Epistemic status: This is a preliminary mechanism model. I developed it mainly while thinking about Chinese dynastic history, and I do not yet know how far it generalizes. I am more confident in the individual mechanisms than in the claim that they form a general historical pattern.
I do not think the usual explanations of dynastic decline are wrong.
Corruption matters. Population pressure matters. Elite competition matters. War matters. Climate shocks matter. Fiscal crises matter. Individual rulers and policy choices matter.
What interests me is a different question:
Can some of these familiar mechanisms be connected into a common temporal path?
More specifically:
Why can centralization increase institutional capacity during one period, while making correction increasingly expensive during a later period?
I originally arrived at this question through Chinese dynastic history. But the broader problem seems relevant to institutional design.
Large institutions face a real problem. They need coordination, standardization, and reliable execution. But they also need local information, independent feedback, redundancy, and the ability to correct mistakes.
So the interesting question is not whether centralization is good or bad.
It is:
How can an institution preserve the gains from coordination without gradually losing its capacity for correction?
My tentative answer has two stages.
The first is what I call the precursor model. It explains why successful centralization can reinforce the institutional path that produced it.
The second is a fiscal-feedback model. It explains how, under sufficient pressure, a mature centralized structure can enter a self-reinforcing decline in effective capacity.
The important point is that these two stages may be connected.
The model does not begin with centralization failing.
It begins with centralization succeeding.
Imagine a political system composed of many semi-independent local authorities.
They maintain different tax systems, military organizations, administrative procedures, legal practices, and standards.
Coordination is expensive.
A stronger center can reduce some of those costs.
It can standardize taxation. It can coordinate military resources. It can build infrastructure across local boundaries. It can enforce common rules. It can move resources between regions. It can reduce conflicts between local political units.
These are real advantages.
So during an early scaling phase, two things can happen at the same time:
formal concentration of power rises
effective state capacity risesThis is important because otherwise the historical success of centralization becomes difficult to explain.
If centralized systems were simply inefficient from the beginning, why would they repeatedly outcompete more fragmented arrangements?
My model therefore starts with the opposite assumption:
Centralization can spread because it solves real coordination problems.
The later problem partly emerges from that success.
One simple explanation for centralization is that rulers want more power.
That can certainly be true.
But I do not think it is sufficient.
Consider a simplified administrative chain:
central government
->
provincial government
->
local government
->
administrative agent
->
taxpayerSuppose the center encounters a new problem.
It could redesign the institutional system, build a new information channel, negotiate with local actors, or use an administrative channel that already exists.
Usually, the existing channel is cheaper.
So it uses that channel again.
Now consider the provincial government.
It receives a target from above. It could challenge the target, build new local capacity, or pass the pressure downward.
Passing the pressure downward may be cheaper.
The local government faces a similar choice.
So does the administrative agent.
Each actor may therefore make a locally reasonable decision.
But:
The sum of locally reasonable decisions does not have to produce a globally efficient system.
The important mechanism is repetition.
A channel that is repeatedly used attracts complementary institutions.
Officials learn how to operate it.
Budgets are designed around it.
Promotion systems reward people who succeed inside it.
Information systems adapt to it.
Alternative channels receive less investment.
Local capacities that are no longer needed begin to decay.
The sequence looks roughly like this:
a solution works
->
the solution is reused
->
institutions adapt around it
->
reuse becomes cheaper
->
alternatives weaken
->
leaving the path becomes more expensiveThis is what I mean by a path of least resistance.
"Least resistance" does not mean the solution with the lowest total social cost.
It means the cheapest available solution from the position of the actor making the next decision.
The center can rationally choose the easiest available tool.
The province can rationally do the same.
The county can rationally do the same.
The resulting institutional trajectory can still become increasingly difficult to change.
This process creates an asymmetry.
Centralization can often be added gradually.
One more reporting requirement.
One more standardized procedure.
One more fiscal responsibility moved upward.
One more form of local discretion removed.
No individual decision needs to look revolutionary.
But reversing the process is different.
After decades of adaptation, decentralization does not simply mean "stop giving orders."
It may require rebuilding capacities that no longer exist.
Local governments may have lost independent fiscal capacity.
Alternative information channels may have disappeared.
Officials may have been selected and promoted for their ability to operate inside the centralized hierarchy.
Organizations originally created as instruments of the center may have become necessary intermediaries.
The center may dislike those intermediaries while simultaneously being unable to govern without them.
So I am not claiming:
centralization is irreversibleI am claiming something weaker:
Under some conditions, the cost of reversing centralization increases over time.
