task is given to a machine if, and only if, the cost of completing it with a machine is lower than with a human. We can map reason 3 into lower local labor costs in non-frontier economies, and reasons 2 and 4 into higher effective machine costs. Reason 1 affects the composition of tasks rather than the cost of a given task, and mapping it into my simple model would require additions making it significantly less simple, so I leave it out.

We can then conceptually model the arrival of TAI as a phase in which AI makes machines more and more effective at performing economic tasks, which is mapped as a steadily falling cost of completing tasks with machines. This single driving force causes the share of automation f to increase in both frontier and non-frontier economies, as more and more tasks flip from being overall cheaper with humans to overall cheaper with machines. I find it neat that a simple model, grounded in the literature, can generate increasing automation shares and growth in both frontier economies and non-frontier economies, but with a constant lag D separating them (check the technical note for details).

The model also gives a hint on how to calibrate the lag. D equals the machine-cost halving period times the (log of the) country’s cost disadvantage. For example, a country facing a 10x less favorable machine-to-labor cost ratio, with machine costs halving every three years, has a lag of about 10 years.

You can play with machine-cost halving periods and machine-to-labor ratios here:

The best way to calibrate D that I found is to use the historical evidence in Comin and Mestieri. The authors measured adoption lags for 25 technologies across 139 countries, over about two centuries. They find poor countries got the internet roughly 6 years after rich ones, and PCs and cellphones roughly 6 to 8 years after. So the time it takes for a new technology to “arrive” in a poor country is shrinking.

However, they find that the gap in how intensively countries use technologies once they arrive has been widening. The smartphone, for example, took about 15 years to reach half of Sub-Saharan Africa. Smartphones are a relatively cheap consumer good, while “AI in productive activities” entails expensive machinery that requires complementary capital. So my central guess is D = 10 years, which seems like a conservative guess to me, since in the model D is better mapped to the intensity (fraction of tasks automated) rather than the availability of AI for productive activities.

Step 2. Now add the lag. Each economy keeps its own baseline growth (adjustable below) and the follower receives the boost D years late; the orange wedge between the curves is the divergence engine:

Tip: Set the two baselines equal to isolate the delay’s own contribution to the gap. That isolated part is exactly what the calculator prices, which is why its headline numbers compare two paths for the same country rather than frontier vs follower.

To the numbers

Step 3. Price the wedge: the headline numbers and the two income paths, live. Drag D and T:

The calculator below compares two paths for the same developing-country bloc: one where the AI boost arrives immediately and one where it arrives D years later. Both start at today’s average non-frontier income ($4,200 per capita), and I apply to both a 4% discount rate to compute the NPV of the accumulated GDP difference between them.

Here is what the central scenario (takeoff length T = 15, lag D = 10, g_full = 18%, a 25-year window) implies for the 5.4 billion people who live in low- and middle-income countries outside China:

• By the end of the window, average income in the developing world is about 5x lower than it would have been with immediate adoption.

• The present value of the income shortfall is about $1.2 quadrillion. For scale, the entire world currently produces about $110 trillion per year.

• In welfare terms, the gap is worth about 6.1 quadrillion Coefficient Giving dollars, using the CG methodology to value income gains (a 1% income increase for one person for one year is worth CG$500). In welfare units, the total gap is equivalent to saving roughly 1.86 billion lives.

• Each year of faster adoption is worth about CG$381 trillion, roughly 115 million lives in welfare units. In my view, this is the most important number because I believe economic policy can change D and philanthropy can help.

My main takeaway is anything that accelerates developing countries’ catch-up has welfare returns at a scale that’s hard to find elsewhere in philanthropy. Conditional on TAI having an important effect on frontier growth, it’s hard to imagine ways in which the rate of catch-up in the aftermath doesn’t matter enormously. In fact, ensuring that developing countries’ growth rates stay close to the frontier growth rate might be the second most important task in terms of human welfare, after preventing AI-induced catastrophic risks.16

Why care about catch-up growth if TAI will create great abundance and raise everyone’s welfare?

Rapidly growing frontier economies would produce a lot of goods. The price of tradable goods could fall significantly in international markets, and new amazing products could become available. If developing countries can import super cheap electric cars and pills that cure all diseases, why care about economic growth?

That’s because growth still matters for two reasons.

First, more money increases access to those new goods. Even if great AI-produced goods become available, poor countries will have little money to pay for them if their income is low. Increasing their GDP is the best way to improve their access to these great new goods.

Second, this is a counterfactual comparison about relative incomes. The welfare gaps here compare GDP in non-frontier economies with what it could be if the adoption lag decreased. The model suggests that AI-induced takeoff in frontier economies eventually benefits everyone — by an enormous amount — but by not trying to accelerate catch-up, we’d be leaving a huge amount of welfare on the table.

