Margin debt: is the fuel of the bull market running out?

Four results frame the study. Margin debt is mostly a coincident indicator: its year-over-year growth moves with the S&P 500’s. Its warning value comes from the rollover after a surge, not from the level. Investor leverage measured against cash balances is at a record. And after every margin debt peak since 1998, the market high has come quickly, a median of 2.2 months later.

None of these is a timing signal on its own. Together they describe a late-cycle leverage profile that has, historically, come with higher drawdown risk over the following year.

Margin debt is FINRA debit balances in customers’ securities margin accounts. It is the amount investors borrow from brokers to purchase securities on margin. The series is published monthly by FINRA as part of its margin statistics and is widely used as a proxy for investor leverage in U.S. equities.

Leverage amplifies rallies. When prices rise, margin capacity expands, collateral values increase, and investors can add exposure with borrowed funds. That dynamic can accelerate upside moves and extend bull markets. The reverse is also true. When prices fall, leveraged positions face margin calls, forcing sales into a declining market. Forced deleveraging can amplify drawdowns and turn orderly corrections into sharper declines.

Because of this asymmetry, margin debt has long been discussed in the context of margin debt and stock market crashes, investor leverage, and the margin debt chart/history. The question is whether the series provides a reliable leading signal or whether it is simply a coincident measure of market sentiment.

FINRA also publishes the free credit balances investors hold in cash and margin accounts: the idle cash at brokers that can absorb losses or fund new purchases. Dividing margin debt by those balances gives a simple measure of risk appetite: how much borrowed money sits against each dollar of cushion.

That ratio reached 3.41× in September 2026, the highest since the series began in 1997. It was 1.79× at the March 2000 peak, 1.09× in July 2007 and 2.19× in October 2021, and bottomed at 0.61× in December 2008. Investors are carrying more leverage relative to their cash than at any previous peak. The ratio is nominal and is not scaled to market value.

Margin debt has grown roughly fifteen-fold since 1997, from about $100 billion to a record $1.502 trillion in June 2026. Much of that growth reflects higher asset values, but the path is not smooth: each prior surge in borrowing ended in a sharp deleveraging.

Margin debt fell 55% after March 2000, 52% after July 2007 and 35% after October 2021. Those declines ran alongside the bear markets that followed, as falling prices triggered margin calls and forced selling.

Year-over-year margin growth and SPY year-over-year returns move together closely. The contemporaneous correlation is 0.806 with n=345. That is a strong coincident relationship, consistent with margin debt rising in bull markets and falling in bear markets.

Lead-lag analysis shows how the relationship shifts with timing. The sign convention used here is positive lag = margin leads. The table below summarizes correlations of margin YoY versus SPY YoY at different lags.

The correlation peaks at zero lag and declines as the horizon moves away. A negative lag means SPY leads margin. The pattern indicates prices lead margin slightly more than the reverse. In practice, margin debt is fuel that follows price action rather than a clean leading indicator at short horizons. The correlation falls to -0.14 at a 12 month lag, consistent with mean reversion over longer periods.

Short-term predictive power is weak. Correlations of margin YoY with forward returns are near zero at 3 and 6 months: forward 3m 0.035, n=342; forward 6m 0.005, n=339. Forward max drawdown correlations are modestly positive at 3m 0.19 and 6m 0.169.

At longer horizons the sign flips. Forward 12m return correlation is -0.14, n=333, and forward 24m return correlation is -0.385, n=321. Forward max drawdown correlation at 24m is -0.154. The negative sign suggests that very high margin growth today is associated with weaker returns over the following one to two years. This is consistent with the idea that extreme leverage builds late in cycles and is vulnerable to reversals.

The result does not imply timing precision. It reflects a statistical tendency across many months, with overlapping samples and a small number of independent episodes.

Regime analysis conditions forward outcomes on the state of margin growth. The table below summarizes average forward 12 month return, average maximum drawdown over 12 months, and the share of months with a drawdown worse than -15%.

The peaked & rolling regime is defined as YoY was greater than 30% within the prior 6 months and has since fallen by more than 15 percentage points. This regime captures rollovers from elevated leverage. Average next-12m return is -3.2% versus +8.9% for all months. The average maximum drawdown over 12 months is -17.6% versus -11.0% unconditional. Fifty-six percent of months in this regime see a drawdown worse than -15%, compared with 26% unconditional.

High YoY growth alone, YoY >30%, also shows deterioration: average forward 12m return 0.5% and 45% of cases with a drawdown worse than -15%. Negative YoY, YoY <0%, shows a higher average forward return 9.3% but still elevated drawdown risk at 38%, reflecting the mixed nature of deleveraging periods.

Margin debt peaked before the SPY peaks in the three largest drawdowns since 1997. The table summarizes SPY peak, trough, maximum drawdown, margin peak, lead time, YoY peak, and margin decline from peak to trough.

Margin debt peaked 2 to 5 months before each SPY peak. In 2000, margin peaked in March 2000 with YoY 80.5%, five months before the August 2000 SPY peak. In 2007, margin peaked in July 2007 with YoY 62.6%, three months before the October 2007 SPY peak. In 2021-2022, margin peaked in October 2021, two months before the December 2021 SPY peak, with YoY peaking earlier in March 2021 at 71.6%.

The subsequent margin declines were large: -53.6% from peak to trough in the dot-com bust, -52.1% in the financial crisis, and -35.2% in 2022. These episodes illustrate how margin debt and stock market crashes can coincide with forced deleveraging after peaks.

