How the American Distress Index Works

The Five Domains

Every input series belongs to one of five domains, and every domain carries the same weight. The weight rule is structural, not fitted: domain_weight = 1/len(DOMAIN_IDS); member_weight = 1/len(domain members); zero hand-typed weight literals. Membership follows a written rule rather than editorial taste: A committed series joins the registry only if (a) it measures a household-distress rate or share — not a spread or derived difference, and not a nominal dollar or employment-count total whose full-history percentile tracks trend growth rather than household condition; (b) it has quarterly-resolvable history spanning 2005Q1 to the present; and (c) within its domain it is neither a population subset or superset of an existing member, nor a data-identical duplicate, nor level-redundant with one, nor distorted by a reporting-regime break. Every committed series under data/indicators/door*/ with a direction field and that span is either a registry member or carries a written disposition in registry_exclusions — enforced by the dispositions build gate, which fails closed when a new candidate appears undispositioned. Membership changes are versioned registry diffs.

The Math

Percentile normalization

Each series is converted to a percentile within its own full quarterly history, in one pass:

That is the Hazen percentile: rank every quarter of the series, average tied ranks, subtract half a rank, divide by the series length. The engine's own statement of the rule: Hazen percentile per series over its entire available quarterly history in one pass: percentile = (average_rank - 0.5) / n * 100, ties averaged. One yardstick: no regime split, no baseline window, no winsorization, no z-cap, no goalpost anchors.

Direction is read from each input's data file, never assumed: Every input's direction field is read from its JSON; lower_is_worse values are negated before ranking so higher always means more distress. A missing or unexpected direction aborts the build.

Aggregation

A domain's score at quarter T is the mean of its member percentiles present at T. The composite is the mean of the five domain scores. There are no fitted weights anywhere in the chain.

What the composite is, and is not

The composite is a mean of input percentiles. It is not itself a percentile of quarters. The two read differently on purpose: the current composite of 44.6 carries the gloss "On average, its inputs sit higher than in 45% of their own quarterly histories since 2005," while the composite's own rank in the published record is stated separately — The composite itself sits higher than 53% of all published quarters since 2005. The two numbers answer different questions, and the index publishes both so neither gets quoted as the other.

The Five Bands

The 0–100 scale splits into five equal bands. The thresholds are uniform — not calibrated to any narrative period — and the labels are an editorial decision, locked with severity ascending by band number. The engine's usage rule: labels apply to the national time axis only and always publish alongside the literal reading; state and county scores use ranks and quintile wording, never these labels.

The Record Since 2005

The published series holds the financial crisis at its peak — 85.3 in 2009-Q3 — and the pandemic stimulus era at its low, 15.7 in 2021-Q4, when support programs left household buffers unusually full. Nothing in the method points at those quarters; they emerge from ranking each series against its own record.

Data Rules

Missing data

• A series-quarter exists only if at least one raw observation is dated within it; the quarterly value is the mean of the observations present. No interpolation, no carry-forward, no imputation.
• Domain score at quarter T = mean of member percentiles over members present at T, minimum one member; members_present is published per domain per quarter so dropout is visible.
• The composite publishes only when all five domains have at least one present member. A lagging member drops out of its domain's trailing quarter rather than stalling the index, and the score restates when it lands.
• A discontinued series stays in history; adding or removing a member is an explicit versioned registry diff, never a runtime fallback.
• Known-stale or broken sources are excluded up front: hud_fha_performance (stale, gappy, starts 2017), data/distress NY Fed snapshots (nyfed_household_debt.json carries no values — all 92 observations null; nyfed_delinquency.json shape inverted vs the published NY Fed transition series), short-history behavioral series (no GFC span).

Publication gate

The engine refuses to publish a quarter rather than publish a partial one. The rule: publish a quarter only when all five domains have at least one member present. Every run also re-proves a set of validation gates — weights, label vocabulary, input dispositions, series-seam integrity, orientation, reproducibility — and writes nothing if any gate fails.

Revisions and vintages

Inputs are revised series (Federal Reserve Board bank-condition rates via FRED, NY Fed Consumer Credit Panel, BLS, DOL, and BEA series via FRED). Historical ADI values are computed on today's revised vintages, not on the data as first published. Every run restates the full history: the percentile yardstick grows by one quarter per refresh, trailing partial quarters fill in as observations land, and upstream revisions re-rank past quarters. No out-of-sample claim is made for any historical reading, including the 2008-2010 stretch.

Each refresh re-ranks every series' full history, so published historical scores restate by small amounts as the yardstick grows, as trailing partial quarters complete, and as upstream sources revise. This is the same revision model as CDI re-ranking counties each refresh.

Limitations

The ADI is a national time-series composite. It does not score individual states or counties — the State Distress Index and County Distress Index rank places against other places using the same five-domain taxonomy. It does not measure household wealth, consumption inequality, or the income distribution; the inputs are distress signals, not welfare metrics. A high reading does not say which households are in distress, only that the population-level signal sits high in its own record.

The percentile method reads each series against its own published history. Quarters early in a series' record have fewer comparison points, and every refresh restates history by small amounts as the yardstick grows. The index treats that restatement as a feature — the alternative, freezing a baseline window, would make the scale depend on a chosen era.

The ADI is not a forecast of recessions, default rates, or any future outcome. It measures current conditions against the historical record and makes no claim about what comes next.

Replication

The full scoring pipeline is open source. The production engine is at scripts/indexes/compute_adi.py, the shared percentile transform at scripts/indexes/family_normalization.py, and the independent validator at scripts/indexes/validate_adi.py. All are Python with no dependencies beyond the standard library and numpy.

Source data

10 series, all freely available federal and Federal Reserve data:

FRED attribution

Per FRED's terms of service, source attribution is required for any FRED-derived series: "[Source Agency], [Full Series Title] [SERIES_ID], retrieved from FRED, Federal Reserve Bank of St. Louis." Users replicating the index should include the appropriate FRED citation per series. FRED data may not be used for AI/ML training, per the same terms.

License

The ADI score, methodology, and underlying scoring code are released under the Creative Commons Attribution 4.0 license. Reuse is free, commercial use is permitted, and the only obligation is attribution to American Default Research with a link to https://americandefault.org/methodology/adi/.

How to Cite

The American Distress Index is an open data product released under Creative Commons Attribution 4.0. If you use it in research, policy analysis, or journalism, please cite American Default Research as the source.

APA

MLA

Chicago (author-date)

BibTeX

@techreport{adi_methodology_2026,
  title       = {{American Distress Index: Methodology and Scoring}},
  author      = {{American Default Research}},
  year        = {2026},
  institution = {{American Default Research}},
  url         = {https://americandefault.org/methodology/adi/},
  note        = {Composite of 10 federal series across 5
                  equal-weighted domains, each ranked against its own
                  quarterly history; published 2005-Q1 through 2025-Q4},
}

News-copy short form

In-line attribution

Change Log

Substantive methodology changes are recorded here with the date applied. Routine quarterly data refreshes (which update scores but not methodology) are not change-log events.

• 2026-06-11 Percentile-of-own-history methodology adopted: Hazen normalization, five equal-weighted domains, refuse-to-publish gates, and the Minimal/Low/Typical/High/Severe band vocabulary. The full published history is computed under this method.
• 2026-04-29 Methodology page /methodology/adi/ first published as the canonical reference for the ADI scoring methodology.