The Wide Coalition as an Adversarial Decision System

1. The Wrong Question: How Wide Can the Coalition Become?

The expression campo largo suggests a spatial optimisation problem. Let the political spectrum be represented by a policy space $X$, and let each party $i$ occupy an ideal point ${\theta }_i\in X$. The intuitive electoral objective is to construct a coalition $C$ large enough to maximise:

where $v_i$ is the expected electoral contribution of party $i$, while $\Lambda (C)$ is the number of votes lost because of ideological incoherence, candidate competition, abstention, tactical defection or the reduced credibility of the alliance.

This is the conventional perimeter problem. It asks whether the marginal party contributes more votes than the coalition loses by including it:

The difficulty is that winning an election and operating a government are two different games with different payoff functions.

Before an election, the coalition maximises votes. After an election, it must generate binding decisions. Its objective function changes from:

to:

where government performance depends on survival probability, decision quality and perceived legitimacy.

A coalition may therefore be electorally additive and operationally subtractive. Its parties can bring votes into the system while simultaneously reducing its ability to process decisions.

This distinction is particularly relevant because the political space conventionally associated with the campo largo contains actors whose positions overlap strongly on social policy but diverge on foreign policy, defence, fiscal discipline, nuclear energy, major infrastructure and judicial reform. The perimeter itself remains politically contested, and the parties are still presenting the programme as the mechanism through which the alliance is to be constructed rather than as an already completed contract.

Source note: Partito Democratico: coalition readiness and programme debate

The correct question is therefore not:

How many actors can be included?

It is:

Under what decision rules can heterogeneous actors produce binding outputs when confronted with events they did not schedule?

2. The Coalition as a Distributed Decision Architecture

A coalition government can be modelled as a distributed system.

Each party is an autonomous processing node. It possesses:

• its own electoral database;
• its own preference function;
• its own reputational constraints;
• its own leadership incentives;
• its own communication layer;
• its own definition of political failure.

The coalition programme acts as an interface specification. Cabinet procedures constitute the orchestration layer. Parliamentary groups execute the transactions. Public opinion provides an external validation environment. Elections are the periodic production release.

In a well-designed distributed architecture, components can be heterogeneous provided that they share:

• a common data model;
• explicit interface contracts;
• deterministic decision rights;
• exception-handling procedures;
• conflict-resolution protocols;
• observability over system status;
• predefined recovery strategies.

A coalition that lacks these elements is not a political architecture. It is a collection of services connected by undocumented APIs.

The system may appear functional during low-load conditions. It fails when an unexpected request reaches an endpoint for which no common response schema exists.

Let the coalition's policy vector be:

where:

• $x_1$ represents foreign and defence policy;
• $x_2$ represents fiscal and distributive policy;
• $x_3$ represents infrastructure and energy;
• $x_4$ represents justice and institutional reform.

Each party $i$ has an ideal vector:

A simple quadratic utility function is:

Here:

• $B_i$ is the benefit of holding office;
• $w_{ik}$ measures the importance of policy dimension $k$ to party $i$;
• $R_i(x)$ is the reputational cost of accepting a compromise;
• $E_i(x)$ is the electoral cost imposed by voters, activists, donors, media ecosystems and internal factions.

The quadratic distance term is not enough. Two parties may be equally distant from a compromise and experience radically different political costs because an issue is peripheral to one and constitutive of the other.

A ten-unit displacement on a technical procurement rule may be tolerable. A two-unit displacement on NATO, criminal justice or nuclear power may be interpreted as ideological surrender.

This means that coalition compatibility depends not only on distances:

but on weighted distances:

It also depends on whether the compromise is observable.

Let $o_k\in [0,1]$ measure the public observability of a decision. The reputational cost can be written as:

where $s_{ik}$ is the signalling sensitivity of party $i$ on issue $k$.

A closed technical negotiation may have low $o_k$. A roll-call vote on weapons, taxation, a nuclear plant or judicial independence has $o_k\approx 1$.

The same substantive compromise that is feasible in private may therefore become impossible when made public.