What happens after the detachment window?

So far I have only talked about what happens in the detachment window – but what happens after can undermine the main point I’m trying to make here. My main point is that accelerating catch-up growth during the detachment window matters a lot. An objection that AI-believers could have is that the initial detachment phase doesn’t matter much because it’s short and what follows is quick catch-up, or a post-scarcity future where money becomes a useless concept and income differences stop mattering. If so, the welfare gap we’ve just estimated is just a blip, a rough couple of years on the way to somewhere great for everybody.

Why spend attention, let alone money, on a blip?

To answer that, let’s be concrete about what could come after the detachment period modeled so far. I can imagine four main scenarios:

• Future 1: the plateau. Growth rates stay high everywhere. The income gap freezes at roughly 5x and stays there. We have a new, spectacularly richer world, exactly as unequal as the window left it.

• Future 2: the fade. The AI boost turns out to be temporary, and frontier growth comes back down. Then something nice happens: The follower rides the tail of the boost after the frontier has slowed, growing faster than the frontier for a while, and a temporary boost plus a pure lag creates only a temporary gap.

• Future 3: the singularity. The boost keeps growing and the gap explodes.

• Future 4: the convergence. Developing countries grow faster than the frontier after the window, the way classic growth theory says capital-scarce economies should. The gap closes while growth stays high.

The “it’s just a blip” objection only works in future 4. In future 1, the window’s gap is permanent. In future 2 it closes, but closing it one year early still creates the same amount of welfare we estimated here. At the central calibration, the detachment window is 15 years. That’s billions of person-years lived at incomes several times below what faster catch-up would have delivered.

Future 3 could go either way. A singularity could be the future where the early gap developed in the detachment window matters most, because whoever ends the window ahead may get to write the rules of everything after. In 2021, then co-CEO of Open Philanthropy Holden Karnofsky wrote that a state in which “key things about society, such as who is in power or which religions/ideologies are dominant, are locked into place indefinitely, plausibly for billions of years,” is possible and even likely. The argument is that, if some groups become much more powerful than others when TAI arrives, they might use a temporary advantage to consolidate a permanent position of advantage, or even dominance, over the groups lagging behind. On the other hand, the singularity could create a post-scarcity world where initial income differences become irrelevant. It’s too hard to think about what happens in this scenario, so I skip it.

Future 4 is the world to aim for. It’s the one that maximizes human welfare, and if you were confident we get it by default, the right policy really would be to accelerate AI and not worry about catch-up growth at all, since faster arrival means faster convergence. But look at what future 4 requires: Post-window growth has to be driven by non-rival ideas, things poor countries can copy for free, rather than by capital accumulation, which has owners, price tags, and shipping delays. The entire argument of this post has been that AI’s gains during the transition are embodied in expensive machines. Betting the welfare of five billion people on that not being true seems hard to justify. So let’s aim for fast convergence, but not assume it.

What to do about it

So the trillion-dollar question: Is there anything poor countries, and the people who care about them, can do to catch up faster in the aftermath of TAI?

I think so. Economic policy decisions will meaningfully speed up or hinder catch-up by poor countries after TAI. I won’t pretend I know exactly what policies. But I have some ideas that I will write in a follow up post.

I expect to learn a lot in the next couple of years. My day job is making grants to accelerate catch-up growth, and I’m keen to understand how our grantmaking should change if a frontier AI takeoff materializes (or in anticipation of that) so we can better help countries. I will be talking with policy makers and observing the data, and I will keep you updated.

In the meantime, if you have reactions or ideas, especially concrete ones about what countries should do, please reach out.

Coda: the main ways this could be wrong

I divide the objections here into two camps, the ones that have an AI-skeptic flavor and the ones that have an AI-believer flavor. The main point of the post is that catch-up growth becomes super important in the aftermath of TAI. Broadly, AI-skeptics object because they think AI won’t cause much growth, and AI-believers object because they think AI will automatically benefit everyone. I go into the main reasons I’ve encountered for skepticism on both sides.

AI-skeptic: The AI-induced takeoff never comes. Most mainstream economists put AI’s growth effects at tenths of a percentage point, and if they’re right, this post is just a thought experiment.

Response: On the question of how powerful AI will be, the AI believers’ track record over the past five years has been better than the skeptics’. I think mainstream economists are a bit too skeptical, and I think this is a good joke.

AI-skeptic: The AI-induced takeoff you assume is too high. In the Forecasting Research Institute’s rapid-progress scenario, AI experts forecast roughly 5% annual growth in 2050, while economists and superforecasters forecast less. The IMF’s high-adoption scenario for sub-Saharan Africa estimates a cumulative GDP gain of about 4% over 10 years. Why does your central scenario reach 18%?

