Each ministry, administration or level of government acts within a defined perimeter, with its own objectives, constraints and indicators. A decision can be rational at that scale and produce, when it interacts with other policies, consequences that no one would have chosen as a whole.

I use the term decoherence here to designate the progressive loss of compatibility between locally justifiable decisions, their timescales, the available means and the effects they produce collectively. The word remains a political metaphor, distinct from the precise mechanism to which it refers in physics.

The energy crisis of 2021 to 2023 offers a concrete illustration. Faced with the abrupt rise in prices, France protected households and businesses through price caps, various forms of aid and a fuel rebate. In its July 2024 report on public finances, the French Court of Auditors puts the net cost, in expenditure, of the support measures against rising energy prices at €20.8 billion in 2022 and €17 billion in 2023.

These interventions answered a strong economic and social constraint. Some of them simultaneously blunted the price signal while other policies sought to reduce fossil-fuel consumption. Protecting purchasing power, supporting economic activity, containing public spending and pursuing the energy transition each answer a defensible logic. Their coexistence imposes trade-offs that a separate reading of public policies grasps poorly.

Sectoral institutions in a networked reality

Public action rests on a division of responsibilities. The State functions through ministries, administrations, budget programmes, territorial competences and specialised indicators. This organisation makes it possible to handle a considerable quantity of problems. It runs into greater difficulty when the effects of a decision cross several domains and reappear far from their starting point.

The distinction between complicated and complex systems illuminates this gap. A complicated system may comprise a great many elements while remaining largely decomposable. In a complex system, interactions take part directly in the behaviour of the whole. They produce feedbacks, delayed effects and sometimes consequences remote from the initial decision.

Contemporary public policies unfold within this kind of network. Energy acts on industry, transport, housing and purchasing power. Employment influences tax revenue and social spending. Agriculture bears on questions of water, soils, biodiversity, food and trade. Budgetary choices determine part of the investment capacity that will in turn weigh on future spending.

Institutions logically continue to carve up these domains in order to administer them. The real continually crosses their boundaries. To a large extent we thus use an organisation chart to govern a network.

This structure favours a highly developed sectoral learning. A ministry can better control its spending, an administration reduce its delays, a service improve its results. Consequences located outside the perimeter remain harder to track. A saving made by the State may increase a local expense. A regulatory change may lighten the burden on one service and create another elsewhere. An indicator may improve while part of the observed phenomenon has simply shifted into another category.

The necessary data often exist, distributed among several bodies. Assembling them remains far more difficult than producing them. A fall in expenditure tells us rather little about the collective result when the need it answered persists elsewhere or reappears a few years later.

When costs change place or period

The maintenance of the French road network illustrates this displacement in time. An audit carried out in 2018 estimated that at least one billion euros of annual investment was needed to halt the deterioration of the non-conceded national road network. The Senate subsequently updated this need to about €1.2 billion a year taking inflation into account. For engineering structures, the trajectory set at €150 million a year between 2023 and 2027 corresponded to about €170 million after adjustment, while maintenance spending stood at around €130 million in 2024.

Deferring part of the maintenance is understandable when a budget must answer several urgent needs at once. A succession of deferrals nevertheless turns the immediate saving into a material debt. The need remains, the infrastructure goes on ageing, and the cost reappears later in the form of repairs or of a catching-up that has become hard to postpone.

Each annual decision may remain defensible while their succession progressively reduces future margins.

Le cost of maintenance makes it possible to observe this evolution. An infrastructure, an administration or a policy may continue to function while mobilising more repairs, procedures, subsidies or resources to preserve a comparable performance. Stability remains visible, while its price rises.

The same logic can apply to environmental costs. An activity can produce an immediate economic benefit while part of its consequences is displaced onto ecosystems, other territories or future generations. These effects rarely appear in the indicators used to evaluate the initial decision. Biophysical limits nevertheless give this displacement an important particularity: some forms of degradation directly reduce capacities of regeneration or recovery and cannot be compensated indefinitely by new spending, new technologies or new displacements of resources.

This environmental extension does not turn political decoherence into a general explanation of how the economy works. It simply shows that the same mechanism of displacement can concern budgetary, social, material or ecological costs.

Compensatory mechanisms sometimes follow a similar trajectory. A policy produces an unwanted effect, and then a subsidy, an exemption or a supplementary rule comes to attenuate its consequences. The correction may answer a real necessity. Over time, it becomes part of ordinary functioning, modifies behaviour and creates its own beneficiaries.

In its work of October 2024, the French Council on Compulsory Levies counted 170 tax expenditures linked to income tax in the 2024 budget bill. The report also takes up the expression« zombie effect », drawn from a 2014 European Commission study, to describe the persistence of a tax break after its initial justification has disappeared, when governance difficulties and the appropriation of the advantage favour its retention.

This example does not allow the finding to be extended to all tax expenditures. It shows how a response created in a precise context can become progressively harder to remove because behaviours, investments and interests have organised themselves around its existence.

