Chapter 6: Economic Tensors and Scalar Irrecoverability

What the Theorem Says

The theorem establishes three things:

First, scalar utility is not merely an imprecise approximation – it is a lossy compression that maps entire families of distinct economic states to the same number. Maria’s coffee shop valued at $200,000 occupies a level set – an eight-dimensional surface of all possible business configurations that happen to generate the same discounted cash flow. Some of these configurations involve exploitative labor practices, broken supplier relationships, and community-destroying operations. Others involve fair wages, trusted relationships, and community enrichment. The scalar cannot distinguish them.

Second, the information destroyed by the projection is mathematically irrecoverable. This is not a claim about measurement difficulty or data availability. It is a topological impossibility: you cannot continuously embed a nine-dimensional space in a one-dimensional space. The “missing dimensions” do not exist in the scalar representation and cannot be reconstructed from it by any means.

Third, the economic consequences are unbounded. The gap between what the scalar recommends and what the full manifold recommends can be arbitrarily large – infinite, in cases where the scalar-optimal path crosses a moral boundary that the scalar cannot see.

The Allais and Ellsberg Paradoxes

The theorem immediately resolves two classical paradoxes that have troubled expected utility theory since the 1950s.

Corollary 15 (The Allais and Ellsberg Paradoxes Are Not Paradoxes). The Allais paradox (violation of the independence axiom) and the Ellsberg paradox (violation of subjective expected utility under ambiguity) are both consequences of Scalar Irrecoverability.

1. Allais: The agent’s preference reversal between “certain $1M” and “89% chance of $1M, 10% chance of $5M, 1% chance of $0” reflects the activation of d_9 (epistemic certainty has value beyond expected monetary value) and d_7 (identity: “I am prudent”). On the scalar projection, independence is violated. On the full manifold, preferences are consistent: the attribute-vector difference between certainty and risk includes non-zero components on d_7 and d_9 that the expected-value calculation discards.

2. Ellsberg: Ambiguity aversion (preferring known to unknown probabilities) reflects d_9 directly: known probabilities have higher epistemic-status scores than ambiguous ones. The Mahalanobis distance from the ambiguous gamble to the goal is larger than from the known-probability gamble, even when expected values are equal, because \Delta d_9 \neq 0 for the ambiguous case.

The resolution is not to add “certainty preference” or “ambiguity aversion” as ad hoc modifications to the utility function. It is to recognize that the utility function was computing on the wrong manifold. The “paradoxes” are artifacts of the d_1 projection. On the full manifold, the behavior is consistent, transitive, and predictable.

From Scalars to Tensors

A scalar is a single number. A vector is an ordered list of numbers with transformation rules. A tensor is a multi-dimensional array of numbers with specific transformation rules that preserve geometric content under change of coordinates. The key property is that tensors encode relationships between dimensions – exactly the information that scalars destroy.

In physics, the stress-energy tensor T_{\mu\nu} encodes the full state of matter and energy at a point in spacetime: energy density, momentum flux, pressure, shear stress. Contracting this tensor (summing over indices) produces scalars like total energy or total pressure, but the contraction destroys the directional information: you cannot recover the full stress state from the total energy alone.

The same mathematical structure applies to economics.

Economic Obligations as Vectors

Consider an economic obligation – Maria’s lease, for instance. The lease is not a scalar (a monthly rent payment). It is a vector on the decision manifold:

\mathbf{L} = (L_1, L_2, L_3, L_4, L_5, L_6, L_7, L_8, L_9)

where:

• L_1 = -7{,}000 (monthly rent payment, d_1)
• L_2 = +1 (the lease grants occupancy rights, d_2)
• L_3 = -0.6 (the 40% increase is perceived as unfair, d_3)
• L_4 = -0.4 (the increase constrains business autonomy, d_4)
• L_5 = -0.7 (the increase eroded trust with the landlord, d_5)
• L_6 = +0.8 (the lease preserves the community gathering space, d_6)
• L_7 = -0.3 (accepting the increase conflicts slightly with self-image, d_7)
• L_8 = +0.5 (the lease is a legitimate legal contract, d_8)
• L_9 = -0.5 (uncertainty about future increases, d_9)

The monthly rent payment L_1 = -7{,}000 is the scalar projection. It captures one component of a nine-dimensional object. The remaining eight components – the rights, the fairness, the trust, the community value, the identity, the legitimacy, the epistemic status – are all real, all measurable (at least ordinally), and all relevant to Maria’s decisions about the lease.

