Correction, 30 October 2026. Proposition (Subsidy Incidence, corrected). In the precise-signal limit with ordered thresholds x_A* < x_C* < x_R*: (a) the partial derivative of x_j* with respect to s_k is 1/delta_j if j = k and zero otherwise; (b) the full-cascade threshold x_R* is lowered only by a subsidy to Type R, by Delta/delta_R; (c) the weighted mean threshold sum_j w_j x_j* is lowered by a subsidy to type k by w_k Delta/delta_k. A summary of this correction is on the papers page at https://www.susarb.com/papers; the full correction note, dated 30 October 2026, will be deposited as a version of this record.
| title: "The Scalar Lobby: Political Economy of the Transition from Scalar to Multi-Dimensional Observation" |
| author: "Gregory Caldwell Beier" |
| institute: "Susarb LLC, Cambridge, Massachusetts" |
| date: "March 2026" |
Computer Science, Mathematics, and Physics
by Gregory Caldwell Beier (forthcoming 2026).}
{
--
This paper models the political economy of transitioning from scalar to multi-dimensional observation as a heterogeneous global game. Three stakeholder types - asset managers, compliance professionals, and scalar producers (credit rating agencies, index providers) - receive noisy private signals about a Fragility Gap measuring the cost of remaining with scalar systems. Applying the heterogeneous global games framework of Sakovics and Steiner (2012), the paper derives type-specific adoption thresholds and three central results: (i) the transition is sequential, with asset managers adopting first; (ii) the most cost-effective policy subsidizes early adopters, not resisters, because asset managers impose the largest externalities on other types; and (iii) a partial-adoption equilibrium exists in which early adoption relieves enough systemic pressure to prevent full transition. The economic engine is the Grossman-Stiglitz information rent on constraint dimensions the scalar market cannot price, which enters the model as a competitive disadvantage borne by holdouts that increases with aggregate adoption. The rent lowers adoption thresholds but stabilizes partial adoption through the same self-limiting mechanism Grossman and Stiglitz identified; the cascade to full adoption requires the Fragility Gap to exceed the thresholds the rent has lowered.
Keywords: global games, heterogeneous coordination, political economy, switching costs, scalar compression, observation architecture, tipping threshold, cascade, Beierspace
--
The Prolegomena's technical argument is information-theoretic: scalar channels cannot maintain high-dimensional coordination structures, the unmaintained dimensions dissipate, and supplementary multi-dimensional channels are the structural remedy (Beier, 2026h; 2026i). The adoption problem is different. A vast institutional ecosystem has been optimized around scalar information over the past century. The literature on socio-technical transitions (Geels, 2002; Geels and Schot, 2007) identifies this as a regime-level lock-in: the observation architecture is a configuration of technologies, institutions, user practices, and regulations that co-evolve and mutually reinforce.
This paper models the transition as a heterogeneous global game. The global games methodology (Morris and Shin, 1998, 2003) resolves the multiplicity of equilibria in coordination games by introducing incomplete information, yielding a unique equilibrium in threshold strategies with sharp comparative statics. The heterogeneous extension (Sakovics and Steiner, 2012) allows agents with different switching costs and network benefits to have different adoption thresholds, and characterizes which agents matter most for the coordination outcome.
The combination gives the political economy of the observation-architecture transition a formal structure with quantitative predictions: who adopts first, who resists longest, where subsidies are most effective, and under what conditions the transition cascades from early adopters to holdouts.
The transition from scalar to multi-dimensional observation affects stakeholders differently depending on their institutional function, revenue model, and professional expertise. We identify three types, ordered by their switching costs.
Definition (Type A: Asset Managers and Long-Horizon Investors).
Sovereign wealth funds, endowments, pension funds, and active asset managers whose institutional function is capital allocation. Their switching cost \(s_A\) is low because they can reallocate research budgets incrementally, add constraint-dimension data to existing analytical frameworks, and integrate multi-dimensional risk factors without restructuring their institutions. Multi-dimensional information directly improves portfolio construction and reduces tail risk from scalar-masked fragility (Beier, 2026k), giving Type A the highest network benefit \(\bar{g}_A\) of any type.
Definition (Type C: Compliance Professionals, Regulators, and Auditors).
Financial regulators, accounting firms, compliance officers, and standards bodies whose institutional function is monitoring and enforcement. Switching requires workforce retraining, legal framework redesign, enforcement system adaptation, and cross-jurisdictional coordination (comparable in scope to implementing Basel III), placing \(s_C\) in the moderate range. The network benefit \(\bar{g}_C\) is moderate but diffuse: multi-dimensional observation improves regulatory effectiveness and risk detection, though the improvement is hard to attribute to any individual regulator's adoption decision.
Definition (Type R: Scalar Producers).
Credit rating agencies, index providers, standardized testing organizations, and sell-side analysts whose primary output is a scalar compression of multi-dimensional information. These actors face the highest switching cost \(s_R\): franchise depreciation (the regulatory moat of the NRSRO/ECAI designation), human capital loss (decades of consensus-dimension expertise), and revenue model restructuring. In a fully multi-dimensional world, demand for scalar compression declines, making \(\bar{g}_R\) low or negative at the margin.
The types differ not only in their private costs and benefits but in the externalities they impose on each other when they adopt. Define \(E_{jk}\) as the marginal benefit to type \(k\) from type \(j\)'s adoption.