Margin debt is not a universal early warning system. Exogenous shocks can override leverage dynamics.

The 2020 COVID crash is a clear example. Margin YoY was already negative at -6.2% in February 2020, so the signal gave no warning of the abrupt drawdown that followed. The shock was external and rapid, not driven by a leverage build-up.

The 2018 Q4 selloff also shows limited signal. YoY margin growth in September 2018 was +6.9%, not extreme, and the subsequent decline was modest relative to major bear markets.

Early 2025 provides a more recent illustration. YoY margin growth reached +33.5% in January 2025 before the spring 2025 selloff. The level was elevated but not unprecedented, and the timing was not precise.

These cases reinforce that margin debt leading indicator properties are statistical tendencies, not deterministic rules.

As of September 2026, margin debt stands at $1.48 trillion. Year-over-year growth is +31.4%, down from a +53.7% YoY peak in the last 12 months. The record nominal level was $1.502 trillion in June 2026. The three-month change has turned negative at -1.5%.

This configuration matches the peaked & rolling regime described above: a prior surge in YoY growth above 30% followed by a meaningful deceleration. Historically, such rollovers have been associated with weaker forward returns and higher drawdown incidence, with average next-12m return -3.2% and 56% of cases experiencing a drawdown worse than -15%.

It is important to state the limitations plainly. The historical tendency is based on a small number of independent episodes. Monthly samples overlap, which inflates statistical significance. Margin debt is reported in nominal dollars and is not scaled by market capitalization or adjusted for inflation. The series is not a timing signal on its own and should not be used in isolation for decisions.

To test what happens after a margin debt peak, we identified every record high that was followed by at least three months without a new record and a decline of at least 8%. Since 1998 that gives seven episodes. For each we measured when the S&P 500 (SPY) made its final high, how far above the peak-month level it got, and how far it fell afterwards.

The pattern is consistent on timing and loose on magnitude. The market high came 0 to 11 months after the margin peak (median 2.2 months) and never more than 8.5% above the peak-month level (median +2.4%). The decline that followed ranged from 14% to 57%, with a median of 24%. In 2018 the measured decline includes the 2020 crash, which came after the market recovered.

The seven episodes split into two groups. In four of them (2000, 2007, 2018 and 2021) margin debt went on to fall more than 25% from its peak. The market topped a median 3.7 months after the margin peak, 1–9% higher, and then fell 24–57%, with a median decline of 37%.

In the other three (1998, 2015 and 2025) margin debt fell only 9–14%. The market corrected 14–19% within months and then, as borrowing resumed, made new highs. At the moment of the peak the two groups look the same. The difference becomes visible only as successive monthly FINRA releases show how far deleveraging goes.

Applying the historical ranges to the S&P 500 (converted from SPY): the provisional margin peak is June 2026, when the index closed the month near 7,494. History puts the market high 0–8.5% above that level, about 7,494–8,130, within 0–11 months. The high so far is about 7,818 on 6 October 2026, 4.3% above the June level after 3.2 months, inside the historical range.

If 6 October proves to be the high, the historical declines translate to about 6,720 (−14%, the mild cases), 6,330 (−19%, as in 2025), 5,940 (the −24% median) and 4,930 (the −37% median of the terminal episodes). These are scenario ranges from seven episodes, not forecasts. A new margin debt record would void the peak.

Three observations would confirm the peak. First, the October FINRA release, due in late November, coming in below the June record of $1.502 trillion: four months without a new record. Second, the S&P 500 losing its 200-day moving average and failing to recover it. Third, margin debt falling 8% or more from the record, below about $1.38 trillion, the threshold used to define the peaks in this study.

The signal is cancelled if margin debt sets a new record. That is what happened in 2025: the January peak was followed by a 19% correction, then borrowing and prices both made new highs.

Margin debt is one lens. We cross-check it against breadth, macro and liquidity conditions, a bull/bear regime model and the calendar. In mid-October 2026, leverage, breadth and liquidity point the same way: margin debt has peaked and its growth is rolling over, only about a quarter of stocks trade above their 50-day averages while the index sits near highs, and credit, financial conditions and liquidity are tightening without signs of stress.

The bull/bear regime, defined on price and time following Ned Davis Research’s rules, has not broken, and the midterm-year calendar is neutral. The combination describes a late-cycle market whose price trend has not yet confirmed the leverage warning. A break of the 200-day average would be that confirmation.

This study uses monthly FINRA margin statistics from January 1997 through September 2026, paired with SPY month-end prices for price return calculations. The sample covers 357 months. All correlations and regime statistics are computed on year-over-year changes in margin debt versus SPY year-over-year returns, unless otherwise noted.

The analysis focuses on margin debt leading indicator properties, lead-lag relationships, and regime-dependent forward outcomes. The data are nominal and not scaled by market capitalization or inflation. The period includes multiple bull and bear cycles, including the dot-com peak, the global financial crisis, the COVID shock, and the 2022 drawdown.

• Monthly FINRA margin statistics from January 1997 to September 2026, paired with SPY month-end price returns. Sample size 357 months.
• Correlations are sample statistics and do not imply causation.
• Margin debt is nominal and not adjusted for market cap, inflation, or changes in reporting coverage.
• Regime definitions are ex-post and may be subject to look-ahead bias in practice.
• Overlapping monthly observations increase serial correlation and can overstate precision.
• Results are specific to U.S. equities and FINRA margin statistics. They do not generalize to other markets or leverage measures.

Disclaimer: This is research for informational purposes only and does not constitute investment advice.