Chapter 12: Coalition Building as Pareto Optimization

The Geometry of Coalition

If voting collapses the manifold to a scalar, coalition building partially preserves it. A coalition is a subset of voters or political actors who agree to coordinate despite disagreeing on some dimensions. They share a connected submanifold on which their positions are close, even if their positions on other dimensions are far apart.

Coalition building is the search for this shared submanifold — the process of identifying which dimensions unite potential partners and which dimensions divide them, then constructing a political program that emphasizes the uniting dimensions and brackets the dividing ones.

Definition 6 (Agreement Submanifold). For a coalition C = \{v_1, \ldots, v_k\} of k agents on the preference manifold \mathcal{P}, the agreement submanifold A_C is the subspace of \mathcal{P} on which all coalition members’ positions are within \epsilon of each other:

A_C = \{d_i : \max_{v_a, v_b \in C} |v_a^{(i)} - v_b^{(i)}| \leq \epsilon\}

The dimensionality of A_C measures the depth of the coalition: a coalition that agrees on one dimension (\dim(A_C) = 1) is fragile; a coalition that agrees on four dimensions (\dim(A_C) = 4) is robust.

Definition 7 (Pareto-Optimal Coalition). A coalition C is Pareto-optimal on the manifold if there is no alternative coalition C' of the same size that is strictly better for all members — closer to each member’s ideal point on the manifold. Formally, C is Pareto-optimal if there is no C' with |C'| = |C| such that d_\mathcal{P}(v, \mu_{C'}) \leq d_\mathcal{P}(v, \mu_C) for all v \in C and strictly less for at least one v.

The Coalition Geometry of American Politics

The two major American party coalitions, analyzed geometrically, have surprisingly thin agreement submanifolds.

Coalition Fragility and the Manifold

A coalition fractures when external events or internal dynamics increase the variance of members’ positions on a dimension that was previously within A_C.

Theorem 6 (Coalition Fragility). A coalition C with agreement submanifold A_C of dimension \dim(A_C) = k is vulnerable to fracture on any dimension d_i \notin A_C for which the within-coalition variance \sigma_i^2(C) exceeds the tolerance threshold \epsilon^2. The fragility index of the coalition is:

F(C) = \frac{d - \dim(A_C)}{d} \cdot \max_{d_i \notin A_C} \frac{\sigma_i^2(C)}{\epsilon^2}

A coalition with F(C) > 1 is geometrically fragile: it contains more disagreement energy on its worst dimension than its agreement structure can contain.

For the American party coalitions with \dim(A) \approx 2 and d = 6, the fragility coefficient (d - \dim(A))/d = 4/6 \approx 0.67, meaning that the fragility index reaches 1 (the fracture threshold) when the within-coalition variance on the worst disagreement dimension exceeds 1.5 \cdot \epsilon^2. Both coalitions are operating near this threshold, explaining the persistent anxiety within both parties about coalition stability.

Cross-Partisan Coalitions on the Manifold

The most geometrically interesting coalitions are cross-partisan: they unite voters from opposing parties who agree on a suppressed dimension.

In the 1D party system, cross-partisan agreement is invisible. Voters who agree on d_3 (environment) but disagree on d_1 (economics) are placed on opposite sides of the left-right axis. Their agreement on d_3 cannot be expressed through the voting system, because the voting system projects onto d_1 (or the d_1-d_2 diagonal), and the d_3 agreement is orthogonal to this projection.

On the manifold, the cross-partisan coalition is visible. Voters from both parties who share a region of d_3 space form a connected submanifold on the environmental dimension. The coalition exists; the voting system simply cannot see it.

Historical Realignments as Manifold Restructuring

The major party realignments in American political history — the New Deal realignment of the 1930s, the civil rights realignment of the 1960s, the Reagan realignment of the 1980s, and the populist realignment of the 2010s — can be understood as restructurings of the manifold’s covariance matrix.

The New Deal realignment (1930s): Before the New Deal, the d_1-d_2 correlation was weak: economic conservatism and social conservatism were not strongly linked, and the parties were internally diverse on both dimensions. The New Deal activated d_1 as the dominant axis: the Democratic party became the party of economic progressivism (labor unions, welfare state, regulation), and the Republican party became the party of economic conservatism (business, tax cuts, deregulation). The covariance matrix shifted: \Sigma_{12} increased as economic and social positions began to align along party lines.

The civil rights realignment (1960s): The Civil Rights Act of 1964 and the Voting Rights Act of 1965 activated d_2 (social values) and d_6 (identity) as major axes of partisan differentiation. Southern Democrats — economically progressive but socially conservative and racially segregationist — found their d_2-d_6 positions increasingly at odds with the national Democratic party. They migrated to the Republican party over the following three decades, dramatically increasing \Sigma_{12} and \Sigma_{16} — the correlation between economics, social values, and identity.

The Reagan realignment (1980s): Reagan fused economic conservatism (d_1), social conservatism (d_2), and foreign policy hawkishness (d_4) into a single coalition identity. The three-dimensional fusion increased \Sigma_{14} (economic-foreign policy correlation) and cemented the high \Sigma_{12} created by the civil rights realignment. The Reagan coalition’s agreement submanifold was approximately three-dimensional: economics + social values + foreign policy.

