Looking at China from the Inside: How a Strong State Can Lose Feedback

By Gorrrrr @ 2026-10-12T14:29 (+2)

Epistemic note: This is a diagnostic hypothesis based on personal observation and a limited reading of the literature, not a causal or representative account of China as a whole. The parts about bureaucratic incentives and policy experimentation have some support in existing research. My discussion of social “exit” is substantially more speculative.

I am not writing this because I want to tell Western readers what political system China should adopt.

Nor am I trying to reduce China to a simple success story or a simple failure story.

What I want to do is more modest, but in some ways more difficult: to translate what institutional life can look like from inside a country like China.

From across the Pacific, some features of China are easy to see. A large central government. Massive infrastructure. Long-term planning. Rapid resource mobilization. Industrial policy on a scale that few countries can attempt. An administrative system capable of transmitting decisions from the top through provinces, cities, counties, townships, and sometimes all the way to individual communities.

All of this is real.

So when people in Europe or the United States tell me that one of their political frustrations is that government cannot get things done—that courts constrain it, legislatures obstruct it, local governments resist it, political parties veto each other, and large projects take years to approve—I understand the frustration.

But living in China makes another kind of institutional risk much easier to see.

A state can fail because it lacks the ability to coordinate and act.

It can also run into serious problems because its capacity to act becomes so concentrated that the system gradually loses the ability to receive accurate feedback from below.

If I had to compress the contrast into two phrases, I would say that many Western political debates worry about coordination failure.

What increasingly worries me about China is feedback failure.

I am not claiming that one is categorically worse than the other. I do not have a metric that could establish that. My point is simply that the second failure mode is easier to underestimate when looking at a high-capacity state from the outside.

China is not a low-capacity state, and this is easy to miss if it is understood only through the binary language of democracy versus authoritarianism.

For decades, one of the interesting features of Chinese governance was that strong hierarchy did not mean that every policy detail was designed in Beijing and mechanically imposed nationwide. A significant part of the reform-era system relied on what Sebastian Heilmann called “experimentation under hierarchy”: the center retained ultimate political authority while local governments could generate and test policy options, some of which were later incorporated into national policy.

That mattered because Beijing cannot possibly know the conditions of every locality in a country of China’s scale.

Guangdong is not Gansu. Shanghai is not an inland county. An export-oriented industrial city does not face the same problems as an agricultural region.

One source of resilience in the Chinese system was therefore a combination that may look contradictory from the outside: strong political hierarchy together with meaningful local experimentation.

That is why I do not think the interesting question is simply whether centralization “works.”

Of course it works in some domains.

The more important question is what happens when increasing amounts of judgment, responsibility, and political risk move upward inside a system that once depended not only on central coordination, but also on local adaptation.

From the outside, people often discuss political feedback as a binary condition. A society either has channels for expression or it does not.

Reality is more complicated.

China has many formal feedback channels: government hotlines, petition systems, online complaint platforms, local administrative portals, internal reporting systems, media reporting within permitted boundaries, legislative proposals, and digital public-service systems.

So the problem is not simply that the state has no way of knowing when people are dissatisfied.

The harder question is what happens to information after it enters an organization.

In what form does it travel upward?

How much is filtered?

What kinds of information are rewarded?

What kinds are dangerous for the person reporting them?

And what finally reaches the people who are actually able to change policy?

A 2025 study of China’s online petition system, based on quantitative analysis and interviews conducted between 2022 and 2024, found that digital oversight could accelerate the processing of public complaints while also producing what the authors call “mechanical responsiveness”: under strong top-down political pressure, local officials could become more concerned with satisfying superior authorities than with actually resolving the underlying disputes.

I think that phrase captures something important.

A system can produce more data while producing less useful information.

A complaint is recorded.

A case is processed.

A response is uploaded.

A performance target is met.

The database becomes more complete.

But whether the original problem has actually been solved is a separate question.

This is not uniquely Chinese. Any large bureaucracy can behave this way.

