The Great Debates: Can Economies Be Planned?

Follows: Capitalism: How Markets, Prices and Private Property Organise an Economy · The Mixed Economy · The Political Spectrum, Fairly Presented

In this deep dive

1. The question, and why it never went away
2. The socialist calculation debate, round by round
3. Try it: a toy market against a toy plan
4. Incentives and information: what planning actually did
5. The natural experiments, and their caveats
6. Inequality and growth: Piketty and his critics
7. Markets and the environment
8. Can computers plan now?

Part 1 — The question, and why it never went away

Every economy has to answer three questions: what to make, how to make it, and who gets it. A market economy answers them through millions of separate decisions guided by prices; a planned economy through a central authority that sets targets, allocates materials and often fixes prices. Every real economy uses some of both: inside every large company there is planning, and around every planned economy there have been black markets.

So the real argument is not whether anyone should plan. It is how much can be planned, and by whom. Can a central authority, with enough information and good intentions, do better than markets at producing what people need? Or is there something about an economy that no planner can know?

The earlier pieces in this series laid the groundwork. Capitalism set out Friedrich Hayek's argument that prices carry knowledge no planner can gather; Socialism: The Many Kinds, From Utopians to Social Democracy and Communism: The Idea, the States, and What Happened set out the case for collective ownership and what happened when it was tried; The Mixed Economy traced how rich democracies settled on a mixture. This piece takes the arguments head-on: the theory first, then a toy model you can run, then the evidence from history, then the two big modern questions (inequality and the environment), and finally the question that computers have reopened.

One warning: the evidence is mostly natural experiments, not controlled ones, and every natural experiment has confounding factors. Where the evidence points clearly, this piece says so. Where it doesn't, it says that too.

Part 2 — The socialist calculation debate, round by round

Round 1: Mises throws down the gauntlet (1920)

In 1920 the Austrian economist Ludwig von Mises published an essay in a German social-science journal, "Die Wirtschaftsrechnung im sozialistischen Gemeinwesen", known in English as "Economic Calculation in the Socialist Commonwealth". Its argument was short and sharp.

In a market economy, a factory manager deciding whether to build with steel or concrete can compare their prices. Those prices summarise how scarce each material is and what else it could be used for. In a socialist economy where the state owns all the factories, mines and land, nobody buys or sells them. So there are no market prices for the means of production, and no way to compare the costs of different ways of making something.

"Where there is no free market, there is no pricing mechanism; without a pricing mechanism, there is no economic calculation." — Mises, 1920

Without calculation, Mises concluded, a socialist economy could not tell whether a project created or destroyed value. "There is only groping in the dark. Socialism is the abolition of rational economy."

Mises was not the first to see the problem. In 1908 the Italian economist Enrico Barone had shown that a planning ministry could, in principle, reach the same result as a competitive market by solving a vast system of equations, but he thought solving it in practice impossible. Mises went further: the problem was not difficulty but impossibility, because without markets the data the equations need would not exist.

Round 2: Lange and Lerner fight back (1936–1944)

The most famous reply came from the Polish economist Oskar Lange (1904–1965), then teaching in the United States, in a two-part article "On the Economic Theory of Socialism" (Review of Economic Studies, 1936 and 1937). Lange was a committed socialist, but he used the tools of mainstream economics, and he began by thanking his opponent. Socialists, he wrote, had "good reason to be grateful to Professor Mises, the great advocatus diaboli of their cause", and he suggested that "a statue of Professor Mises ought to occupy an honorable place in the great hall of the Ministry of Socialization or of the Central Planning Board".

Then he proposed what came to be called the Lange model, or market socialism:

1. The state owns the major industries. Consumer goods and jobs are still bought and sold, so households choose freely.
2. A Central Planning Board sets the prices of producer goods. Managers are told to follow two rules: produce at the lowest cost, and expand output until the price equals the cost of making one more unit.
3. The board then adjusts prices by trial and error. If stocks of a good run down (a shortage), it raises the price; if they pile up (a surplus), it lowers it.