Adoption by: | Type A | Type C | Type R
Type A (Asset Mgrs) | \(E_{AA}\) (moderate) | \(E_{AC}\) (high) | \(E_{AR}\) (high)
Type C (Compliance) | \(E_{CA}\) (low) | \(E_{CC}\) (moderate) | \(E_{CR}\) (moderate)
Type R (Scalar Prod.) | \(E_{RA}\) (low) | \(E_{RC}\) (low) | \(E_{RR}\) (low)
The asymmetry is the key structural feature. When asset managers adopt multi-dimensional observation, they demand constraint-dimension data from portfolio companies. This demand creates disclosure pressure that regulators codify into standards (high \(E_{AC}\)) and that rating agencies must incorporate to remain relevant (high \(E_{AR}\)). The flow runs predominantly in one direction: Type A's adoption reshapes the information environment for Types C and R, but Types C and R's adoption does comparatively little for Type A's portfolio construction.
The asymmetry is asserted based on the institutional logic of capital market information flows, not derived from a micro-model of inter-type interaction. A competing argument - that rating agencies set benchmarks constraining asset managers, making \(E_{RA}\) high - is plausible in the current scalar regime but less so in a multi-dimensional regime where the demand signal originates from investors. The model's qualitative predictions (sequential adoption, subsidy targeting) are robust to moderate perturbations of the externality structure but would change if the asymmetry reversed. The ordering should be tested empirically, and the results below are conditional on it.
Sakovics and Steiner's (2012) central result is that optimal subsidies target agents who impose high externalities on others and on whom others impose low externalities. In this setting, the subsidy should target Type A - a prediction the model formalizes below.
We provide order-of-magnitude estimates of switching costs over a ten-year transition horizon, discounted at 7
Type A. Research budget reallocation ($3 -- 5B/year for the global institutional investor base, PV \(\approx\) $21 -- 35B), plus data infrastructure ($5 -- 10B one-time). Net: \(s_A \approx\) $26 -- 45B.
Type C. Standards development ($5 -- 8B, extrapolating from ISSB spending), workforce retraining ($10 -- 15B across global regulatory and audit professions), enforcement redesign ($10 -- 20B across major jurisdictions). Net: \(s_C \approx\) $25 -- 43B.
Type R. Franchise depreciation for rating agencies ($15 -- 30B, 30 -- 50
The ordering \(s_A < s_C < s_R\) confirms the type classification. The aggregate switching cost across all types is $111 -- 180B over a decade, or roughly 3 -- 5
A continuum of stakeholders drawn from three types \(j \in \{A, C, R\}\) with population weights \(w_A, w_C, w_R\) (summing to 1) simultaneously choose whether to adopt multi-dimensional observation (\(a_i = 1\)) or remain with scalar systems (\(a_i = 0\)). Let \(\ell_j = \int_{i \in j} a_i di / w_j\) denote the adoption fraction within type \(j\), and \(\ell = \sum_j w_j \ell_j\) the aggregate adoption fraction.
The Fragility Gap \(F\) is drawn from an improper uniform prior on \(\mathbb{R}\). \(F\) is measured in the same units as the switching costs (dollar-equivalent cost per unit of non-adoption over the transition horizon), so that the net payoff equation is dimensionally consistent. Each stakeholder \(i\) of type \(j\) observes a private signal:
\(x_i = F + \varepsilon_i, \varepsilon_i \sim \mathcal{N}(0, 1/\beta)\)
where \(\beta > 0\) is the common signal precision. Signals are conditionally independent given \(F\).
Type \(j\)'s payoff from adopting is:
\(\pi_j(1, \boldsymbol{\ell}, F) = \sum_{k} E_{jk} w_k \ell_k + \bar{g}_j \ell_j - s_j\)
The terms capture, respectively, cross-type externalities, within-type network effects, and the switching cost.
Type \(j\)'s payoff from remaining with scalars is:
\(\pi_j(0, \boldsymbol{\ell}, F) = -F^+ \delta_j - c_j \ell\)
where \(F^+ = \max(F, 0)\) is the effective Fragility Gap, \(\delta_j > 0\) is the type-specific idiosyncratic fragility exposure, and \(c_j \geq 0\) is the competitive disadvantage of operating on scalar methods when a fraction \(\ell\) of the system has adopted multi-dimensional observation.
The idiosyncratic exposure \(\delta_j\) captures the direct operational cost of using scalar methods in a high-fragility environment: inaccurate ratings, mispriced portfolios, ineffective regulation. This cost accrues regardless of others' adoption decisions. The competitive disadvantage \(c_j\) captures the relative position loss as others adopt: clients migrate to better-informed managers, regulators scrutinize scalar holdouts more intensely, and the market share of scalar products erodes. This term is the mirror image of the Grossman-Stiglitz information rent (Grossman and Stiglitz, 1980): what appears as a rent to the early adopter appears as a competitive disadvantage to the holdout. The parameter \(c_j\) is type-specific. For asset managers (\(c_A > 0\)), the pressure comes through client migration and relative performance benchmarks. Compliance professionals (\(c_C > 0\)) face growing regulatory expectations as multi-dimensional standards become the norm. Scalar producers (\(c_R > 0\)) confront franchise erosion as demand shifts toward structured multi-dimensional alternatives.
