With a Cooler Head and a Warmer Heart

I trust the workings of the market economy quite strongly.

The market coordinates, through prices and competition, dispersed information that no single central planner could hold; competition presses firms on their choice of production methods; and profit provides the economic incentive to respond to consumer demand. Hayek saw the economic problem not as one of gathering all relevant information in one place and computing, but as one of using knowledge scattered throughout society; and Coase explained that while the price system coordinates the economy, using the price system itself carries a cost.

So I do not regard the mere fact that government does more things as progress in itself.

The role of government should be judged not by the size of the organization but by the social results it produces.

Indeed, the OECD itself identifies one of the core challenges facing governments today not as the size of government but as improving productivity by raising efficiency and effectiveness.

So if the state does something, that thing must actually improve people's lives.

More civil servants, more regulations, and more budget do not automatically produce better results.

OECD data also show that the share and structure of general-government employment differ greatly across countries: in some, the public sector directly provides healthcare, education, and emergency services, while in others these are provided through the private or nonprofit sectors. That is, which organizational structure should produce a given social service is a separate design question.

So when I evaluate a state, what I look at most is not the size of the state but its productivity.

When scarcity changes, the possibilities of policy change too

At the starting point of economics lies scarcity.

Because human wants and the goods and services people desire exceed the available resources, not everything can be had at once, and so choice and opportunity cost arise. Because factors of production such as labor, land, and capital are limited, a society must choose what to produce and how much.

So expanding healthcare requires resources to produce healthcare services; expanding education requires resources such as teachers and facilities; and running a welfare system requires administrative resources to raise funds, identify recipients, make payments, and enforce.

This is not simply a matter of political will.

It is a matter of cost.

As Coase pointed out, in the real economy transactions themselves carry costs. Discovering prices, negotiating, drafting contracts, verifying performance, and resolving disputes all require resources. North likewise explained that institutions and technology determine the transaction and transformation costs required for production.

So if some social policy was not adequately implemented in the past, the reason cannot necessarily be attributed to the value of that policy alone.

It is also possible that the productivity and administrative capacity needed to implement it were insufficient.

This is an important distinction.

Which policy is desirable, and which policy is feasible with present technology, finances, and administrative capacity, are different questions.

Technology moves the boundary of economic possibility

And this is exactly where technological progress becomes important.

Because technology can change the production function itself.

The central finding of research on automation is that automation can substitute capital for labor in specific tasks, lowering costs while raising productivity, and at the same time can impose adjustment costs — job losses — on the workers who performed the automated tasks.

Nor does AI remain a merely abstract possibility.

In an empirical study, the introduction of generative AI to 5,179 customer-support workers raised productivity, measured as issues resolved per hour, by about 14% on average, with the effect especially large for novice and lower-skilled workers.

The OECD's 2024 analysis likewise assesses AI as a general-purpose technology with the potential to spur long-run productivity growth, while noting that considerable uncertainty remains about the size of that effect and its distributional consequences.

In scenarios the OECD estimated by synthesizing several studies, AI was projected to raise annual total factor productivity by roughly 0.25 to 0.6 percentage points and labor productivity by roughly 0.4 to 0.9 percentage points over the next decade. These are projections, not confirmed results, but they give concrete quantitative grounding to the hypothesis that AI can change the economy's production possibilities.

So what I am saying is not that “technology creates socialism.”

The more precise statement is this.

Technology changes the set of economic possibilities from which politics can choose.

What was too expensive in the past can become cheaper; what people had to do in the past can be automated; and information on a scale that was hard to manage in the past can become processable.

As a result, policies that were financially or administratively hard to bear in the past may enter the future as new options.

But there is an important caveat here.

Technology lowering costs does not mean all costs disappear.

AI, too, depends on physical infrastructure such as data centers, electricity, semiconductors, and networks, and AI's economic effect is determined not only by the technology itself but by actual adoption rates and organizational change. So it is more accurate to see technological progress as a process that changes the composition of scarce resources rather than one that abolishes scarcity.

So the question of politics can change, too

In the past, when any policy was discussed,

“Who will bear the cost?”

was the most important question.

That question will not disappear in the future.

But if technology raises productivity and administrative capacity, a new question is added to it.

“How much can technological progress lower that cost?”