Response: Eighteen percent is the growth rate under full automation. The forecasters and the IMF paper are not assuming full automation; they allow for automation to get bottlenecked by regulation and a myriad of frictions. To get closer to those scenarios, you can slide the f_max parameter and reduce the share of automated tasks at the end of the detachment window. I assume full automation as default because I see the exercise here as conditional on accepting the AI-believer point-of-view. In that case, 18% as the final growth rate might be too low. In a recent paper, some of the most prominent economists in the field show that, once machines can produce the next generation of machines and automated R&D, exploding growth rates (unbounded) are possible. So I do not regard 18% as the maximally aggressive scenario.

AI-skeptic: The AI-induced takeoff comes, but it stops at relatively low levels of total automation f because AI only automates cognitive tasks.

Response: I admit that it is a (so far) implicit assumption in my modeling scenario that the AI-induced growth acceleration includes automation of most physical economic activities, so the maximum f is not too low. That is, I want you to take as part of the modeling scenario that robots are advanced enough that they can automate most physical economic activities, from cooking to construction.17 You don’t have to believe that nobody will pay for fancy human-chef meals anymore, just that robots can do physical things almost as well as humans in almost all economically relevant areas.

If we don’t assume that, then the AI takeoff peters out at a relatively low f, the scenario becomes less interesting and the implications for non-frontier economics change a lot (because they can specialize in tradable physical activities, given that they have a lot of relatively young humans). If you are skeptical how much AI and robots can automate in the next 15 years, the only thing I can do now is to promise a future post about robots, and concede that without automation of physical tasks the main point here becomes much less important.

AI-skeptic-believer: D=10 is too low relative to historical evidence, and maybe TAI itself causes the lag to grow. Rasmus Andersen, author of the quite interesting AGI Growth Atlas, commented on a draft of this post:

“The model says followers are on the same S-curve shifted by D years. But the dominant historical outcome in many ways was non-catch-up: your own opening (Mali, 5x to 25x) is not the exception, and you can argue full adoption of second and third industrial revolutions still hasn’t happened (electricity and effective ICT use remain incomplete across much of the world two generations on). Historically the four reasons you listed didn’t produce a fixed lag; they produced follower curves that many times didn’t complete.

The few who did reach the frontier used low wages as the entry ramp into tradables, reinvested earnings as physical capital and eventually made it there. Foreign capital generally came later, I would argue. Given TAI may foreclose that path to capital accumulation outside the frontier by automating the low-wage tasks themselves, doesn’t that mean catch-up and convergence in adoption are even less likely this time around?”

Response: In the model, the follower is assumed to trace the same growth path as the frontier, only shifted. This implies that, after D years, even Mali will reach the same level of automation as the US and get the same boost to growth.

This is probably too optimistic. Persistent differences in infrastructure, energy, conflict, or institutions could instead leave poorer economies with a lower automation endpoint or a lower growth rate even after AI-induced automation has run its course.

A way to incorporate this in the model would be to set a lower growth rate after maximum automation (g_full) or lower maximum rate of automation (f_max) to poor countries. These changes would make the main point of the post stronger, and I wanted to keep things simple and not create more free parameters.

AI-believer: TAI causes the lag D to decrease. Maybe AI will diffuse to poor countries faster than other technologies, and maybe AI itself will help increase the speed of technological diffusion. The main reason I can imagine for this is lowering technical barriers for complex economic production. If a firm has trouble operating a chemical plant in Nigeria now for lack of chemical engineers, Claude 5.1 might dissolve this barrier.

Response: I’m actually hoping that non-frontier economies do everything they can to lower the adoption lag, and hoping to work in and help fund initiatives aiming at that. I think AI itself lowering technical barriers for running complex economic activity is a reason for hope. But I don’t think it’s strong enough to overcome the four barriers I listed above.

AI-believer: Trade boosts catch-up. A big missing ingredient in the model is trade. AI adoption in frontier economies can increase the demand for non-frontier exports a lot, which would accelerate their growth.

Response: I think this is real but not enough to meaningfully decrease the gap. In a scenario where the AI-induced frontier take-off includes the automation of physical production, manufacturing becomes more competitive in frontier economies than in non-frontier ones. In fact, China (which for our AI-adoption purposes I classify as frontier) is already rapidly automating manufacturing, and I don’t see why Ethiopia’s garment industry would become more viable after TAI than it is now.

I think the exports that will see more demand are extractives (especially the ones used to build the machines doing the automation, or the energy to power them) and tourism. Non-frontier economies should try to maximize this opportunity, but the opportunities are likely to be narrow and to require clever economic strategy – we shouldn’t just assume that trade automatically causes catch-up.