The net payoff from adopting is:
\(\Delta\pi_j(\boldsymbol{\ell}, F) = \sum_{k} (E_{jk} + c_j) w_k \ell_k + \bar{g}_j \ell_j - s_j + F^+ \delta_j\)
Strategic complementarity. The net payoff is strictly increasing in \(\ell_k\) for all \(k\) (since \(E_{jk} + c_j > 0\)), guaranteeing global strategic complementarities. The Morris-Shin uniqueness result and the Sakovics-Steiner threshold derivation both require this property. The competitive disadvantage \(c_j\) ensures complementarity even when the cross-type externality \(E_{jk}\) is small: the holdout's position deteriorates as others adopt, making adoption increasingly attractive regardless of direct network effects.
For \(F\) sufficiently large: at \(\ell = 0\) (nobody else has adopted), the net payoff is \(-s_j + F^+ \delta_j\). For \(F > s_j/\delta_j\), adoption is dominant. At \(\ell = 1\), the net payoff is \(\sum_k (E_{jk} + c_j) w_k + \bar{g}_j - s_j + F^+ \delta_j\), which is also positive for \(F > s_j/\delta_j\) (and in fact positive for lower \(F\) since the network and competitive terms are positive). The upper dominance region is \(F > \max_j s_j/\delta_j\).
For \(F < 0\): \(F^+ = 0\). At \(\ell = 0\), the net payoff is \(-s_j < 0\). Not adopting is dominant.
We assume throughout that \(s_j > \sum_k (E_{jk} + c_j) w_k + \bar{g}_j\) for all \(j\) - the network effects and competitive disadvantage at full adoption do not cover the switching cost without fragility pressure. This ensures the lower dominance region is non-empty at \(\ell = 1\).
The coordination region is \(F \in [0, \max_j s_j/\delta_j]\).
We look for a Bayesian Nash equilibrium in type-specific threshold strategies: type \(j\) adopts if and only if \(x_i > x_j^*\).
Conditional on \(F\), the adoption fraction within type \(j\) is:
\(\ell_j(F) = \Phi\left(\sqrt{\beta}(F - x_j^*)\right)\)
In the limit \(\beta \to \infty\), the Laplacian property (Morris and Shin, 2003) applies within each type: an agent of type \(j\) at the threshold \(x_j^*\) assigns a uniform distribution to \(\ell_j\) over \([0, 1]\).
For the cross-type terms, the adoption of type \(k \neq j\) as perceived by a type-\(j\) agent at the threshold depends on where \(x_k^*\) sits relative to \(x_j^*\). In the precise-signal limit:
If \(x_j^* > x_k^*\) (type \(k\) has a lower threshold), then at \(F = x_j^*\) type \(k\) has fully adopted: \(\ell_k \approx 1\).
If \(x_j^* < x_k^*\) (type \(k\) has a higher threshold), then at \(F = x_j^*\) type \(k\) has not adopted: \(\ell_k \approx 0\).
If \(x_j^* = x_k^*\), then \(\ell_k \sim \text{Uniform}[0,1]\).
The ordering property of heterogeneous global games drives the sequential structure: the type with the lowest threshold adopts first, and each subsequent type's indifference condition incorporates the adoption decisions of all lower-threshold types.
Conjecture the ordering \(x_A^* < x_C^* < x_R^*\) (asset managers adopt first, scalar producers last). We verify this after solving.
Type A's threshold. At \(F = x_A^*\), Types C and R have not adopted (\(\ell_C = \ell_R = 0\)), and \(\ell_A \sim \text{Uniform}[0,1]\). The indifference condition is:
\(\int_0^1 \left[(E_{AA} + c_A) w_A \ell_A + \bar{g}_A \ell_A - s_A + x_A^* \delta_A\right] d\ell_A = 0\)
Evaluating:
\(x_A^* = \frac{s_A - \tfrac{1}{2}\bigl((E_{AA} + c_A) w_A + \bar{g}_A\bigr)}{\delta_A}\)
The competitive disadvantage \(c_A\) lowers the threshold (the holdout penalty makes adoption attractive at lower fragility). The idiosyncratic exposure \(\delta_A\) appears in the denominator: higher personal exposure to fragility increases the urgency of adopting.
Type C's threshold. At \(F = x_C^*\), Type A has fully adopted (\(\ell_A = 1\)), Type R has not (\(\ell_R = 0\)), and \(\ell_C \sim \text{Uniform}[0,1]\):
\(\int_0^1 \left[(E_{CA} + c_C) w_A + (E_{CC} + c_C) w_C \ell_C + \bar{g}_C \ell_C - s_C + x_C^* \delta_C\right] d\ell_C = 0\)
Evaluating (the \((E_{CA} + c_C) w_A\) term is constant in \(\ell_C\) and integrates to itself):
\((E_{CA} + c_C) w_A + \tfrac{1}{2}\bigl((E_{CC} + c_C) w_C + \bar{g}_C\bigr) - s_C + x_C^* \delta_C = 0\)
\(x_C^* = \frac{s_C - (E_{CA} + c_C) w_A - \tfrac{1}{2}\bigl((E_{CC} + c_C) w_C + \bar{g}_C\bigr)}{\delta_C}\)
Type R's threshold. At \(F = x_R^*\), Types A and C have fully adopted (\(\ell_A = \ell_C = 1\)), and \(\ell_R \sim \text{Uniform}[0,1]\):
\(x_R^* = \frac{s_R - (E_{RA} + c_R) w_A - (E_{RC} + c_R) w_C - \tfrac{1}{2}\bigl((E_{RR} + c_R) w_R + \bar{g}_R\bigr)}{\delta_R}\)
The denominator is simply \(\delta_R\): the threshold sensitivity to switching cost is \(1/\delta_R\), bounded by the idiosyncratic exposure. The competitive disadvantage terms \((E_{RA} + c_R) w_A\) and \((E_{RC} + c_R) w_C\) lower the threshold because prior adopters make the holdout position increasingly untenable.
