Is the recession alarm broken?

The gap between 10-year and 3-month US Treasury yields inverted in late 2022 to depths not seen in four decades, pushing the New York Federal Reserve's recession probability model towards 38 per cent. The yield curve is the bond market's most trusted forecaster, and it has been flashing red. Equity investors and executives now wonder whether a decade of central bank balance-sheet intervention has broken the alarm. Trillions of dollars of quantitative easing have bent the term structure of interest rates, and some doubt that an inverted curve still tells anyone much about recession.

The New York Fed's model is deliberately simple. It takes the spread between the two yields and turns it into a probability of recession twelve months ahead, estimated on decades of past data. Its reputation rests on that record: an inversion of this spread has come before every American recession since the late 1960s, with only rare false alarms.

A bent curve

Sceptics argue that the deep inversion of the 10-year against the 3-month rate reflects technical quirks more than insight into the economy. The Fed's huge holdings of long-dated bonds hold down term premia and push 10-year yields lower even when the economy is doing fine. Its balance sheet peaked at nearly $9 trillion this year, and although it has begun shrinking, the run-off is slow relative to the stock it accumulated.

Short-term rates, meanwhile, are temporarily high because the Fed has tightened so fast. On this view the inversion is a mechanical gap between overnight policy rates and long-run equilibrium rates, and says little about any coming collapse in demand.

Structural demand matters too. Corporate pension schemes buying bonds to match their liabilities create a steady, price-insensitive bid for long-dated Treasuries, which pushes the long end down whatever the state of the cycle. Foreign central banks and insurers add their own buying for reasons unrelated to the American business cycle, and the New York Fed's own estimates of the term premium have spent long stretches of the past decade below zero.

Heard this in 2006

The argument that the curve has been distorted is not new. In 2006, as the curve inverted, Ben Bernanke suggested in a speech early in his term as Fed chairman that unusually low term premia, partly the product of heavy foreign buying of Treasuries, might mean the inversion signalled less than usual. The explanation was plausible and well argued. The recession began at the end of 2007, and the financial crisis followed.

The sceptics of 2006 were thoughtful people, and they still missed the turn. A distorted long end changes the level of the spread without changing what an inversion does to lenders, and that the transmission runs through credit whatever the reason the long yield is low.

Banks lend less

Writing the curve off as broken, however, ignores the main way it reaches the economy, through the profits of lending. Banks and non-bank lenders borrow short and lend long. When the curve inverts, the margin on new loans turns negative.

When funding costs more than long-term commercial loans pay, lenders mostly respond by cutting how much they lend, and only secondarily by raising its price. Senior loan officer surveys show banks tightening standards at a pace usually seen only in deep recessions.

The squeeze is slower for banks than for other lenders, because many depositors still accept rates well below those on Treasury bills. That lag will not last. Deposits have begun to drift out of banks towards money market funds offering higher yields, and every dollar that leaves must be replaced with dearer funding or matched by less lending. Non-bank lenders, which fund themselves in wholesale markets at close to the policy rate, feel the inversion at once.

What the sceptics get right

The honest objection concerns timing, not direction. The lag between inversion and recession has ranged from several months to about two years, which makes the curve a poor tool for anyone trying to time a quarter. The spread also inverted in 2019, and the recession that followed in 2020 was caused by a pandemic that no bond market foresaw, so that episode proves less than either side claims.

Fed economists have also argued that a spread at the very short end, comparing where markets expect the policy rate to be in eighteen months with where it is today, is a cleaner signal, and Jerome Powell pointed to that measure this year. The objection cuts both ways. Futures now price rate cuts in the second half of next year, which is exactly the pattern that shorter measure is designed to detect.

A second objection is that the labour market remains strong, with unemployment near its lowest in half a century. Unemployment has, however, usually been low at the moment the curve inverts, because inversions happen late in expansions when the central bank is still tightening. A tight labour market at this stage is consistent with the signal, not evidence against it.

Signals that would settle it

The next senior loan officer survey, due in early February, will show whether banks are still tightening standards and whether demand for loans is falling. Weekly jobless claims are the most timely sign of whether tighter credit is reaching employers. A sustained rise in claims, alongside further tightening in the survey, would confirm that the curve's signal is working through the usual channel.

An inverted curve does its own damage by choking off bank credit, and with lending standards already this tight, the recession it points to will probably arrive within the next year or so, whatever the distortions in the bond market.