MacroVision v96: macro sensitivity and completion boundaries

29 October 2026. Reproducible hypothetical simulations, not estimates, probabilities or validation of the thesis.

Main finding

Across 576 fixed-grid actual scenarios, 202 complete 20 years. 144 stop because existing government commitments lack cash financing, and 230 stop because corporate liquidity, real capital or equity redemption capacity is exhausted. Failure is a modeled financing/closure boundary, not a predicted crisis probability.

For each group, preservation means a nonhousing-consumption index of at least 90 of its own initial level in every annual observation, conditional on completion. This is a diagnostic threshold, not an empirically justified adequate living standard. It does not measure total real disposable income, housing quality, food security, utility or a common absolute standard.

Preservation relative to initial consumption does not imply being as well off as the evolving control. Cohorts without consumer debt may have mortgages; "no debt" is not a universal debt-free label.

Model and grid

Reuse the v77 credit-resolution household simulation and v74 reduced-form capital feedback, without changing their economic rules. V78/79 partial mortgage/arrears anchors are not imported into this grid. All household and production parameters here remain illustrative.

Full factorial:

• productivity gain: .25, .50, 1.00;
• price pass-through: 0, .50, 1.00;
• terminal profit tax rate: .25, .35;
• accessible profit share: .50, 1.00;
• initial productive ownership: concentrated, broad;
• housing: renter, owner without mortgage, fixed mortgage, reset mortgage;
• stress: stable or rate shock.

Fix 20-year horizon, ten-year transition, terminal unemployment .20, capital-output elasticity .30 and capital-labor elasticity zero. Thus unemployment reaches exactly 20% in completed actual paths, with no feedback-induced employment change. Fixed saving rule consumes 90% of modeled spendable income; productive equity has a constant real quote. Resolution uses inherited two-year rules, cost share .10 and zero recourse. Housing/credit valuation, endogenous investment and cash-financing constraints remain inherited assumptions, not equilibrated market prices or optimized decisions.

Comparison and failure handling

Each actual case has a control that removes displacement and productivity gain but retains its tax/capture schedule, ownership, housing and stress. Because the source control retains policy interpolation, call it a no-displacement/no-productivity control under matched policies, not a pure no-automation/no-policy-change counterfactual.

Only 72 controls and 42 pairs complete. For incomplete paths all terminal outcomes and preservation classifications are null, never the last observed year masquerading as year 20. Prefix minimum consumption and credit loss are explicitly labeled observed-prefix diagnostics. No imputation of failed paths.

Every completed pair has a negative terminal real nonhousing-consumption gap for both displaced cohorts. This applies only to 42 selected feasible pairs; it cannot characterize all 576 cases. No complete pair exists for either mortgage tenure or the .35 profit tax setting in this grid. Control failures prevent assessing those comparisons.

Factor-conditioned diagnostic counts

Each row averages over/counts all other factor combinations, not a causal effect or probability.

Higher modeled productivity or price pass-through is not uniformly protective under these closures. It changes nominal revenues, profits, saving/issuance and capital financing jointly. The counts identify sensitivity and feasibility dependence; they do not isolate which structural channel causes a specific failure.

These totals are design dependent. Adding more grid points changes shares, and no distribution or likelihood is assigned to inputs. Cross-tenure ownership comparisons also change initial housing wealth. Do not interpret owner-versus-renter counts as a causal housing intervention.

Implication and next step

Within this model, cheaper goods alone do not establish preservation of displaced households' consumption. Budget financing, initial asset distribution, debt/housing obligations and corporate liquidity must be checked together. The grid does not prove a general theorem or falsify the broad thesis: many paths expose limitations of the assumed saving, equity issuance and residual-investment closure.

Priority next: audit the corporate financing failure boundary, separating insufficient residual investment financing from real-capital depletion and equity redemption, with explicit alternatives fixed before new result comparisons. This is more informative than adding a favorable scenario or treating the 374 truncated actual paths as economic forecasts. Retain the original model unchanged for comparison. Quantifying a boundary across an input grid is not solving for a unique critical threshold or a global sensitivity index.

Reproduction, output and checks

Run from repository root:

```bash python scripts/macro_sensitivity_v96.py python -m unittest discover -s scripts -p test_macro_sensitivity_v96.py -v

``` The script generates the full 576-case paired output locally at data/processed/macro_sensitivity_v96.json. The checked-in summary is data/processed/macro_sensitivity_summary_v96.json; it is the generated output with only the large cases array removed. It retains every factor-conditioned aggregate and code hashes. Full case results can be rebuilt without network, optional libraries or participant data.

Three meaningful tests pass: the explicit zero-transition intervention matches its control; money and output-use identities hold with the fixed unemployment path; failed paths cannot expose terminal outcomes/preservation. The full grid's maximum money residual is checked separately before publication. Accounting consistency does not establish empirical correctness or macroeconomic equilibrium.

Sources: existing repository scripts v71–77, pinned by SHA-256 in the generated summary. No new empirical observations, source estimates, causal approvals or calibration updates.