The Korean Levered ETFs Shaking Markets All Around the World

Korean retail investors have driven an unusually large share of the growth in Asia Pacific levered ETF markets, with regional AUM rising from roughly 12 to 13 billion at the start of the year to approximately 50 to 55 billion. Ninety-three percent of levered ETFs in Korea are owned by retail investors, compared to around 75 percent in the US, and single stock levered ETFs in Korea appear to be having a measurable impact on the movement of the underlying shares themselves. Global levered ETF AUM stood at roughly 250 to 270 billion dollars, with US AUM peaking just north of 200 billion, most of that growth driven by price performance rather than new inflows. Alex Altman of Barclays described the notional exposure coming from these products as terrifying.

Levered ETFs obtain their exposure synthetically through swap agreements with prime brokerages, placing notional exposure directly on bank balance sheets. With AUM north of 200 billion, levered ETFs are generating over 400 billion dollars of additional synthetic exposure in swap format across the banking system. In a triple-levered ETF with 30 billion in AUM, a 10 percent drop in the underlying requires reducing exposure by over 10 billion dollars to maintain the stated leverage ratio, illustrating the mechanical daily rebalancing pressure these products create. Altman argued this introduces a short-gamma dimension to the market that was relatively small only a couple of years ago, and because overwriting ETFs that sell volatility have not grown at the same pace, the net gamma profile of the market is effectively becoming more negative.

Levered ETFs are a contributing factor to tighter financing rates but should not be singled out as the primary cause. Higher spot market levels raising asset prices is the primary driver, and multi-manager hedge fund platforms, which now represent approximately one trillion dollars of AUM having roughly tripled since COVID, also consume significant bank balance sheet. Bank balance sheet capacity is expected to remain scarce as long as the S&P trades around 7,500 to 7,600. A regulator might decline to approve an individual five-times or hundred-times levered product, but the aggregate of all such products across the market could represent a systemic concern that no single regulator is positioned to address.

The bulk of levered ETF AUM growth has been concentrated in NASDAQ-related products, semiconductors, and single-name products tied to AI and tech themes. The largest single-name levered ETF in the US is tied to Micron, while the largest single-name levered ETF in the world is tied to Heinecks. A 2x levered ETF on the MAU was trading at 23 a year ago and recently peaked at over 1,200. An attempt to list 5x levered ETFs was effectively blocked when the regulator made clear there are limits to acceptable leverage, and the SEC has approved all levered ETFs currently in the market.

US household wealth held in equities stands at 34 percent, the highest on record and above dot-com era levels, while real estate sits at approximately 26 percent, making the gap between the two the widest on record. A 20 percent structural impairment to the S&P 500 would destroy approximately 16 trillion dollars of wealth, roughly half of US GDP, and no econometric model currently accounts for this scenario. AI is now driving macro growth, stock prices, and consumption simultaneously, which Altman describes as AI impact squared, making the economy more sensitive to equity market outcomes than at any prior point. Joe Weisenthal adds the caveat that the median household is likely still more exposed to their home than to equities, meaning aggregate figures obscure significant distributional differences.

The Barclays Equity Timing Indicator, called Betty, has 19 inputs, was first created in 2018, and excludes sentiment indicators in favor of quantifiable data including real yields, spot-vol correlation, mutual fund flows, CTA positioning, vol control flows, and levered ETF flows. On a random two-month holding period the average S&P return is approximately 190 basis points with a 73 percent hit rate. When Betty is in elevated warning territory of six to eleven, the probability of making money in the S&P over a two-month horizon falls to around 35 percent or lower with a negative average return. Betty has been in record warning territory driven primarily by momentum crowding, though a recent momentum pullback has partially reset conditions, and the model first flashed a warning in late May with the S&P having done nothing since then.

Real yields on a 10-year basis are running at approximately the 95th percentile, around 230 basis points, while the S&P is currently trading at approximately 20.2 to 20.3 times earnings against a historical average multiple of 14 to 15 times at comparable real yield levels. Using the post-COVID environment as a reference, the comparable multiple is around 18.5 times, still implying roughly a 10 percent multiple contraction from current levels. Higher margins and rising return on equity justify some premium, but the model is not predicting a crash, only that asymmetry in the equity market is not favorable given that earnings remain good and a large fiscal deficit continues to support the economy.