This also explains why past success can become misleading evidence.
An institution that failed from the beginning is easy to criticize.
An institution associated with decades or centuries of expansion has a strong argument in its favor:
"It worked before."
And that argument is correct.
The difficulty is that the institution's own success may have changed the environment in which it operates.
The distinction between formal authority and effective capacity is not original to this model.
Michael Mann distinguishes between "despotic power" and "infrastructural power": roughly, the ability of state elites to make decisions without negotiation, and the state's ability to actually penetrate society and implement decisions.
My question is not whether these dimensions are different.
My question is dynamic:
Why might a process that initially increases both eventually cause them to diverge?
Early centralization can raise both formal authority and effective capacity.
Standardization makes decisions easier to implement.
But successful scaling also creates a larger and more heterogeneous system.
More territory has to be governed.
More information must move upward.
More instructions must move downward.
More agents stand between the center and the activities being governed.
This requires information compression.
A village, county, province, or ministry cannot transmit every relevant detail upward.
Reality is converted into categories, reports, quotas, statistics, and performance indicators.
That is unavoidable.
But compression loses information.
Agents also have incentives.
Bad news can be delayed.
Metrics can be optimized.
Ambiguous instructions can be interpreted strategically.
Local knowledge can disappear during aggregation.
So a mature system may reach a condition where:
formal authority continues to rise
while
the marginal gain in effective capacity fallsIn a stronger case:
formal authority rises
while
effective capacity declinesThis gives us an important possibility:
A state can become more centralized institutionally while becoming harder to govern operationally.
The precursor model explains why a successful institutional path can become expensive to leave.
It does not yet explain dynastic decline.
For that, I want to add a simple fiscal mechanism.
Let:
B_t = productive tax base
q_t = gross extraction rate
G_t = gross extracted resourcesThen:
G_t = q_t * B_tBut gross extraction is not the same as resources actually available to the state.
Resources must be assessed, collected, transported, protected, monitored, administered, and converted into something the government can actually deploy.
So a crude accounting identity would look like:
R_t =
G_t
- administrative cost
- information cost
- enforcement cost
- transfer cost
- agency losswhere:
R_t = net deployable state resourcesThis is not intended as a finished empirical model.
Its purpose is conceptual.
It forces a distinction between:
economic output
and
resources that the state can actually coordinate and use.
This produces a counterintuitive possibility.
The burden experienced by people at the bottom can increase while the resources reaching the center decrease.
In other words:
gross extraction pressure risescan coexist with:
net state capacity fallsThere is no contradiction if the transmission mechanism itself is becoming more expensive.
Corruption may be part of this loss.
But corruption is only one possible component.
Administrative overhead, enforcement, information failure, transportation, bargaining with intermediaries, local extraction, and implementation failure can all reduce the conversion of social resources into deployable state resources.
This is why I am reluctant to treat corruption as a sufficient explanation of dynastic decline.
The deeper problem may be declining conversion efficiency.
Suppose the productive base weakens.
The original trigger could be war, demographic pressure, ecological disruption, trade decline, land concentration, political conflict, or something else.
This model does not require one universal trigger.
Suppose the tax base falls while military and administrative obligations remain high.
The state now faces fiscal stress.
Deep reform is one possible response.
But deep reform is expensive.
There is often a cheaper short-run response:
Use the existing extraction machinery more intensively.
That can create the first feedback loop:
fiscal pressure
->
greater extraction from accessible groups
->
damage to productive capacity
->
smaller future tax base
->
greater fiscal pressureA second loop involves intermediaries:
fiscal weakness
->
greater dependence on existing intermediaries
->
greater agency and transmission losses
->
lower net resources
->
greater fiscal weaknessA third loop involves monitoring:
net resources fall
->
monitoring capacity falls
->
agency loss rises
->
administrative cost rises
->
net resources fallA fourth loop involves public goods:
net resources fall
->
public goods and security weaken
->
productive capacity weakens
->
net resources fallAt this point the system is no longer facing several independent problems.
The problems have become coupled.
The system's response to weakness can begin producing more weakness.
That is the transition from ordinary stress to self-reinforcing decline.
One historical example does not prove a large model.
Still, an eighteenth-century Qing fiscal reform illustrates the type of transition I have in mind.
Yu Hao and Kevin Zhengcheng Liu study a reform that formalized county-level informal surtaxes and centralized control over them in the hands of provincial governors.
The reform initially improved fiscal capacity. Their results indicate that the new provincial revenues increased famine relief during exceptional disasters and improved intertemporal smoothing and interregional risk sharing.