3. Programmable Agreement and Event-Driven Conflict

The most important distinction is between scheduled policy and event-driven policy.

Scheduled policies can be negotiated through ordinary coalition bargaining. Their implementation time $T$ is known, their content can be decomposed, and side payments can be arranged.

A healthcare package, for example, can often be divided into:

• staffing;
• capital investment;
• waiting-list management;
• digital infrastructure;
• regional transfers;
• pharmaceutical expenditure.

The multidimensional structure creates room for logrolling. One party can concede on component $a$ in exchange for influence over component $b$.

Suppose the policy package contains $m$ components. Party $i$'s utility is:

When $m$ is sufficiently large, compensating exchanges become possible:

The party accepts a loss on one component because it receives a gain elsewhere.

Exogenous crises work differently. Let shocks arrive according to a stochastic process:

where ${\lambda }_k$ is the arrival rate of shocks in domain $k$.

A military escalation, financial-market crisis, European fiscal constraint, industrial closure, judicial ruling or infrastructure emergency is not chosen by the coalition. It changes the state of the system from $s_t$ to $s_{t+1}$:

where ${\epsilon }_{t+1}$ is an exogenous event.

The government must then choose an action:

within a compressed decision window $\tau$.

As $\tau \to 0$, the possibility of bargaining, package deals and internal consultation declines. Complex policy space collapses into a binary choice:

Send weapons or do not send them. Authorise the plant or reject it. Raise the tax or cut the programme. Support the judicial reform or oppose it.

Coalitional ambiguity survives while policy is continuous. It becomes unstable when events discretise it.

We can define the crisis pressure associated with issue $k$ as:

where:

• $S_k$ is public salience;
• $O_k$ is vote observability;
• $I_k$ is ideological intensity;
• ${\tau }_k$ is the time available for bargaining;
• $C_k$ is compromise elasticity.

High salience, high observability, high ideological intensity, low time and low compromise elasticity create the ideal conditions for coalition fracture. The four domains examined below exhibit precisely this configuration.

4. Game One: NATO, Ukraine and European Defence

4.1 A Game of Chicken with Audience Costs

Foreign and defence policy creates the clearest form of strategic incompatibility because the relevant decisions are often binary, externally timed and difficult to compensate through side payments.

Within the potential coalition space, the Democratic Party has maintained support for Ukraine while combining it with opposition to indiscriminate national rearmament and advocacy of a common European defence. AVS has opposed continuing military supplies and emphasised diplomacy, while Azione has explicitly supported economic, humanitarian and military assistance to Ukraine and a stronger European defence capability.

Source note: Partito Democratico: support for Ukraine and European defence

These are not merely different rhetorical emphases. They imply different action functions after an external shock.

Consider two coalition actors:

• $A$: an Atlanticist or strongly pro-European defence party;
• $P$: a pacifist or anti-rearmament party.

Each can either Concede or Resist during negotiations over a military decision.

Payoffs for actor A and actor P under Concede and Resist strategies

Strategy	P: Concede	P: Resist
A: Concede	(2, 2)	(1, 4)
A: Resist	(4, 1)	(−5, −5)

The two asymmetric cells are Nash equilibria. Each party prefers the other to concede. Mutual resistance produces coalition crisis, government defeat or collapse. This is a classic game of Chicken.

The danger increases when leaders publicly precommit. Let $c_i$ be the cost of backing down after a public declaration. The utility of concession becomes:

A party can attempt to gain bargaining power by increasing its own concession cost:

Public declarations, party resolutions and maximalist interviews serve as commitment devices. By telling its electorate that compromise is impossible, a party reduces its own strategic flexibility and attempts to force the flexibility cost onto its allies.

This is rational at the party level but destructive at coalition level.

The coalition faces a commitment race:

Each party strengthens its negotiating position by making compromise more politically expensive. The aggregate result is a shrinking bargaining set.

Let $Z_A$ and $Z_P$ denote the sets of outcomes each party can accept. Coalition agreement exists only if:

Public commitment can move the reservation thresholds until:

At that point no negotiated statement can solve the problem. A parliamentary vote merely reveals that the feasible set had already disappeared.