This question is already appearing in actual policy debate.

The IMF assesses that generative AI has the potential not only to raise productivity but to improve the delivery of public services, and at the same time analyzes that adjustments to social protection and tax systems may be needed to respond to labor-market shocks and inequality.

That is, technological progress is not simply a matter of “more production.”

It simultaneously creates the policy problem of how to convert productivity gains into social outcomes.

And this is exactly where the politics of the future, as I imagine it, begins.

More services, at lower administrative cost

None of this means I have come to like state bureaucracy.

Quite the opposite.

What I want is to expand social outcomes while lowering the administrative cost of producing those outcomes.

This is not idle fantasy.

The OECD analyzes that AI can be used to raise the efficiency of government's internal work, deliver public services faster and in more tailored ways, support decision-making and policy design, and detect fraud.

In particular, tasks such as repetitive data entry, payroll processing, basic handling of citizen inquiries, and information verification and classification are identified as areas with high potential for AI and automation.

Among the 200 government AI use cases the OECD analyzed, AI adoption is spreading mainly around public services and internal operations, with raising productivity, responsiveness, and accountability as the main goals.

So if technology develops far enough, a state that provides more services with less administrative labor may become possible.

But here, too, there is one distinction I want to make.

AI automating tasks does not mean the number of civil servants automatically falls.

The staff that technology frees up may be redeployed to new work, or used to raise the quality of existing services. Indeed, the OECD itself points out that the effect of government AI adoption depends not on the technology itself but on organizational, data, workforce, and institutional conditions.

So what I am arguing is not “adopt AI and fire the civil servants.”

More precisely,

it is to build an administrative system in which the same workforce produces more value, and to eliminate the structure that keeps putting people into work that could be automated.

Not a bigger state,

but a smarter state.

Rather than increasing the number of civil servants,

raising the social value a single civil servant can create.

Rather than piling on regulation indiscriminately,

using data and technology to enforce the regulations that are needed with more precision.

And providing better services to more citizens with the same resources.

This direction is not a merely ideological claim; it largely coincides with the core direction of digital government the OECD proposes. The OECD recommends that data, digital tools, and AI have the potential to raise government efficiency and productivity, and that regulatory simplification, data use, AI adoption, and improvement of budgeting and procurement systems be pursued together.

Asking the old market-versus-state question again

So I do not think the politics of the future can be explained simply as a contest between left and right.

Rather, one of the important questions will be which institutions produce more social value from given resources.

Both market and government are ways of allocating resources, and real institutions are never made of only one of them.

As Coase emphasized, the real economy contains not only markets but planning inside organizations and administrative coordination, and which mode of coordination is efficient depends on the transaction costs and organizational structure of each.

So what matters is not only the abstract question of “market or state?”

Which method produces better results at lower cost?

Which method creates stronger incentives?

Which method uses more information?

Which method corrects itself faster when it fails?

And which method is sustainable over the long run?

This is how I, as a believer in markets, look at the state.

So I rate highly the chance of “policies that look socialist”

From this perspective, it is no contradiction that I, a believer in markets, rate highly the chance that the society of the future will accept policies that look far more socialist than today's.

Because what matters here is not the name of the policy but its economic feasibility.

If AI and automation raise productivity, if digital infrastructure lowers administrative costs, and if data and algorithms make it possible to manage complex policy populations with greater precision, then some social guarantees that were hard to implement in the past because of cost may become more feasible. Both the IMF and the OECD analyze AI's potential to improve productivity and public services alongside its distributional and employment shocks.

But this should not be read straight away as “technology proves socialism.”

The effects of AI do not automatically bring equality.

The IMF's 2025 research analyzes that AI may reduce wage inequality by replacing some of the tasks of high-income workers, but may at the same time widen wealth inequality as high-skilled workers and owners of capital gain more through AI's complementarity and rising returns to capital.

That is, technology does not automatically choose a particular political system.

Technology widens the space of policies that can be chosen, but which choice is made within that space remains a matter of politics and institutions.

Not abandoning the market, but using the market's success

So I do not want to abandon the market.

On the contrary, I want the market to become more powerful.

Firms must produce more with fewer resources, competition must raise productivity, and new technology must keep replacing existing methods of production.