But the effects weakened after the central government began appropriating provincial fiscal revenues.
The authors interpret the short-lived effect through a credible commitment problem between the central and provincial governments.
For my purposes, the interesting pattern is:
formalization and centralization
->
better coordination
->
greater effective capacityfollowed later by:
greater legibility of resources
->
greater ability of higher levels to appropriate them
->
weaker local fiscal capacityThis does not mean that centralization caused the later problem in some simple sense.
The same institutional change produced different effects at different stages.
It first improved coordination.
Its success then changed the incentives surrounding the resources it had made more visible and controllable.
That is exactly the kind of transition I am interested in.
A second Qing example concerns local tax collection in Henan.
Xiaoqun Xu describes local "equal sharing" arrangements used to satisfy fixed tax quotas under limited administrative capacity.
The case is useful here because it shows local actors adapting collection methods to the practical constraints of fulfilling higher-level revenue demands.
The general mechanism is:
Under institutional constraints, actors choose relatively cheap local adjustments to satisfy immediate targets.
Those adjustments can be rational.
Repeated rational adjustments can still move the larger system toward a less desirable long-run equilibrium.
The full path now looks roughly like this:
fragmentation
->
centralization lowers coordination costs
->
centralization produces real gains
->
successful administrative channels are reused
->
complementary institutions accumulate around them
->
alternative channels and local redundancy weaken
->
reversal becomes more expensive
->
scale and heterogeneity increase
->
information is compressed through longer agency chains
->
formal authority begins to diverge from effective capacity
->
fiscal or external pressure arrives
->
existing administrative and extraction channels are used more intensively
->
local actors respond with the cheapest available adaptations
->
agency, information, and extraction costs rise
->
productive and local administrative capacity weaken
->
net deployable resources fall
->
monitoring and public-goods capacity fall
->
the system becomes even more dependent on channels whose marginal effectiveness is decliningThe last step is the part I find most interesting.
The system does not simply become weaker.
It can become more dependent on the mechanism that is becoming less effective.
That is what I mean by institutional self-locking.
It also explains why a single-variable explanation is unlikely to be enough.
Suppose corruption is reduced.
That should help.
But eliminating corruption does not automatically rebuild local administrative capacity.
It does not shorten information chains.
It does not recreate redundant information channels.
It does not restore a damaged tax base.
It does not automatically reduce the cost of institutional reversal.
So both of the following can be true:
anti-corruption worksand:
anti-corruption does not reverse the larger trajectoryThe same logic applies to other familiar explanations.
Population pressure can be real.
Elite competition can be real.
Military costs can be real.
Climate shocks can be real.
The model is not trying to replace them.
It asks a different question:
Why can a mature institutional system become progressively worse at absorbing them?
A model that explains every historical outcome after the fact is not useful.
So this one needs predictions that can fail.
Late-stage weakness should not always look like decentralization.
Some systems should continue concentrating formal authority while information quality, implementation quality, or fiscal conversion efficiency deteriorates.
People at the bottom can experience increasing burdens while the center receives decreasing net resources.
If this pattern is impossible or historically rare, an important part of the fiscal mechanism is wrong.
A declining system should sometimes require increasing administrative, political, or coercive effort to obtain each additional unit of deployable resources.
Nominal revenue alone would therefore be an incomplete measure of fiscal health.
Reducing corruption or correcting one administrative failure should help.
But it should not necessarily reverse the larger trajectory when information loss, a damaged tax base, institutional lock-in, and weakened redundancy remain.
Independent local information channels and administrative capacity may look wasteful during stable periods.
But they should sometimes improve adaptation to heterogeneous shocks.
Failure provides evidence against an institution.
Success provides evidence for it.
Institutions strongly associated with previous expansion may therefore persist even after their marginal returns begin to decline.
A reform that improves coordination, legibility, or fiscal capacity may also change the strategies available to actors at different levels.
So the relevant question is not only:
"Did the reform work?"
It is also:
"What new strategy became cheaper because the reform worked?"
There are several observations that would substantially reduce my confidence.
If mature centralized institutions routinely maintain high-quality bottom-up information, low agency loss, inexpensive policy reversal, strong independent local capacity, low fiscal conversion costs, and effective adaptation to heterogeneous shocks as they scale, then the precursor mechanism is much weaker than I suspect.
I would also update against the fiscal part if declining states generally show no deterioration in their ability to convert economic resources into deployable political resources.