AI-believer: Regulatory barriers in frontier economies reduce the gap in adoption. We can imagine that relatively strong governments, unions and pressure groups could lead to high and well-enforced regulation against AI in frontier economies. If non-frontier economies have weaker regulatory and regulation enforcement capacity, these barriers could be lower in them, which would compress the adoption lag D.

Response: This is a plausible argument, but it doesn’t seem strong enough to overcome the four barriers I listed above. Frontier economies would really have to regulate away AI use in the economy to overcome the barriers, and competition between US, China, and Europe has been pushing them to race and suppress regulation. I expect this race dynamic to continue well into the detachment period. Even if rich democracies regulate AI a lot, the Gulf countries, or richer middle-income countries like Russia, would be in a better position to automate their economies than the truly poor countries.

Learn more

• Full calculator

• Technical note

• Comin and Mestieri (2018), the main source for the adoption-lag calibration

I’d like to thank Oliver Kim, Justin Sandefur, Joseph Levine, and Rasmus Andersen for very thoughtful comments.

Bob Solow’s original 1956 formulation of the neoclassical growth model predicted that poorer countries would grow faster than rich ones. Capital flows—from rich to poor— would be the great leveler. Three decades later, the advent of the Penn World Tables gave economists access to comparable cross-country data to test that hypothesis rigorously. It proved false. From the 1960s through the 1990s, the opposite was true, per capita GDP levels were not converging as predicted—and if you rolled the clock back even further in history, it was obvious rich countries had been getting richer for quite a while.

But then something changed. No sooner had Lant Pritchett announced “Divergence: Big Time” in 1997 than developing countries, for the first time in two centuries, started catching up with the living standards of advanced economies. Belatedly observing this phenomenon, we wrote a fairly giddy piece in 2018 declaring that “Everything You Know about Cross-Country Convergence Is (Now) Wrong.” Convergence had finally arrived.

Alas, the facts have changed again. In this post, we return seven years later to extend our earlier results to the latest data. Just as we heralded the new era of unconditional convergence, as we called it in the journal-length version of our piece, reality came back to undermine our optimistic finding. It turns out, everything you once knew is right again. Soumaya Keynes noted signs of this (second) reversal of fortune last year in the Financial Times. And of course the growth slowdown in emerging markets has been well-documented, as in this note on divergent post-Covid recoveries by Eduardo Olaberria and Carmen Reinhart and this piece by Jonathan Hartley. In sum, unconditional convergence appears to have come to an end – at least by the metric originally used to test Solow’s hypothesis in the 1990s, and it appears to be slowing down although not necessarily reversing by other measures too.

The harder question is, why? Our answers are pretty speculative at this point, but first, the basic facts.

Everything we told you about convergence is (now) wrong

The figure below presents our headline result. We use the newest releases from three different sources of income data: the Penn World Tables (PWT), the World Bank’s World Development Indicators (WDI), and the Maddison series.

Following the same non-linear specification as Barro and Sala-i-Martin (1992) and as our own earlier paper, each dot captures a convergence coefficient for per capita income, essentially just regressing growth rates on initial incomes. The time period varies along the horizontal axis, always covering a 10-year span. In this formulation, positive numbers mean that poor countries are catching up to rich ones, or in other words, that the lower your initial income, the faster you grow.

The original story of divergence then convergence remains the same, despite the scope for the revised data to alter results. For the three decades leading up to the empirical literature on economic growth, the gap between developed and developing nations expanded.

That all started to change around the fall of the Berlin wall, and poor countries went from lagging behind richer ones to maintaining pace with and then ultimately to growing faster resulting in convergence.

The latest data updates show that that golden era of relative prosperity has come to a halt. In the past decade and a half, there is no evidence that poor countries have grown faster than richer ones.

The share of developing countries growing faster than the G7 is falling fast

So far we’ve shown that poorer countries only grew faster than richer ones on average for about 20 years. But that is not the only way to think about the notion of economic convergence. Roy et al. (2016) proposed an alternative metric of convergence that focuses less on averages, drawing inspiration from Oscar Wilde: “We are all in the gutter but some of us are looking at the stars.” "Wilde convergence”, as Roy et al call it, simply measures how many poor countries are catching up with the frontier, namely the richest countries.

That share remains quite high, though it is falling fast.

To see this, define the frontier as the average growth rate among the G-7. Also, because World Bank definitions change arbitrarily over time, we categorize countries as "low income" or "middle income" if they fall below the 25th and 75th percentile, respectively, of U.S. income in each year.

