Snowball autocallables and the CSI 500/1000 knock-ins

A bond-like coupon for a market short on yield

A "snowball" (in Chinese, xueqiu, 雪球) is a structured note, sold mainly by Chinese brokerages to retail and high-net-worth clients since around 2016-2017, and marketed as a bond-like income product: hold it, and so long as a reference index stays roughly within a set range, the buyer collects a generous coupon, often 10-20% a year. Chinese investors were looking for yield in a market where safe returns had become harder to find, and they bought these notes in growing volume: about RMB 10 billion in 2019, rising past RMB 50 billion in 2020, and reaching an outstanding notional of around RMB 216 billion (about USD 30 billion) by August 2022, referencing mainly the CSI 500 and CSI 1000 small and mid-cap indices.

The product was not what it looked like. Underneath the coupon sat an embedded barrier option. If the index stayed inside its range, the note behaved like a bond. If the index fell hard enough, the note "knocked in", and its payoff at maturity turned into something much closer to a direct long position in the index from a much higher starting level. That downside had mostly stayed dormant through 2018-2021, so the coupon kept arriving and the product kept selling.

The knock-ins arrive together, January 2024

January 2024 is when that dormant risk woke up for a large share of the market at once. Bloomberg reported on 11 January 2024 that about USD 13 billion of snowball notional was nearing loss-trigger levels, including around 30 billion yuan (about USD 4.2 billion) of CSI 1000-linked notes already close to triggering losses at maturity. The market kept falling. On 22 January 2024, the CSI 500 fell 4.73% to close at 4,780.97, its lowest level in more than four years, and the CSI 1000 fell 5.77% to close at 5,000.83; both index futures hit their 10% daily down-limit intraday. Analysts estimated that up to a third of outstanding snowball products at some issuers knocked in that single day. By 23 January, UBS estimated total outstanding snowball notional at around USD 50 billion, with roughly 40% of knock-ins already triggered.

One investor's account

Reuters put a face on these numbers. One investor had put 1 million yuan into a CSI 1000-linked snowball in early 2023, expecting a coupon of at least 8% a year. By the time the note knocked in, that coupon stopped, and the investor needed the CSI 1000 to climb back above 7,000, from below 5,000 at the time, within a year, just to avoid losing part of the principal.

The sell-off deepens, and the regulator moves

The sell-off did not stop there. On 5 February 2024, the CSI 1000 fell a further 8.68% intraday, its seventh consecutive losing session, down 27% for January alone and 38% over the trailing year. Two days later, China's State Council replaced the securities regulator's chairman, appointing Wu Qing, and state-linked funds stepped up purchases of broad-index ETFs in the days that followed. By April 2024, with some issuers reportedly raising coupons towards 40% to keep attracting buyers after the losses, the securities regulator told major brokerages to stop increasing their net exposure to A-share-linked OTC derivatives, snowballs included.

A bond wrapped around a sold put

A snowball is not a bond. It is a package: the buyer's principal, plus a sold down-and-in put on the CSI 500 or CSI 1000 index, with an added up-and-out (autocall/knock-out) feature. The buyer is short volatility and short the tail, dressed in the clothing of a fixed-income coupon.

Two barriers, two very different outcomes

A reference level is set at inception. A knock-out barrier, roughly 3-10% above that level, is checked on periodic (commonly monthly) observation dates. If the index closes above this barrier on an observation date, the note ends early and the buyer is paid the coupon accrued so far. Most snowballs that survive to any observation date, in normal markets, end this way. A knock-in barrier, roughly 20-30% below the reference level (equivalently, near 70-80% of it), is often observed continuously rather than only on set dates. If the index ever trades below this barrier, the note knocks in: the coupon stops accruing, and the payoff at maturity converts to something much closer to a direct long position in the index, so the buyer now bears the index's downside measured from the barrier, not from the original purchase level.

What the coupon is actually paying for

The 10-20% a year coupon is the price the market pays the buyer for writing this embedded put. This is premium income from a sold option: the buyer is the insurance writer, and often does not realise it. The product is legally a brokerage-issued OTC note, not a listed option, but its risk is identical to a retail investor being short a barrier put on a small-cap index, without the margin discipline a professional option seller would be required to post.

Why the knock-ins arrived together

What breaks in January 2024 is the assumption that knock-ins are spread-out, idiosyncratic events. Tens of thousands of snowball contracts were written over 2021-2023 referencing the same two indices, with barriers clustered in similar percentage ranges below similar reference levels. A single sharp index decline pushed a large share of outstanding contracts through their barriers at once. Individually sensible positions become a systemic event once enough of them share the same trigger.

The hedging feedback loop

A second-order feedback effect follows, separate from the retail buyers' own losses. The brokerages that sold these notes typically delta-hedge their short-put exposure by holding index futures. As the index approaches the knock-in barrier, the embedded put's delta rises, so issuers buy more futures to stay hedged. Once the barrier is breached, the delta falls sharply, so issuers must sell futures quickly to rebalance. That forced selling pushes index futures down further and widens the futures-to-spot basis. A large, one-directional hedging flow, with every issuer rebalancing the same way at the same time, can move the basis away from its no-arbitrage level even though no single issuer is trying to move the market.