This, Lange argued, is exactly what a market does, but with the board in the role of the auctioneer. The prices are real numbers that can be used for calculation, so Mises's impossibility disappears. The British economist Abba Lerner developed the same rules, and in The Economics of Control (1944) argued they could guide any economy, socialist or not. Many economists at the time thought Lange had won.

Round 3: Hayek changes the question (1940–1945)

Friedrich Hayek replied in "Socialist Calculation: The Competitive 'Solution'" (Economica, 1940) and, more famously, "The Use of Knowledge in Society" (1945). Capitalism in this series explains the 1945 essay. The point that matters here is how it answered Lange.

Lange's model, Hayek argued, assumed the board could know the "data": what each factory could produce, at what cost, and what people wanted. But the knowledge that matters is mostly "the knowledge of the particular circumstances of time and place": a machine sitting idle, a surplus of stock in one warehouse, a skill not being used, a shortage developing somewhere. Such knowledge "by its nature cannot enter into statistics and therefore cannot be conveyed to any central authority in statistical form." It is known only to the "man on the spot", and it changes daily.

Hayek also asked practical questions. How often would the board revise millions of prices? Who would check whether a manager really was producing at the lowest possible cost? What would make managers want to? And where would new products and new methods come from, if nobody could risk their own capital on an untested idea?

Round 4: Kornai looks at the real thing

For decades the debate was mostly theory. The Hungarian economist János Kornai changed that by studying how socialist economies actually behaved. In Economics of Shortage and later The Socialist System (1992), he argued that chronic shortages, queues and hoarding were not accidents of bad planning but built into the system. His most influential idea was the soft budget constraint: socialist firms knew that if they lost money they would be rescued with subsidies, cheap credit or forgiven debts, rather than closed. So they had little reason to control costs and every reason to demand more inputs than they needed.

This cut against Lange: his board needs managers to follow its rules honestly, and Kornai found that real managers, facing soft budgets and political bosses, did not.

Round 5: the theorists find a common language

The debate had a remarkable second life inside economics itself. From 1960, Leonid Hurwicz began treating any economic system, market or plan, as a "mechanism": a set of rules by which people send messages (bids, reports, orders) and an outcome follows. The question becomes which mechanisms get people to reveal what they know and act on it. Hurwicz, Eric Maskin and Roger Myerson shared the 2007 Nobel prize "for having laid the foundations of mechanism design theory".

Mechanism design gave both sides something. The Nobel committee noted that some results "support Friedrich Hayek's (1945) argument that markets efficiently aggregate relevant private information". But the same theory shows where markets fail, as with public goods such as clean air, where it is "in the selfish interest of each person to give false signals" (Paul Samuelson, 1954). It also shows that a planner's problem is not just computation but incentive compatibility: designing rules so that people want to tell the truth.

The Soviet mathematician in the table deserves a note. In 1938, while a professor at Leningrad University consulting for the laboratory of a plywood trust, Leonid Kantorovich worked out how to allocate materials between machines to get the most output, and in 1939 he published the main ideas of what is now called linear programming. His methods produce, as a by-product, prices that measure how valuable each scarce input is to the plan. He shared the 1975 Nobel prize with Tjalling Koopmans "for their contributions to the theory of optimum allocation of resources". Planning, in other words, gave the world one of the most useful tools in economics. Airlines, delivery firms and factories all use it today.

Mises said planners couldn't calculate. Lange said they could copy a market. Hayek said the problem was knowledge, not arithmetic. Kornai said the problem was incentives. All four were partly right.

Part 3 — Try it: a toy market against a toy plan

A small simulation can show what the argument is about, and also what it leaves out. Below is a town that eats about 1,000 loaves of bread a week. Demand wobbles from week to week, and in two scenarios something big happens in week 13. You can run the same year under three systems:

• Free price. Bakers decide how much to bake from last week's price. The price then moves until buyers take exactly what was baked.
• Fixed price and quota. A planner sets the price at 5 coins and the output at 1,000 loaves a week, the right numbers for an ordinary week, and leaves them there.
• Planning board (Lange). The board sets a price; bakeries are told to bake where that price equals their cost of one more loaf. The board watches the queues and the unsold stock and revises the price every 1, 4 or 13 weeks.