The results are striking. Prior to 1990, all three data sources show that less than half of developing countries are exceeding this frontier growth. But in the same period that Solow-convergence begins, “Wilde-convergence” takes off with a fury. By 2000, nearly all low- and middle-income countries grow faster than the G-7. While the last 15 years have exhibited a decline in this rate, nearly three out of four developing countries still are achieving gains in income above the rich-country frontier.

This poses a puzzle: if 70 percent of poor countries around the world are growing faster than the G7, why are we not seeing Solow convergence? One answer is that although the number of developing countries still converging to the rich world is high, the pace of catch up has declined sharply. At the peak, Asia outpaced the G-7 by over 4 percentage points annually which declined to just below 3 percent recently. In contrast, countries in Africa and Latin America that were outpacing the G-7 by nearly 2.5 percentage points are now doing so at a much slower pace, resembling the period of divergence. It is slowing growth in these 2 regions that accounts for the demise of Solow convergence.

Another part of the answer is that Solow-convergence is an average result across all poor countries scaled by initial income: if some middle income countries grow faster than low income countries convergence could be obscured, even though both grow faster than rich countries. Indeed, the data show middle income countries grow faster than either other group – a middle income “trampoline” rather than a trap – which dampens Solow convergence, as shown by the 2000s spike among the black diamonds representing middle-income growth rates in the graph below.

A third explanation is that rich countries are growing faster than in the 2010s so that the average gap between rich and poor as a whole has declined.

The figure below illustrates both the second and third of these mechanisms, showing the raw data on growth between the start year on the x-axis and the latest available year. They illustrate that after the mid-2000s, global financial crisis growth in low and middle income countries declined while still substantially exceeding growth in high income countries. But after the 2010s, the decline has been dramatic, which combined with a slight improvement in HIC growth rates has led to the elimination of convergence. What happened?

Why did convergence slow down?

We'll leave a detailed econometric investigation of that question for another day, but we can't resist a couple of preliminary conjectures. In a Foreign Affairs piece last year, we argued that "hyperglobalization" was a key driver of convergence.

“In the era of rapid growth, nearly every country expanded trade. Most famously, China and India—followed by Vietnam and Bangladesh—experienced economic-growth miracles on the back of rapid growth in their exports and trade, in manufacturing in the case of China and East Asia and in services in the case of India. Commodity exporters, especially in sub-Saharan Africa, benefited from a surge in commodity prices, itself induced by China’s rapid growth and its voracious demand for oil, copper, iron, and other minerals.”

With hyperglobalization stalling, especially for low and middle income countries, it is not surprising that so too has convergence.

Convergence was accompanied by rapidly rising trade (hyperglobalization) until around the 2008 Global Financial Crisis. Thereafter, trade as a share of GDP has either levelled off or declined in both low- and middle- income countries. This decline was dramatic for some larger developing economies such as China (which say a 20 percentage point from 2008 to 2023) and Indonesia (a 17pp drop). Indeed, high income countries do not seem seriously affected by the end of hyperglobalization which seems to have disproportionately affected developing countries.

But is there something much bigger going on? In that same piece, we argued that another key driver of convergence was a turn toward markets but more fundamentally that a majority of developing countries began to follow a variant of U.S. President Barack Obama’s famous foreign policy dictum: “Don’t do stupid shit.” The question is whether that has now been reversed. We do not have the space for identifying a long list of what would constitute such a reversal but a more common global change in institutions, especially political ones, seems evident in the past decade. The world has become more illiberal, majoritarian and nationalistic, reflected in perceptible democratic backsliding. One commonly cited indicator of democracy, the V-dem index, has stopped the gains it enjoyed at the onset of the Solow-convergence era, with particularly marked declines in the middle- income countries that have seen the sharpest decline in growth and globalization.

Of course, the links between weak political institutions/illiberalism and economic outcomes are tenuous at least over short horizons, but there are connections that are striking. In China, the nationalism and illiberalism under Xi Jinping has been associated with a crackdown on private sector entrepreneurship that produced frontier technologies and services. In India, the weaponization of the state against political opposition has also been directed at domestic and foreign investors, leading to elevated investor risk; economic nationalism has led to promoting national champions at the expense of investment more broadly. In Turkey, defiance of macro-economic orthodoxy has been a handmaiden of political authoritarianism. In Ethiopia, a Nobel peace prize and an industrialization boom gave way to a bloody civil war and a fiscal crisis. Across the Sahel's "coup belt", military takeovers in Mali, Niger, and Burkina Faso have come with economic volatility, withdrawal from regional trading blocs, and fiscal contraction in the face of lost aid revenue. Not all of these connections are causal, but a combination of declining trading opportunities externally and deteriorating institutions domestically is a cocktail that is not conducive for long-run economic growth.

