# Stage 127 — matched corporate cash attribution

Compare 324 identical scenario keys (capital elasticity, investment rule, consumption floor, owner spending response and collection ceiling) for each 25%, 50%, 75% and 100% retention policy against 0%. Keep the original 1,620-path experiment intact.

Report transitions in completion and qualification. A failure-year cash decomposition is comparable only if **both** matched paths fail in the **same year**; compare opening cash after debt service, retained payout, net household equity subscriptions and investment outflow. The delta of closing cash must reconcile. Different failure years are deliberately excluded from cash-flow attribution; a sum of unmatched shortfalls is not causal evidence.

This is a declared, uncalibrated scenario experiment. The identity explains the composition of cash, not the ultimate structural cause of behavioral changes. Results are populated only after CI execution.

Reproduce: `python scripts/matched_cash_attribution_v127.py --output data/processed/matched_cash_attribution_v127.json`

## Executed findings — GitHub Actions run 393

All 1,296 matched comparisons were executed and CI passed. For each nonzero retention rate there are 324 pairs.

| Retention | Completed→failed | Failed→completed | Qualified successes lost | Qualified successes gained | Same-year failed pairs |
|---|---:|---:|---:|---:|---:|
| 25% | 27 | 0 | 57 | 0 | 0 |
| 50% | 27 | 0 | 96 | 0 | 0 |
| 75% | 27 | 0 | 156 | 0 | 0 |
| 100% | 27 | 0 | 213 | 0 | 0 |

At each retention rate, 270 pairs completed under both policies, 27 completed at 0% but failed under retention, and 27 failed under both policies. **No pair of failed paths failed in the same year.** Consequently no failure-year cash-delta attribution is identified by the prespecified same-year rule. The zeroes in the output's cash-delta sums indicate an empty comparison set, **not** a zero cash-flow effect.

The matched-grid results establish a worsening in completion and floor qualification for the declared scenarios. They do not identify which of the cash-bridge components caused the transitions. Next research should compare pre-failure years shared by matched paths, with care around selection and timing. These are uncalibrated model scenarios, not real-world probabilities or causal policy estimates.