4.2 Incomplete Information and Strategic Misrepresentation

Before elections, the parties do not know with certainty how rigid their allies will be after victory.

Let party $i$ possess a private type:

A pragmatic actor may compromise in exchange for coalition survival. A doctrinal actor may prefer government collapse to an identity-threatening concession.

Because types are private, coalition formation is a Bayesian game. Each party estimates:

where $M_j$ consists of public statements, voting history, leadership incentives and activist pressure.

Pre-election moderation may be cheap talk. A party can signal governability without accepting a binding future constraint.

The coalition programme therefore risks becoming a pooling equilibrium in which pragmatic and doctrinal actors issue the same ambiguous message:

Support peace, uphold international law, strengthen Europe and protect national interests.

The formula pools incompatible types because every actor can endorse it. It transmits almost no information about behaviour under a concrete contingency. The ambiguity disappears only when the state of the world requires a specific action.

4.3 The Opposition's Optimal Attack

A rational opposition will not ask the coalition whether it supports peace. It will generate high-resolution queries:

• Will the government approve the next military aid package?
• Will it authorise the use of supplied systems under specified conditions?
• Will it accept a NATO expenditure commitment?
• Will it participate in a European procurement mechanism?
• Will it deploy personnel to a deterrence or monitoring mission?

The adversary seeks to maximise the variance of coalition responses:

where $q$ is the wording of the parliamentary or political question.

A vague question produces low variance. A binary, time-bound and publicly recorded vote produces high variance.

The opposition's weapon is not necessarily persuasion. It is preference revelation.

5. Game Two: Fiscal Policy as a Common-Pool Resource Problem

5.1 The Budget Constraint Cannot Be Coalition-Negotiated Away

The second battlefield is fiscal policy.

Progressive coalition actors can agree that healthcare, education, wages, welfare and ecological transition require additional resources. The conflict begins when expenditure preferences must be mapped onto taxes, borrowing, cuts or reallocations.

AVS has advocated a progressive tax on very large fortunes to finance healthcare, education and ecological transition. The PD has called for a more progressive tax system and a statutory minimum wage. M5S programme documents emphasise minimum wages, income protection and public intervention, while liberal-centrist actors tend to place greater emphasis on investment selection, competitiveness and expenditure discipline.

Source note: Partito Democratico: progressivity and fiscal justice

The government budget identity is:

where:

• $G_t$ is primary public expenditure;
• $T{R}_t$ is transfers;
• $i_t{B}_{t-1}$ is debt-service expenditure;
• $T_t$ is tax revenue;
• $\Delta B_t$ is new borrowing.

No coalition narrative can invalidate this identity.

If each party proposes expenditure $g_i$ targeted at its constituency, total expenditure is:

Suppose party $i$ captures political benefit ${\alpha }_i{g}_i$, while the fiscal cost is distributed across the whole coalition. A simplified utility function is:

with:

Party $i$ chooses $g_i$ such that:

The coalition optimum, however, requires:

Because each actor internalises the electoral benefit of its own spending while externalising much of the fiscal cost, the Nash equilibrium tends toward excessive aggregate expenditure or excessive promises. This is a common-pool resource problem.

The resource being depleted is not simply money. It is the coalition's shared fiscal credibility.

5.2 The Pre-Election Auction

Before elections, parties compete not only against the government but against their prospective allies.

Let party $i$'s vote share depend on promised distributive benefit $g_i$:

where $T_i$ is the visible tax burden associated with its proposals.

If taxes and cuts are left unspecified, then $T_i\approx 0$ in the electoral narrative. The marginal electoral return on promising expenditure becomes artificially high:

This generates an internal auction. Each party has an incentive to increase promises before a shared fiscal framework has been agreed.

The coalition programme may then contain:

The inconsistency remains latent until the first budget law.

At that point, the system must allocate losses. Every budget is not merely a spending document; it is a loss-allocation algorithm.

