The Fed raised interest rates by a quarter of a percent last week. The move was widely expected. More surprising was the accompanying press release, which echoed the Fed’s prior communication but dropped the phrase “some additional policy firming may be appropriate”. In excising these words, the central bank indicated an openness to winding down one of the most rapid and aggressive hiking cycles on record.

In the press conference following the hike, Chairman Jerome Powell pointed to a number of developments that might slow growth and warrant the Fed taking its foot off the brake. The Treasury is approaching the debt ceiling. Job vacancies are declining. The housing market is slowing. But he returned repeatedly to one economic risk: credit conditions are tightening. In plain english, banks are lending less money. 

Credit is the lifeblood of the economy. That’s why the main tool the Fed uses to speed or slow growth is the interest rate—the cost of credit. Higher rates make investment and consumption financed through borrowing less appealing, and the economy slows.  

But cost is not the only relevant gauge of how freely credit is flowing. The quantity on offer matters, too. And ultimately it is banks—not the Fed—that make lending decisions: who gets to borrow, and how much.

At the time of the press conference, Powell had access to data about bank lending—the Fed’s Senior Loan Officer Opinion Survey (SLOOS)—that was only released to the public moments ago. The survey results go a long way towards explaining the Fed’s willingness to pause here.

The SLOOS is a quarterly survey of 80 domestic commercial banks and 24 US-based branches of foreign banks. The survey asks loan officers at each bank about the demand for loans they are seeing, and about their willingness to make loans, on everything from credit cards to commercial real estate. 

But one part of the survey—the one that gauges bank appetite to make loans to large companies—has attracted almost all of the attention. And for good reason: the proportion of banks tightening their lending standards for large firms—for example, asking for more collateral to secure a loan, or lending only for a shorter term—has been perhaps the single-best leading economic indicator for over 50 years. 

Today’s release shows that, on net, 46% of loan officers are making it harder for companies to borrow. Banks are as unwilling to make loans as they were during the early stages of the 2008 financial crisis and the tech crash in 1999.

The Fed started conducting this survey in 1967. Since then, there have been 8 instances where banks evinced an unwillingness to lend. All but one of these instances—in late 1998 after Russia defaulted on its debt—were followed by recessions. And every recession since 1967 has been preceded by banks tightening lending standards.

Other indicators provide a timely signal of an impending slowdown, but also cry wolf more often. For example, the yield curve—the difference in compensation for lending to the government for longer versus shorter periods of time—has inverted ahead of every recession in the last 70 years. But it has also inverted four times without a recession occurring.

Aside from its accuracy, the Fed’s survey also stands above almost all other economic data for its prescience. Most “leading indicators”—like surveys of manufacturing confidence or changes in employment like nonfarm payrolls—are better understood as timely indicators of the economy’s current state, rather than where it’s going. They are useful as “nowcasts” rather than forecasts. 

By contrast, bank loan officers are canaries in the coal mine. Their lending signal is genuinely predictive of recessions four to ten quarters down the road. This shouldn’t be entirely surprising: bank loan officers have on-the-ground information from companies across sectors and geographies. They are well placed to spot risks. Moreover, their views are self-fulfilling: when loan officers think they see trouble on the horizon, they restrict lending, making a slowdown all but inevitable. 

But maybe where SLOOS shines brightest is in a comparison with another very good leading indicator: financial markets. The prices of stocks and bonds represent the totality of investor views. Individual investors have limited insight into the future, but there is enormous wisdom in the crowd. In theory, prices reflect all available public information. It’s therefore remarkable that lending standards move closely in line with security prices, and sometimes ahead of them. 

This is, perhaps, what is most interesting about today’s data. The S&P 500 has climbed 15% since bottoming last October. The high yield credit spread—the financial reward for lending to the riskiest US companies—is stable and close to the long run average. But banks are retrenching from lending in a way that is only consistent with falling stock prices and rising credit spreads. 

Maybe today’s survey is a red herring. Medium-sized banks—those with assets between $50bn and $250bn (SVB and First Republic fell into this category)—have been experiencing deposit outflows and declining stock prices. More of them reported tightening standards than the small or very large banks in the Fed’s sample. If medium-sized banks are biasing the reading, the survey could be less indicative of credit conditions than in the past. 

But an equally likely explanation for the disconnect between markets and today’s survey is that, for stocks and bonds, the shoe has yet to drop.

This is not the first time the signals from bank loan officers and markets have diverged. In the lead-up to the financial crisis, the Fed’s survey showed banks growing increasingly risk-averse in the summer of 2005—a full two years before markets started to decline. 

Banks may have only temporarily lost their appetite for risk. Loan officers tightened lending standards sharply in the third quarter of 2020, only to loosen them rapidly by the end of the year.

But a turnaround like that seems unlikely. In a special question included in this quarter’s survey, the Fed asked banks how they expect their lending standards to evolve. Their response was, essentially, more of the same—tighter lending across all categories for the rest of the year.

Moreover, because of the survey’s predictive power, regardless of how lending evolves, the damage is already baked-in. In fact, if the historically strong relationship between lending standards and employment held today, payrolls would be shrinking (they are still growing relatively quickly).

If there is any reason to think today’s data was a non-event, its because the trend is not new. Banks started tightening lending in 2021. The question we may be asking down the road is not why Powell stepped off the economic brake last week, but why he didn’t do it sooner.