What the toy shows

Try the demand jump first. These are the results for that year, computed by the model:

Toy model, demand-jump scenario, 52 weeks. "Queued" counts loaves people wanted at the going price but could not get.

Three things stand out.

The fixed quota can't respond. When demand jumps, nobody in the system has a reason or a signal to bake more. Queues appear in 49 of the 52 weeks and add up to more than 12,000 missing loaves. This is the shortage economy Kornai described, in miniature.

The free price responds, but not gently. In week 13 the price leaps from about 5 to about 8.5 coins. That high price does two jobs at once: it tells bakers to bake more, and it rations the bread that exists to the people most willing to pay. Then the market overshoots: in the bad-harvest year the price crashes to about 3.7 coins when the harvest returns, because bakers had over-expanded. Markets are not smooth. Economists call this back-and-forth the "cobweb".

The Lange board does almost as well, if it is fast. With weekly reviews it bakes nearly as much as the market and keeps queues small. That is Lange's point, and in this toy it holds. With a review every 13 weeks it does much worse. The real-world question is how fast a board responsible for millions of prices could actually move.

Now notice what the toy leaves out, because that is Hayek's reply. The model hands the board the true cost of baking one more loaf. In reality the board would have to ask the bakeries, and Kornai's managers had every reason to overstate their costs. The model has one product; a real economy has millions. The model's bakers never invent a better oven or a new kind of loaf. And the model's "free price" rations by willingness to pay, which means the poorest families are the ones priced out. A queue rations by time and luck instead. Which is fairer is a question of values, not of the model. Many real economies use both: market prices for most bread, plus a benefit that makes sure everyone can buy some.

A toy can show that prices carry signals. It cannot show who knows the curves, who has an incentive to lie, or where tomorrow's bread comes from. That is where the real debate lives.

Part 4 — Incentives and information: what planning actually did

The Soviet Union ran the largest planned economy in history, from the first Five-Year Plan in 1928 until 1991. Its record is the most important evidence in this debate, and it is more mixed than either side likes to admit.

The case for the planned record

For four decades the USSR grew faster than the United States. On the Maddison Project's estimates (2011 dollars), Soviet output per person grew about 3.4% a year from 1928 to 1970, against about 1.8% in the US. It turned a largely peasant country into an industrial power that built the tanks and planes to defeat Nazi Germany, launched the first satellite, and achieved near-universal literacy. The Stanford Encyclopedia of Philosophy's entry on socialism credits the Soviet model with "its successes in industrializing the country (making it capable of mobilizing in a war effort to defeat Nazi Germany)". The economic historian Robert Allen argued in Farm to Factory (2003) that without the revolution and the plans, Russia would have ended up somewhere between India and Argentina.

Source: Maddison Project Database 2023 (GDP per capita, 2011 dollars). Soviet-era figures are reconstructions and carry wide uncertainty.

The case against

The same table shows the problem. After about 1975 the Soviet economy stopped catching up and began falling back: from 1970 to 1989 its output per person grew about 1.3% a year, against 2.3% in the US. Planning was good at mobilising labour and capital to build steel mills and dams, the things a planner can count. It was bad at what Hayek said it would be bad at: innovation, quality, and producing the variety of consumer goods people actually wanted. The Stanford Encyclopedia puts it this way: the model "failed to generate sufficient technical innovation and intensive growth to deliver differentiated consumer goods of the kind available within advanced capitalist economies."

The costs of the fast early growth were also enormous. Allen himself called collectivisation a disaster that "did perversely accelerate industrialization by driving people off the land", and reviewers of his book have argued that his consumption estimates are too rosy. The famines, the labour camps and the repression are covered in Communism: The Idea, the States, and What Happened in this series. A defender of planning can fairly reply that these were political crimes, not features of planning as such. A critic can fairly reply that concentrating all economic power in the state is exactly what made them possible.