Someone must receive:

• less expenditure than promised;
• more taxation than expected;
• a delayed reform;
• a cancelled subsidy;
• a reduced investment;
• a stricter eligibility threshold.

The politically relevant variable is therefore not the coalition's total policy surplus but the vector of deviations from party promises:

The first budget crisis occurs when one or more actors conclude that:

meaning that the benefit of remaining in government is lower than the combined policy and reputational loss.

5.3 A Two-Sided Opposition Trap

The opposition can attack from mutually inconsistent directions because it does not need to solve the government's optimisation problem.

Suppose the coalition chooses fiscal discipline $d$, where:

High $d$ means tight expenditure control; low $d$ means expansion.

The opposition's attack function can be defined as:

where:

represents attacks from voters demanding greater social expenditure, while:

represents attacks based on debt, taxation and fiscal irresponsibility.

Whatever the government chooses, the opposition selects the narrative with the highest expected political return:

If the government is fiscally cautious, the opposition tells progressive voters that the coalition has betrayed its social mandate. If it expands spending, the opposition tells businesses, creditors and centrist media that the coalition is financially reckless.

The opposition possesses an option portfolio. The government possesses a budget constraint. That asymmetry is structurally advantageous to the attacker.

6. Game Three: Infrastructure, Energy and the Veto-Player Trap

6.1 Binary Projects and Non-Compensable Preferences

Major infrastructure and energy decisions are especially dangerous because they generate concentrated and geographically visible effects.

A railway, nuclear programme, regasification terminal, waste-to-energy facility, industrial site or high-voltage interconnection cannot be implemented at 43 per cent intensity as a rhetorical compromise. In the relevant territory, the project is authorised or it is not.

Azione remains explicitly committed to restoring nuclear energy to the Italian energy mix. AVS supports a renewable and de-nuclearised energy system, while recent PD positions prioritise renewables, storage and grids and reject reliance on commercially immature nuclear solutions. The M5S 2026 programme process calls for a careful assessment of fission and fusion while maintaining a broader ecological-transition framework.

Source note: Azione: proposal on nuclear energy

For each project $j$, define:

where $q_j=1$ means authorisation.

Party $i$ supports the project when:

where:

• $E_{ij}$ is expected economic value;
• $S_{ij}$ is strategic value;
• $I_{ij}$ is electoral or territorial benefit;
• $C_{ij}$ is environmental, financial or social cost;
• $R_{ij}$ is reputational cost.

Different parties do not merely assign different values to the same variables. They may use different ontologies.

One party treats nuclear energy as a decarbonisation technology. Another treats it as an intergenerational environmental liability.

One treats a high-speed railway as network infrastructure. Another treats it as territorial extraction.

One treats a regasification terminal as strategic resilience. Another treats it as fossil-fuel lock-in.

The disagreement therefore occurs at the data-model level. The actors do not share the same definition of the object being evaluated.

6.2 The Veto-Player Theorem

Suppose a decision requires approval from all politically indispensable coalition actors. Let $A_i$ be party $i$'s set of acceptable projects.

The coalition's feasible project set is:

As the number of veto players increases, the intersection tends to shrink:

This is not universally true in a strict mathematical sense, but it is the normal outcome when additional players possess heterogeneous preferences.

Coalition width therefore has an operational cost. Every additional actor may contribute votes while adding another acceptance filter to the decision pipeline.

A project becomes impossible when:

The relevant power of a small party is not proportional to its vote share. It depends on whether its support is pivotal.

Let $v_i$ be the parliamentary weight of party $i$, and let $M$ be the majority threshold. Party $i$ is pivotal when:

A small actor may possess high effective power if the coalition cannot survive its defection. This can be measured using a Shapley–Shubik or Banzhaf power index rather than raw seat share.

The political consequence is predictable: a party with 5 per cent of coalition seats may exercise considerably more than 5 per cent of bargaining power on a project it can veto.

6.3 Project-by-Project Evaluation as Deferred Conflict

"Case-by-case assessment" appears moderate, but it can merely defer disagreement.