Another important counterexample would be a mature centralized state that clearly enters something resembling the proposed feedback loop, but escapes it cheaply without rebuilding information channels, fiscal relations, institutional redundancy, or local administrative capacity.
Those cases are especially interesting.
A strong counterexample may identify a stabilizing mechanism that this model currently lacks.
Almost none of the individual components here are new.
Principal-agent problems are not new.
Path dependence is not new.
Fiscal stress is not new.
The distinction between formal authority and effective state capacity is not new.
Administrative information loss is not new.
Local optimization around institutional targets is not new.
The possible contribution is narrower.
I am proposing that these mechanisms can sometimes form a directional sequence:
coordination gains
->
institutional concentration
->
path dependence
->
loss of alternative capacity
->
rising correction costs
->
divergence between formal authority and effective capacity
->
fiscal feedback
->
self-reinforcing declineThis is not a theory that says:
"Centralization causes collapse."
It is a theory about a transition.
A mechanism can be useful in one regime and damaging in another because its own success changes the environment in which it operates.
That is why I started with dynastic cycles but ended up more interested in another question:
Why do successful systems become expensive to correct?
My tentative answer is:
Successful institutions attract complementary institutions, weaken alternatives, and alter the incentives of the actors operating through them.
When conditions later change, the old path is no longer simply one option among many.
It is often the cheapest path still available.
Sometimes the system responds to the declining effectiveness of that path by using it even more.
I do not think this model is complete.
The first uncertainty is generality.
I developed it mainly from Chinese dynastic history. Some mechanisms may apply to other centralized states or even non-state organizations, but I do not yet know which parts depend specifically on imperial Chinese fiscal and administrative structures.
The second problem is the transition point.
I have described a mechanism by which centralization can move from raising capacity to raising correction costs.
I have not identified the threshold at which that change occurs.
The third problem is endogeneity.
I do not know how much decline this structure can generate by itself.
A centralized system may remain stable for a very long time unless war, demographic pressure, ecological disruption, fiscal stress, or another shock pushes it into the feedback regime.
The fourth problem is redundancy.
I have treated several things as forms of redundancy that may actually behave very differently: local fiscal autonomy, independent information channels, duplicated administrative capacity, political competition, and informal institutions.
These probably need to be separated.
The fifth problem is recovery.
The model currently explains failure much better than escape.
If a system becomes locked into a declining path, what allows it to recover?
Does it need decentralization?
Can the center rebuild local capacity without losing coordination gains?
Can independent information channels be recreated after they have disappeared?
Can redundancy be deliberately maintained even when it appears inefficient?
Or does recovery normally require a crisis large enough to change the incentive structure itself?
These questions matter because a stronger version of the model should not merely explain decline.
It should explain the boundary between systems that remain corrigible and systems that become increasingly dependent on the mechanisms that are failing.
So the question I am left with is not:
"Does centralization eventually fail?"
That is too broad.
The question is:
Under what conditions does successful centralization remain corrigible?
And conversely:
When does the cost of correcting a successful system begin to rise faster than the system's ability to adapt?
I do not yet have a good answer.
That is the part of the model I would most like to see challenged.
If this mechanism exists, its relevance is not limited to explaining the decline of historical states.
The practical institutional question is:
How do we preserve coordination gains without allowing correction costs to compound?
Centralization, standardization, scale, and legibility can all create real value.
The problem is not to eliminate them.
The problem is to prevent successful coordination from destroying the alternative channels that make later correction possible.
This changes how I think about robustness.
Robustness is not only the ability to survive an external shock.
It is also the ability to revise the mechanisms that produced earlier success.
A system that performs extremely well under familiar conditions but gradually loses the ability to detect and correct its own mistakes may be efficient without being robust.
So perhaps the institutional design problem should not only be:
How do we build systems that coordinate well?
It should also be:
How do we build systems that remain corrigible after coordination succeeds?
That is the broader question behind this model.
Hao, Yu, and Kevin Zhengcheng Liu. 2020. "Taxation, Fiscal Capacity, and Credible Commitment in Eighteenth-Century China: The Effects of the Formalization and Centralization of Informal Surtaxes." The Economic History Review 73(4): 914-939.
Mann, Michael. 1984. "The Autonomous Power of the State: Its Origins, Mechanisms and Results." European Journal of Sociology 25(2): 185-213.
Xu, Xiaoqun. 2025. "State Revenue Extraction and Local Collection Expediency in Qing China: 'Equal Sharing' of Tax Quota on 'Renamed Lands' in Henan Province." Journal of Chinese History.