What went wrong, from the inside

Accounts from inside planned economies describe the same problems, and they map onto the theory:

• Information was distorted. Factories reported what helped them meet their targets. As the Stanford Encyclopedia notes, firms "might tend to exaggerate the resources they need to produce and mislead about how much they can produce."
• Targets were gamed. A target in tonnes encourages heavy products; a target in units encourages small ones. Whatever the planner measured, managers optimised.
• Budgets were soft. Loss-making firms were rescued, so there was little pressure to cut waste (Kornai).
• The scale was crushing. The economist Alec Nove, a sympathetic critic, noted that the Soviet economy produced some 12 million distinct products, far more than any planning office could track item by item.

Even sympathetic socialist philosophers draw a hard conclusion. The Stanford Encyclopedia quotes the economist Giacomo Corneo: "a system that only works with exceptional individuals only works in exceptional cases." That is why many modern socialists favour market socialism, worker cooperatives or democratic participatory planning instead of the Soviet model. Socialism: The Many Kinds, From Utopians to Social Democracy in this series sets out those alternatives in their own terms.

Part 5 — The natural experiments, and their caveats

History ran three experiments that look almost designed for this debate: a country split in two with one half planned and the other not. Here is what each shows, and why none is as clean as it looks.

East and West Germany

Same language, same culture, same industrial traditions, split in 1949 into a Soviet-style planned economy in the East and a market economy with a strong welfare state, the "social market economy", in the West. When the two were reunited in 1990, the gap was enormous. In 1991, the first year for which unified regional statistics exist, output per person in the eastern states (excluding Berlin) was about 33% of the western level, according to Germany's federal and state statistical offices. By 2024, after more than three decades of transfers and investment, it had risen to about 72%.

The caveats. The East started with real handicaps that had nothing to do with planning. The Soviet Union took reparations, including dismantled factories. Division itself cut eastern firms off from their traditional suppliers and markets in the West; the economic historians Albrecht Ritschl and Tamás Vonyó found that the West pulled ahead in industrial productivity even before its 1948 market reforms, mainly because division distorted the East's industrial structure. And the West received Marshall Plan aid while the East did not. On the other side of the ledger, the people voted with their feet: by August 1961, according to the Berlin Wall Foundation, East Germany had lost a sixth of its population, at least four million people, to the West. The Wall was built to stop them.

North and South Korea

Korea was split in 1945, and after the Korean War of 1950–53 the North became one of the most centrally planned economies on earth while the South became a market economy. The North actually started ahead. Under Japanese colonial rule, industry was concentrated in the north: on the Maddison Project's estimates, output per person in 1938 was about $2,560 in the territory that became North Korea against about $1,940 in the South (2011 dollars).

GDP per person, 2011 dollars. Source: Maddison Project Database 2023, using estimates from Cha et al., Historical Statistics of Korea (2020). North Korean figures are rough estimates.

The caveats. North Korean data are close to guesswork. An older version of the Maddison data simply set the North equal to the South in 1950 and held the North's figure flat from 1975 to 1990 for lack of information. The North lost its main source of aid and subsidised trade when the Soviet Union collapsed, and suffered a famine in the 1990s: life expectancy there fell from 70.5 years in 1990 to 60.9 in 2000 on UN estimates. The North is also a hereditary dictatorship that spends heavily on its military. And the South was not a laissez-faire economy either. For much of its fast-growth period it was ruled by authoritarian governments whose state steered credit and industry; the economist Alice Amsden's Asia's Next Giant: South Korea and Late Industrialization (1989) is the classic account of that state-led development. So Korea compares totalitarian central planning with state-guided capitalism, not pure plan against pure market.

China before and after 1978

China is the experiment in time rather than space: the same country under two systems.

Output: Maddison Project Database 2023, which uses Harry Wu's revised estimates for 1952–2008 (lower than official figures). Life expectancy: UN World Population Prospects via Our World in Data.