The populist realignment (2010s): The Trump era activated d_5 (institutional trust) and d_6 (identity) as the dominant axes, displacing d_1 (economics) from its New Deal-era primacy. The covariance matrix shifted again: \Sigma_{56} surged, as institutional distrust and identity salience became strongly correlated. Meanwhile, \Sigma_{14} weakened, as the Republican party’s traditional free-trade and interventionist positions were displaced by economic populism and isolationism.

Each realignment is a restructuring of the covariance matrix — a change in which dimensions are correlated and which are independent. The new correlations define new coalition structures: the agreement submanifolds of the parties shift to accommodate the new covariance structure. The old coalitions, built on the old covariance structure, fracture when the structure changes. The fracture is not a failure of party management; it is a geometric consequence of manifold restructuring.

The Bond Geodesic Equilibrium for Coalitions

Coalition formation has the structure of a multi-agent equilibrium problem, connecting to the Bond Geodesic Equilibrium from Geometric Economics (Ch. 7).

Each potential coalition member faces a decision: join the coalition (and compromise their position to the coalition’s agreement region) or stay independent (and maintain their ideal position but forfeit the coalition’s collective benefits). The BGE for coalition formation is the stable partition of the electorate into coalitions where no voter can improve their representation by switching coalitions.

Formally, let \{C_1, \ldots, C_m\} be a partition of voters into m coalitions (including the possibility of single-member “coalitions” for independent voters). The partition is a coalition BGE if no voter v has a lower behavioral friction in any alternative coalition C_j than in their current coalition C_i:

\text{BF}(v, C_i) \leq \text{BF}(v, C_j) \quad \forall v \in C_i, \forall j \neq i

where \text{BF}(v, C) is the cost to voter v of membership in coalition C — the manifold distance between v’s ideal position and the coalition’s agreed-upon position on the agreement submanifold.

The coalition BGE predicts that voters will sort into coalitions that minimize their representational cost on the manifold — not on the 1D axis. The two-party system is a stable equilibrium only if the two-party partition is the BGE — which requires that no voter can achieve lower manifold-distance representation by switching parties or joining a third party. When the manifold structure changes (new dimensions become salient, old correlations weaken), the two-party equilibrium can become unstable, and realignment — the formation of new coalitions around new agreement submanifolds — becomes possible.

District 7: The Bond Measure Coalition

The environmental bond measure in District 7 assembles a cross-partisan coalition:

• Progressive urban voters (30% of district): Support on d_3 (environmental priority) and d_1 (public investment). Manifold distance to the bond measure’s position: 0.8.
• Moderate suburban parents (25%): Support on d_3 (clean air for children, reduced commute) and d_1 (infrastructure investment). Manifold distance: 1.2.
• Libertarian exurban voters (10%): Support on d_3 (conservation easement protects property values). Oppose the tax increase (d_1) but accept it as the price of development restriction. Manifold distance: 2.0.
• Retired pragmatists (8%): Support on d_1 (infrastructure as sensible investment) and moderate d_3. Manifold distance: 1.5.

Total support: approximately 73% — well above the 60% threshold, if the coalition holds.

On the left-right axis, this coalition is incoherent: it includes voters from every point on the spectrum. The political consultant, analyzing the district through the 1D lens, advises that the measure “can’t pass because it doesn’t have a partisan base.” The consultant is thinking one-dimensionally.

On the manifold, the coalition occupies a connected region of d_3 space. The agreement submanifold A_C has dimension 1 (agreement on environmental priority) with supplementary agreement on d_1 for most members. The coalition is fragile on d_2 (social values divide the urban progressives from the exurban libertarians) and d_5 (institutional trust divides the suburban moderates from the exurban populists). A campaign that reframes the bond measure as a social issue (d_2) or an institutional issue (d_5) could fracture the coalition by activating the disagreement dimensions.

The bond measure’s advocates, intuitively understanding the manifold geometry, keep the campaign focused on d_3: clean energy, clean air, open space, property values. They avoid d_2 (no social commentary) and d_5 (no institutional arguments — no “trust the government” messaging). The campaign holds the coalition together by holding the projection axis on the agreement dimension.

The measure passes with 62% support.

DISTRICT 7 — CHAPTER SUMMARY

We have analyzed coalition building as a search for shared submanifolds on the political preference manifold. A coalition’s depth is measured by the dimensionality of its agreement submanifold; its fragility is measured by the variance of its members on disagreement dimensions. Both major American party coalitions have agreement submanifolds of dimension approximately 2, making them geometrically fragile.

Cross-partisan coalitions — united on suppressed dimensions that the 1D axis cannot see — are visible on the manifold and can succeed when the electoral process activates the agreement dimension. In District 7, a cross-partisan environmental coalition spanning the full left-right spectrum passes a bond measure by holding the campaign’s projection axis on the shared environmental dimension.

Part IV turns to democratic design: how can voting systems be redesigned to preserve more manifold structure? Chapter 13 begins with ranked-choice voting and the RCV Recovery Theorem.