But vertical administrative systems create a particularly clear incentive problem because local officials often face two audiences.

One is the public affected by the policy.

The other is the superior authority that evaluates their performance and determines much of their career trajectory.

If the demands of those two audiences diverge, we do not need to assume that officials are unusually immoral to predict which audience they will often prioritize.

This creates another paradox.

Stronger supervision sounds like an obvious good.

The center discovers local inaction, so it strengthens assessment.

It discovers delay, so it introduces deadlines.

It discovers concealment, so it creates digital reporting systems.

It discovers weak implementation, so it strengthens accountability.

Each move can make sense on its own.

But when these mechanisms accumulate, something can change.

If the cost of making a mistake keeps rising while the space for independent judgment does not expand at the same time, “avoid making mistakes” can gradually replace “solve the problem” as the safest organizational strategy.

The story is more complicated than saying that local experimentation has simply disappeared. Research on contemporary China describes something closer to “experimentation under pressure”: local governments can still be required to innovate, but increasingly do so under stronger top-down steering and political risk. In some cases, not experimenting may itself carry a penalty.

That distinction matters because being ordered to innovate is not the same thing as possessing autonomous room to experiment.

And even a very large experimental system can learn incorrectly.

A 2025 Journal of Political Economy study of Chinese policy experimentation since 1980 documents three recurring distortions: positive selection into pilot locations, unusually high local effort during experiments that cannot be replicated after national rollout, and imperfect interpretation of results by the central government. The authors’ conclusion is not that experimentation is useless, but that political incentives can systematically bias policy learning.

This distinction has become increasingly important to me.

State capacity and state learning capacity are not the same thing.

A government can be extremely capable of implementing a policy and still learn the wrong lesson.

It may be able to scale a decision across the country very quickly.

But if the information at the beginning of that process is distorted, stronger implementation capacity can also allow error to spread faster.

This is where institutional life inside China can look almost like the mirror image of some Western debates.

In many Western political systems, people worry about courts, local governments, opposition parties, legislatures, media, organized interests, and multiple veto points preventing government from acting.

The fear is paralysis.

From inside China, I often see a different risk.

When more and more issues become matters of upward responsibility, local actors can become less willing to take real responsibility themselves.

Real responsibility requires judgment.

Judgment creates the possibility of error.

And when the rewards for independent judgment are limited while the penalties for failure are high, the safest strategy may become waiting for a document, waiting for instructions, following standard procedure, and minimizing exposure.

A paradox then appears.

The center accumulates more authority, but it also accumulates more problems.

Power and responsibility move upward together.

If local governments are primarily seen as executors, local failures increasingly become failures of central policy.

If an industry develops a serious problem, responsibility moves upward.

If a local governance failure becomes politically salient, it can become a reputational problem for the whole system.

The center then has an incentive to increase control in order to reduce those failures.

But increased control can make local actors even less willing to make independent judgments.

That, in turn, sends still more problems upward.

I do not claim that this loop explains every change in Chinese governance. I do think it is a useful way to understand one recurring institutional pressure.

There is another kind of feedback that is harder to see.

When people believe that some public problems are difficult to change through expression, they do not necessarily become politically confrontational.

Often, exit is cheaper than voice.

Here I need to be especially cautious. China is an enormous society with major differences across class, region, generation, occupation, and political attitude. I am not claiming that 1.4 billion people think in the same way.

What I am describing is a pattern I increasingly notice.

When people cannot easily change a public problem, they may try to privatize it.

If a workplace is intolerable, leave.

If urban life becomes too expensive, move elsewhere.

If marriage, housing, child-rearing, or eldercare become too burdensome, reduce commitments.

If public discussion seems unlikely to change much, participate less.

If long-term investment feels uncertain, hold more cash or save more.

If an organization feels impossible to change, minimize dependence on it.

None of this looks like a dramatic political movement.

That is precisely why it is easy to underestimate.

From a systems perspective, exit can also contain information.

