Does falling household debt mean families are financially healthier?

Reported US household debt fell $13 billion to $18.8 trillion in the second quarter of 2026, while credit-card balances rose $21 billion and auto balances increased $28 billion. These are end-June balances, published by the New York Fed on August 11. The report summary immediately complicates the apparent improvement: it attributes the $74 billion mortgage decline to a temporary credit-reporting gap during a servicing transfer.12

This decline does not establish that families reduced the debt they owe. A mortgage missing from the records does not demonstrate repayment. The headline total answers how much debt was recorded at a particular date; the reporting interruption limits what its change can tell us about families’ financial progress.2

Repayment statistics get closer to the health question, but what their percentages measure matters. At end-June, 4.7% of outstanding reported debt was in some stage of delinquency, or overdue repayment, a slight improvement. That is a percentage of dollars owed, not families. Larger balances receive more weight, so the figure does not tell us how widely repayment trouble is spread or which households improved.1

The Consumer Credit Panel’s serious-delinquency stock answers: how much reported debt is at least 90 days overdue? Its flow answers: how much debt newly crossed that threshold? The flow divides balances becoming at least 90 days late during the quarter by the previous quarter’s balances not already that late. Older seriously overdue debts therefore remain in the stock without counting as new transitions.3

A charge-off helps explain how those measures diverge. The Federal Reserve Board defines it as a loan removed from bank books and charged against loss reserves: the bank records a loss.53

New York Fed researchers found that new credit-card delinquency stabilized in early 2024, while accumulated charged-off balances drove nearly all the recent stock increase. About 40% of charged-off debts remained reported a year later during 2004–2012, versus 80% by 2024. Longer reporting keeps older overdue balances visible alongside newly troubled debt, allowing the stock to grow without new transitions accelerating. This historical example illustrates the difference between measures; the researchers did not establish why reporting lasted longer.3

Timing adds another distinction. Flow annualization uses a four-quarter moving sum, combining four quarterly transition shares. The resulting percentage describes transitions across four quarters, rather than one quarter alone.3

Bank statistics draw a different boundary. The Federal Reserve Board’s series covers insured US-chartered commercial banks. Its delinquency measure includes loans at least 30 days overdue that still accrue interest, plus loans in nonaccrual status, where interest is no longer being accrued. Those balances are divided by loans outstanding at period-end. This differs from the panel’s 90-day serious-delinquency threshold. The bank measure therefore describes repayment problems among the loans remaining on those banks’ books.53

For borrowers, a write-off generally does not eliminate the debt. The New York Fed’s FAQ explains that banks may continue reporting charged-off balances while seeking repayment. Bank delinquency rates exclude them, but the panel retains those still reported in both the overdue amount being counted and the total debt it is divided by. The same obligation can therefore remain visible in consumer credit records after leaving bank books.4

Stabilization also does not mean a low level of distress. The August release says new card and auto delinquencies remain elevated despite broad stability over two years. In Q2, the share of balances newly entering early delinquency rose slightly for autos and mortgages, while transitions into serious delinquency were mostly unchanged. Re-reporting defaulted student debt continued to distort student-loan measures, limiting their usefulness as a clean signal of changing repayment.12

Coverage adds further limits. The panel samples people with credit reports, and many buy-now-pay-later loans go unreported. Balances are nominal: their changes do not account for inflation. Card figures capture statement balances, including people who pay in full monthly; they cannot distinguish those payers from people carrying debt from month to month. The New York Fed release, summary, FAQ and research interpret related evidence from the same panel, rather than supplying independent confirmations of household health.413

For the next headline, check whose records it covers, what the percentage divides by, and whether it measures balances at a date or transitions over time. Then ask how missing reports and written-off debts are treated. Those checks identify the question the statistic can actually answer.2451

Report an error · Corrections policy