From a functional-analysis perspective, such a formula defines no business rule. It lacks:

• evaluation thresholds;
• weighted criteria;
• authoritative data sources;
• decision ownership;
• time limits;
• tie-breaking procedures;
• escalation paths.

Suppose a project is evaluated on economic return $r$, environmental impact $e$, strategic relevance $s$ and territorial consent $c$:

Unless the coalition agrees ex ante on the weights:

each party can produce a different score from the same project data.

The conflict is not caused by missing evidence. It is caused by different weighting functions.

A data-rich process does not automatically produce consensus. More data may simply increase the number of variables over which actors can disagree.

6.4 Adversarial Project Selection

A sophisticated opposition will select the project $j$ that maximises internal coalition dispersion:

where $S_j$ is salience and $O_j$ is observability.

The ideal wedge project has four properties:

1. it is materially concrete;
2. it requires a yes-or-no decision;
3. it activates different party identities;
4. it affects a recognisable territory or constituency.

The opposition can then frame the government's dilemma in opposite ways:

• authorisation proves ecological betrayal;
• rejection proves hostility to growth;
• delay proves incompetence;
• internal review proves paralysis.

The governing coalition controls the decision. The opposition controls the interpretation menu.

7. Game Four: Justice as an Identity and Constitutional Signalling Game

7.1 Why Justice Is More Explosive Than Ordinary Policy

Justice policy has an unusual payoff structure because it operates simultaneously on legal, constitutional, moral and symbolic levels.

In the 2026 referendum on the separation of judicial careers, the PD, M5S and AVS campaigned for No, presenting the reform as a threat to judicial independence, whereas Azione supported Yes and defended the separation of careers and the reform of judicial self-government.

Source note: Movimento 5 Stelle: referendum mobilisation on judicial reform

The disagreement is therefore not reducible to administrative efficiency.

Actors attach incompatible constitutional meanings to the same reform:

• judicial independence versus political accountability;
• equality before the law versus protection from prosecutorial power;
• anti-corruption capacity versus due-process guarantees;
• democratic control versus separation of powers.

Let the utility of party $i$ on justice policy be:

where:

• $P_i(x)$ is ordinary policy utility;
• $M_i(x)$ is moral-identity utility;
• $S_i(x)$ is signalling utility.

On many economic issues:

On justice issues, the inequality may reverse:

A technically efficient compromise can therefore remain politically unacceptable because it sends the wrong signal about the party's identity.

7.2 A Nearly Zero-Sum Symbolic Game

Suppose one actor frames a reform as protecting citizens from state power, while another frames opposition to it as protecting citizens from political impunity.

The symbolic gains are mutually exclusive.

Let $z$ represent public acceptance of the "guarantee" frame. Then:

where $G$ is the guarantee-oriented actor and $J$ the justice-and-legality actor.

The game is close to zero-sum:

Compromise language does not necessarily create joint value. It may instead blur both identities and impose losses on both parties.

This is why justice reform can produce harsher coalition conflict than a distributive measure of comparable material importance.

7.3 Repeated Interaction and Historical Memory

Justice games are repeated, not one-shot.

Parties remember previous conflicts over:

• corruption;
• statutes of limitation;
• abuse of office;
• wiretapping;
• pre-trial detention;
• relations with prosecutors;
• judicial appointments;
• criminal investigations involving political actors.

Let $h_t$ denote the accumulated history of interactions. Current utility is:

where ${\delta }_i$ is the discount factor applied to future reputational consequences.

A concession today may be interpreted as a precedent constraining the party tomorrow.

If the party believes that agreeing to reform $x$ weakens its future credibility on reform $y$, the cost is not local:

This path dependency makes compromise harder.

7.4 Judicial Shocks as Coalition Stress Tests

Justice policy is particularly exposed to endogenous and exogenous shocks:

• a politically sensitive investigation;
• a high-profile acquittal;
• a controversial pre-trial detention;
• publication of intercepted communications;
• a Constitutional Court ruling;
• disciplinary proceedings;
• a conflict between government and prosecutors.

Each event changes public salience and party incentives.

