# V108 — A treasury buffer with a lender, interest and terminal repayment

29 October 2026. Controlled extension of the frozen v107 settlement experiment. No empirical calibration, estimated sovereign borrowing rate or endogenous macro feedback.

## Question and financing
Can a cash buffer prevent intrayear government arrears once its financing cost and final repayment are recorded?

An identified model lender transfers cash to treasury before the first flow month, receiving a matching principal claim. Government gross debt rises by the advance; cash is initially the offsetting asset. The loan has fixed principal until the final December of the twenty-year horizon. Interest is due monthly at the declared annual rate divided by twelve. Idle buffer cash earns no return. No refinancing, capitalized interest or debt write-off is introduced.

The same lender and taxpayer cash endowments are used in all eight cases. Unlent lender cash remains with the lender. The common taxpayer cash reserve is large enough for the largest tested tax bill; paying a surcharge reduces this account's ending cash rather than adding money to that policy's initial endowment.

This is still a synthetic external lender/taxpayer sector. Their spending behavior, preferences and actual tax capacity are not estimated. Equal opening cash makes the financial comparison consistent; it does not establish that the financing is available in a real economy.

## Experimental design
The underlying annual source remains v100's default complete path and v107's full government obligations, with only the displaced-free/owner-free household cohort replayed. Monthly benefits and source spending remain fixed. Test tax collection at quarter-end and December, each with:
1. no buffer;
2. a financed buffer with zero interest as a costless control;
3. the same buffer with 4% interest and no added revenue;
4. a slightly resized buffer with 4% interest and an explicit cash-tax surcharge funding the interest.

Zero and 4% are scenario assumptions, not current market rates. The surcharge is buffer times .04 per year, collected on the same tax schedule as the base receipts. Source private income, prices, desired spending and asset-purchase plans remain frozen. Original government interest remains separate from the new loan's interest allocation.

Goods and old/new interest retain v107's payment priority ahead of benefits. Paid interest is allocated pro rata between old and buffer-lender interest obligations, with unpaid amounts carried explicitly. Original scheduled bond principal is paid before the buffer's terminal principal. Outstanding principal remains a claim if payment fails; it is not silently forgiven.

## Conditional buffer sizing
Without interest, size the buffer from the largest cumulative gap between scheduled source obligations and cash-tax receipts. With interest funded by the surcharge, the financing-cost timing gap itself depends on the buffer. Solve the resulting linear bound month by month. This gives conditional lower bounds for the **declared deterministic profiles**, not an optimized or empirically recommended reserve.

| Tax timing | Zero-cost buffer | Buffer with 4% interest funded by surcharge |
|---|---:|---:|
| Quarter-end | 3.8523 | 3.8781 |
| December | 21.1876 | 21.9941 |

Units are normalized aggregate cash in the original annual-output-100 model, not currency amounts. A test reducing the cost-funded bound by .01% produces arrears. Final principal solvency is checked separately in the full cash replay.

## Results
All cases cover 240 artificial months over twenty source years. Government-arrears counts include overdue terminal loan principal when applicable.

| Tax timing | Financing case | Arrears months | Selected-consumption shortfall months | Unpaid terminal buffer principal | Added tax over twenty years |
|---|---|---:|---:|---:|---:|
| Quarter-end | No buffer | 160 | 0 | 0 | 0 |
| Quarter-end | Zero interest | 0 | 0 | approximately 0 | 0 |
| Quarter-end | 4%, cost unfunded | 107 | 0 | 3.0818 | 0 |
| Quarter-end | 4%, cost tax-funded | 0 | 0 | approximately 0 | 3.1025 |
| December | No buffer | 220 | 11 | 0 | 0 |
| December | Zero interest | 0 | 0 | approximately 0 | 0 |
| December | 4%, cost unfunded | 99 | 9 | 16.9501 | 0 |
| December | 4%, cost tax-funded | 0 | 0 | approximately 0 | 17.5952 |

The unfunded-cost cases pay interest by eroding the buffer. A liability remains at maturity even though interest payments were made. In the tested cost-funded closures, added taxes equal total loan interest, all obligations are paid on time and principal is repaid. The zero-cost cases are an accounting control, not evidence that costless finance is available.

These are different fiscal closures: the cost-funded result uses additional tax receipts. It must not be described as eliminating arrears with the same revenue or without a cost. The table does not establish welfare improvement, causal liquidity protection or sustainability under permanent automation-driven unemployment.

Counts are deterministic scenario outcomes, not default probabilities or sample frequencies. A consumption shortfall is relative to the frozen target, not the 90%-of-reference benchmark. Material residual/arrears counts use tolerance 1e-9.

## Accounting and validation
The complete transaction tape begins with the lender-to-treasury advance and includes subsequent interest and principal payments to that lender. The wrapper splits v107's omnibus creditor receipts into actual old-holder and buffer-lender payments; neither creditor spends in this replay, so that split does not change treasury feasibility.

Every month's cash tape reconciles to all closing accounts, including the initially idle lender and taxpayer reserves. Government bonds match old-holder plus buffer-lender principal claims. Paid plus unpaid loan principal equals the original advance, and paid plus unpaid interest equals cumulative obligations. January openings match December closings. Maximum reported cash identity residual is below 1e-12; unpaid terminal balances in the successful cases are below 3e-13.

Twelve new tests passed with the previous thirty-eight component tests. They check common initial cash, complete payment tapes, unchanged no-buffer controls, matching creditor claims, recorded interest, terminal repayment failure, tax incidence on the external taxpayer cash account, the conditional bound, rejected unavailable finance, calendar carry and exact archive reproduction. The new suite is added to CI.

## Limits and next gate
A surcharge is debited from the common external taxpayer cash account, but its distribution across the modeled household cohorts and its effect on their disposable income are not yet represented. Lender liquidity/cash receipts are tracked, but changes in lender consumption or investment are also absent. Source production, supplier solvency, prices, dividends and household spending targets remain frozen.

Thus the government cash-flow comparison now includes financing cost and terminal principal; the broader macro and household welfare comparison remains incomplete. Historical v105–v107 outputs and source model rules are unchanged.

Next, allocate the funding tax to explicit household groups and reconcile their disposable income, cash and consumption. A policy that preserves benefit recipients' spending by reducing another group's resources needs that tradeoff shown before any claim of general purchasing-power preservation. Wider household debt and firm feedback remain subsequent gates. No restricted records or new empirical estimates were used.

Reproduce:
```
python scripts/test_financed_fiscal_buffer_v108.py
python scripts/financed_fiscal_buffer_v108.py --output data/processed/financed_fiscal_buffer_v108.json
```

The artifact contains source fingerprints, common endowments, all annual results, first-year months, every December and all selected-consumption-shortfall months. Full monthly accounts and transaction tapes are reproducible through build().

[Results and fingerprints](data/processed/financed_fiscal_buffer_v108.json) · [V107 fiscal settlement](MONTHLY_FISCAL_BRIDGE_V107.md) · [V106 cohort bridge](ANNUAL_MONTHLY_BRIDGE_V106.md)