# Harm pricing — the 42-plaza shadow-software scam (2026-09-06)

Model anchored on the one measured number in the record: **STF's ₹45,000/day cash siphon at Atraila Shiv Gulam** (Delhi HC judgment, 2026-07-03, para 5 — STF press note 36), which over 730 days gives **₹3.29 cr for Atraila alone**, matching Bhaskar's '₹3.28 crore embezzled at one location'.

## Inputs

- **Anchor intensity:** ₹45,000/day at Atraila (car fee ₹115, Rajmargyatra DB id 675). Implies ~390 car-equivalents/day suppressed ≈ a 4-5% cash-lane share at a ~10k/day plaza — consistent with the HC record's cash-share spike.
- **Fee scaling (Model 1):** suppressed volume uniform, revenue scales with each plaza's car fee vs Atraila's. 19/42 plazas fee-matched (median ratio 0.70, range 0.17-1.26); 23 unmatched at median fee ₹80 (ratio 0.70).
- **Uniform intensity (Model 2):** every plaza runs at Atraila's absolute ₹45k/day — the ceiling assumption (Atraila was the accused's home plaza; franchise plazas likely ran cooler).
- **Coverage f and window:** conservative 25%/12 months; central 50%/18 months; aggressive 100%/24 months (mid-2023 → STF raid 19 Jan 2025).

## Results (₹ crore, 42 plazas)

| Scenario | Model 1 (fee-scaled) | Model 2 (uniform) | Take |
| --- | --- | --- | --- |
| **Conservative** (f=25%, 365d) | 12.6 | 17.2 | 25% lane-shift coverage, 12-month window |
| **Central** (f=50%, 548d) | 37.8 | 51.8 | 50% coverage, 18-month window |
| **Aggressive** (f=100%, 730d) | 100.8 | 138.0 | 100% coverage, 24-month window |

**Best estimates:** Conservative **₹13-17 cr**, Central **₹38-52 cr**, Aggressive **₹101-138 cr**.

## Network scaling (ED's ~100-plaza footprint)

At the 42-plaza network median fee ratio (0.696), scaling the same scenarios outward — linear in plaza count, holding intensity models fixed:

| Scenario (coverage / window) | 42 plazas — M1 / M2 | 100 plazas — M1 / M2 | 200 plazas — M1 / M2 |
| --- | --- | --- | --- |
| Conservative (25% / 12 mo) | 12.6 / 17.2 | 29.2 / 41.1 | 57.8 / 82.1 |
| Central (50% / 18 mo) | 37.8 / 51.8 | 87.6 / 123.3 | 173.4 / 246.6 |
| Aggressive (100% / 24 mo) | 100.8 / 138.0 | 233.4 / 328.5 | 462.1 / 657.0 |

(₹ crore. The script prints the full grid; the report's section 13C carries the rounded table.)

## Sanity checks

- **STF's ₹120 cr network claim** ⇒ ₹2.86 cr/plaza ⇒ ~635 days at full Atraila intensity — sits in the upper-central band; internally consistent.
- **Early press ₹750 cr** ⇒ ₹17.9 cr/plaza ⇒ 5.4× Atraila at every plaza for 2 full years — implausible; treat as unverified inflation.
- **ED 'receipts running in crores'** — consistent with any scenario; no number to check against.
- Per-plaza Aggressive (Model 1) ranges ₹0.5 cr (Nainsar, fee ₹20) to ₹4.1 cr (Pratappur/Madanpur, fee ₹145).

## Caveats

- Traffic per plaza is **not public** (Lok Sabha USQ 1017 hosts unreachable; NHAI publishes no plaza-level daily traffic). Fee-scaling proxies revenue share, not vehicle count — real distribution is wider.
- The 42-plaza list is the accused's confession (press note 36); ED's '~100 plazas' is larger and unenumerated. Harm scales roughly linearly: at ~₹1.2 cr/plaza central, 100 plazas ⇒ **~₹90-125 cr**.
- 'Harm' here = toll revenue diverted from remittance (state+lender side). Consumer harm is additive but separate: double-fee penalties wrongly levied, receipt-less cash demands, and systemic distrust — not priced here.
- Cited as section 13C of the published report.

## Reproduce

`python3 harm_model_2026-09-06.py` (same dir; needs DuckDB + Rajmargyatra DB + rmy-scam-crosswalk CSV).