# V104 — Monthly transfer-response contract and annual-model audit 29 October 2026. Design and code audit only. No new empirical estimate, calibrated parameter or macro result. ## Access decision The researcher reports that the provider declined access because the current application does not meet eligibility conditions. The current UAS/UASFin access route is therefore closed for this project. The provider message has not been independently read in this update. This records the project decision, not a claim about every provider product or every applicant. No participant records were acquired. No collaborator, new application or follow-up contact is presumed. [V95](UAS_ACCESS_CONCEPT_V95.md) remains the historical enquiry concept; its status notice is updated. Public documentation omits personal eligibility particulars and correspondence. Denied access is a feasibility constraint, not evidence for or against the thesis. ## What the code actually supports Audit anchored to main commit 953e8c3767bf57ac63f5933f350442a262dafee7; file blob identifiers are recorded in the JSON. | Code | Finding | Consequence | |---|---|---| | scripts/coupled_thesis_v71.py — simulate | Annual loop; annual fiscal benefit calculation | Payment month and intrayear cash financing cannot be recovered | | scripts/credit_resolution_v77.py — simulate_households | Benefits enter gross income before .90 consumption rule; cohort totals later divided by mass | Adding an MPC response to unchanged consumption double-counts the transfer | | scripts/corporate_workout_model_v100.py — simulate_households | Same annual spending rule with firm-credit/workout accounting | Monthly debt service cannot be inferred from annual default streaks | | v77/v100 year-zero branches | Reference consumption shown; stocks are not settled as a normal year | Year zero must not become twelve cash-flow months | The existing annual macro and household equations remain unchanged. Annual liquidity feasibility does not establish monthly feasibility: a receipt in December cannot pay a January bill without opening liquidity or explicit financing. Equally, annual cohort-average employment does not supply individual separation dates. ## Measurement contract Unit: fixed modeled cohort, with aggregate nominal flows and an explicit cohort mass. Per-person values divide by that mass once. Currency remains normalized model units; no conversion to survey dollars is implied. For each month record opening and closing cash, paid transfers, net wages, distributions, consumption cash, debt interest/principal, asset purchases/sales, borrowing and lending. Distinguish actual cash from accrued interest, write-offs and valuation changes. All transfers and financial exchanges require the matching payer/counterparty entry. Define a paired baseline with the same job loss, opening stocks and price path. Specify the fiscal source of the incremental transfer and all counterpart effects; holding prices fixed is a partial-equilibrium diagnostic, not a claim that aggregate effects disappear. For a first nonnegative transfer-only prototype, a possible desired response is: ``` desired_delta_C[m] = sum(k[j] * delta_T[m-j], j = 0..H) ``` H and k are deliberately unspecified. Nonnegative k and sum(k)<=1 would be scenario restrictions of this first prototype, not empirical laws or fitted findings. Do not reuse the same kernel for anticipated changes, negative transfers or different transfer persistence without a new specification. The baseline spending rule must exclude the incremental transfer before adding this response. Keep desired and realized spending separate: cash limits, asset sales and declared credit rules can constrain actual spending. Report those constraints rather than interpreting a constrained realized response as the unconstrained kernel. Monthly cash must reconcile: ``` closing cash = opening cash + cash receipts + asset sales + borrowing - consumption cash - interest paid - principal paid - asset purchases - lending ``` Cash receipts include transfers exactly once. Principal repayments are financial exchanges; interest is recorded separately from the target consumption basket. Housing maintenance, rent, mortgage interest and principal need separate fields before mapping any source spending definition. For each year, sum nominal flows, sum monthly real consumption after monthly deflation, and use December closing stocks. January must open at the previous December close. A December receipt's response carries into the next year. Extend the reporting window H months beyond the last receipt or disclose the omitted tail. Do not sum stocks, average monthly consumption flows, or multiply a monthly MPC by twelve. Year zero supplies reference stocks and consumption indices only. The proposed incremental module cannot by itself fund permanent benefits. Changes in spending eventually affect firm cash/investment; integration must reconcile households, treasury, firms and lenders before interpreting the original preservation criterion. ## Required acceptance cases before integration The JSON lists ten cases: zero shock; delayed response without anticipation under this kernel; December carryover; early bills versus late receipts; principal classification; negative-cash rejection/explicit financing; year-zero exclusion; cohort-mass reconciliation; finite tail; monthly price deflation. These are declared future acceptance cases, not tests claimed to have run. Pass accounting first on synthetic inputs with no empirical label. Only then integrate the module with a frozen annual reference path and explain any change in completion, consumption and cash failures. Numerical agreement with an imported calibration target is not independent validation. ## Public-data fallback Existing feasibility audits remain the starting point, not a newly verified data inventory: - SIPP/SHED can inform public household financial-resilience associations already investigated; they do not automatically provide the required transaction spending outcome. - CE is a spending-oriented candidate from v92, but event timing, strictly pre-event liquidity and adequate precision remain unresolved. - Published transfer studies supply contextual comparison estimands, subject to v103's population, horizon and basket limits. Choose by the required observable, timing and access conditions. Do not silently substitute food insecurity for consumption, annual resources for pre-event liquidity, or permanent unemployment for a temporary shock. No new data access is asserted. ## Decision The annual simulator is not ready for empirically calibrated monthly responses. The next implementation is an isolated synthetic monthly cash ledger with the acceptance cases above, followed by a separate source-feasibility decision. The thesis gains a checkable distinction between desired demand, funded transfers and spending that cash actually permits. [Machine-readable contract](data/metadata/monthly_transfer_contract_v104.json) · [V103 estimands](TRANSFER_ESTIMANDS_V103.md) · [V101 thesis synthesis](THESIS_SYNTHESIS_V101.md) Validation: inspected the source functions at the recorded revision; parsed the contract JSON and checked its unresolved gates and publication links. Existing CI checks publication and regressions, not empirical correctness.