After market reforms began in 1978, output per person grew roughly twice as fast as under the plan, and the World Bank credits the reform era with lifting almost 800 million people out of extreme poverty. Before 1978, the Great Leap Forward of 1958–62 produced the deadliest famine in recorded history: the World Peace Foundation's estimate used by Our World in Data is 36 million deaths, and scholarly estimates range from about 15 million to over 45 million. Life expectancy in 1960 was 33.4 years.

The caveats. The planned era is not only a story of failure. Between 1950 and 1978, Chinese life expectancy rose from about 44 to about 63 years, far faster than in India, where it went from about 41 to 52 over the same period. Literacy and basic health care spread widely in those decades, and some economists argue that this healthy, literate workforce is part of why the reforms worked so well. After 1978 China was also not a free market: the state kept the banks, the commanding heights of industry and the power to direct investment, as The Mixed Economy describes. So China, too, shows a move from heavy planning to a mixture, not to laissez-faire.

The natural experiments agree on one thing: comprehensive central planning lost badly to economies where most prices were free. They don't settle how much state there should be in a market economy, because the winners all had plenty.

Part 6 — Inequality and growth: Piketty and his critics

If the calculation debate was the twentieth century's big argument, inequality is the twenty-first's. Markets won the efficiency contest; the question now is who gets the gains.

From Kuznets to Piketty

For a long time the standard view came from Simon Kuznets, who won the 1971 Nobel prize for his studies of growth. He suggested in the 1950s that inequality rises in the early stages of industrialisation and then falls as countries grow rich: an inverted U, the "Kuznets curve". The falling inequality of the mid-twentieth century seemed to confirm it.

Then inequality rose again. In Capital in the Twenty-First Century (French 2013, English 2014), Thomas Piketty, working with long series of tax records, argued that the mid-century fall was an exception caused by wars, depression and high taxes, not a law of development. His central mechanism is the inequality r > g: when the rate of return on capital (r) exceeds the growth rate of the economy (g), wealth that is already accumulated grows faster than incomes from work, and inherited fortunes become ever more important. The data his team assembled, now in the World Inequality Database, show the rise vividly:

Source: World Inequality Database (2026 release), via Our World in Data. Income before taxes and transfers.

The critics

Piketty's data work is widely admired, and his theory widely disputed. Three lines of criticism matter most.

Housing. Matthew Rognlie, in the Brookings Papers on Economic Activity (2015), found that the long-run rise in capital's share of income comes mostly from housing. That points to planning rules and land supply more than to a general law of capitalism.

Institutions. Daron Acemoglu and James Robinson, in the Journal of Economic Perspectives (2015), argued that "general laws" of capitalism ignore politics and technology, and found that the gap between r and g does not explain historical movements in top income shares. Piketty himself, in a 2015 reply, clarified that he saw r > g as central to wealth inequality, while the rise in top labour incomes, such as executive pay, needs other explanations.

Measurement. The sharpest fight is over the American numbers. Gerald Auten and David Splinter, in the Journal of Political Economy (2024), argue that top income shares estimated from tax returns are biased by changes in tax law and by income that never appears on returns. Correcting for these, they find top shares lower and rising less since 1980, and once taxes and government transfers are counted, "little change in aftertax top income shares." Piketty, Emmanuel Saez and Gabriel Zucman dispute their method; as of September 2026 the two sides were still exchanging replies. This is a live argument among careful scholars, and readers should be wary of anyone who treats either set of numbers as settled.

Does inequality help or hurt growth?

Here the old view was clear. In Equality and Efficiency: The Big Tradeoff (1975), Arthur Okun argued that redistribution has a cost: money moved from rich to poor is carried in a "leaky bucket", losing some to administration and weaker incentives along the way. Societies must choose how much leakage to accept for how much equality.