Proposition (Ordered Thresholds).
Under the parameter ordering \(s_A < s_C < s_R\) and \(\bar{g}_A > \bar{g}_C > \bar{g}_R\), and with the externality structure \(E_{AC}, E_{AR} \gg E_{CA}, E_{RA}\), the thresholds satisfy \(x_A^* < x_C^* < x_R^*\). The transition is sequential: asset managers adopt at the lowest Fragility Gap, compliance professionals at an intermediate level, and scalar producers only when the Gap is large.
Proof.
From and , \(x_A^* < x_C^*\) if and only if:
\[ \frac{s_A - \tfrac{1}{2}((E_{AA} + c_A) w_A + \bar{g}_A)}{\delta_A} < \frac{s_C - (E_{CA} + c_C) w_A - \tfrac{1}{2}((E_{CC} + c_C) w_C + \bar{g}_C)}{\delta_C} \]
When \(\delta_A \approx \delta_C\) (comparable idiosyncratic exposures), the ordering reduces to comparing numerators. The switching cost differential \(s_C - s_A > 0\) dominates when it exceeds the net externality and competitive disadvantage terms. For the parameter ranges in Section 2.3 and the externality structure in Section 2.2, the condition is satisfied with margin. Analogous reasoning gives \(x_C^* < x_R^*\).
The type-specific thresholds yield immediate comparative statics.
Corollary (Augmentation Multiplier).
Reducing Type \(j\)'s switching cost by \(\Delta s_j\) lowers Type \(j\)'s threshold by:
\(\frac{\partial x_j^*}{\partial s_j} = \frac{1}{\delta_j}\)
The multiplier is the same functional form for all types: \(1/\delta_j\). Types with smaller idiosyncratic exposure have more sensitive thresholds. The ordering of multipliers depends on the \(\delta_j\) ordering, which is an empirical question independent of the switching cost ordering.
Remark.
The uniform denominator \(\delta_j\) is a consequence of the complementarity-preserving payoff structure. The idiosyncratic exposure \(\delta_j\) serves two formal functions: it guarantees the upper dominance region for all types (preventing the free-rider equilibrium) and provides the state-dependent sensitivity that connects fragility to the adoption threshold. Economically, \(\delta_j\) reflects the fact that a stakeholder operating on obsolete methods in a high-fragility environment bears direct operational costs regardless of whether others have upgraded.
A policymaker with a fixed budget for transition subsidies must choose which type to subsidize. The Sakovics-Steiner (2012) result applies: the most cost-effective subsidies target the type whose adoption imposes the highest externalities on others and on whom others impose the lowest externalities.
Proposition (Subsidy Targeting).
A marginal subsidy \(\Delta\) reducing \(s_A\) produces a larger reduction in the aggregate adoption threshold than the same subsidy reducing \(s_C\) or \(s_R\).
Proof.
The subsidy operates through three channels, and the asymmetry among them determines the ordering.
Direct channel. A subsidy to Type \(j\) lowers \(x_j^*\) by \(\Delta / \delta_j\). Since \(\delta_j\) is the denominator for all types, the direct multiplier is identical in form across types. The direct channel alone does not determine cost-effective targeting.
Externality channel (finite \(\beta\)). In the precise-signal limit, the threshold formulas for later movers condition on earlier movers having fully adopted. At finite \(\beta\), adoption is a smooth function \(\ell_j(F) = \Phi(\sqrt{\beta}(F - x_j^*))\), not a step function. Lowering \(x_A^*\) shifts Type A's adoption curve to the left, raising \(\ell_A(F)\) at every \(F\). This increases the externality \((E_{CA} + c_C) w_A \ell_A(F)\) in Type C's payoff and \((E_{RA} + c_R) w_A \ell_A(F)\) in Type R's payoff, effectively lowering their thresholds. The externality channel from Type A is large (\(E_{AC}, E_{AR}\) are high and \(c_C, c_R > 0\)) and propagates forward to both later types. The externality channel from Type R is zero (no later type exists) or near-zero (\(E_{RA}, E_{RC}\) are small, and Types A and C have already adopted).
The externality channel makes a dollar of Type A subsidy more effective at lowering the aggregate tipping point than a dollar of Type R subsidy. The Sakovics-Steiner framework formalizes this: the agents who matter most in the coordination process are those whose adoption reshapes the payoff environment for others, which in the present model means the type with the largest forward externalities.
