Portfolio and Concentration Model

Assumptions

Everything the model believes, stated before the math. If you do not buy one of these, do not buy the output.

1. Concentration is measured, not estimated: indices are computed from the record, never modeled.
2. Concentration dimensions are fixed (spec Section 58): sponsor, utility, node, geography, contractor, equipment, technology, offtaker, transmission, water.
3. Common-mode exposure is the share of book exposure sharing a correlation group from the Dependency Model.
4. The capital-problem ranking is computed only where deal terms exist; elsewhere it is UNKNOWN, not approximated.

Derivation

Herfindahl-Hirschman indices per dimension over the customer's covered book, with s_i the share of book exposure in bucket i:

The five projects most likely to create a capital problem are ranked by expected capital impact:

with the probability from the Delivery Model and the severity from the Financial Bridge. Where deal terms are missing, severity does not exist and the project does not receive an invented ranking - the gap is reported.

Identification strategy

Why the estimate supports the claimed meaning - and where the model refuses to claim more than the data carries.

• HHI is exact arithmetic on the book as recorded; there is no estimator to go wrong. The identification question is whether the book's bucket assignments are real - which is why the inputs are record fields, not tags.
• Common-mode exposure is identified through the Dependency Model's correlation groups (shared utility, supplier, node): projects that share a group are exposed to the same failure, and the measure counts that overlap directly.

Inputs and outputs

Calibration and gate evidence

Validation design

• Concentration arithmetic is exact; the ranking inherits the Delivery and Financial Bridge backtests.

Evidence on record

• No separate calibration applies to the exact arithmetic; inherited backtests are tracked on the component models' pages.

Failure modes

The ways this model declines, degrades or goes UNKNOWN - and what it prints when it does.

• Missing deal terms on a project: it is excluded from the capital-problem ranking with the reason reported.
• Unbucketed exposure (a project with no recorded utility, say) lands in an UNKNOWN bucket rather than being dropped, so indices do not flatter the book.

Sensitivity

• The ranking is most sensitive to the Delivery probability: a project moving from 30% to 60% material-delay probability doubles its expected impact at fixed exposure and severity.