How the American Distress Index works

The five parts of the American Distress Index

Each part has the same weight. Within a part, the available inputs have equal weight too. So an input’s share of the score depends on how many inputs share its part. With every input present, each late-payment input counts for 5%, each lender-loss or jobs input for 10% and each debt-payment or saving input for 20%. The index draws on 10 public data series, or sets of figures recorded over time, listed below.

Equal weights keep the calculation easy to follow: no part gets extra weight because it happened to match a past crisis. Choosing which inputs belong in each part still requires judgment.

Student loan delinquency isn’t an input. The federal payment pause held reported delinquency near zero, and the return of credit bureau reporting brought it back as a sudden jump, so that series shows reporting policy as much as household behavior.

How the American Distress Index score is calculated

The calculation puts rates and counts on the same scale before combining them. Each input is compared with its own available history, which can begin earlier or later than the published index.

1. Group the data by quarter. We average the reported values dated within each quarter. Missing values are not filled in or copied from an earlier quarter. National unemployment has no October 2025 figure, so the fourth quarter of 2025 averages November and December.
2. Make higher mean more strain. For most inputs, a higher rate or count means more strain. Saving works the other way: a lower saving rate means more strain, so its direction is reversed before ranking.
3. Compare each input with its history. We rank its quarterly values from least to most strain. A percentile is a rank on a percentage scale: a higher percentile means more recorded quarters had less strain. Equal values share an average rank.
4. Average the inputs within each part. Only inputs with data for that quarter are included.
5. Average the parts. We publish a score only when every part has data from at least one input.

The index score averages those historical comparisons. It is not the percentage of households in trouble. It also differs from the percentage of past quarters with a lower index score; that comparison is published separately.

For the second quarter of 2026, the score is 47.0. On average, its inputs sit higher than in 47% of their own quarterly histories.

How the American Distress Index score ranges work

The 0–100 scale is divided into five equal ranges, called bands. Their labels describe the national score’s position relative to the inputs’ histories. They are not labels for states or counties.

A score exactly at a dividing point goes into the higher band. The latest score is in the Typical band. The composite itself sits higher than 44% of all published quarters since 2005.

The American Distress Index through history

The published record runs from the first quarter of 2005 through the second quarter of 2026, covering 86 quarters. The highest score in that record is 90.0 in the third quarter of 2009. The lowest is 22.7 in the fourth quarter of 2021.

Not every input reaches back that far. The credit card late-payment and charge-off rates are ranked only from the first quarter of 2010, when an accounting change brought loans that banks had securitized back onto their balance sheets. Rates before and after that change cover different loans. So scores before then have no credit card data: late payments come from three inputs and lenders’ losses from mortgage charge-offs alone. That includes the peak in the third quarter of 2009. The debt-payment input is ranked from the first quarter of 2005, when the Federal Reserve Board’s current method begins, and that sets the start of the published record.

We check how the method scored during the 2008 financial crisis by applying it to past data. This is called a backtest: a test using historical data. These scores use the data as revised today, rather than the figures available at the time, and each quarter is ranked against the whole record, including later quarters. The history shows how the method describes past conditions; it does not prove that it could have predicted them.

The table shows the last published quarter of each year. The full quarterly record is available with the downloads linked at the end.

Sources and updates for the American Distress Index

Most inputs come from public agencies. One, the share of auto loan balances that are seriously late, comes from the Federal Reserve Bank of New York’s Consumer Credit Panel, which is built from Equifax credit records. Federal Reserve Economic Data, or FRED, is the Federal Reserve Bank of St. Louis service that distributes many of these agencies’ figures. The table names the original publisher and the latest quarter available for each input. In the source names, delinquency means late payments and debt service means required loan payments.

The published score covers the second quarter of 2026. A later quarter in this table means newer data has arrived for that input. That quarter gets its own score only in an approved release, once all five parts have a reading. Until then, the newer reading may already count in the history that input is ranked against. A quarter still under way is only partly reported.

Some agencies publish weekly or monthly; others publish quarterly. We recalculate a candidate when inputs refresh, then review it before an approved release changes the public scores. A late input can be left out of its part while the other inputs are available. When the missing data arrives, the candidate is recalculated.

Past scores can change in a later approved release because agencies revise their figures, more data arrives within a quarter and new quarters extend the history used for ranking. A discontinued input keeps its place in past scores. Changing which inputs the index uses requires a documented change to the method.

The calculation checks equal weights, approved labels, input selection, changes in reporting methods and whether more strain moves an input’s score upward. It also checks that repeating the calculation produces the same output. Those checks confirm that the method was followed; they do not establish that the score predicts future events.

FRED data use is subject to the Federal Reserve Bank of St. Louis terms of use. Anyone rebuilding the index should cite each FRED series by its agency, title and series ID. Those terms do not allow FRED data to be used to train AI models.

This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.

The current method has been in use since Sep 23, 2026, and the whole published record is calculated with it.

We write with assistance from AI tools. Figures come from the data and sources named on this page.

What the American Distress Index can't tell you

It is one national number. It doesn't say which households are struggling, and it doesn't score states or counties; the state and county scores do that. It doesn't measure wealth or how income is shared out. Its inputs are signs of strain, not a full picture of how people are doing.

Each input is compared with its whole record, so a past score reflects what we know today, and every new quarter can shift older scores a little. Those records differ in length, from 1948 for the longest to 2010 for credit cards. That is a choice: a fixed comparison period would make the scale depend on which years we picked.

Other financial strain data from American Default

The economic indicator catalog covers more data than the inputs used in this index. Each indicator page explains what is counted, who publishes it and how often it updates.

Separate research into early warning patterns tests whether a change in one indicator tends to come before a change in another. Those historical relationships are not part of the national index calculation.

The State Distress Index method and the County Distress Index method explain how state and county scores compare places at the same time. The national score compares conditions with their own past and is not an average of county or state scores.

How to cite the American Distress Index

Name American Default Research as the publisher, give the quarter and score used, and link to this page. Include the date you accessed the data, because past scores can be revised. The press citation tools provide more formats.

American Default Research. (2026). American Distress Index [Data set]. https://americandefault.org/methodology/

For developers

Data downloads and API provides machine-readable files, including JSON for software and CSV for spreadsheets, along with access and citation guidance.

The scoring code is public: the engine that computes the index, the shared percentile step and the separate checker. They are Python and need only the standard library and numpy.

The score, this method and the scoring code are free to reuse under the Creative Commons Attribution 4.0 license, including for commercial work. Credit American Default Research and link to this page. The license covers our scores, not the agencies' own data.