The policy implication is counterintuitive to those who frame the transition as requiring the resisters to be persuaded. The resisters (Type R) are the last to adopt precisely because their switching costs are highest and their network benefits lowest. Subsidizing them is expensive and produces no cascade. Subsidizing the early adopters (Type A) is cheap (their switching costs are already the lowest) and produces a cascade through the externality channel. However, as Section 6 establishes, lowering thresholds is necessary but not sufficient: the Grossman-Stiglitz mechanism may stabilize partial adoption at the lowered thresholds. Completing the cascade requires the Fragility Gap to exceed \(x_R^*\), which may depend on complementary policy instruments - disclosure mandates that raise the competitive disadvantage \(c_j\) for holdouts, or regulatory frameworks that make the Fragility Gap visible rather than masked by scalar reporting.
The transition cascades when Type A's adoption triggers Type C's, which triggers Type R's. The cascade requires that the Fragility Gap, once it exceeds \(x_A^*\), also exceeds \(x_C^*\) after accounting for the externality from Type A's adoption, and similarly from Types A and C to Type R.
Definition (Cascade).
A full cascade occurs at Fragility Gap \(F\) if \(F > x_A^*\) implies \(F > x_C^*\) implies \(F > x_R^*\). Since each threshold is a constant determined by the payoff parameters, the cascade occurs whenever \(F\) exceeds the highest threshold \(x_R^*\).
From the threshold equations, Type C adopts once \(F\) exceeds \(x_C^*\) as given by . The externality from Type A's full adoption (\(E_{CA} w_A\)) is already embedded in this expression. The cascade from A to C occurs when:
\[ x_A^* > x_C^* \text{fails} \Longleftrightarrow x_C^* > x_A^* \]
which is the ordering we already established. But what matters for the cascade is whether \(x_C^*\) is close enough to \(x_A^*\) that the same Fragility Gap triggers both. Define the cascade gap:
\(\gamma_{AC} = x_C^* - x_A^*, \gamma_{CR} = x_R^* - x_C^*\)
Proposition (Cascade Condition).
The full cascade occurs at any \(F > x_A^*\) if and only if \(\gamma_{AC} = 0\) and \(\gamma_{CR} = 0\) - that is, all types adopt at the same threshold. In general, the cascade requires \(F > x_R^*\), and the cascade gaps determine how much the Fragility Gap must exceed the first mover's threshold before the last mover joins. The cascade gaps are decreasing in the cross-type externalities \(E_{AC}\), \(E_{AR}\), \(E_{CR}\): stronger externalities compress the cascade and bring all types' thresholds closer together.
Proof.
From and :
\(\gamma_{AC} = x_C^* - x_A^* = \frac{s_C - (E_{CA} + c_C) w_A - \tfrac{1}{2}((E_{CC} + c_C) w_C + \bar{g}_C)}{\delta_C} - \frac{s_A - \tfrac{1}{2}((E_{AA} + c_A) w_A + \bar{g}_A)}{\delta_A}\)
\(\gamma_{AC}\) is decreasing in \(E_{CA}\) (the externality from A to C), decreasing in \(c_C\) (Type C's competitive disadvantage from holdout), and decreasing in \(\delta_C\) (higher exposure compresses the gap). Analogous expressions give \(\gamma_{CR}\) decreasing in the cross-type externalities and competitive disadvantage terms.
The improper uniform prior used for the threshold derivations produces equilibrium thresholds that are independent of signal precision \(\beta\) in the precise-signal limit. The aggregate effect and the dispersion effect of increased precision cancel exactly under the Laplacian property (Szkup, 2020). The thresholds derived in equations -- confirm this: \(\beta\) does not appear.
A proper prior (e.g., normal) would make the threshold generically dependent on \(\beta\), and increased adoption that improves information quality would lower the thresholds for subsequent adopters, creating a bootstrapping dynamic. The present model, built on the standard uniform-prior global game for tractability, does not deliver this channel. The cascade depends entirely on the direct externality channel (\(E_{jk}\)) and the competitive disadvantage channel (\(c_j\)). Whether this understates or overstates the ease of transition depends on whether multi-dimensional adoption in fact improves signal precision - an empirical question the ESG divergence literature (Berg, Koelbel, and Rigobon, 2022) suggests should not be taken for granted.
The preceding analysis assumes the Fragility Gap is fixed. In practice, early adoption may reduce \(F\) by improving the maintenance of coordination dimensions. The partial-adoption equilibrium arises when this reduction is large enough to remove the incentive for further adoption. The heterogeneous model gives the trap a sharper characterization.
If Type A's adoption reduces the effective Fragility Gap (by improving the maintenance of coordination dimensions through better capital allocation), then \(F\) itself falls as a function of \(\ell_A\). Let \(F(\ell_A) = F_0 - \eta_A \ell_A\) where \(\eta_A > 0\) is the fragility reduction per unit of Type A adoption. (Note: \(\eta_A\) is distinct from \(\delta_A\), the idiosyncratic fragility exposure defined in Section 3.2.)
Proposition (Partial-Adoption Stability).
The partial-adoption equilibrium (Type A has adopted, Types C and R have not) is stable if the fragility reduction from Type A's adoption pushes \(F\) below Type C's threshold:
\[ F_0 - \eta_A > x_C^* \text{fails, i.e.,} \eta_A > F_0 - x_C^* \]
where \(F_0\) is the initial Fragility Gap and \(\eta_A\) is the fragility reduction from full Type A adoption. The trap is stable when Type A's adoption reduces fragility faster than it creates pressure for further adoption through the externality and competitive disadvantage channels.
