# Stage 122 — Failure-year goods and cash accounts

## Finding

The 40 halted stage-121 policy paths have two distinct first-stop mechanisms:

- **13 inventory failures:** planned productive investment and consumption exceed current output plus opening inventory. Corporate cash is still positive.
- **27 corporate cash failures:** investment outlays exceed opening corporate cash and equity issuance. Goods balance holds, no supplemental tax or grant is paid, and the matched no-policy control fails identically.

The 27 cash cases repeat one underlying no-policy trajectory across floor/response/ceiling labels. They are not 27 independent observations that redistribution causes distress.

This stage adds diagnostics only. It does not alter the paths or simulate a repair.

## Inventory example: year 10

Configuration: zero capital-output elasticity, productive investment target of 7.5% of output, 90% floor, owner q=0 and 10% dividend ceiling.

| Failure-year account | Nominal model units |
| --- | ---: |
| Current output | 102.0000 |
| Value of opening inventory | 1.2098 |
| **Goods available including inventory** | **103.2098** |
| Government goods | 17.1700 |
| Nonhousing consumption | 76.5608 |
| Housing upkeep | 2.0000 |
| Planned productive investment | 7.6500 |
| **Planned goods uses** | **103.3808** |
| **Goods gap** | **0.1711** |

Attempted real inventory is -0.2013. The maximum productive investment consistent with those current consumption/output/inventory values is 7.4789 rather than 7.6500. Corporate cash after the proposed transactions is still about 30.8598, so this failure is not explained by a cash deficit.

The matched control has the same output and planned productive investment. Its nonhousing consumption is 2.0138 lower, and its opening real inventory is 3.5863 higher. Prior consumption differences have already depleted the policy path's inventory cushion. The control can continue.

Across all 13 inventory failures, the first-stop nominal goods gaps range from 0.0023 to 0.7404 model units. Some paths can also have unmet grant requests; the inventory label identifies the first model stop, not an assertion that every other gate passed.

## Cash example: year 20

Configuration: capital elasticity 0.45, residual productive-investment rule. The following accounts are identical in all 27 policy labels and in the failed control.

| Failure-year account | Nominal model units |
| --- | ---: |
| Corporate cash before investment | 3.0489 |
| Net equity cash raised | 11.6225 |
| New corporate borrowing | 0.0000 |
| **Available financing** | **14.6714** |
| Investment outlay | 25.8923 |
| **Additional cash financing required** | **11.2210** |

Attempted corporate cash is -11.2210. Productive capital remains positive, and the goods gap is zero. Supplemental tax, grants and requested grants are all zero. The floor does not require redistribution in this trajectory, so the failure cannot be attributed to its supplemental transfers.

This reveals a mismatch between investment specified by the residual rule and funding produced by the inherited equity/cash rules. The gap is a financing requirement, not evidence that an external lender, new shareholder or public fund would supply it.

## Exact diagnostics and interpretation

For every failed proposal:

- Goods resources = current output + current price × opening real inventory.
- Goods uses = government goods + consumption + upkeep + planned productive investment.
- Corporate funding = corporate cash before investment + net equity cash + actual new credit.
- Cash gap = max(0, total investment − corporate funding).

The local investment reduction shown for inventory failures is exactly the goods gap. It would bring the proposed ending inventory to zero **holding all other current values fixed**. It is not applied, and future output, grants or financing are not recomputed under such a change.

Likewise, the cash gap is not inserted as a loan or unexplained cash balance. There is no hidden rescue.

The diagnostics distinguish a model closure that demands too many goods from one that lacks transaction funding. Neither establishes an empirical limit on an automation economy. Prices, investment targets, production, equity demand and credit could adjust in reality; these adjustments are absent or fixed in the current experiment.

## Reproduction and validation

The diagnostic variant replays exactly the 40 halted paths from the published stage-121 grid. Failure year and reason must match. Proposed failure-year accounts are captured before commitment; the last valid state remains the previous year. Successful control accounts at the same calendar year, or the control's own failed proposal, provide the comparator.

Six tests cover unchanged committed states and failure outcomes, failure-year dating, exact goods and financing identities, local inventory arithmetic, identical no-policy cash failure, feasible inventory controls and archive reproduction.

```sh
python scripts/failure_accounting_audit_v122.py --output data/processed/failure_accounting_audit_v122.json
python scripts/test_failure_accounting_audit_v122.py
```

[All forty audits and matched accounts](data/processed/failure_accounting_audit_v122.json) retain current flows, opening stocks, attempted closing stocks, gaps and source hashes. Original simulator files were unchanged. Repository CI remains the merge gate.

## Next step

Test explicit, declared investment-adjustment rules that respect goods availability and financing, with the same rule in each policy and control. Track any reduction in productive capital and future income rather than treating removal of a stop as proof of preservation. Any credit or equity remedy must identify its payer and matching claim.