A coalition agreement written in abstract institutional language may fail when applied to a named individual or live investigation.

The opposition can exploit this by submitting motions, amendments, parliamentary questions and roll-call votes that force parties to choose between coalition discipline and identity preservation.

Its purpose is not necessarily to pass a law. It is to generate an observable conflict record.

8. The Opposition as an Adversarial Optimiser

8.1 Wedge Politics as Mechanism Design

The opposition should not be modelled as a passive beneficiary of coalition mistakes. It is an active mechanism designer.

It chooses:

• which issue enters the agenda;
• how the question is worded;
• when the decision is demanded;
• whether the vote is public;
• which constituency is activated;
• which internal contradiction is highlighted.

Let $q$ be an adversarial query submitted to the coalition.

The opposition maximises:

where:

• $D(q)$ is internal preference dispersion;
• $S(q)$ is public salience;
• $O(q)$ is observability;
• $U(q)$ is urgency;
• $K(q)$ is the cost of activating the issue.

The best wedge issue is:

This is analogous to adversarial testing in machine learning or software security. The attacker does not feed the system average inputs. It searches for inputs close to the decision boundary.

If the coalition classifier is:

the attacker seeks a query $q+\eta$ such that small contextual changes produce divergent classifications across coalition components:

The objective is to locate political adversarial examples.

8.2 The Fragmentation Index

A practical coalition-risk metric can be defined as:

where:

• ${\sigma }_k^2$ is the variance of party ideal points on issue $k$;
• ${\bar{w}}_k$ is average preference intensity;
• $o_k$ is observability;
• $s_k$ is public salience;
• $c_k$ is compromise elasticity;
• $p_k$ is the probability of an external shock requiring a decision.

The four domains discussed here all exhibit high values on several dimensions.

Foreign policy has high shock probability, urgency and observability.

Fiscal policy has unavoidable recurrence and a hard accounting constraint.

Infrastructure has low compromise elasticity and high territorial visibility.

Justice has high symbolic intensity and strong historical path dependence.

The opposition's rational strategy is to allocate attention proportionally to $F_k$, not merely to the intrinsic popularity of the issue.

8.3 Agenda Control as Attack-Surface Management

In cybersecurity, attack surface is the total set of points at which a system can be challenged.

A broad coalition has a large political attack surface because it contains more:

• preference discontinuities;
• historical contradictions;
• leadership rivalries;
• competing media networks;
• pivotal actors;
• veto thresholds.

Let coalition complexity be:

where:

• $n$ is the number of parties;
• $d_{ij}$ is pairwise ideological distance;
• $v$ is the number of veto points;
• $h$ is the accumulated weight of historical conflict.

Coalition complexity increases faster than party count because the number of bilateral relationships is:

Adding one party does not create one new relationship. It creates $n$ new relationships with the existing actors.

This is why coalition width can generate combinatorial governance complexity.

9. Pre-Election Conflict and Post-Election Conflict Are Not the Same Game

9.1 Before the Election: Differentiation Under a Shared Brand

Before voting, each party faces two objectives:

The party wants the coalition to win, but it also wants the largest possible share of seats, ministries and bargaining power inside the winning coalition.

These objectives are only partially aligned.

A party benefits from coalition unity, but it also benefits from differentiation. If all parties converge perfectly, voters may migrate toward the largest actor.

The pre-election equilibrium therefore requires optimal distinctiveness:

where $d_i$ is the distance between party $i$'s message and the coalition average.

Too little distance produces electoral cannibalisation. Too much distance destroys coalition credibility.

Each party seeks:

The result is a continuous struggle over how much public disagreement is compatible with a shared electoral project.

The four high-risk policy domains become ideal differentiation assets. Parties can reassure their core voters by refusing compromise precisely where coalition credibility requires it.

9.2 After the Election: The Portfolio Allocation Game

After victory, electoral differentiation is transformed into bargaining over portfolios, agenda control and veto authority.

Let ministries be resources $m\in M$, each with policy jurisdiction $J_m$.

