Independent index administrator
Some risks are real, growing, and difficult to transfer for one reason: there is no number credible counterparties can agree to use. Pharometric designs that number and the governed release, versioning, and calculation infrastructure required to use it.
Our first instrument, the Rung Index, measures contraction in entry-level hiring opportunity. Backtested over 43 quarters of federal data; live publication remains in pre-contract validation.
“Any parameter or index that is used as the basis for a parametric solution must be objective, transparent, and readily available from an independent reporting agency.”— Swiss Re, on what a parametric trigger requires
That sentence is our job description. Where no such agency exists for a risk, the risk cannot be transferred at all — and that is the gap we are built to close.
01 — What we do
A parametric contract needs an index that is objective, independent of both parties, and computed by someone with no stake in the payout. When that index does not exist, the risk stays uninsurable. We build the index, we operate it under published governance, and we sit as the party that determines settlement.
01
We build a contract-ready index for a peril that has none — fusing public and licensed sources into a single fidelity-scored value, with the falsification tests registered before they are run.
Delivered as a methodology, a backtest, and a basis-risk analysis you can attack before anyone writes paper on it.
02
We compute the official number that settles the contract, and stand behind it. Independent of both parties, reproducible from a disclosed method, with a fixed vintage rule so a later data revision can never reopen a settled payout.
Signed releases, a permanent key ledger, and a published correction policy.
03
The live, full-resolution index, under commercial terms. Coarse public releases stay free and openly licensed; the calibrated instance is the product.
Machine-readable vintages on a disclosed schedule, with fidelity attached to every emitted cell.
02 — How it is licensed
Every serious index administrator gives away the thing that makes the standard spread and sells the thing adoption forces you to need. So do we, deliberately.
The schema, the field definitions and the reference client are openly licensed, because an index nobody can independently read cannot become the language a market transacts in. What is not open is the calibrated live instance and the settlement authority — the two things that cannot be forked, because they are made of accumulated track record, not code.
02A — Why this institution exists
Pharometric begins with contract-ready measurement. The longer path is to let qualified sponsors pre-arrange protection for places and communities whose conditions are deteriorating—or whose conditions remain dangerously under-measured.
A rising average is not sufficient. Every in-scope community must remain visible, persistent shortfalls must create an accountable response obligation, and Pharometric's compensation must remain independent of index direction, trigger, payout, or intervention outcome.
See the public mission architecture →03 — The Rung Index · our first instrument
Pharometric is a standard for expressing measurement under uncertainty, and a reference instance that publishes under it. Every value carries a fidelity score — an explicit statement of how far that number can be trusted, given how many independent sources constrain it, how recent they are, and how much they agree once their correlation is accounted for.
Fidelity is intended to be an underwriting control surface, not a badge: high-fidelity cells eligible, intermediate cells requiring a wider attachment corridor, low-fidelity cells excluded. Its empirical interpretation — what a given score means as interval coverage, and its reliability under backtest — is specified and calibrated ahead of the first live vintage, and will be published with it. It is not yet calibrated, and we do not describe it as if it were.
Where evidence is weak, the design requires an explicit low-fidelity state—never zero and never false precision. Completely source-less cells are not yet emitted by the current recomputation pipeline; closing that gap is a stated implementation requirement before universal-coverage claims are warranted. An underwriter can widen, buffer, or decline an emitted low-fidelity cell explicitly instead of absorbing basis risk nobody measured.
The first index measures entry-level hiring contraction — the rate at which early-career workers are hired into an occupation and place, against that cell's own pre-2020 baseline and hiring cycle. Whether AI causes that contraction is a question we test in the open and report against. It is not a premise we sell.
04 — The falsification record
The tests below were committed to a public, timestamped record before any result was computed, along with the exact thresholds each had to clear. Then we ran them on 43 quarters of federal administrative data. The index measured a large, youth-specific, cycle-robust contraction — and its own falsification layer rejected premature attribution of that contraction to AI at coarse occupation resolution.
| Test | What it had to show | Verdict |
|---|---|---|
| V1 | A sustained collapse appears in high-exposure occupations, with onset inside the registered window | ✕ Failed |
| V2 | The contraction is large enough to matter — ≥5pp mean cycle-adjusted deficit | ✓ Passed — 16.4pp |
| F1 | Low-exposure control occupations stay quiet — the signal is not just "everything fell" | ✕ Failed |
| F2 | Mid-career workers in the same cells do not show the effect — it is youth-specific | ✓ Passed |
All registered outcomes are published, including results that invalidate the original hypothesis. An index that reports only its confirmations cannot be used to settle anything. The peril that survives the evidence is contraction in entry-level hiring opportunity, cause-agnostic. AI attribution remains an open research question with a public scoreboard, and the scoreboard currently reads: not established at the resolution tested.
Read Research Note 001 — Gate 1: what we found, and what we disproved →
05 — Governance
A contract that settles on an index inherits whatever discipline the administrator does or does not have. So the discipline is stated first, in public, in advance — and it binds us whether or not it is convenient.
These are commitments, not aspirations. They are a floor we can raise and may not quietly lower.
06 — Provenance
The backbone is the U.S. Census Bureau's Quarterly Workforce Indicators — administrative hiring records, not a survey sample, which makes them hard to game and slow to arrive (a publication lag of two to three quarters). BLS labor-turnover series carry the hiring cycle. O*NET task data carries occupational exposure.
Sources are fused by generalized least squares against an explicit source-covariance matrix, because sources are not independent witnesses: when two of them republish the same underlying survey, naive averaging mistakes an echo for a confirmation. The covariance structure corrects for that. The estimator is open; the calibrated weights are the reference instance's own.
Not every input is fused. Some are observations of the latent quantity; others are covariates, controls, or external validators. The full measurement model — the latent variable, the observation equation for each source, which errors are treated as correlated, and how the covariance is estimated rather than asserted — is specified in the technical appendix, published with the first live vintage.
Cadence and the settlement information set. The administrative backbone is quarterly and lags roughly two to three quarters. A release is therefore not a monthly re-measurement of the backbone: it is a re-estimate against the information available at that release date, with faster series updating the estimate between administrative confirmations. Which vintages are provisional, which are final, what a contract may settle on, and the maximum settlement delay are stated in the release specification and fixed before the first live vintage — not left to a marketing cadence.
07 — Contact
Joshua DeFord
Methodology authority
Evaluate a measurement problem
outreach@pharometric.aiPharometric is operated by Clattershop LLC (Wisconsin, USA), which holds the methodology and signs releases.
Underwriters and MGAs evaluating trigger eligibility and historical behaviour. Brokers assessing whether a client's exposure can be represented by this index. Reinsurers and alternative capital pricing frequency, severity, correlation and administration risk. Actuaries and model validators conducting independent technical review before contractual adoption.
No index values are published yet. Until the first live vintage, the substantive conversation is basis risk — what loss this index proxies, and where it would fail to.