Economic Interests as Covectors

If obligations are vectors (they “point” in the direction of their impact on the agent’s state), then economic interests are covectors (dual vectors) – they assign a value to each direction.

Maria’s interest in her business is a covector:

\mathbf{I} = (I^1, I^2, I^3, I^4, I^5, I^6, I^7, I^8, I^9)

where I^k represents how much Maria cares about changes along dimension k. The satisfaction of an obligation-interest pair is the contraction (inner product):

S = \mathbf{I} \cdot \mathbf{L} = \sum_{k=1}^{9} I^k L_k

This is a scalar – a single number. But it is a derived scalar, obtained from the contraction of a vector and a covector. The full tensor structure is preserved in \mathbf{I} and \mathbf{L}; the scalar S is a specific summary of their interaction.

GDP as Tensor Contraction

The most consequential scalar in economics – GDP – is a specific contraction of the full economic tensor.

Define the national economic tensor \mathbf{T} as the aggregate of all economic activity in a country over a period, with components T_{k} representing the total change along each dimension k:

\text{GDP} = T_1 = \pi_{d_1}(\mathbf{T})

GDP is the d_1 component of \mathbf{T} – the projection onto the monetary dimension. The Stiglitz-Sen-Fitoussi Commission (2009) was arguing, in the language of this framework, that economic welfare should be assessed using \mathbf{T} (or at least a higher-rank contraction of it), not T_1 alone.

The Scalar Irrecoverability Theorem tells us that the Commission was mathematically correct: the information in \mathbf{T} that GDP discards is irrecoverable from GDP alone. Two economies with identical GDP can have radically different values on d_3 (fairness), d_5 (trust), d_6 (social cohesion), and d_8 (institutional legitimacy). No post-hoc adjustment to GDP – no “GDP plus inequality adjustment,” no “green GDP” – can recover this information, because the projection has destroyed it.

What would recover it is not projecting in the first place: reporting the full vector \mathbf{T} (or a sufficient subset of its components) rather than the scalar T_1.

The Tensor Structure of Business Valuation

Maria’s coffee shop valuation of $200,000 is a contraction of the full business tensor:

\text{Valuation} = \phi(\mathbf{B}) = \phi(B_1, B_2, \ldots, B_9)

where \mathbf{B} is the attribute vector of the business and \phi is the discounted-cash-flow function. The theorem says: this valuation maps many distinct businesses to the same number. Some businesses worth $200,000 in DCF terms are thriving community institutions with deep trust networks and ethical supply chains. Others worth $200,000 are extractive operations with high turnover, adversarial supplier relationships, and no community value. The number cannot distinguish them.

RUNNING EXAMPLE – MARIA’S COFFEE SHOP

What would a tensor valuation of Maria’s coffee shop look like?

Dimension	Component	Assessment
d_1: Consequences	B_1 = \$200{,}000	DCF valuation (the only thing the appraiser reported)
d_2: Rights	B_2 = 0.7	Strong: clear lease, business licenses, supplier contracts
d_3: Fairness	B_3 = 0.8	High: fair wages, fair prices, equitable supplier terms
d_4: Autonomy	B_4 = 0.6	Moderate: constrained by landlord’s pricing power
d_5: Trust	B_5 = 0.9	Very high: six-year supplier relationship, loyal customers
d_6: Social impact	B_6 = 0.85	High: community hub, four jobs, neighborhood identity
d_7: Identity	B_7 = 0.9	Very high: Maria’s identity as ethical business owner
d_8: Legitimacy	B_8 = 0.7	Moderate: standard regulatory compliance
d_9: Epistemic	B_9 = 0.4	Low: uncertainty about rent, neighborhood trajectory

The full tensor \mathbf{B} = (\$200K, 0.7, 0.8, 0.6, 0.9, 0.85, 0.9, 0.7, 0.4) contains information that the scalar \$200K cannot. A buyer who acquires the business and replaces the fair-trade supplier with a commodity supplier, fires the long-term employees, and converts the community space to additional seating would preserve B_1 (maybe even increase it) while destroying B_3, B_5, B_6, and B_7. The scalar valuation would say: “Business maintained or improved.” The tensor valuation would say: “Business gutted on five of nine dimensions.”

