The Index Says Typical. Its Inputs Do Not Agree.

The American Distress Index closed Q1 2026 at 43.8. That put the quarter in the band the index calls Typical. The gloss the index publishes beside that label reads: On average, its inputs sit higher than in 44% of their own quarterly histories.

10 federal series feed that number. Sort them by where each one sits against its own recorded history and the middle one lands at the 31st percentile.

The composite and its typical input are not describing the same quarter.

What the band covers

The band the composite fell into is 20 points wide. Exactly 1 of the 10 inputs sits inside it. 8 of the 10 sit below the midpoint of their own histories.

On its own that proves nothing. An index built as a mean of percentiles is under no obligation to resemble its parts, and across every quarter the ADI has published the median count of inputs landing inside the composite’s own band is 2. A summary that matches 2 of 10 is the normal case here, not an anomaly.

What is less ordinary is how far apart the parts are.

Where the noise is

Auto-loan serious delinquency, the Repo Line, entered the quarter at the 99th percentile of its own record, according to the Federal Reserve Bank of New York’s Household Debt and Credit Report.

The personal saving rate, the Buffer, entered at the 92nd on the index’s distress-oriented scale, which flips a series where lower is worse. That is U.S. Bureau of Economic Analysis data retrieved via FRED.

The other 8 ran from the 7th to the 49th.

That is 2 series near the ceiling of their own records and 8 in the lower and middle registers of theirs. The average of that is a middle number, and the middle is the one place none of the evidence is.

How far apart is far

Take the standard deviation of the input percentiles and the quarter reads 29.54. Across the whole published record the median is 22.81. The most recent earlier quarter whose inputs were at least this scattered was Q2 2021.

Among the 63 quarters the index has placed in this band, only 2 scattered wider.

The domain view says the same thing in fewer numbers. Safety Net & Buffer scored 92.04. Labor scored 16.28. That is a gap of 75.76 points between the highest and lowest of the five, the widest since Q4 2022, and the domains have now sat further apart than their own record median for 24 consecutive quarters, since Q2 2020.

A run that long means “the domains disagree” has stopped being an event and started being a description.

It is not the weighting

The obvious suspect is the construction. 2 of the 5 domains carry a single series each, and every domain carries the same weight. So the personal saving rate alone accounts for 20% of the composite, while auto-loan delinquency, the highest-reading input in the quarter, accounts for 5% as one of 4 members inside Delinquency.

That looks like the whole story until you test it.

Throw out the domain structure and average every input equally. The result is 40.8, against the published 43.8. Same band. The alternative that was supposed to expose the distortion lands in the same place.

The weighting is not what makes this quarter read the way it reads. The disagreement is.

Which way the disagreement runs

Rank each domain against its own history and the split has a shape. Labor sits at the 14th percentile of its own record. Safety Net & Buffer sits at the 85th.

Labor here is the unemployment rate and initial jobless claims. Both count people arriving at the door. Neither says anything about how long the room behind it holds them.

Step outside the index and that distinction has some support. The Long Haul, the share of unemployed workers who have been out of work twenty-seven weeks or longer , sits at the 88th percentile of its monthly record back to 1948. The unemployment rate, on the same transform over the same span, sits at the 19th. Both are U.S. Bureau of Labor Statistics series retrieved via FRED.

Read that as few people falling in and a hard climb out, and it is a mechanism hypothesis, not a finding. Two things have to be said against it.

The first is that the Long Haul is registered in our own catalog as a bidirectional series. A high reading is not automatically a distress reading, because the share rises when short-duration unemployment drains away as well as when long-duration unemployment builds. The percentile above is a rank, not a verdict.

The second is that the other duration-adjacent series we publish decline to cooperate. The Short Shift, involuntary part-time work as a share of employment, sits at the 21st percentile once oriented toward distress. The Prime Years, the employment rate for workers aged twenty-five to fifty-four , sits at the 11th on the same orientation. Both belong in the calm column.

One series pointing one way is a question worth carrying into the next release. It is not a second labor market.

What the spread does not do

Here is where the piece would like to turn, and where the record refuses.

If a wide spread were an early warning, quarters that opened wide would be followed by quarters that moved. Take every published quarter through the anchor, pair its domain spread with the change in the composite 4 quarters later, and the correlation across 137 windows is r = 0.25.

Weak, positive, and built on overlapping windows, which manufactures the appearance of structure where none has to exist. Grade it at association and leave it there.

The decile test says less than that. The widest-spread tenth of those windows saw the composite fall an average of 0.07 points over the following year. Every window taken together saw it fall 0.57. The wide quarters and the ordinary quarters are not separable by what came next.

So the spread is a fact about the quarter it describes. It says the index was summarizing a disagreement rather than a consensus. It does not say which side of the disagreement wins.

The honest position

Two things are true about Q1 2026 and both belong in the same sentence.

The composite is a fair summary of its inputs. It was built to be one, the method is fixed in advance, and no series was reweighted to produce that reading.

The composite is also a poor description of any of them. Of the 10 series that went in, 1 came out in the band the label names.

The number to quote is not the composite alone. It is the composite and the range behind it, and anyone quoting the first without the second is repeating the part of the sentence that carries the least information.

One vintage note, because it changes what this piece can claim. The index’s own record for auto-loan delinquency stops at the anchored quarter. The Repo Line itself has since printed Q2 2026, a quarter the composite has not yet published. The scatter described here is the scatter the index recorded, not the scatter the later observation would produce.

We will know more when the next quarter lands. We do not know it now.