Who Gets to Predict the Future?
The State cannot out-think its citizens.
A society cannot sustain a highly agentic State and highly agentic citizens at the same time. One always crowds out the other. What we usually call “good governance” is simply a decision about where agency is allowed to live.
In Why Nations Fail, the authors make a precise claim: States with extractive political and economic institutions can, at times, generate bursts of prosperity—often by enriching elites while preventing the rest from competing with them. By centralising power, forcibly assembling resources, aggregating information through authority, and collapsing degrees of freedom, such States can push a system in a chosen direction. This is why the Soviet Union could execute Five-Year Plans that rapidly expanded steel, rail, and heavy industry, and why colonial administrations could mobilise labour for plantations, mines, and infrastructure with administrative speed. Concentrated agency simplifies coordination. When only one actor is allowed to model the future, alignment becomes trivial. Growth engineered through authority eventually runs out of steam; once mobilisation gains are exhausted, only creative destruction can produce genuinely novel growth.
But this prosperity is brittle. It relies on sustained pressure and the suppression of competing agency. When pressure relaxes or incentives drift, the system stalls or fractures—as centrally planned economies repeatedly did once growth required innovation rather than mobilisation. This is precisely what entrenched elites fear, because creative destruction does not just replace firms or technologies; it redistributes power. Long-term prosperity, by contrast, appears only when the State steps back—when it limits its own agency and allows inclusive political and economic institutions to emerge. The book returns to cases like England after the Glorious Revolution, where constraints on the Crown strengthened Parliament, stabilised property rights, and lowered borrowing costs for private actors; or Botswana after independence, where diamond revenues were channelled through rule-bound fiscal institutions instead of discretionary patronage. In each case, the State traded control for predictability, and command for rules.
Reading the book felt quietly validating because it mapped cleanly onto how I had earlier defined agency: not as freedom in the romantic sense, but as the ability to calculate the future rather than merely react to the present. In that sense, agency is rivalrous. When multiple agents coexist—here, the State and its citizens—they compete to build more accurate models of the future and to decide which trajectories are worth betting on.
Agency itself is not scarce. What is rivalrous is the authority to decide which imagined futures are allowed to materialise. When agents are free to speculate, choose, and learn from failed bets, they pursue divergent paths—sometimes even learning to hack their own rewards. What looks like disorder from above is often just the system exploring its search space. Seen this way, Why Nations Fail reads less like a new theory and more like an application of my understanding of agency at civilisational scale: constrain agency and outcomes narrow; distribute agency and outcomes compound.
When the State exercises high agency, short-term success is often predictable. This is the intuition Seeing Like a State develops through the idea of high modernism—the belief that complex societies can be simplified, made legible, and then directed from the top. Consider scientific forestry in Prussia, where diverse forests were replaced with uniform, measurable tree grids optimised for timber yield; or Tanzania’s villagisation programme, which reorganised rural populations into planned settlements to make agriculture administratively visible. By suppressing local variation, the State reduces uncertainty and coordination costs. Large projects become easier to execute. Direction becomes clear.
But clarity comes at a cost. High-modernist schemes disregard local knowledge and competing models of the future. The State becomes the primary predictor; citizens become executors. The system works only as long as control is continuously maintained. Once that pressure weakens, fragility surfaces. Monoculture forests become ecologically vulnerable. Planned villages fail to produce expected yields. Feedback degrades because information must travel upward through bureaucratic layers that reward compliance more than accuracy. Errors are concealed, incentives drift, and reality arrives late.
When citizens exercise high agency and the State remains automatic, outcomes look very different. Progress is slower because coordination emerges from agents reacting to one another’s bets rather than from central instruction. Failure is more visible because mistakes are not hidden inside administrative hierarchies. In the short run, this can appear inefficient compared to centralised direction. But the system is doing something else: it is distributing responsibility. When the State confines itself to calculating its present—enforcing rules rather than predicting outcomes—its behaviour becomes a constant in the citizen’s model of the future. Citizens must calculate their own paths, make informed bets, and update their strategies subject to consequences. Over time, this tends to cultivate both humility and realism, because forecasts collide directly with results.
The State, in such a system, does not compete with private models; it stabilises the environment in which they operate. Scott’s cases repeatedly show that arrangements preserving local autonomy and experiential knowledge outlast those designed primarily for legibility and control. Read through the lens of agency, the trade-off is simple: coherence now versus resilience over time.
This reframes what we should expect from government. Its primary virtue is not benevolence, intelligence, or even efficiency, but predictability. A predictable State is one whose behaviour can be treated as a constant in the citizen’s model of the future. Rules are defined in advance. Procedures are applied mechanically. Discretion is minimised. In such an environment, individuals and firms do not need to interpret intent; they can plan around constraints. Contracts extend further into the future. Investments tolerate longer horizons. Private models compound because the boundary conditions stay fixed.
History suggests that when States instead try to be adaptive, visionary, or strategically omnniscient, they inject uncertainty into every calculation below them. Industrial policy that changes with each administration, tax regimes rewritten annually, licensing systems dependent on approvals rather than criteria—these do not merely alter incentives; they alter whether citizens can model the future at all. When prediction shifts upward, foresight disappears downward.
The uncomfortable conclusion follows. A civilisation that wants agentic citizens must accept an unglamorous State. No grand visions. No constant intervention. No attempt to out-think its own population. Just rules that hold, constraints that don’t move, and behaviour that can be relied upon. Agency, once clearly located, compounds. When repeatedly reclaimed, it decays.
If this claim feels too rigid, that discomfort is the point. The argument is not moral or ideological, but structural. If you disagree, the challenge is straightforward: point to a civilisation where both the State and its citizens independently and continuously calculated the future, over long stretches of time, without one displacing the other. Not temporarily. Not rhetorically. Structurally. Until such a case appears, the trade-off stands.