This is not a difference of opinion. It is a difference of dimensionality. The scalar and the tensor are making different mathematical claims, and the tensor’s claim is richer, more informative, and – by the irrecoverability theorem – strictly non-recoverable from the scalar.

Projection 1: Pure DCF (\phi = \pi_{d_1})

The standard discounted-cash-flow projection considers only the monetary dimension. Inputs: annual revenue $226,800 (from the monthly figure), annual costs $210,000 (rent + supplies + wages + utilities), discount rate 10%, assumed 10-year horizon with 2% growth.

\text{Value}_1 = \sum_{t=1}^{10} \frac{(226{,}800 \times 1.02^{t-1}) - 210{,}000}{1.10^t} \approx \$200{,}000

This is the appraiser’s number. It captures one dimension.

Projection 2: DCF + Community Value (\phi = \pi_{d_1, d_6})

Suppose we attempt to monetize the community value. The shop serves as a gathering place for approximately 120 neighborhood residents who use it as their primary social space. If the closure of the shop forced even 20% of these residents to use alternative venues (coworking spaces, other cafes farther away), the additional cost to them would be approximately $50/month each – $14,400/year in monetized community value. Adding this stream:

\text{Value}_2 \approx \$200{,}000 + \$88{,}000 = \$288{,}000

But this monetization is itself a projection – converting d_6 to d_1 at an exchange rate that may not be well-defined. The community value of a neighborhood gathering place is not commensurable with monetary flow. The $88K figure is an estimate, not a measurement.

Projection 3: DCF + Supply Chain Value (\phi = \pi_{d_1, d_3, d_5, d_7})

The fair-trade supply chain represents six years of relationship-building. The cost of re-establishing an equivalent chain from scratch (finding reliable fair-trade cooperatives, building trust, achieving quality consistency) can be estimated at 18-24 months of procurement effort, or approximately $35,000-$50,000 in search costs, quality failures, and relationship development. Adding the supply chain’s relational value:

\text{Value}_3 \approx \$200{,}000 + \$88{,}000 + \$42{,}000 = \$330{,}000

Projection 4: Full Tensor Assessment

The full tensor does not produce a single number. It produces a vector:

\mathbf{B} = (\$200K, 0.7, 0.8, 0.6, 0.9, 0.85, 0.9, 0.7, 0.4)

This representation preserves the fact that the business is strong on trust (B_5 = 0.9) and identity (B_7 = 0.9) but weak on epistemic status (B_9 = 0.4 – uncertainty about the future). A buyer can see which dimensions are strong and which are fragile. A scalar valuation hides this structure entirely.

The progression from $200K to $288K to $330K is not convergence toward a “true” scalar value. Each additional dimension reveals additional value, but the scalar still destroys the structure of the value. The $330K number does not tell you that the business is trust-rich and certainty-poor. Only the tensor does.

The Irrecoverability in Action

Here is the critical point: given only the scalar $200,000, can you determine whether this is a community-embedded, trust-rich, fair-trade coffee shop or a sterile chain location with no community ties, adversarial supplier relationships, and high employee turnover? No. Both could generate identical cash flows. The scalar maps them to the same number. The information needed to distinguish them has been destroyed, and the Scalar Irrecoverability Theorem proves that no procedure can reconstruct it.

This is not an academic curiosity. It is the mechanism by which real economic damage occurs. When the REIT evaluates whether to buy Maria out and install a chain tenant, the DCF analysis says: “The chain tenant pays higher rent and generates higher cash flow. The acquisition is value-creating.” On d_1, this is correct. On the full manifold, the acquisition destroys community value (d_6), trust networks (d_5), fair-trade supply chains (d_3, d_7), and neighborhood identity (d_7). The REIT cannot see what it is destroying because its valuation methodology operates on a scalar projection of a nine-dimensional reality.

For Utility Theory

The theorem strikes at the foundation of expected utility theory. Von Neumann and Morgenstern (1944) proved that if preferences satisfy four axioms (completeness, transitivity, independence, continuity), then there exists a utility function u: X \to \mathbb{R} such that the agent’s behavior is equivalent to maximizing \mathbb{E}[u].

This is a representation theorem, not a behavioral theory. It says: if preferences satisfy the axioms, then they can be represented as scalar utility maximization. The behavioral economics program has shown that preferences systematically violate the axioms – particularly independence (Allais) and transitivity (Tversky).

