# Resource-constrained investment and financing guard — v123

This declared, uncalibrated annual scenario experiment changes only productive-investment allocation. It is not evidence that a real economy will preserve purchasing power.

## Prespecified rule
Cap planned productive investment at max(0, Y − government goods − household consumption − housing maintenance + nominal opening inventory). Keep total investment equal to the goods residual; inventories receive the difference. Check corporate cash against total investment including inventories using actual household equity subscriptions. No new money, loan, or retained-dividend contract is introduced. A financing-infeasible plan still stops before commitment.

The same 324 scenarios and 12 controls from v121 are replayed with the rule enabled and disabled. Elasticities: 0, .15, .30, .45; investment plans: residual, 5%, 7.5% of output; consumption floors: .90, .95, 1; owner spending responses: 0, .5, 1; collection ceilings: .05, .10, .20. Original matched control year-1 consumption references are held fixed. Qualified success requires a completed constrained control, completed policy, all annual recipient floors and fully funded current grant requests.

## Results
- Completed policy paths rise from 284 to 297 of 324.
- All 13 inventory stops are avoided. Ten of those newly completed paths qualify; three still fail the grant/floor criteria.
- Qualified successes rise from 257 to 267. These grid counts are not probabilities or independent observations.
- All 27 corporate-cash stops remain at year 20, as in the matched no-supplement control. They cannot be attributed to supplements.
- Productive investment is cut in exactly 13 paths, all with zero capital-output elasticity and a 7.5% output plan.
- Cumulative nominal investment cuts range from 0.975109 to 20.503278 model units. Forgone year-20 real capital relative to the SAME realized-output investment plan ranges from 1.147187 to 24.087929 units. This comparison is an accounting plan benchmark, not an executable unconstrained economic path.
- At the last year committed by both variants, capital, output and cohort real incomes coincide in those 13 paths: the baseline stops before the first cut. There is no valid baseline year-20 income comparison.
- With elasticity zero, output and aggregate payout do not respond to capital cuts by construction. This cannot establish that cuts have no future income cost. None of the positive-elasticity paths triggers this resource cap, so this grid does not identify that cost under capital feedback.
- Maximum money-identity error is 1.78e-13; goods-identity error is zero.

## Implication for the thesis
Goods-feasible investment allocation can extend the horizon over which funded transfers and consumption floors are assessed. Completion alone is insufficient. A financing guard exposes the remaining closure problem rather than creating funding. A next experiment should explicitly fund investment through a retained-dividend or household-financed claim contract, and test capital feedback where the resource cap actually binds. Such a contract changes household disposable income and must be evaluated against its own matched controls.

Four tests cover disabled-rule recovery, binding resource-cap ledger identities, preservation of the corporate-financing stop, and the forgone-capital calculation. Annual timing, exogenous prices/employment, no depreciation and no inventory carrying cost remain limitations.

Reproduce: `python scripts/resource_constrained_investment_v123.py --output data/processed/resource_constrained_investment_v123.json`