The data series we analyze here stop in 2023, well before Donald Trump's "liberation day" ratcheted up tariffs in the world's largest importer to levels not seen since the Great Depression. In the wake of that tariff announcement in April, the IMF quickly revised its global growth forecast for 2025 down by about half a percentage point. But the immediate impact of this policy shock on income gaps between rich and poor countries is technically ambiguous: the IMF's downward revision for advanced and emerging markets was the same, and largest for the U.S. at -0.9%. Nevertheless, if we're right that hyperglobalization and the spread of liberal democratic norms were catalysts for economic convergence between the fall of the Berlin Wall and the Global Financial Crisis, we see little reason to hope that the world's poorest countries will fare well in this new era of political and economic nationalism. A falling tide is unlikely to lift the flimsiest boats.

--

Download replication code from GitHub.

Summary: the U.S. and the World Bank are putting more and more foreign aid money into private sector deals: for-profit investments designed to stimulate economic growth. But most of that money gets absorbed by banks in the BRICs, very little goes to poorer countries, and almost none of it to manufacturing or exports. As this style of foreign aid grows, the focus on industrialization is declining. If development finance institutions are serious about economic growth, they need to rediscover export discipline.

Since Donald Trump returned to Washington, the hottest trend in foreign aid is making deals. USAID’s model of doling out free HIV drugs and emergency food relief to poor countries has been DOGE’d. Meanwhile, Congress is set to grant new authorities to the U.S. Development Finance Corporation (DFC), which directs foreign aid to companies, especially American companies, looking to turn a profit overseas.

A block down the road from the White House, the World Bank – eager to stay in the good graces of the new administration – has doubled down on a similar strategy. Five years ago the bank's private-sector divisions (IFC and MIGA) did about $14 billion in new business, comprising about a fourth of the bank's portfolio. Last year that number hit $40 billion, and at the current pace, private sector investments will eclipse the World Bank's traditional sovereign lending portfolio in a few years time.

The result of all this is an interesting hybrid of foreign aid and industrial policy, combined with a Wall Street-style compulsion to make money and "do deals."

In theory, that's not all bad. Poor countries need to grow, and for that they need investment. If rich countries are embracing industrial policy at home – with subsidies for chosen firms, and public money taking equity stakes in private companies – why shouldn't foreign aid do the same for poor countries?

The problem is that the money is not being used to industrialize poor countries. Both the verb and the object in that last phrase are wrong. Instead, it's being invested in the banking sectors of places like Mexico and Turkey, i.e., financial intermediation for the domestic service sector (not industrialization) in middle-income BRICs countries (rather than the poorest countries in need of jobs). The result is industrial policy without any industry, and foreign aid with surprisingly few poor people involved.

Historically, poor countries have grown rich by exporting labor-intensive manufactures 

From the 1950s through the 1980s several East Asian economies – South Korea, Japan, Taiwan, and the city-states of Hong Kong and Singapore – achieved unprecedented economic growth rates on the back of export-oriented manufacturing. In the decades that followed, China repeated their feat on a monumental scale.

These economies didn’t just export manufactures. As Joe Studwell documents in his bestseller “How Asia Works”, these "Asian Tigers" structured their financial systems to direct cheap capital to firms that successfully exported. This model of carefully targeted subsidies – designed to prioritize making stuff, not real estate booms and financial bubbles – stands in stark contrast to the scattershot approach taken by Washington today.

So what is it about manufacturing, specifically, that made it central to the East Asian growth experience – not to mention almost every major development success story from the British Industrial Revolution to Bangladesh’s 21st century boom?  

Let’s look at three elements that distinguish manufacturing, and then later we can judge the World Bank's current IFC private-sector portfolio on these same metrics:

1. Labor intensity (which I'll define as workers employed per million dollars of output, L/Y)

2. Tradability (defined as the value of exports as a share of output, X/Y)

3. Productivity growth (the historical rate of growth in value added per worker, Δln(VA/L))

While it’s a bit dated and doesn’t cover today’s poorest countries, the World Input-Output Database (WIOD) provides data to estimate each of these dimensions at the sector level for some larger middle-income countries, essentially BRICs+: Brazil, Russia, India, China, Indonesia, and Mexico. (Here I’m collapsing data from ISIC Rev. 4 industries to the 9 sectors listed in the IFC data for comparability).

In terms of job creation, “agribusiness and forestry” is the most labor intensive sector in the WIOD data, employing about 380 people per $1 million of output, while finance is the least labor intensive, at 31 workers per $1 million. Manufacturing is sort of middle-of-the road here at 61 workers. 

Where manufacturing really stands out is on the other two dimensions: historical growth in labor productivity and export orientation. The sector that performs the worst on all these metrics is the sector where the IFC devotes the largest share of its investment: finance.