The geometric framework provides the diagnosis: the axioms are satisfied on the full manifold but violated on the scalar projection. A preference relation that is transitive in nine dimensions can appear intransitive when projected onto one dimension (Theorem 16). A preference relation that satisfies independence on the full attribute vector can violate independence when the projection discards dimensions that are differentially activated across comparisons (Corollary 15).

The fix is not to abandon the axioms or to catalog the violations as “biases.” The fix is to recognize that the axioms hold on the correct manifold – and that the correct manifold has nine dimensions, not one.

For Welfare Economics

If scalar utility is irrecoverably lossy, then welfare functions defined on scalar utilities inherit the loss. Arrow’s impossibility theorem (1951) proves that no social welfare function can simultaneously satisfy unrestricted domain, Pareto efficiency, independence of irrelevant alternatives, and non-dictatorship when aggregating ordinal rankings.

The geometric framework suggests that Arrow’s impossibility is an artifact of the scalar domain. Arrow’s theorem operates on one-dimensional ordinal rankings. When agents evaluate alternatives as attribute vectors in \mathbb{R}^9, a component-wise social welfare mapping W: (\mathbb{R}^9)^n \to \mathbb{R}^9 can satisfy analogues of all four Arrow conditions simultaneously; the impossibility reappears only when W is composed with a projection \phi: \mathbb{R}^9 \to \mathbb{R}. A full treatment of multi-dimensional social choice is deferred to future work, but the geometric perspective suggests that the impossibility is in the projection, not in the aggregation.

For Policy Analysis

Cost-benefit analysis requires converting all costs and benefits to a common monetary unit. The Scalar Irrecoverability Theorem says this conversion is necessarily lossy: the monetized value of environmental damage, social displacement, or trust erosion is a projection of a multi-dimensional cost onto d_1, and the projection cannot be inverted.

This does not mean cost-benefit analysis is useless. It means that the single number it produces should be understood as a component of the full evaluation, not as the evaluation itself. The geometric alternative: report the full cost-benefit vector, one component per active dimension, and make the trade-offs between dimensions explicit rather than hiding them inside an exchange rate that converts everything to dollars.

Cross-Dimensional Interactions

The covariance matrix \Sigma encodes how the nine dimensions interact. The scalar discards \Sigma entirely.

For Maria’s coffee shop, the critical interactions include:

• d_1 \leftrightarrow d_5 (money-trust interaction): The monetary terms of the lease affect the trust relationship with the landlord, and the trust relationship affects the monetary stability of the business. These are not independent.

• d_3 \leftrightarrow d_6 (fairness-community interaction): Whether the shop’s pricing is perceived as fair directly affects community engagement and loyalty, which in turn affects revenue. The d_3-d_6 cross-term in \Sigma captures this coupling.

• d_5 \leftrightarrow d_7 (trust-identity interaction): Maria’s identity as an ethical business owner depends on maintaining trust with her suppliers and employees. A betrayal of trust (d_5) produces an identity crisis (d_7) that is larger than either dimension’s marginal change would suggest. The cross-term amplifies the effect.

A scalar valuation that monetizes each dimension independently and sums the results would capture the diagonal terms of \Sigma but miss the cross-terms. The cross-terms are often the most economically significant: they represent synergies and conflicts between dimensions that drive the decisions that matter most.

Boundary Information

The scalar discards all information about moral-economic boundaries. The boundary between “fair pricing” and “exploitation” (d_3), between “trustworthy partner” and “unreliable counterparty” (d_5), between “legal transaction” and “criminal act” (d_2) – these boundaries define the phase structure of the decision manifold. They determine which paths are available and which are blocked by infinite-weight edges.

A scalar valuation cannot represent boundaries. A continuous function \phi: \mathbb{R}^9 \to \mathbb{R} maps points near a boundary to values that are continuous with points far from the boundary. The discontinuous jump in decision regime – the phase transition from “normal pricing” to “price gouging,” from “tough negotiation” to “exploitation” – is smoothed away by the continuity of \phi.

This is why scalar economic models consistently underestimate the severity of boundary crossings. The model sees a continuous increase in price; the agent (and the community) sees a discontinuous transition from fair dealing to exploitation. The scalar model predicts gradual response; the actual response is a sharp phase transition – boycott, regulation, reputational destruction – because the boundary penalty \beta_k is not a continuous function of the scalar change. It is a discrete event: you cross the boundary or you don’t.