In 2014 an IMF team led by Jonathan Ostry tested this across countries and found that lower inequality after taxes and transfers goes with faster and more durable growth, and "not a lot of evidence that redistribution has in fact undercut economic growth (except in extreme cases)." Sceptics reply that cross-country regressions cannot prove cause and effect, that the "extreme cases" exception matters, and that the countries doing the redistributing were already rich and well-governed. The defensible summary is that moderate redistribution in rich democracies does not seem to cost much growth, while the effects of very high taxes or very low ones remain argued over.

Part 7 — Markets and the environment

Capitalism in this series explains why pollution is a classic market failure: a price tells you what something costs the buyer and seller, not what it costs the neighbours or the future. Arthur Pigou proposed taxing the damage; Ronald Coase proposed clear property rights and bargaining. The question here is what the evidence says about which system protects the environment better.

The planned record

Planned economies did not escape the problem; arguably they had it worse, because the polluter and the regulator were the same state. The starkest example is the Aral Sea, once among the largest lakes in the world. From the 1960s Soviet planners diverted its two feeder rivers, the Amu Darya and the Syr Darya, to grow cotton in the desert. NASA's satellite record shows what followed: by 2000 the sea was already "a fraction of its 1960 extent", split into separate lobes, and in 2014 the eastern lobe of the southern sea disappeared completely. Market economies have their own disasters, but their pollution could at least be exposed by a free press, challenged in court and voted on.

Markets as a tool

The most interesting development is that environmental policy now often uses markets on purpose. A government sets a cap on total pollution, issues permits, and lets firms trade them, so the cuts happen wherever they are cheapest.

• The US Acid Rain Program, created under the Clean Air Act Amendments of 1990, was the country's first national cap-and-trade scheme. Its final cap on power-station sulphur dioxide, from 2010, was 8.95 million tons a year, about half of the sector's 1980 emissions.
• The EU Emissions Trading System, launched in 2005 as the world's first carbon market, covers about 40% of the EU's greenhouse-gas emissions. The European Commission reports that by 2023 it had helped cut emissions from power and industry by about 47% compared with 2005.
• According to the World Bank's 2026 State and Trends of Carbon Pricing, nearly 30% of global greenhouse-gas emissions are now covered by a direct carbon price, across 87 policies.

Notice that these are hybrids: the state sets the quantity (the cap), and the market sets the price. That is neither Hayek's spontaneous order nor Lange's planning board, but something that borrows from both.

The commons

A third answer comes from Elinor Ostrom, who won the 2009 Nobel prize "for her analysis of economic governance, especially the commons", the first woman to win it. Studying fisheries, forests and irrigation systems around the world, she found that communities often manage shared resources well for centuries without either privatising them or handing them to the state, using rules they make and enforce themselves. Her work is a reminder that "market or state" leaves out a large part of how humans actually cooperate.

The open question

Climate change is the hardest case, because the damage is global and long-term. The UK's Stern Review (2006) concluded that "the benefits of strong and early action far outweigh the economic costs of not acting." Economists broadly agree a carbon price helps; they disagree on how high, how fast, and whether governments should also plan the transition directly, with subsidies and rules. That brings the calculation debate back to life: can a state pick the technologies of the future better than a carbon price and a market can? The answer may differ by technology.

Part 8 — Can computers plan now?

Hayek's arguments were written in an age of index cards and slide rules. Ever since the first computers, people have asked whether machines could do what the planning ministries could not. The history is more interesting than either side usually admits.

The Soviet network that never was

In 1962 the Soviet cybernetician Viktor Glushkov proposed OGAS, a nationwide computer network that would link factories to regional centres and a central computer in Moscow and run the plan in close to real time. As the historian Benjamin Peters tells it in How Not to Network a Nation (2016), the project was never funded at scale. It foundered not on technology but on bureaucratic politics, as ministries resisted sharing information and control: information is power, and organisations guard it.

Cybersyn: the socialist internet

The most famous experiment happened in Chile. In 1971 Fernando Flores, an official in Salvador Allende's socialist government, contacted the British management cybernetician Stafford Beer for help managing the newly nationalised industries. The result, Project Cybersyn (1971–73), linked factories to Santiago by telex machines. The Guardian's account reports that some 500 telexes were found unused in a warehouse; other sources suggest fewer were ever in use. Factories sent daily production figures, software flagged problems, and a futuristic operations room was designed for managers to see the whole picture.