Compare that to where the World Bank puts its private sector investments

Let's start with a positive example that bucks the overall trend here: the exception that proves the rule.

In 2024, the World Bank’s private sector arm, the International Finance Corporation or IFC, made a $15 million loan to Star Garments, an American-owned, Sri Lanka-based garment manufacturing company to open up a new “cut-make-trim” factory outside Lomé, Togo. 

Lots of economist would tell you that that kind of export-oriented manufacturing is, historically, the best way for poor countries to graduate up to higher levels of development. Star Garments is private-sector development finance done right: Togo is a low-income country, garment manufacturing is highly labor intensive, and the factory will export products to high-income markets.

The problem with the IFC investment in Star Garments is not that it has any secret flaw, it's that it is almost literally one of a kind: one of only a handful of cases globally of an IFC investment in export-oriented manufacturing in a low-income country in the last five years: $15 million out of a portfolio of hundreds of billions of dollars. 

The problem is two-fold:

1. Geographic: Development finance institutions like IFC and DFC make very few investments in the places that really need subsidized capital, i.e., low-income countries. So the reallocation of aid money from traditional public sector projects to "blended finance", private-sector oriented models means a reallocation away from poorer countries toward wealthier, faster-growing emerging markets.

2. Sectoral: In the quest for financial returns, these institutions also focus heavily on financial sector development, neglecting investments in the real economy that will be necessary for economic growth.

While there are some older counter-examples –a slaughterhouse in Ethiopia in 2019, a mango plantation in Malawi in 2014, some cashew processing plants in Guinea and Mozambique from the 1990s – these are rare exceptions in an overall portfolio heavily weighted toward banking.

Over time, development finance is focusing less and less on jobs, growth, and exports

I'll continue to focus on the World Bank's IFC here, because it's the big player, but the same exercise could be done for the US DFC or the UK's BII.

To calculate the labor intensity, growth orientation, and export propensity of the IFC portfolio I take the allocation of IFC investments across sectors and combine that data with the sector statistics above from the average BRICs+ country. Imagine the two charts above as matrices. In the first matrix, the rows are industries (j=1…J), the columns are countries (k=1…K), and the values are dollar amounts. In the second matrix the rows are again industries, but the columns show industry characteristics (l=1…L) like labor intensity. If we multiply those two matrices together, we get a JxL matrix describing the implied labor intensity, growth orientation, and export propensity of the IFC’s portfolio in each country. We can do the same thing with time periods instead of countries, and repeat the exercise each year. 

The pattern that emerges is disconcerting.

Over the last 20 years, from 1996 to 2025, what we see is a remarkable decline in what I’d argue is the developmental- or structural transformation-focus of the IFC portfolio. The implied labor intensity of IFC investments has declined from about 92 workers per $1 million in output in 1996 to about 72 in 2025. Whereas the IFC used to invest in sectors that had historically grown at nearly 6% per annum in the BRICs benchmark countries, that figure has fallen to nearly 3% annual growth – and an almost identical trend is observed for the ratio of exports to output in the sectors IFC chooses. 

As the IFC portfolio has grown – and it has grown dramatically over the past five years – any focus on labor intensive, export oriented, fast-growing sectors has not really kept up.

So far this statement is based on implied performance, applying BRICs+ parameters to the IFC portfolio. As an alternative, I also construct a direct measure of export performance for the IFC portfolio by scraping the IFC disclosures on the web for any mention of exporting (and then humanly checking a sample, weighted toward the larger projects). 

The picture confirms how unique the Star Garments story is: manufacturing projects mentioning exports are a small, and declining share of the overall portfolio. If you restrict the sample to low-income countries, in many years there are zero investments in manufacturing, and other times one or two. Never mind manufacturing much less exporting: low-income country projects as a whole remain pitifully small compared to IFC’s overall booming portfolio. 

Admittedly, things look a little better for lower-middle income countries, where manufacturing investments reached about $1.6 billion in 2024, of which about $700 million were in export-oriented firms – though these were largely in low-value added industries like cement and fertilizer with exports to regional rather than international markets.

Don't banks create jobs too?

Skeptical readers who are still reading at this stage are perhaps mentally screaming that all of these numbers miss the point. The IFC doesn’t lend to banks to create more jobs as bank tellers. The goal is often very explicitly to enable local banks to lend more to the small- and medium-enterprise sector. 

This is reasonable on one level: the IFC probably shouldn’t be directly involved in the retail banking business, making micro loans to SMEs in Tanzania from its headquarters in Washington. But without denigrating this good-faith effort at local economic development, it does leave open two questions.

