The Cohort Earnings Divergence — Q2 2026

The Second Quarter Alone

The under-$100M tier holds every credit union with total assets under $100 million at quarter-end: 2,439 at June 30, 2026, 57.9% of the credit unions that filed and 3.0% of system assets. For the first half of 2026 its annualized ROAA was 43‱, against 76‱ for the first half of 2025, the comparison System Signals reported for the quarter. A half-year figure blends two quarters. Taken apart, as the half-year total less the first quarter's, the tier's second quarter alone returned 73‱, 60‱ more than its first quarter and 12‱ less than in Q2 2025 alone.

Each larger tier earned more in Q2 2026 alone than in Q2 2025 alone, from the $100M–$500M tier's 86‱, up 3‱, to 113‱ for the over-$10B tier, up 29‱. In Q1 2026 alone the smallest tier trailed the $100M–$500M tier by 62‱; in the second quarter, by 13‱. The second quarter narrowed the gap without closing it.

One Quarter Out of Line: the table gives each tier's annualized ROAA for four single quarters. The smallest tier's Q1 2026 figure stands apart from every other in the table, and in Q2 2026 it is the only tier below its own figure of a year earlier.

Below the Margin Line

ROAA is the net interest margin less what the lines below it take: operating expense and provisions, net of non-interest income. The report does not separate those lines by tier and measures them together as margin less return. In Q1 2026 alone the under-$100M tier's net interest margin was 373‱ of average assets, 8‱ wider than in Q1 2025, and its net charge-off rate was 44‱ of average loans, 3‱ lower. Margin less return was 360‱, up 62‱ from a year earlier, more than the whole 54‱ fall in ROAA. By that arithmetic the first quarter's shortfall sat below the margin and outside charge-offs.

Most of that shortfall was one credit union's. Of the 2,466 credit unions in the tier at March 31, 2026, one reported a first-quarter loss of $92M, while the tier as a whole earned $24M. Without it, the tier's Q1 2026 ROAA would have been 63‱ rather than 13‱, against 67‱ a year earlier: that one loss accounts for 50‱ of the 54‱ fall. For the first half the tier's ROAA was 43‱ with it and 71‱ without it, against 76‱ for the first half of 2025. Of the 33‱ gap that System Signals reported, 5‱ remains without that credit union.

Most of that cost did not recur. In Q2 2026 alone margin less return fell back to 311‱, still 18‱ above Q2 2025, while the margin widened 6‱ to 385‱, the widest of the five tiers. Those two moves account for ROAA's shortfall against Q2 2025. The credit union that carried the first quarter lost another $11M in the second; without it, the tier's Q2 2026 return would have been 79‱, still below the 85‱ of Q2 2025. Credit cost rose as well: the net charge-off rate was 50‱ of loans, 7‱ above a year earlier. Loans grew 0.67% in the quarter, against 0.82% in Q2 2025, after contracting 1.10% in the first quarter.

Each larger tier did better than the smallest on all four lines. Each widened its margin against Q2 2025 by more than the smallest tier's 6‱, and in each margin less return rose less or fell: up 10‱ in the $100M–$500M tier and down 23‱ in the over-$10B tier. Among the five, the smallest tier's net charge-off rate rose the most, and its loans grew the least: the next slowest, the over-$10B tier, grew loans 1.91%.

The Smallest Tier, Line by Line: the table follows the under-$100M tier through six single quarters. Q1 2026 keeps a margin and a charge-off rate in line with the quarters around it; of the lines that make up the return, only margin less return breaks from them.

The Scale Account

The Cohort Earnings Divergence — Q1 2026 (June 27, 2026) read the tier's first-quarter return as a lack of scale: costs that change little with size, spread over a small balance sheet, consume a wide margin, and shrinking loan books move these credit unions further from scale. If scale set the 2026 gap, the gap would show in the years before it.

Earlier first halves show the gap in some years and not in others. In each first half from 2019 through 2022, the under-$100M tier earned the least of the five tiers, 7‱ to 24‱ less than the $100M–$500M tier. From 2023 through 2025 the two were within 2‱ of each other, and the smallest tier earned more than at least one larger tier each year. In the first half of 2026 the gap reopened to 37‱, most of it from the first quarter.

Scale Without the Gap: the chart gives each tier's annualized ROAA for the first half of each year. The gap that the Q1 2026 edition read as scale is absent from 2023 through 2025, and we read the scale account as not holding for 2026: a 62‱ gap that opened in one quarter and mostly closed in the next does not fit costs that change little with size.

We read scale in the tier's costs rather than in its 2026 return. Its margin less return in Q2 2026 alone, 311‱, was the widest of the five tiers, against 235‱ for the $1B–$10B tier. The one credit union's second-quarter loss does not account for that difference: it is 6‱ of the tier's average assets. We read the tier's matching return from 2023 through 2025 as a wide margin covering that cost. We do not read the loan book as what separated the quarters either: the tier's loans also contracted in Q1 2025, by 0.54%, in a quarter that returned 67‱.

What Settles It

System Signals — Q1 2026 (June 27, 2026) set a watch on whether the smallest tier's loan contraction was a first-quarter effect or a trend, and the record of calls carries it as "The smallest credit unions' earnings would turn," undecided after Q2 2026. The item named no threshold and no comparator, so it was a watch item rather than a test. Against Q1 2026 alone, the tier's ROAA rose 60‱, and the call is met; against Q2 2025 alone it fell 12‱, and on that year-over-year comparison, the one the size divide thread uses, the call is not met. The year-over-year reading does not rest on the one credit union: without it, the second quarter alone fell 6‱ short of Q2 2025. We read the first comparison as a recovery from one quarter, not a turn in the tier's earnings, and so read the call as not met.

Two Readings of One Call: the table sets the under-$100M tier's Q2 2026 return beside each comparator the call could have named. It is above the first and below the second.

Test H-2026Q2-breakdowns-cohort-01. Whether the under-$100M tier's ROAA for Q3 2026 alone, annualized over average assets, exceeds its Q3 2025 figure of 97‱. Settles with the Q3 2026 data; made in October 2026, after the quarter ended and before its filings are published. The population is credit unions with total assets under $100 million at September 30, 2026, and its membership floats with each quarter-end; the result is scored in whole ‱. A reading above would be consistent with the tier's earnings back at last year's pace once the first quarter is past; a reading at or below would be consistent with a year-over-year shortfall that outlasts it. We take no side: the second quarter alone fell short of a year earlier by less than the first did, and a quarter derived from two year-to-date totals moves when a credit union enters or leaves the tier.

Connecting the Lines

This edition continues the size divide thread from The Size Divide — Q3 2025 (December 20, 2025) and The Cohort Earnings Divergence — Q1 2026 (June 27, 2026). It takes apart the half-year figure in Signal 3 of System Signals — Q2 2026 (October 5, 2026), which found the smallest tier the only one earning less than a year earlier and left its first quarter unexplained. That finding holds for the second quarter alone. We now place the first quarter's shortfall below the margin, outside charge-offs, within one quarter and mostly in one credit union, and we withdraw the Q1 2026 edition's reading of the 2026 gap as scale. The thread's open question is whether the year-over-year shortfall outlasts the first quarter, and the Q3 2026 test above is its next step.

QuarterCall publishes analysis of publicly available NCUA Call Report data. Nothing here is investment, financial, or membership advice.