Hall and Reis Ten Years On: The Federal Reserve's Losses at the Midpoint

In 2015 Hall and Reis asked whether paying interest on large reserve balances could make the Federal Reserve financially unstable, and concluded that the risks were real in theory but remote in practice. The episode that began in September 2022 is the first full test of their framework. Their conclusion on solvency holds: the deferred asset that records the Federal Reserve's losses peaked at $245.9 billion in January 2026, about 0.8 percent of GDP and well inside the bounds they assumed, and has been falling since. Three features of the episode differ from their scenarios. The losses were larger than both of their adverse cases, $141 and $209 billion. They came through interest on reserves exceeding portfolio income under hold-to-maturity accounting, the variant they modelled in passing, rather than through realized capital losses at exit. And because the Federal Reserve is twelve Reserve Banks rather than one, remittances to the Treasury resumed in part while the aggregate deferred asset was still near its peak. A quarterly projection calibrated to the Federal Reserve's June 2026 statements implies that system net income will cover the deferred asset four to five years from late 2026, close to Hall and Reis's baseline payback, and that this result is insensitive to the path of policy rates. The aggregate deferred asset itself clears later, in six to nine years, because part of system income is remitted before it is extinguished. The framework treats interest on reserves as neutral for solvency and silent on who receives it; a companion paper shows that the recipients are where the economics lies.