# MacroVision v98: cash-funded corporate credit channel
29 October 2026. Illustrative extension, not estimated credit supply, a solvency model or proof of sustainable consumption.

## Finding
Adding an explicit, cash-funded household-to-firm loan channel raises completed actual paths from **202 to 220 of 576** under the larger tested limit. It completes 18 previously failing actual paths and makes no previously completing path fail in this grid. Most financing stops remain.

| Terms (normalized nominal limit, annual rate) | Actual complete | Control complete | Complete pairs | Actual corporate stops | Government stops | Displaced consumer-debt-free preservation | Displaced consumer-debt preservation |
|---|---:|---:|---:|---:|---:|---:|---:|
| Disabled | 202 | 72 | 42 | 230 | 144 | 83 | 75 |
| Limit 20, rate 4% | 212 | 144 | 88 | 220 | 144 | 93 | 85 |
| Limit 60, rate 4% | 220 | 144 | 88 | 212 | 144 | 98 | 90 |
| Limit 60, rate 8% | 220 | 144 | 88 | 212 | 144 | 98 | 90 |

Preservation retains v96's >=90% of each cohort's own initial nonhousing consumption throughout all annual observations, conditional on completion. This is not a common living-standard threshold. Complete-pair samples and actual-completion denominators differ across rows. Equal aggregate counts at 4% and 8% do not establish identical paths or rate irrelevance.

Actual paths with any capitalized interest among their recorded observations: **39** at limit 20/rate 4%, **127** at limit 60/rate 4%, and **127** at limit 60/rate 8%. Counts include both completing and later-failing paths. Capitalized amounts are not treated as cash receipts. Completion must not be interpreted as full cash servicing or solvency.

## Who lends and where the money comes from
Each household cohort may lend out of cash remaining after its existing consumption, housing, consumer-debt and equity decisions. This is a mechanical pro-rata cash sweep, **not voluntary optimized credit supply**. No bank, deposit creation, external sector or new government money is introduced. Lenders' cash falls by exactly the cash received by the firm, and each lender receives a matching corporate loan claim.

There is no new initial endowment. Existing initial household/firm balances remain unchanged. Loan claims are illiquid in this extension and are not counted as spendable cash or used to buy consumption. The sweep may reduce households' desired cash buffers; future effects reflect that imposed behavior.

The tested aggregate nominal credit ceilings are 0,20,60, alongside the existing initial output of 100 and real capital of 200. These are not currency amounts, per-household limits, observed LTVs or calibrated debt capacities. Rates and 10% annual opening-principal amortization are hypothetical. The zero-limit variant is the original v96 benchmark.

## Annual sequence and ledger
1. Retain the original government-feasibility gate and household fiscal calculations.
2. Before household spending, service existing corporate loans using opening firm cash only. Pay interest first, then up to 10% of opening principal from remaining firm cash. Allocate cash receipts pro-rata to opening lender claims. Principal receipt is an asset exchange and is added to available cash, not income.
3. Unpaid interest increases both corporate liability and lender claims. It is not cash, taxable household cash income or new money. Missed scheduled principal remains outstanding; no additional arrears penalty or forced acceleration is imposed.
4. Retain the inherited household consumption/equity decisions and residual investment calculation.
5. When candidate firm cash is negative, draw the minimum of the cash gap, available post-decision household cash, and remaining nominal credit ceiling. Allocate lender cash transfers pro-rata. If the gap remains, stop the path using the original financing constraint; do not add money or hide the stop.
6. Record closing claims, draw, paid interest, paid principal, capitalized interest and loan-balance residual in every accepted annual observation.

The credit ceiling limits **new advances**, not involuntary accrued interest. Capitalization can in principle raise outstanding claims above the ceiling, in which case no new credit is available. No rollover/default loss is simulated. Failed attempted steps are discarded like the benchmark and not counted as completed year-20 results.

Identities:
```
cash transferred by lenders = firm loan advance
closing corporate debt = opening debt + advances + capitalized interest - principal paid
cash interest receipt = firm cash interest payment
cash principal receipt = firm cash principal payment
household cash + firm cash + Treasury cash = unchanged initial money
```

## Fiscal and valuation closures
Tax and benefit rules remain inherited. Corporate financing interest is paid from cash, with no interest deduction in the hypothetical profit-tax base and no new household interest tax. Corporate interest is added to household cash income; the inherited transfer formula is not redesigned to means-test this new income. These are explicit policy assumptions.

Equity transactions retain the exogenous source-model quote. Reported household wealth adds corporate claims to lenders and subtracts corporate debt pro-rata from equity holders, avoiding counting internal loan claims as new aggregate wealth. This is a separate consolidated valuation adjustment, **not** an estimated market equity price or arbitrage-free trading equilibrium. Negative residual equity values are possible; no limited-liability restructuring or corporate default resolution is implemented. Treat wealth rankings and capitalized loan values cautiously; claims remain at par without recovery estimates.

The original consumer/mortgage resolution rules continue. This extension adds corporate financing, not a calibrated bank balance sheet, corporate leverage choice or lender risk model.

## Grid, reproduction and checks
Same 576-factor grid, fixed 20% terminal actual unemployment, capital-output elasticity .30, zero employment-capital feedback, saving/quote rules and matched-policy controls as v96. Four fixed terms variants yield **4,608 paths**. Controls remove displacement/productivity but retain tax/capture and housing schedules.

Run:
```bash
python scripts/corporate_credit_v98.py
python -m unittest discover -s scripts -p 'test_*v9[678].py' -v
```
The full local case file is data/processed/corporate_credit_v98.json. The checked-in summary is data/processed/corporate_credit_summary_v98.json. Code hashes include the separate v98 model fork, ledger and inherited dependencies. Original v77 and v96 model files remain unchanged.

Nine combined tests pass (four new): cash-conserving bounded advances, interest/principal versus capitalization, zero-credit parity, cash and loan-balance identities, and netting of corporate claims/debt across households. All **1,152 disabled actual/control compact paths reproduce v96 exactly**. The full grid checks monetary and corporate loan-balance residuals below 1e-8 in every accepted observation. A test initially used a non-borrowing example; it was corrected to exercise an actual positive draw.

## Implication and priority
An explicit loan counterparty completes some financing-deficient paths without creating money, but it does not solve most corporate stops and does not prove debt sustainability. Prioritize a **corporate debt-servicing and lender-risk audit** next: separate full cash servicing from interest capitalization, locate persistent unpaid interest and debt-equity valuation problems, and define loss/default handling before treating more complete paths as economic improvement. Changing the pass condition to favor credit would not establish the thesis.

References: [v96 sensitivity](MACRO_SENSITIVITY_V96.md), [v97 first-stop diagnosis](FINANCING_DIAGNOSTIC_V97.md). No external empirical calibration or causal approval is added.