The trap holds when adoption fixes problems faster than it reveals them. Mandates that improve capital allocation (high \(\eta_A\)) may paradoxically stabilize the trap by reducing fragility enough to remove the urgency for further adoption. Mandates that increase the competitive disadvantage of scalar holdouts (high \(c_j\)) destabilize it.
The coordination game has a public-good structure: the benefits of multi-dimensional observation (reduced systemic fragility) are diffuse, while the switching costs are private. Public goods are under-provided (Olson, 1965). What funds the transition?
The answer is the Grossman-Stiglitz information rent (Grossman and Stiglitz, 1980), which enters the formal model through the competitive disadvantage term \(c_j \ell\) in the non-adoption payoff. The scalar market prices assets on the consensus dimension and assigns the unconditional expectation to every constraint dimension. An investor who observes constraint dimensions the market cannot see holds an information advantage. In the model, this advantage manifests on the holdout's side of the ledger: the competitive disadvantage \(c_j \ell\) captures the market share loss, client migration, and performance gap that a scalar-only operator suffers as more competitors adopt multi-dimensional observation.
A structural feature of the information rent connects private incentive to public transition. Pursuing the rent helps eliminate it, because the trading activity that generates the rent simultaneously closes the mispricing that sustains it.
Proposition (Self-Limiting Competitive Pressure).
The competitive disadvantage \(c_j \ell\) borne by non-adopters increases with the adoption fraction \(\ell\). As more investors adopt multi-dimensional observation, the holdout position deteriorates further, increasing the pressure to adopt. In the limit \(\ell \to 1\), the competitive disadvantage is maximal: the last holdout faces the full weight of the market's shift. Simultaneously, the mispricing that generated the information rent for early adopters has closed, because the market now reflects constraint dimensions. When the observation architecture is complete, the rent has been competed away and only adopters remain.
Proof.
The competitive disadvantage \(c_j \ell\) is increasing in \(\ell\) by construction. The economic mechanism: as more investors adopt multi-dimensional observation and trade on constraint-dimension information, the price incorporates that information through the standard GS mechanism. The residual mispricing on each dimension shrinks as more informed traders participate. A non-adopter who cannot observe constraint dimensions finds their portfolio construction, risk assessment, and client offerings increasingly inferior relative to the improving market standard. At \(\ell = 1\), the competitive disadvantage \(c_j\) is maximal per unit, but the population bearing it is zero - the last holdout either adopts or exits.
The self-limiting property means the information rent lowers the adoption threshold - but it does not guarantee the cascade. Type A stakeholders (asset managers) adopt multi-dimensional observation because it generates returns. Their adoption creates externalities that lower the threshold for Types C and R. Asset managers' demand for constraint-dimension data funds the data providers (ESG analytics firms, climate risk modelers, structured disclosure platforms) whose products reduce switching costs for everyone. The resulting informed trading pushes constraint-dimension information into prices, making the mispricing visible to less-sophisticated observers and raising the salience of scalar inadequacy. Meanwhile, the holdout's relative performance, client retention, and market standing deteriorate through the competitive disadvantage \(c_j \ell\) as adoption spreads.
The rent decays as the transition proceeds, and its disappearance signals that the observation architecture has caught up with the coordination structure it must maintain.
A connection to the partial-adoption trap (Section 5) deserves emphasis. The Grossman-Stiglitz mechanism predicts a stable interior equilibrium: the rent exactly compensates the marginal investor for the cost of observation, so that some investors are informed and some are not (Grossman and Stiglitz, 1980). If applied directly, the self-limiting property stabilizes partial adoption rather than driving full adoption. The partial-adoption trap of Section 5 is the GS interior equilibrium applied to the observation architecture: Type A adopts, the rent draws in enough informed capital to partially close the mispricing, and the remaining rent falls below the switching cost for Types C and R. The transition stalls. What pushes past this equilibrium is not the rent itself but the Fragility Gap \(F\) - the exogenous state variable that reflects the accumulation of unobserved constraint-dimension stress. If \(F\) remains below \(x_C^*\), the GS equilibrium holds and partial adoption is the long-run outcome. If external forces - physical climate damage, systemic financial events, regulatory mandates, or the compound interaction of concurrent shocks - push \(F\) above the cascade threshold \(x_R^*\), the transition completes regardless of the rent's decay. The rent makes the transition cheaper by lowering the thresholds. It does not make the transition inevitable.
The framework generates a specific empirical prediction that connects the private-return mechanism to the ESG premium debate. Existing evidence shows mixed capital-market effects of sustainability disclosure (Christensen, Hail, and Leuz, 2021), while ESG ratings diverge across agencies because each compresses different dimensions into incompatible scalars (Berg, Koelbel, and Rigobon, 2022). The framework predicts:
Constraint-specific, multi-dimensional disclosure should show stronger capital-market effects than aggregate scalar ESG scores. TCFD-aligned physical risk scenarios, which specify climate exposure by geography, time horizon, and pathway, carry multi-dimensional structure that a single-letter ESG rating compresses away. The mixed evidence in the literature is consistent with the partial-adoption trap if most disclosure studied is scalar-compressed. A test separating constraint-specific from aggregate disclosure would evaluate both the information rent's magnitude and the trap's stability.