Directional Information

Perhaps most importantly, the scalar discards direction. A scalar change of +\$50K in business value could mean:

• Revenue grew 25% while costs remained stable (pure d_1 improvement)
• Revenue grew 40% but only because the owner switched to exploitative suppliers (d_1 up, d_3 and d_7 down)
• Revenue was flat but a major liability was settled (d_1 up from liability resolution, d_9 up from uncertainty reduction)
• Revenue grew modestly from expanded hours that also increased community engagement (d_1 up, d_6 up)

Each of these represents a different direction of movement on the manifold. They have different implications for future decisions, different stability properties, and different effects on the other dimensions. The scalar +\$50K erases all directional information.

A tensor representation preserves it. The change vector \Delta\mathbf{B} = (50K, \Delta B_2, \ldots, \Delta B_9) specifies not just the magnitude of the change but its direction on the manifold. This directional information is precisely what investors, regulators, and communities need to assess whether an economic change is genuinely value-creating or merely value-extracting.

Brouwer’s Invariance of Dimension

[Established Mathematics.] Brouwer’s theorem (1911) states that \mathbb{R}^m and \mathbb{R}^n are homeomorphic if and only if m = n. Equivalently: there is no continuous bijection between Euclidean spaces of different dimensions. This is the topological foundation of the Scalar Irrecoverability Theorem.

The theorem is non-trivial (Cantor showed that \mathbb{R}^m and \mathbb{R}^n have the same cardinality for all m, n, so bijections exist – they just cannot be continuous) and requires algebraic topology for its proof (typically via singular homology or the Borsuk-Ulam theorem).

The Rank-Nullity Argument

[Established Mathematics.] For differentiable \phi: \mathbb{R}^9 \to \mathbb{R}, the derivative d\phi_x: \mathbb{R}^9 \to \mathbb{R} is a linear map with rank \leq 1 and kernel dimension \geq 8 at every point x. By the inverse function theorem, \phi cannot be locally injective at any regular point (where the rank is 1), because the kernel is non-trivial. At critical points (where the rank is 0), \phi is constant to first order.

This means that every level set \phi^{-1}(c) is either empty or has dimension \geq 8 in a neighborhood of any regular point. The pre-image of the appraiser’s $200K figure is an eight-dimensional surface in the nine-dimensional attribute space – a vast family of businesses that all produce the same cash flow but differ on every other dimension.

Tensor Contraction in Index Notation

[Established Mathematics.] A rank-2 tensor T_{ij} in n dimensions has n^2 independent components (or n(n+1)/2 if symmetric). The contraction T_{i}^{i} = \sum_i T_{ii} produces a scalar – the trace. For the 9-dimensional economic tensor:

• The full tensor T_{ij} has 81 components (45 independent if symmetric), encoding all pairwise interactions between the nine economic dimensions.
• GDP is the (1,1) component: T_{11} (or more precisely, the trace of T restricted to the d_1 subspace).
• The trace \text{tr}(T) = \sum_{i=1}^{9} T_{ii} would be a “total economic activity” scalar that at least includes the diagonal contributions from all nine dimensions – a strictly more informative summary than T_{11} alone, though still lossy (it discards the 36 independent off-diagonal components).

The hierarchy of summaries, from most informative to least:

Each step down the hierarchy destroys information that cannot be recovered from the lower representation. The Scalar Irrecoverability Theorem is the formal statement that the bottom two rows are separated from the top two by an unbridgeable mathematical gap.

Connection to Sen’s Capability Approach

[Modeling Axiom.] Amartya Sen’s capability approach (1999) argues that welfare should be evaluated not as scalar utility but as a vector of capabilities – substantive freedoms to achieve various “functionings.” Sen’s capabilities correspond to dimensions of the decision manifold: d_4 (autonomy/freedom), d_1 (material capability), d_9 (epistemic capability – the ability to make informed choices).

The Scalar Irrecoverability Theorem provides the mathematical backbone that Sen’s approach lacked: it proves that the dimensional collapse Sen warned against is not merely inadequate but mathematically irreversible. Sen argued qualitatively that scalar measures miss important aspects of welfare. The theorem shows quantitatively that no scalar measure, however carefully constructed, can avoid missing them.