Not participating can contain information.

Not consuming can contain information.

Not investing can contain information.

Not entering an industry can contain information.

Refusing long-term commitments can contain information.

But this is the most speculative part of my argument.

I do not know how much of these behaviors can actually be attributed to weak institutional feedback rather than housing costs, demographic change, labor-market conditions, economic cycles, technological change, or other causes. I currently have no clean way to separate those explanations.

I include the possibility because, if it matters, it creates an unusual problem for governance: a system may receive very little dramatic political resistance while social expectations are nevertheless deteriorating beneath the surface.

By the time those changes become clearly visible in macroeconomic or demographic data, the expectations producing them may have been shifting for years.

At this point, a natural question appears.

If feedback is failing, why not rebuild the system from the ground up?

This is why revolution has always had political appeal.

I do not think revolution is meaningless.

On the contrary, revolution has one extremely powerful function: it can flush out entrenched structures.

An old order accumulates bureaucracies, property arrangements, status hierarchies, elite networks, and mutually reinforcing interests.

Revolution can break those structures very rapidly.

Old elites can be displaced.

Groups previously excluded from power can enter politics.

Property and organizational relationships can be reconstructed.

Path dependence can be broken.

So I do not think it is fair to say that revolution merely replaces one group of rulers with another.

It can genuinely transform a society.

The reason I remain cautious is different.

To destroy an old organization, a revolutionary movement usually has to construct an organization capable of defeating it.

An organization operating under high-intensity political competition faces severe selection pressures.

It needs discipline.

Command.

Resource concentration.

Internal loyalty.

Information networks.

Member screening.

Systems of reward and punishment.

If violent conflict is involved, it also needs reliable coercive capacity.

This does not mean revolutionaries are morally defective.

It means organizations adapt to the environments in which they must survive.

Political-science research provides at least some support for this mechanism. Levitsky and Way argue that identities, norms, solidarity, and organizational structures forged through sustained violent conflict can produce unusually cohesive ruling parties. A broader comparative study of authoritarian regimes likewise finds that regimes founded in violent social revolution have tended to be unusually durable, in part because revolutionary struggle helps create cohesive ruling parties, powerful security apparatuses, and the destruction of alternative centers of power.

This is the deeper paradox of revolution.

To defeat a highly centralized old system, one often needs to build another organization with strong capacity for coordination.

After victory, that organization does not disappear.

It has members.

Cadres.

Internal hierarchies.

Promotion systems.

Networks built during struggle.

Shared memories of sacrifice.

Security interests.

And reasons to preserve itself.

Old elites may indeed be destroyed.

Old classes may be restructured.

But around the new organization’s control of appointments, resources, coercion, and political access, new interests can begin to form.

This does not require us to assume that revolutionaries were lying from the beginning.

Organizations generate organizational interests.

So the question I keep returning to is not whether revolution can renew society.

It obviously can.

The question is whether the underlying problem has really been solved if the answer to an over-centralized old system is to build a more organized new center capable of destroying the old one.

At some point, we still have to ask why so much power keeps returning to the center.

None of this means that gradual reform is inherently safe.

Reform has characteristic failure modes of its own.

It can be captured by the groups it was supposed to constrain.

It can be partially implemented and then reversed.

It can be delayed indefinitely while the language of reform absorbs pressure without changing the underlying distribution of authority.

And if old constraints dissolve faster than new institutions acquire legitimacy and enforcement capacity, reform itself can increase the risk of political rupture.

So my argument is not that reform is low-risk while revolution is high-risk.

It is that both paths generate their own tail risks, and that the institutional discontinuity created by rupture should not be treated as costless simply because the old structure has serious defects.

This is why I increasingly care about the placement of power, rather than only the ownership of power.

The problem is not simply who controls the state.

It is also why so many qualitatively different forms of authority need to be concentrated within the same hierarchy.

Some powers clearly require centralization.

National defense cannot be independently managed by every city.