Cybersyn's finest hour came in October 1972, when a strike by conservative business owners, with secret support from the CIA according to the Guardian's account, threatened food and fuel supplies. The telex network helped the government find where shortages were worst and route the trucks it still had. Less glamorously, a researcher interviewed by the Guardian in 2003 reported that, according to people she had spoken to, getting the factories to send their statistics was "like pulling teeth". The historian Eden Medina's book Cybernetic Revolutionaries (2011) is now the standard account. After the military coup of 11 September 1973, the network was dismantled.

Cybersyn is a Rorschach test. To its admirers it shows that real-time economic coordination by computer was possible fifty years ago. To its sceptics it shows a small system that helped in an emergency, covered a fraction of the economy, depended on factories honestly reporting their numbers, and never had to decide what to produce, only to track what was being produced.

The modern case for computer planning

In Towards a New Socialism (1993), the computer scientist Paul Cockshott and the economist Allin Cottrell took the computation objection head-on. Nove had said the Soviet economy had some 12 million products, and quoted an estimate that balancing a full plan for Ukraine would take the world's population ten million years. Cockshott and Cottrell replied that this assumed the wrong algorithm. Solving a full table of 12 million equations by the textbook method is indeed hopeless. But each product uses only a few dozen or hundred inputs, so the table is mostly empty, and methods that exploit that scale roughly in proportion to the number of products. Even on early-1990s workstations, they estimated, a plan for 10 million products could be balanced in under two weeks. Computers are now many thousands of times faster.

In The People's Republic of Walmart (2019), Leigh Phillips and Michał Rozworski added a sharper point: giant firms such as Walmart and Amazon already plan enormous internal economies without using prices inside the firm. The idea goes back to Ronald Coase, who won the 1991 Nobel prize partly for explaining why firms exist at all: sometimes it is cheaper to coordinate by command than by market transactions. If Walmart can plan the supply of millions of products, the argument goes, why not an economy?

The replies

Defenders of Hayek have answers, and they go to the heart of what he meant.

• The problem was never arithmetic. Hayek's point was that the relevant knowledge "cannot enter into statistics". A computer can only process the data it is given, and the data come from people with their own interests. The economist Don Lavoie, in Rivalry and Central Planning (1985), argued that much of the knowledge markets use is tacit and is created by competition, as rival firms try things, so there is nothing for a planner to collect in advance.
• Walmart plans inside a market. Walmart's planners use the prices of the world outside the firm, for wages, freight, land and the goods it buys, and they face competitors and customers who can walk away. Remove the market around it and the internal plan loses its yardstick.
• Incentives remain. Even a perfect algorithm needs honest inputs. Mechanism design shows how hard it is to get people to report truthfully when the report determines what they get, which is Kornai's soft-budget problem again.
• What to make, not just how much. Planning can balance quantities of known products. The hardest thing an economy does is decide which new products should exist at all.

The planners answer that many prices are already set by algorithms, and that democratic planning would let people state their needs directly rather than through purchases weighted by wealth. The critics reply that the data platforms gather work only because they sit inside a market.

Computers solved Lange's problem, the arithmetic. Whether they can solve Hayek's problem, the knowledge, and Kornai's, the incentives, is the open question of this century's version of the debate.

Where the debate stands

Settled, or close to it: comprehensive central planning of a whole economy, with prices fixed by authority, performed far worse than market economies in every natural experiment we have, especially on innovation and consumer welfare; prices carry information that no planner has yet been able to replace; and planning methods such as linear programming work very well inside organisations. Still open: how much the state should own or steer in a market economy; how much redistribution costs in growth, if anything; whether inequality is driven by a law of capital or by politics and housing; how far carbon markets alone can handle climate change; and whether new computing and data change Hayek's knowledge problem or only the arithmetic. On these, well-informed people disagree, and the evidence is still coming in.