Party $i$'s value for ministry $m$ is:

where $c_m$ is the ministry's effective control capacity.

Parties will value portfolios associated with their identity-sensitive issues disproportionately.

Foreign Affairs, Defence, Economy, Infrastructure, Environment and Justice become not merely administrative assignments but institutional control points.

A coalition may therefore reach agreement on policy only to reopen the same conflict through portfolio allocation.

Who controls the ministry?

Who appoints the deputy minister?

Who writes the implementing decree?

Who represents Italy in the European Council?

Who controls the parliamentary timetable?

Who has access to the relevant information first?

The government contract must specify not only policy outputs but process ownership. Without that layer, apparent programme agreement may conceal an unresolved control-rights game.

10. What a Robust Coalition Contract Would Require

A conventional political programme is insufficient. The coalition would require something closer to an operational governance framework.

10.1 A Foreign-Policy State Machine

The parties should define ex ante the admissible responses to specified scenarios.

Let the foreign-policy state be:

For each state, the coalition should define:

where $\pi$ is the agreed decision policy.

The agreement should specify:

• authorised categories of assistance;
• parliamentary approval thresholds;
• conditions for European missions;
• expenditure ceilings;
• reporting obligations;
• exit clauses.

Without a state-dependent policy, the coalition has not agreed on foreign affairs. It has agreed on a vocabulary.

10.2 A Fiscal Data Model

Every programme item should be represented as a structured record:

The programme should satisfy:

Promises without a funding source should be treated as invalid records rather than political prose.

A shared coalition control tower should monitor:

• baseline expenditure;
• forecast variance;
• implementation delay;
• revenue underperformance;
• debt-service sensitivity;
• European-rule compliance.

The key variable is not whether each party has a fiscal plan. It is whether all plans are reconciled into one canonical dataset.

10.3 A Named Infrastructure and Technology Registry

The coalition should classify major projects before elections:

For conditional projects, conditions must be machine-readable in political terms:

• maximum public cost;
• minimum expected return;
• environmental threshold;
• territorial consultation procedure;
• completion deadline;
• responsible decision authority.

The same principle should apply to technologies:

Without a named registry, "case-by-case assessment" becomes an unbounded exception process.

10.4 A Justice and Constitutional Protocol

The coalition should define a controlled scope of admissible reform.

It must specify common positions on:

• separation of judicial careers;
• judicial self-government;
• disciplinary jurisdiction;
• wiretapping;
• pre-trial detention;
• limitation periods;
• publication of investigative material;
• anti-corruption offences;
• political and administrative liability.

A generic commitment to "efficient and independent justice" is semantically compatible with almost every position and therefore operationally useless.

10.5 Breach Rules and Enforcement

No agreement is credible without consequences.

Let $b_i$ denote a breach by party $i$. The sanction must satisfy:

where $G_i(b_i)$ is the political gain from violating the coalition agreement.

Otherwise defection remains rational.

Possible enforcement mechanisms include:

• loss of agenda control;
• portfolio reallocation;
• mandatory coalition arbitration;
• confidence-vote rules;
• predefined free-vote domains;
• public breach reporting;
• escalation to coalition leadership.

The problem is political: parties rarely want to sign sanctions that may later be applied to themselves.

This produces a constitutional version of the commitment problem. Everyone wants allies to be bound, but no one wants to eliminate their own option to defect.

11. Coalition-Proofness and the Limits of the Agreement

A stable coalition outcome must be more than a Nash equilibrium. It should be coalition-proof.

A Nash equilibrium is stable against unilateral deviation. A coalition-proof Nash equilibrium is stable against deviations by subgroups that can form credible alternative agreements.

Let $x^{\ast }$ be the governing agreement. It is coalition-proof only if no subset $S\subset C$ can obtain:

through a feasible deviation $x_S$.

This matters because a broad coalition contains potential internal subcoalitions:

• a progressive-social bloc;
• an ecological-pacifist bloc;
• an Atlanticist-reformist bloc;
• a liberal-institutional bloc;
• a welfare-oriented populist bloc.