The information rent is subject to the standard limits to arbitrage (Shleifer and Vishny, 1997): capital constraints, short-selling costs, and career risk can prevent informed investors from correcting mispricings even when the information advantage is real. The timing risk is acute - constraint-dimension decay accumulates invisibly and may take years to surface as a repricing event, during which the informed investor bears carrying costs. The counterparty question is also relevant: some market participants (fundamental analysts, activist investors, credit analysts, insurance underwriters) already conduct multi-dimensional analysis on specific assets. The scalar market is not uniformly scalar. The rent is the residual mispricing after existing multi-dimensional analysis - smaller than the theoretical maximum, positive whenever the existing analysis is incomplete.
Four concurrent shocks are raising the Fragility Gap for all stakeholder types simultaneously, pushing the system toward the cascade threshold.
Climate risk repricing is the most developed of the concurrent shocks. Physical and transition risks are multi-dimensional, geographically specific, and interact through constraint channels the scalar price system cannot reach. Central banks have acknowledged the observability problem (Carney, 2015; NGFS, 2019; Bolton et al., 2020), and the TCFD (2017) and ISSB (2023) frameworks are supplementary observation channels arrived at by practitioners confronting scalar inadequacy. Carbon Tracker (2011) established the concept of stranded carbon assets; subsequent estimates through 2023 place the exposure in the trillions.
The AI transition compounds the problem differently. AI is a high-Beierspace-fraction technology (The Beierspace fraction (Beier, 2026k) measures the proportion of an asset's or technology's value that depends on maintained coordination rather than physical attributes. A high Beierspace fraction means the asset's value is predominantly coordination-dependent and therefore vulnerable to observation withdrawal.): alignment risk, regulatory risk, labor displacement risk, and concentration risk are all coordination-dependent dimensions the scalar price system does not transmit, yet capital allocation for AI infrastructure is being made on one dimension of a problem with dozens. Supply chain reconfiguration - the post-pandemic, post-Russia-Ukraine reorganization across geopolitical risk, infrastructure investment, and environmental compliance - adds further Beierspace dimensions the scalar system cannot process. Geopolitical realignment from unipolar to multipolar order fragments the coordination structures underlying trade, finance, and institutional governance, and scalar country-risk scores cannot distinguish stable equilibria from coordination structures one shock from collapse.
The shocks compound through constraint channels: climate policy affects energy costs, which affect AI training costs, which affect the competitive terrain for deployment, which affects labor markets, which affects political coalitions for climate policy. Each interaction is a Beierspace dimension the price system cannot maintain.
Countervailing forces raise thresholds rather than lower them. Political backlash against ESG-adjacent mandates (the anti-ESG legislative movement in US states since 2022), the defunding of climate disclosure initiatives, and the possibility that AI improves scalar compression (better single-score models) rather than creating demand for multi-dimensional observation all work against the transition. The model accommodates these as increases in \(s_j\) (switching costs raised by political risk) or decreases in \(\bar{g}_j\) (network benefits reduced by regulatory hostility). Whether the concurrent shocks dominate the countervailing forces is an empirical question the model structures but does not resolve.
Shock | Effect on \(F\) | Effect on \(c_j, \bar{g**_j\)} | **Effect on \(x_j^*\)
Climate risk | Stranded carbon assets | Early movers avoid losses | Lowers threshold
AI transition | Alignment/regulatory risk | Better allocation to high-Beierspace assets | Lowers threshold
Supply chains | Cascading disruption | Resilience pricing | Lowers threshold
Geopolitics | Sovereign repricing | Multi-dim. analysis | Lowers threshold
Compounding | Cross-shock tail risk | Multi-dim. tools serve all | Lowers faster than each alone
Anti-ESG backlash | Raises political \(s_j\) | Reduces \(\bar{g}_j\) visibility | Raises threshold
AI scalar improvement | Reduces perceived \(F\) | Better scalar tools | Raises threshold
The paper draws on four literatures. The global games and coordination literature provides the formal apparatus. Morris and Shin (1998, 2003) develop the equilibrium refinement; Sakovics and Steiner (2012) extend it to heterogeneous agents, proving that optimal subsidies target agents who impose high externalities on others. Angeletos and Werning (2006) show that endogenous public information can restore multiplicity; the present model avoids this source of multiplicity because thresholds are independent of signal precision under the Laplacian property. The broader coordination-game tradition traces to Schelling (1960) on focal points and tipping, and the collective action problem underlying the transition - diffuse public benefits, concentrated private switching costs - is the setting Olson (1965) formalized.
The socio-technical transitions literature (Geels, 2002; Geels and Schot, 2007) provides the regime-level perspective. North (1990) on institutional path dependence and Acemoglu and Robinson (2012) on elite resistance explain why the scalar regime persists. Christensen (1997) on disruption predicts the sequential adoption pattern the model formalizes.
The multi-task incentive literature provides the microeconomic foundation. Holmstrom and Milgrom (1991) prove that agents observed on a subset of dimensions underinvest in unmeasured dimensions - the incentive-theoretic ancestor of the Prolegomena's mechanism. The present framework adds the rate-distortion structure (quantitative decay bounds, phase transitions) and the coordination-game dynamics.
