K-shaped wealth concentration — a Decision Pack for a fiduciary allocator

Simulation-backed decision support for fiduciary allocators

A stock-flow-consistent (SFC) two-cohort simulator inside Comuvia's Dyno-Sim platform (24 scenario types across 23 model families as of 2026-07), used to answer a pre-allocation strategic question — how durable is the K-shaped wealth-concentration regime under plausible policy and shock paths, and which intervention class actually bends the trajectory. Every output traces to a run id, an accounting identity, and a validation test.

The question

"Over a 30-year horizon, how durable is the K-shaped wealth-concentration regime — and which class of intervention (flow redistribution vs. stock ownership-broadening) actually bends the trajectory enough to change a fiduciary's strategic asset-allocation thesis, with what confidence under plausible shock paths?" That is the kind of pre-allocation strategic question a fiduciary allocator (pension, sovereign, large endowment) brings to Comuvia. It is not a portfolio-construction question, and it is not a stock-picking question. It is a question about which regime the next 30 years live in — and Comuvia answers it with a simulation, not a memo.

SFC family chosen — and why

Dyno-Sim today (2026-07) hosts 24 runnable scenario types across 23 model families — spanning demographics, ecological economics, financial instability, central-bank operations, settlement mechanics, systemic-risk contagion, and composed cross-domain runtimes — each SFC family with publishable accounting identities and an owned verifier. For the K-shape question we use k_shaped_economy_v1 (alias K-Shaped Inequality Model) — a two-cohort SFC built for exactly this regime:

• Upper cohort: capital-income earners, low marginal propensity to consume.
• Lower cohort: wage-income earners, high MPC, exposed to automation.
• Production: Cobb-Douglas Y = tfp · K^capital_elasticity, with the capital share rising over time at automation_rate toward a capital_share_max ceiling.
• Two policy levers that map cleanly to real-world interventions: flows — redistribution_rate (tax-and-transfer); stocks — ownership_broadening and ownership_transfer_rate (employee-ownership funds, citizen wealth funds, baby bonds).

We chose this family because the allocator's decision hinges on a stock-vs-flow distinction that flow-only macro models structurally cannot answer: redistribution moves income, ownership-broadening moves the claim on future capital income. The two have very different 30-year shapes.

The K-shaped family also now extends across borders: a Cross-Border K-Shaped Model (shipped 2026-07) adds tax competition between jurisdictions, paper-profit base erosion, herding capital-flight sudden stops, and a Pillar-Two-style coordination floor — so the same stock-vs-flow question can be asked for a multi-jurisdiction allocation, where capital mobility changes which interventions actually hold.

Parameter sweep

The sweep varied four parameters across two scenario files (sample-k-shaped-economy.json baseline, sample-k-shaped-economy-ownership-intervention.json intervention) plus the platform-wide 20,000-replica shock Monte Carlo:

1. automation_rate — pace at which the capital share rises toward capital_share_max (low / central / high).
2. redistribution_rate — flow-side tax-and-transfer intensity (zero, moderate, aggressive).
3. ownership_broadening + ownership_transfer_rate — stock-side claim transfer (off, year-8 onset, year-12 onset).
4. Shock surface overlay — the same scenarios run through the platform's 20,000-replica Monte Carlo on a Minsky 1990–2020 (FRED-UNRATE-calibrated) debt-cycle + TFP-shock surface, to measure how often a plausible shock path flips the recommended intervention class.

Compute path: vectorised CPU replicas at ~22.8× speedup over the sequential loop, scaling to 100,000 replicas in ~2.3 s; the CuPy GPU path on server-large is available for higher-fidelity sweeps.

Publishable run outputs

All four numbers below are real, dated 2026-06-21 in the Dyno-Sim change log, and reproducible from the cited scenario files:

• Baseline (no intervention). Over 30 years the model concentrates to a ~47× upper/lower wealth ratio with a ~0.96 Gini-proxy — a textbook K-shape, and the regime an allocator would be sizing for if they took today's trajectory at face value.
• Ownership-broadening intervention (year-8 onset on the same parameter base) bends the trajectory to ~2× wealth ratio / ~0.37 Gini-proxy — a different macro regime entirely. Flow-only redistribution at comparable fiscal cost does not reach this shape; the lever that matters is the stock lever.
• Accounting identities hold to ~1e-14. Income decomposition (labour + capital + transfers) and wealth conservation (cohort balance-sheet sum to economy-wide stocks) close to numerical precision in every run. This is the credibility floor: no hidden flows, no money created off-balance, no Gini drift coming from a leak in the books.
• Tail-risk shock surface. Re-running the recommendation under the platform's 20,000-replica shock Monte Carlo (Minsky 1990–2020 + TFP-shock surface) produces a 47.0% recommendation-flip [46.3–47.7%] Wilson CI versus the no-shock baseline — the same engine and the same uncertainty discipline Comuvia ships for the recession case.

Decision the allocator would make

The Decision Pack gives the allocator a regime call, not a portfolio.

• If the K-shape compounds (baseline trajectory), the 30-year strategic posture is consistent with persistently elevated tail risk on consumer-facing cash flows, a rising share of returns accruing to capital-owning balance sheets, and Minsky-Keen fragility on private-sector debt — i.e., the regime in which the recession-fragility case study's 47% recommendation-flip is the steady-state shock surface, not a one-off.
• If stock-side ownership-broadening policy materially deploys (intervention trajectory), the regime mean-reverts toward a wage-income-led growth path; the same allocator-side tail-risk thesis weakens substantially.
• The Decision Pack states which observable policy signals (legislation, ownership-fund scale, transfer-rate thresholds) move the trajectory from one regime to the other, so the allocator can monitor a small number of leading indicators rather than re-running the simulation.

Each line of the Pack traces to a run id, an SFC equation, and a validation test — auditable by a fiduciary investment committee, defensible to a board, and re-runnable in-house under licence.

What this doesn't fit

This is not portfolio construction. It does not recommend asset classes, weights, tickers, factors, or hedges.

This is not stock-picking, sector rotation, or market timing. The horizon is 30 years and the unit of analysis is the regime, not the security.

This is not a calibrated point forecast of inequality in 2056. It is a structural model of which intervention class bends which trajectory, with honest uncertainty quantified by the 47% recommendation-flip — the simulation states what it does not know as clearly as what it does.

Research / paper-only. Not investment advice. Not a forecast.

Why it matters for a buyer

For a fiduciary allocator, the value is decision-grade and auditable: a transparent, vendor-neutral, stock-flow-consistent model whose accounting identities close to machine precision, whose uncertainty is quantified in a single defensible number, and whose outputs trace to runs, equations, and tests an investment committee can audit. It is the opposite of a black-box macro score — and it lives at the pre-allocation strategic layer where the largest decisions are actually made.

Engagement

The engagement is a Simulation Decision Pack: scoping call, parameter and scenario design with the allocator's strategy team, sweep and shock-surface run, written Pack with every claim traceable to a run id, and a working session to walk the investment committee through the audit trail. First-proof Decision Packs start at $50K; follow-on Packs against the same model family typically $100K+.

Request a 60-min Decision Pack scoping call

← All case studies