The dynamic can follow a fairly simple sequence:

initial difficulty → response → side effect → compensation → dependence

Modifying the original policy then requires taking account of everything that has developed around it. The cost of correction becomes more visible than the cost of retention, which is widely spread across budgets, taxpayers or future years.

When the trade-off is postponed

Political time accentuates this tendency. The benefits of a decision may appear rapidly, while part of its costs emerges only after several budget years or several terms of office. Deferred spending immediately frees resources. A compensation reduces the visible losses of a reform. A transfer to another administration eases the balance of the budget that decided it.

Part of the decoherence appears without any particular intention, as interactions become hard to track. Another part may be sustained when a known trade-off would cost more, politically, than its postponement.

The parliamentary fragmentation in France following the 2024 elections raises the cost of certain compromises by making durable majorities harder to assemble. It acts on the conditions of decision without explaining the administrative silos, the maintenance backlogs or the tax provisions accumulated well before that period.

The institutional reaction after a contradiction has been brought to light gives a useful indication. A relatively rapid correction suggests that a problem of information or coordination played an important role. When a difficulty remains documented for several years without substantial change, constituted interests, created dependencies, legal constraints, budgetary costs and postponed trade-offs carry more weight in the analysis.

Complexity can thus explain certain incoherences and, in other situations, supply a convenient vocabulary for prolonging a choice that no one wishes to own explicitly.

Recognising a trajectory of decoherence

The concept would quickly lose its usefulness if it served to rename every political error. Three elements make it easier to identify an established trajectory.

The first concerns effects outside the perimeter. A policy continues to be evaluated mainly against its initial objective while significant consequences appear in other budgets, sectors or timescales.

The second bears on the cost of correction. Renewal becomes easier in the short term than modification, while the necessary spending, constraints or compensations continue to accumulate.

The third concerns the traceability of responsibilities. Transfers of burden, exemptions and successive corrections progressively make it harder to identify the actor able to modify the trajectory. The collective result ends up arising from a series of decisions none of which, on its own, suffices to explain the whole.

This dilution reinforces inertia. A significant reform requires reopening several decisions taken at different moments, by different institutions and sometimes for reasons that have since disappeared. Those liable to lose an advantage are immediately identifiable, while the beneficiaries of a future improvement remain diffuse.

The legal possibility of change may persist for a long time. The margins allowing it without budgetary, social or political rupture diminish progressively. A trajectory becomes particularly hard to correct when continuing it remains the least costly short-term option for almost all the actors able to modify it.

This reading makes it possible to distinguish coherence from uniformity. A democracy lives on disagreements, compromises, experiments and competing priorities. The coherence envisaged here rests on a more limited requirement: preserving a sufficient legibility of interactions and trade-offs.

A policy may favour one objective at the expense of another. The debate gains in quality when those responsible can specify, within the limits of available knowledge, which consequences are expected, who is likely to bear them and over what horizon they may appear.

Tracking displaced costs

One simple measure would improve this visibility. As a proposal, any reform announcing an annual budgetary saving above €500 million could be accompanied by a public statement of the main costs liable to be displaced outside the perimeter that records that saving.

The €500 million threshold is an order of magnitude intended to focus the exercise on major reforms. A cumulation rule would be needed to prevent a single policy direction from being artificially split into several smaller measures.

The statement would identify the expected effects on other administrations, local authorities, households, public operators and deferred investment and, where the field warrants it, on certain environmental objectives. Consequences hard to quantify would still be presented with their degree of uncertainty. Institutions liable to bear a transferred burden could publish their own assessment where it diverges from that of the administration behind the reform.

Such a procedure would never produce a perfectly neutral measure. The choice of the effects examined, of the period adopted and of the method of calculation necessarily influences the result. Confronting several assessments would at least make those choices visible.

Follow-up over time would complete the arrangement. A review after three years, and possibly after five for policies with durable effects, would allow the anticipated consequences to be compared with those actually observed. This memory should remain accessible despite changes of government and administrative reorganisations. Existing oversight institutions already hold part of the competence needed to preserve and confront this information.

Information alone decides nothing. Those who would immediately bear the constraint of such transparency are often easier to organise than those who would benefit from it several years later. The statement would be worth something only if it remained public, comparable over time and usable by Parliament, the financial jurisdictions and public debate.

Its function would remain limited: to make certain displacements of cost more visible and to allow the announced effects to be confronted, a few years later, with those actually observed.

A society can accept contradictory policies when their trade-offs are known and owned. The difficulty increases when costs change budget, territory or generation, when compensations accumulate and when responsibility for the whole becomes progressively hard to trace.

The hardest problem does not appear when a bad decision leads to a bad result. It appears when hundreds of reasonable decisions, taken separately, end up producing a result that no one would reasonably have chosen as a whole.

Analysing a complex political system then requires looking beyond each decision taken separately, following what it displaces, what it obliges us to compensate for, and the margins it leaves when the moment comes to alter the trajectory.