The S&P 500 has continued its upward climb in recent months, despite rising fears about the Delta variant and tapering. But one indicator may be warning of a weaker—or even bearish—period ahead for stocks.

Margin debt, which is the money investors borrow from brokers to buy securities, hit a record high in June, at $882 billion per FINRA, and has since slumped. July marked the first time margin debt declined since the pre-COVID days, and that decline comes at a time when the S&P 500 is hitting all-time highs of its own. In other words, investors were borrowing less money to buy stocks last month than they were earlier this summer, even as stocks ticked higher.

So why would that be a bad thing? Think of margin debt as "basically just a statement of confidence in the market," explains Randy Frederick, managing director of trading and derivatives at Charles Schwab. "If someone's willing to borrow money in order to buy stocks, they believe that the return they get on those stocks is going to exceed the interest rate they pay on that borrowing."

Indeed, as Stephen Suttmeier, technical research strategist at Bank of America, wrote in a August 23 note, "Rising leverage tends to confirm U.S. equity rallies. It is not new record highs for margin debt that we worry about." Instead, "we get concerned when margin debt stops rising to suggest that investors have begun to reduce leverage." He points out that the S&P 500 rose 2.3% in July, but margin debt dropped by 4.3% from its high the month prior.

"Although peaks in margin debt don't always coincide with highs for the [S&P 500], they tend to be bearish for equities," Suttmeier argues. In his note, he points to 21 times since 1997 (based on FINRA data) when margin debt peaked and weaker, and even poor, S&P 500 performance ensued (see Bank of America's chart below).

"If June 2021 proves to be another cycle high for FINRA margin debt, the risk is for weaker, and in most cases negative, [S&P 500] returns from one month through 24 months after a margin peak," Suttmeier says. To put numbers on that, he notes that one year after "a peak in margin debt shows the [S&P 500] down 71% of the time with an average return of -7.8% (-10.7% median)."

Frederick suggests the July drop in margin debt is "unusual," since historically margin debt goes up when the market is going up, and goes down when stocks sell off. "We haven't really seen a big downturn in the market, and we really haven't seen a big drop in margin debt since March 2020," he tells Fortune. "That's why the July data is very unusual, because the market is essentially sitting at an all-time high and we had a pretty sizable downtick in margin debt. That's really weird."

On one hand, Frederick noted in a recent tweet that declining margin debt "reduces the risk that a pullback becomes a full correction" (indeed, lately we've seen dip buyers jumping in whenever the market sells off, preventing it from going into a full correction), but that "it also implies investors are getting more cautious."

Margin debt is mostly tied to market performance. But this month it declined for the first time post-COVID, despite new $SPX highs. This reduces the risk that a pullback becomes a full correction, but it also implies investors are getting more cautious.https://t.co/wyfH3UcCAS pic.twitter.com/2i7wMaKrlQ

— Randy Frederick (@RandyAFrederick) August 16, 2021

Certainly there are other factors at play in terms of how investors are feeling about stocks and how the S&P 500 may perform in the short term, including seasonality (August and September tend to be weak months for stocks), the upcoming Jackson Hole summit, and Fed meetings (which could offer clues about tapering), and the continued path of the coronavirus.

Frederick cautions that "one month doesn't make a trend" and that if margin debt data for August comes in higher, July may be considered a somewhat insignificant "anomaly." If August data continues the slump while stocks remain resilient, however, that may be "more significant." (In general, Frederick says, he wouldn't be surprised by a 5% to 7% pullback in the near term.)

Either way, it might be a good idea to keep your eye on the next month's worth of data.