On different issues, different subcoalitions may align with the parliamentary opposition.

The government can therefore lose votes without formally collapsing. Policy-making becomes dependent on variable majorities.

Variable majorities are not necessarily dysfunctional, but they weaken collective responsibility. They also allow every party to claim credit for approved measures and disclaim responsibility for rejected ones.

The government then behaves less like a unified player and more like a legislative marketplace.

12. Data Culture Is Not a Substitute for Political Agreement

There is a temptation to believe that technical analysis can resolve ideological disagreement. It cannot.

Data can estimate:

• the fiscal cost of a tax measure;
• the expected traffic volume of a railway;
• the levelised cost of an energy technology;
• the expected effect of a procedural reform;
• the strategic consequences of a defence decision.

But data cannot determine the political loss function.

Suppose two parties agree on the same empirical estimate:

Party A approves the project if:

Party B approves it if:

If:

the parties disagree despite sharing the same data.

The conflict concerns thresholds, risk tolerance and normative weighting.

A mature data culture does not pretend that values disappear. It makes them explicit.

The purpose of quantitative modelling is therefore not to manufacture artificial consensus. It is to identify:

• which assumptions differ;
• which variables are contested;
• which thresholds are decisive;
• which risks are being transferred;
• which actors bear each cost;
• where no feasible compromise exists.

Good analysis reduces ambiguity. It may reveal that disagreement is deeper than political communication admits.

That is not analytical failure. It is data quality.

13. The Central Strategic Result

The stability of the campo largo can be expressed as:

where:

• $B$ is the benefit of office;
• $K$ is the size of the common policy core;
• $E$ is the strength of enforcement mechanisms;
• $D$ is weighted ideological dispersion;
• $V$ is veto-player complexity;
• $A$ is adversarial pressure.

Coalition survival requires:

The common mistake is to focus almost exclusively on $K$, the policies on which the parties agree.

But under crisis conditions, the other variables may dominate.

An opposition cannot easily reduce the coalition's common support for public healthcare or minimum wages. It can, however, increase $A$ by selecting divisive issues, increase $V$ by making every actor pivotal, and increase the effective value of $D$ by forcing public binary votes.

The opposition's optimal strategy is:

subject to minimal political cost.

The government's defensive strategy must be:

while reducing the number of unresolved decision rules.

The two sides are not playing the same game.

The coalition seeks to preserve a multidimensional bargaining space. The opposition seeks to collapse that space into a sequence of binary decisions.

Conclusion: A Coalition That Can Agree on What It Opposes

The campo largo is not impossible because its constituent parties have nothing in common. They share a meaningful policy area and a powerful negative preference: replacing the Meloni government.

The problem is that shared preferences are concentrated in domains that can often be programmed, negotiated and decomposed. The deepest disagreements concern domains that are frequently activated by external events and require rapid, observable and binary decisions.

This creates four structural vulnerabilities.

Foreign policy becomes a game of Chicken in which parties attempt to force the cost of concession onto their allies.

Fiscal policy becomes a common-pool problem in which every actor internalises the electoral value of expenditure and externalises part of the cost.

Infrastructure and energy become veto-player games in which additional coalition width reduces the set of feasible decisions.

Justice becomes an identity-signalling game in which technical compromise can generate greater reputational loss than open conflict.

A strategically competent opposition would not attack the coalition primarily where it is strongest. It would implement a wedge strategy: identify the variable with the highest fragmentation index, increase its salience, compress the decision time, demand a roll-call vote and publicise the cost of every possible response.

The coalition's survival would therefore depend less on the literary quality of its programme than on the precision of its decision architecture.

It would need binding foreign-policy contingencies, a reconciled fiscal dataset, a named infrastructure and technology registry, a justice protocol, explicit decision ownership and credible breach rules.

Without these mechanisms, electoral victory would merely migrate the conflict from campaign communication to government execution.

The campo largo could win the aggregation game and immediately begin losing the governance game.

It could become a coalition perfectly capable of agreeing on what it does not want, yet structurally incapable of deciding what it must do.