Indeed, research on the productivity effects of automation and AI shows that technology can lower production costs and raise productivity, and at the same time shows that labor substitution and distributional problems can arise in the process.

So the core of technological progress is neither simply “eliminating jobs” nor “making everyone rich.”

It is raising productivity while at the same time creating a new institutional problem: deciding to whom the benefits of that productivity gain accrue.

The IMF, too, analyzes that because AI's productivity gains can simultaneously create labor-market shocks and inequality, tax and social-protection systems can play a role in distributing those gains more widely.

If so, providing more social guarantees need not mean the market's defeat.

Socially guaranteeing part of the abundance that sufficient productivity has created is not the opposite of a market economy. It may be one institutional choice about the productivity a market economy has produced.

What Lee wants, in the end

In the end, the future I imagine is not one in which we choose between capitalism and socialism.

There is something to ask first.

What creates the most value from the fewest resources?

What the market does better, leave to the market.

What the state can do better, the state can take on.

If private and public working together is more efficient, do that.

And if no existing institution is efficient any longer, find a new way.

This is not the abandonment of market principles.

It is closer to understanding market principles not as a religion about means but as a principle about results.

If, in the name of the market, competition is eliminated and monopoly is protected, that may contradict the principles of the market.

Conversely, if the state can provide some service at lower cost and higher quality, there is no reason to exclude it as non-market merely because it is the state providing it.

Coase's central insight, too, was not to push all economic activity into either market or government, but to compare in reality which mode of coordination is more efficient once transaction costs are taken into account.

So what I care about is not the name of the policy.

It is productivity.

It is cost.

It is incentives.

It is information.

And it is results.

And the final question

Here I want to ask a slightly warmer question.

If technology can actually reduce the amount of labor a human being must perform in order to survive, how will we use the abundance created by that productivity gain?

The answer to this question has not yet been decided.

Because the fact that AI can raise productivity by replacing or complementing labor, and the question of to whom the fruits of that productivity will go, are different problems.

It could be used to create more consumption.

It could be used for more capital accumulation.

It could be reinvested in new industries and technologies.

Or it could be used to return more time and more choice to human beings.

I want to see that last possibility.

A society where more people can be educated.

A society where people can access the medical care they need.

A society where no one has to give up the basic choices of life for economic reasons.

And above all, a society that lowers the degree to which the circumstances of one's birth determine the choices of one's whole life.

This is not a story of simply eliminating present inequality.

It is asking how much of the expanded possibility, when technological progress widens production possibilities, can be converted into human freedom and opportunity.

This problem is exactly why the IMF, in discussing AI policy, emphasizes the role of social protection, education, and tax systems alongside productivity gains. Because the economic effects of technology are not automatically converted into social benefits.

A cool head and a warm heart

So I believe in the market.

I believe in technology.

But I do not believe that the results the market and technology produce always automatically flow in a direction good for human beings.

The market holds the power of competition and innovation, but the market can at the same time create monopoly, externalities, and distributional problems.

Technology holds the power of productivity gains, but technology can also expand labor substitution and inequality.

So what is needed is neither blind worship of the market nor blind trust in the state.

It is calculating the constraints of reality precisely.

Which policies are actually possible?

What do they cost?

How much can technology lower that cost?

What incentives do they create?

What side effects do they create?

And to whom will the new possibilities created by productivity gains be returned?

These are the questions I think the political economy of the future must answer.

The state does not need to be bigger.

It only needs to be more productive.

Nor is there any need to permanently protect inefficient bureaucracy in the name of welfare.

Public services only need to get better, and administrative costs only need to get lower.

There is no need to abandon the market either.

On the contrary, the market must be made to produce even more productivity.

And if technology eases the scarcities of the past, then we must decide how to use the options that newly arise at that moment.

I hope those options are used in a direction that returns more freedom to human beings.

More education.

Better medical care.

Wider opportunity.

More choice.

And above all,

more time.

That is the future I am talking about.

Not a future that abandons the market economy.

Not a future that rejects technology either.

Rather, a future that, on the foundation of the productivity the market has created and the abundance technology has created, recalculates the human goals that were impossible in the past because of cost.

Calculating the constraints of reality with a cooler head,
and imagining the world after those constraints are gone with a warmer heart.

Perhaps the political economy of the future will be rewritten somewhere between those two.