First, there is a question of additionality and impact, even on the metrics IFC has chosen for itself. Is cheap capital for a few Mexican banks actually lowering interest rates and expanding credit access for Mexican SMEs? Or is that money getting swallowed up by bank profits? There’s very rarely any evidence provided to answer that kind of question: IFC doesn’t routinely do impact evaluation of the sort other development actors do.

The World Bank's own Independent Evaluation Group has repeatedly bemoaned this failure to demonstrate that "blended finance" – i.e., this model of lending to banks, to lend on to specific target groups like SMEs or women-owned businesses – has any development impact (see reports in 2019, 2021, and 2024), leading to some grumbling from the bank's board. A recent review of the academic literature on the topic noted, "evidence on the impact of these public programs intermediated via commercial banks is limited and rarely extends to estimates of the impact on final beneficiaries."

Second, from a development strategy or industrial policy point of view, small is not beautiful. More prosaically, it’s unclear that SME credit is the path to growth and structural transformation for the middle-income countries which are the IFC’s main clients. There is a large literature documenting the positive relationship between firm size and exporting. In many developing countries, the SME sector is a low-productivity, low-earnings job of last resort. If the IFC’s clients want to pursue outward-oriented industrialization, they’re going to need some big firms with the wherewithal to navigate international markets. Alas, IFC is not financing that kind of project.

What's the constraint? Profit incentives or a missing pipeline of projects

Earlier this year I attended a Chatham House discussion sponsored by British International Investment (BII) with people from various development finance institutions. Per the "Chatham House rule" I can't quote anybody in particular. But the entire premise of the discussion was that the supply of development finance can't find enough demand. This is often referred to as the missing "pipeline" of bankable projects in poor countries.

It is odd to hear people whose job is to give away cheap money complaining that nobody wants it, but the complaint is sincere. Many officials at the IFC, BII, or the U.S. DFC would jump at the chance to back an export-oriented manufacturing venture in (low income) Mozambique or even (lower-middle income) Tanzania. 

This is where profit incentives, for development finance institutions as a whole and for the investment officers within them, militates against long-run development strategy. A recent paper by Sean Cole, Martin Melecky, Florian Mölders and Tristan Reed draws on confidential IFC data to calculate the commercial return to its investments over multiple decades. They find that even after adjusting for implicit management fees, “over the long run the IFC’s portfolio of emerging market private equity has done about as well as the S&P 500 index.” But they also show significant differences between sectors. In short, mining, oil, and gas yield super high returns, financial institutions and infrastructure do pretty well, and manufacturing, agribusiness, and services are at the bottom of the pile.

Figure: Manufacturing is less profitable for the IFC than banking or mining

Source: Cole, Melecky, Mölders, and Reed (Management Science 2025)

Many development finance professionals would respond to everything I’ve written above: “yes, we’d love to lend to an export-oriented agro-processing sector in Malawi, or the manufacturing sector in Ethiopia. But there are no bankable opportunities there. We have an obligation to make money.”

My guess is that Taiwanese or Korean policymakers in the 1960s could’ve said something similar. The private sector didn’t want credit to build manufacturing export businesses. They wanted to buy banks, and build apartment buildings, and open retail import businesses. And when they did build manufacturing plants, they wanted tariff protection to produce import-competing goods for the domestic market. 

This is where export discipline came in. Forward-looking policymakers did not surrender development strategy entirely to market signals: they restricted their subsidized credit to projects they felt were in the long-run developmental interest of their countries.

Fixes: bring back export discipline

The challenges facing the private-sector development finance industry aren’t new, but they’re increasingly urgent as finance swallows up more and more of the global development sector. Let me end by quoting from the list of policy recommendations from my friend Charles Kenny’s 2021 essay on how to get IFC back in the development game, which anticipates many of the arguments made above: 

• “Shift the portfolio mix in [poorer] countries towards investment in tradeable manufacturing, services, agriculture and agro-processing, alongside infrastructure and services prioritized on their contribution to trade competitiveness and global value chain integration. [...]

• “Move further toward tolerance of failure by de-emphasizing returns as a project success measure and including incentives for groundbreaking deals in new sectors/ industries alongside development of metrics that encourage rapid exit from failing approaches or investments.

• “Use subsidies solely in the selective, competitive, support of investments proactively chosen based on potential impact on sectoral development outcomes.”

In short, Washington's new model of development finance needs export discipline. If subsidies for private investors are going to contribute to development, they need to target the industries that are going to drive growth and job creation. Sometimes that will be in tension with maximizing short-run financial returns. So be it. The whole point of public subsidies is to nudge the private sector to do things that are socially optimal even if the market doesn't sufficiently reward them.

There's nothing wrong with development finance helping investors make money. But it also needs to help poor countries make stuff.

--

Download replication code for tables & figures from GitHub.