A unified monetary system requires central coordination.

Large infrastructure networks, major disasters, cross-regional environmental problems, and macroeconomic stabilization all require significant national capacity.

But other decisions do not obviously need to be concentrated to the same degree.

How should local public services be organized?

How should some industry regulations be tested?

How should local social problems be handled?

How should technical judgments be made by professional institutions?

How should citizens communicate implementation failures?

How should a bad policy be stopped?

If all of these questions continuously flow upward, the center receives more than authority.

It also receives information burdens, management burdens, responsibility burdens, and the risk that a single mistake becomes national.

That is why I increasingly prefer the idea of reallocating power to simply calling for “decentralization.”

The question is not whether the central government should become weaker.

The question is where each kind of authority should be located.

Which powers genuinely need to remain central?

Which can be delegated to local governments?

Which should belong to relatively autonomous professional institutions?

Which should be left to social coordination?

Which decisions should be vulnerable to institutional veto?

The objective is not less central power in the abstract.

It is to ask where each form of power can be placed so that the total error cost of the system is lower.

When power is distributed differently, risk is distributed differently as well.

Suppose a locality has real room to test a policy.

It fails.

The failure is initially local.

Other regions can observe it and decide not to copy it.

The center can terminate the experiment.

This is one of the deepest functions of policy experimentation.

It is not merely an expression of local autonomy as a political value.

It is a mechanism for containing error.

If an inadequately tested policy is instead implemented simultaneously across the whole country, a mistaken judgment can immediately become a systemic mistake.

The same principle applies to information.

If there are multiple genuinely independent channels through which information can reach decision-makers, the failure of one channel does not necessarily blind the system.

If nearly all information travels upward through the same administrative hierarchy, the center may receive enormous quantities of information while, in an important sense, still listening through one sensor.

So reallocating power also reallocates responsibility and risk.

This may be one reason why a strong center could rationally accept some degree of decentralization.

Not because it suddenly stops valuing power.

But because, at some point, not all power is worth possessing.

The marginal benefit of additional control can become smaller than the information burden, political responsibility, and systemic risk that come with it.

This is also why I do not think the argument can be reduced to “China should democratize.”

The conclusion I least want a Western reader to draw from this essay is that China should simply become more like Europe or the United States.

Western systems have serious problems of their own.

Too many veto points can create paralysis.

Local autonomy can protect local elites.

Media markets can reward attention rather than accuracy.

Electoral politics can encourage short-termism.

Courts and nominally independent institutions can develop organizational interests of their own.

Every mechanism that constrains one form of concentrated power can itself become a new source of power.

The more general question is not which institutional template is universally correct.

It is what capability a particular system currently lacks.

A weak state may need greater coordination capacity.

A fragmented state may need stronger central authority.

But for a state that already possesses unusually strong powers of coordination and execution, adding more of the same capacity may eventually produce diminishing returns.

What becomes scarce instead may be independent feedback, local experimentation, separation of responsibility, policy exit, and the ability of the center to say—without threatening the stability of the entire political order—we were wrong this time.

There is another part of Chinese political thinking that I think foreign readers can easily underestimate.

Any serious discussion of political change in China eventually encounters one word:

disorder.

I am not claiming that all Chinese people therefore oppose political change.

I mean something narrower.

When I think about political change in China, I find it difficult to treat state breakdown, war, fragmentation, and social disorder as abstract possibilities.

Modern Chinese historical memory is not very far removed from invasion, civil war, revolution, mass campaigns, institutional destruction, and national fragmentation.

So the idea of “tear everything down and rebuild” does not automatically sound like a clean institutional upgrade.

It immediately creates practical questions.

Who collects taxes if the existing structure collapses?

Who pays salaries and pensions?

Who controls local governments?

How does the military remain unified?

What happens to banks and contracts?

Which laws remain valid?

Who has final authority when new political organizations disagree?

Those questions cannot be answered simply by saying that the future regime will be better.