The ESG disclosure literature provides the empirical analog. The mixed capital-market effects of sustainability disclosure documented by Christensen, Hail, and Leuz (2021) are consistent with the partial-adoption trap if most disclosure is scalar-compressed rather than structured. The ESG rating divergence that Berg, Koelbel, and Rigobon (2022) document - different agencies compressing different dimensions into incompatible scalars - is what the framework interprets as Goodhart's Law operating at one level of abstraction higher. Evidence that long-horizon investors engage with climate risk before other institutional types (Krueger, Sautner, and Starks, 2020) confirms the model's prediction that Type A enters positive-adoption territory earliest.
The framework generates a testable prediction: structured, constraint-specific disclosure (TCFD physical risk scenarios) should show stronger capital-market effects than aggregate scalar ESG scores. A test separating these two disclosure types would directly evaluate the partial-adoption trap's stability.
The three-type classification is a simplification of a continuous distribution of switching costs and network benefits. The Sakovics-Steiner framework accommodates arbitrary type distributions; the three-type restriction is for analytical tractability and interpretability. Extending to a continuous type space would require numerical methods but would produce a smooth adoption curve rather than the step-function cascade the discrete model generates.
The payoff specification - adoption payoff with cross-type externalities \(E_{jk}\), within-type network effects \(\bar{g}_j\), and switching cost \(s_j\); non-adoption payoff with idiosyncratic fragility exposure \(-F^+ \delta_j\) and competitive disadvantage \(-c_j \ell\) - provides closed-form thresholds under the linear specification. Convex failure costs (which the Prolegomena's phase-transition results suggest are realistic) would change the exact threshold values without affecting the qualitative ordering or the cascade logic. The comparative statics on switching costs, externalities, and competitive disadvantage hold directionally for any payoff structure with strategic complementarities.
The payoff specification models the holdout's deteriorating competitive position (\(c_j \ell\)) but does not model the systemic risk-sharing channel: the observation that others' adoption stabilizes the coordination structure and reduces the failure cost for non-adopters. Including a systemic term \(-F^+ \lambda_j(1-\ell)\) alongside the competitive disadvantage would capture this public-good externality, but would reintroduce the strategic substitutability that the current specification avoids. The model's policy implications - that the transition is a public good warranting subsidy - rest on the externality matrix \(E_{jk}\) rather than on systemic risk reduction through the failure cost. A richer model incorporating both channels while preserving tractability is a direction for future work.
The switching cost estimates are order-of-magnitude approximations based on publicly available industry data. They should be refined through detailed analysis of MiFID II unbundling disclosures (for sell-side research), Big Four annual reports (for audit costs), and NRSRO financial filings (for franchise value). The estimates are sufficient to establish the ordering \(s_A < s_C < s_R\) and the aggregate magnitude ($111 -- 180B), which are the inputs the model requires.
The improper uniform prior produces equilibrium thresholds that are independent of signal precision \(\beta\). A precision channel - in which early adoption improves information quality, lowering thresholds for subsequent adopters - would require a proper prior (e.g., normal) and is absent from the present specification. The cascade therefore depends entirely on the direct externality and competitive disadvantage channels. Whether this leads the model to overstate or understate the difficulty of the transition is an empirical question.
The model treats the Fragility Gap as exogenous to the adoption decision for the threshold analysis. The partial-adoption analysis (Section 5) introduces endogeneity through \(F(\ell_A) = F_0 - \eta_A \ell_A\), but a fully dynamic model with strategic interaction between adoption and fragility reduction would produce richer predictions. Hirschman's (1970) insight - that partial reform can prevent full reform by relieving pressure - motivates the stability analysis but is not modeled dynamically.
The political economy of the observation-architecture transition is a heterogeneous coordination game, not a persuasion problem. Asset managers adopt first - their switching costs are lowest and their network benefits highest - while scalar producers resist longest because their franchise value depends on the scalar system's primacy. The transition cascades through the externality channel (asset managers' demand for constraint-dimension data forces disclosure that regulators codify and rating agencies incorporate) and the competitive disadvantage channel (holdouts face deteriorating market position as more competitors adopt multi-dimensional observation).
The most cost-effective policy subsidizes the early adopters, not the resisters. The formal model confirms Sakovics and Steiner's general result in this specific setting: the agents who matter most are those who impose high externalities on others and on whom others impose low externalities. A dollar spent reducing asset managers' switching costs produces a larger aggregate threshold reduction than a dollar spent on rating agencies or regulators, because the asset managers' adoption reshapes the information environment for everyone else.
The aggregate switching cost for the global observation architecture is $111 -- 180B over a decade. The concurrent shocks of climate repricing, AI deployment, supply chain reconfiguration, and geopolitical realignment are raising the Fragility Gap simultaneously, pushing the system toward the cascade threshold. The private return to multi-dimensional observation - the information advantage over scalar-only competitors on constraint dimensions the market cannot price - lowers the adoption thresholds: asset managers adopt because it is profitable, and their adoption creates the externalities and competitive pressure that lower the threshold for everyone else. But the Grossman-Stiglitz mechanism stabilizes partial adoption, not full adoption. The cascade completes only if the Fragility Gap exceeds the thresholds the rent has lowered - a condition that depends on the accumulation of unobserved constraint-dimension stress, not on the rent alone.