# Presidential Economic Pressure Index (PEPI)
## Methodology Overview

*Aka TACO Index*

**A quantitative behavioural index of the gap between presidential pressure, follow-through, and economic results**

— By Kurt Sather, API Automations
Methodology version 1.2.17 | Classification: Public

This overview describes what the index measures, how events move through it,
what kinds of evidence it uses, and how it is validated. It is written for
traders, researchers and journalists who need to know what a PEPI score means
and how far to trust it. The full methodology specification — dimension
weights, formulas, thresholds, source configuration and the field-level data
dictionary — is maintained as an internal document and is available to API
customers and research partners under agreement (`pepi@apiautomations.com`).

---

## 1. What the index measures

PEPI tracks the lifecycle of presidential threats, ultimatums and policy
announcements and measures the distance between the initial rhetoric and what
actually happened. Every event is detected, monitored to a resolution, and
scored from **0 (carried out in full) to 100 (fully reversed or abandoned)**.
The daily index is a recency-weighted blend of recently resolved events, with
live pressure from overdue open threats, so it moves when follow-through
changes — not when posting volume does.

Three audiences use it: traders exploiting the overreaction-to-threat /
relief-on-reversal pattern; policy researchers studying executive signalling;
and journalists who need a defensible number behind "chickened out".

## 2. What counts as an event

A statement or formal action by the President qualifies only if it is
**specific** (a concrete action, target or consequence), **actionable**
(within presidential authority or influence), **measurable** (independently
verifiable as done or not done) and **material** (economic, diplomatic or
policy consequences worth tracking). General opinion, vague rhetoric and
restatements of an existing threat do not mint new events — a restatement is
evidence *about* the original threat and is attached to it.

Events fall into four domains: **Trade & Tariffs**, **Foreign Policy**,
**Domestic Policy** and **Rhetorical**. Each domain has its own sub-index.

## 3. Lifecycle

| Stage | What happens |
|---|---|
| Threat | Detected from a presidential statement or a Federal Register document; target, consequence and any stated deadline are captured |
| Escalation | Subsequent restatements, intensifications or de-escalations are attached to the event, not counted as new threats |
| Deadline | The stated deadline — or a bounded implicit one — passes; overdue threats accrue live pressure in the daily index |
| Action | Implementing documents, court rulings, legislative actions or independent reporting are linked as evidence |
| Outcome | The event resolves to one of a fixed outcome vocabulary and is scored; unresolved events are closed after a bounded monitoring window |

Outcome vocabulary (an API field): `full_implementation`,
`partial_implementation`, `delayed`, `partial_reversal`, `reframed`,
`full_reversal`, `abandoned`, `indeterminate`, `pending`.

## 4. Scoring

Each resolved event is scored on **six behavioural dimensions**, blended into
the composite with fixed weights:

- **Reversal magnitude** — how much of the threatened action was actually
  delivered. For trade events with both an announced rate and an official
  applied rate, this is measured quantitatively from the gap; otherwise it
  uses a categorical scale keyed on the outcome type.
- **Timeline slippage** — how far past the effective deadline the action
  landed, on a diminishing scale.
- **Rhetorical dilution** — how much the stated severity softened between the
  original threat and its later restatements.
- **Scope reduction** — how many of the original targets were dropped or
  exempted.
- **Framing shift** — whether a retreat was presented as a win.
- **Repetition pattern** — whether the threat re-makes an earlier threat that
  was itself reversed (the serial re-threat lineage).

Two rules matter for reading a score. A dimension that could not be measured
for an event **contributes nothing** rather than a guessed value, so a score is
never moved by an assumption. And an outcome supported only by the President's
own claim is **discounted** relative to one corroborated by a court, Congress,
the Federal Register or independent reporting; every event carries an
`outcome_evidence_source` label so you can see which.

## 5. Evidence

PEPI uses public records wherever a public record exists:

| Category | Used for |
|---|---|
| Presidential statements | Threat detection and follow-up attachment |
| Federal Register presidential documents and agency notices | Threat detection (formal actions) and implementation evidence |
| Federal court opinions — the trade courts, the Supreme Court and the courts hearing executive-order litigation | Outcome resolution and the judicial reaction sub-index |
| Legislative records (Congress.gov) | Outcome resolution for bill-linked threats and the congressional reaction sub-index |
| Official trade statistics (USITC) | Applied tariff rates for the statutory-vs-applied gap and sector impacts |
| Official price and employment statistics (BLS) and macro-financial series (FRED) | The economic-impact layer and market context |
| Equity market data | Broad-market, volatility and sector reaction around threat and outcome dates |
| Code of Federal Regulations snapshots (eCFR) | Verification that a claimed regulatory outcome actually codified |
| Independent-reporting reconciliation | Outcome resolution for non-trade events the record-based resolvers cannot reach |
| Independent nonprofit harm records | Corroboration of domestic-domain consequences |

Classification uses large-language-model extraction with a confidence floor
and human review; an operator queue settles low-signal detections and any
verification contradictions. Statistical inputs are published at **latest
vintage**: when an agency revises a series, the affected economic measures
update on the next cycle and the revision is ledgered.

## 6. Economic impact

For trade events the index pairs each threat with an economic-impact stack
built on the retail price pass-through methodology of Cavallo, Llamas &
Vazquez (2026): the announced-versus-applied tariff gap, retail pass-through,
CPI contribution, consumer-versus-firm incidence and sector-level effects.
Aggregate applied rates are import-value weighted. Forward-dated and
non-trade events legitimately lack some of these measures.

## 7. Validation

- **Calibration benchmarks.** Seventeen hand-curated, externally documented
  events from 2018–2026 (first-term Section 232/301 actions, the 2025
  reciprocal and China tariffs, the birthright-citizenship executive order
  resolved by the Supreme Court in 2026) are frozen and republished with the
  live index. Their scores are visible on the site and the API.
- **External benchmark.** The foreign-policy domain is checked against the
  ECFR force-threat database (2 of 22 threats carried out, 9.1%).
- **Court verification.** Outcomes confirmed by rulings carry the `court`
  evidence source; the court paper trail is published as public records.
- **Inter-rater reliability** target for classification: Cohen's kappa above
  0.80, with human review of the queue.
- **Revision ledger and status page.** Upstream revisions, source freshness and
  post-deploy endpoint health are published at `/status/`.

## 8. Limitations

The congressional reaction sub-index reads near zero because few tracked
threats cite a bill; the rhetorical domain currently rests on calibration
benchmarks only; dilution and framing coverage is partial for events that
resolved before follow-up attachment existed; and the judicial sub-index
measures court *activity*, not influence. These are documented on the
validation page and revisited each release.

## 9. Access

| Surface | Tier |
|---|---|
| Headline index, stats, branch sub-indices, source status, teaser snapshot | Free |
| Event lists and detail, index history, domain sub-indices, tariff rates, indicators, sector reactions, courts | Standard |
| Per-event economic impact and sector reaction | Premium |

The OpenAPI contract is versioned per methodology release with a 30-day
breaking-change window. Pricing and endpoint documentation: `/api/` and
`/pricing/`.

---

*Methodology overview v1.2.17 | © API Automations. Written by Kurt Sather, 2026.
The full methodology whitepaper, its changelog and the data dictionary are
available to customers and research partners on request.*