This is why I do not think stability is merely a slogan invented by centralized governments to frighten people.

Stability is a real public good.

The problem begins when the need for stability becomes an argument against all institutional correction.

China therefore faces a difficult dual constraint.

We cannot assume that the current system can remain unchanged indefinitely.

But we also cannot assume that the cost of breaking it apart would be manageable.

That, to me, is the real difficulty of political transformation in China.

Perhaps the hardest reform is not building a better center.

It is building a system that does not require a heroic center.

A country that depends on top leaders being unusually wise is structurally fragile.

Even highly capable leaders make mistakes.

They age.

They operate inside information environments shaped by subordinates.

They receive filtered reports.

Institutions learn what superiors want to hear.

A robust system should not assume that leaders will avoid error.

It should assume that everyone makes errors.

Then it should ask different questions.

How are mistakes detected?

How far can one mistake spread?

How long can it continue?

Who can tell the highest authority that it is wrong without first risking personal destruction?

Can different regions test different approaches?

Can a failed policy be abandoned without first turning disagreement into a question of loyalty?

This is not, in my mind, primarily a romantic argument about liberty.

It is a systems problem.

How do you design a machine in which one component can fail without forcing the entire machine to fail with it?

I do not know how much room China still has for this kind of institutional adjustment.

I do not know what China will look like in twenty or thirty years.

I do not know the second- and third-order effects of meaningful decentralization.

And I certainly do not believe that all decentralization produces good outcomes.

Local governments create their own interest groups.

Social organizations can become rent-seeking structures.

Capital can capture regulation.

Media systems can create different forms of distortion.

Institutions do not move from bad people to good people.

They redistribute power, incentives, information, and error.

There is also a harder question that this essay does not solve.

Even if reallocating authority would improve feedback, who would actually do it, and why would actors who benefit from the existing arrangement accept it?

Describing a more robust error-correction structure is easier than explaining the political coalition capable of creating it.

Pressure for change could come from within the state, from local actors, from social organizations, from economic constraints, or from some interaction among them. But I do not currently have a reliable basis for saying which mechanism could produce meaningful institutional change in China today.

That limitation matters.

A desirable institutional arrangement is not automatically a politically reachable one.

Still, this uncertainty is exactly why I no longer think China’s deepest need can be described as another supposedly correct center, another final ideology, another new elite, or another organization confident that it can make the right decisions for everyone indefinitely.

The question that interests me is simpler and harder.

Can a country that has already built extraordinary state capacity learn not to concentrate every important capacity in the same place?

If there is one thing I want readers across the Pacific to see, it is this.

China should not be understood only as a country that lacks Western-style democracy.

Nor should it be understood only as an impressive machine of state capacity.

Both descriptions are too simple.

From inside, what makes China fascinating and troubling at the same time is that it has already demonstrated how much a large state can accomplish when it possesses strong coordination, organizational, and implementation capacity.

It has also developed sophisticated traditions of local experimentation and policy learning.

But precisely because it became so good at concentrating resources and implementing decisions, a harder problem becomes visible:

What happens when centralization itself begins to interfere with feedback?

Revolution offers one answer: construct a new organization powerful enough to break the old structure.

That answer has real transformative force.

But it does not automatically explain why the new organization will not become another center of concentrated power.

Preserving the status quo offers another answer: use the existing center to solve more and more problems.

That avoids the immediate risks of rupture.

But it may also concentrate more information, responsibility, and systemic risk in the same place.

The harder path is to reconsider where power should be located.

Not in order to destroy state capacity.

Perhaps in order to preserve it.

Because a genuinely strong state should not only be able to carry a decision all the way through.

It should also be able to discover that the decision was wrong, stop, allow another place to try something different, allow another voice to enter, separate failure from disloyalty, and reverse policy without requiring the entire political system to declare itself fundamentally mistaken.

For me, that may be the form of state capacity China now most needs to think about:

not stronger execution, but the ability to recover feedback without losing the capacity to act.