Autocallables & Cliquets

You can already read a vanilla payoff, you’ve met the Greeks, and you’ve just dissected barriers and digitals — options that switch on or off when the underlying crosses a level. Now we assemble those parts into the single best-selling structured product on Earth. The autocallable is everywhere: it dominates retail structured-product sales across Europe and Asia, sold by the trainload to investors who think they bought a high-coupon bond. They didn’t. They bought a barrier option in a bond’s clothing — a position that behaves like short volatility and short a deep put. By the end of this lesson you’ll read one straight off its term sheet and know exactly which side of the risk you’re standing on. Then we’ll meet its stranger cousin, the cliquet. Let’s decode the brochure.

Anatomy of an autocallable

Analogy. Imagine a high-interest savings account with two cruel twists. First, the bank — not you — can call the money back early, and it will do so the moment markets are calm and it no longer wants to pay you. Second, if markets crash hard enough, the account quietly stops being a savings account and turns into a leveraged bet on the very stock that just collapsed. That’s an autocallable: generous while things are boring, yanked away when things are good, and a trapdoor when things go bad.

The structure. An autocallable is a note linked to an underlying (a stock or, more often, an index or basket). On a fixed observation schedule — typically once a year — it checks the underlying’s level against its starting value and applies three barriers:

• Autocall barrier (usually 100% of the initial level). If, on an observation date, the underlying is at or above this level, the note redeems early: you get your principal back plus the coupon, and the trade is over. Early, clean, done.
• Coupon barrier (e.g. 70%). If the note doesn’t autocall but the underlying is at or above this barrier, you’re paid the coupon for that period. Below it, no coupon (for now — see memory next section).
• Protection / knock-in barrier (e.g. 60%). This is the trapdoor. If at maturity the underlying has fallen below this barrier, your principal protection is gone: redemption tracks the underlying 1:1, so a 40% drop in the underlying is a 40% loss of your capital.

Three barriers, one schedule, and a headline coupon that looks like yield. The island below walks the same 4-year note (8% coupon, autocall at 100%, coupon barrier at 70%, protection at 60%) through its three possible lives. Click all three buttons and watch the story change:

• “Autocalled early” — the underlying is at 105 on Year 1, clears the autocall barrier, and the note redeems immediately for 108 (principal 100 + one 8 coupon). Over in a single year. This is the most common outcome, and notice the catch in the note: your cash comes back exactly when markets are calm and you have nowhere good to reinvest it.
• “Coupons, par at maturity” — the underlying sags but mostly stays above 70%, paying coupons, skipping Year 3, then recovering principal at maturity for a total of 132. (The skipped coupon comes back — that’s the next section.)
• “Barrier breached → loss” — the underlying slides to 52 by maturity, below the 60% protection barrier, so the note converts to the underlying 1:1: you get back 60 total, a brutal loss barely softened by one early coupon.

Stare at that third scenario. The same note that looked like a 132-over-four-years income machine in scenario two can hand you a near-halving of capital in scenario three — and the only thing that changed was how far the underlying fell. That asymmetry is the whole product.

The memory coupon (and why coupons aren’t free)

The definition. A memory coupon (also called a snowball coupon) means a missed coupon isn’t forfeited — it’s deferred and paid retroactively the next time the underlying clears the coupon barrier on an observation date. Miss Year 3 because the underlying dipped to 65%? Not lost. If Year 4 comes back above 70%, you collect the Year-4 coupon and the stored Year-3 one.

Look again at the “Coupons, par at maturity” scenario in the island above and read its observation list. Years 1 and 2 pay their 8 coupons. Year 3 the underlying sits at 65% — below the 70% coupon barrier — so no coupon… but with memory, it’s only parked. At maturity the underlying recovers to 95%, clearing 70%, so the note pays the Year-4 coupon plus the deferred Year-3 coupon — 16 of coupon in one go — and because 95% holds above the 60% protection barrier, the full 100 principal returns. Total: 132. Here’s the arithmetic laid bare:

That 132-on-100 over four years looks like a tidy 8%-a-year income stream that even survives a rough patch. The brochure loves this row. But here’s the reframing that the whole rest of the lesson hangs on.

The fat coupon is not yield. It is the premium you collect for selling an option — specifically, for being short a deep down-and-in put struck around the initial level with its knock-in at the protection barrier. You are not a lender earning interest. You are an insurance seller: you’ve sold the market protection against a big crash in the underlying, and the 8% is your premium. When you remember from the volatility lessons that implied vol usually sits above realized — the volatility risk premium — the autocallable is just a retail-friendly wrapper for harvesting that premium. Which means it shares the premium-seller’s signature payoff: lots of small wins, one rare catastrophe.

What you’re really short: the knock-in put

Let’s do the financial-engineering autopsy. A long position in an autocallable decomposes, roughly, into three familiar pieces:

1. A funding / bond leg. In the friendly outcomes (autocalled, or par at maturity) you get your principal back. That’s the bond-like skeleton — it’s where the “feels like a bond” illusion comes from.
2. A strip of digital coupons. Each coupon is cash-or-nothing: you get the fixed 8 if the underlying is above the coupon barrier on the observation date, else nothing. That is exactly a digital option from the previous lesson — a binary bet on a level, one per observation date, with the memory feature linking them.
3. A short deep down-and-in put. Here’s the dangerous leg. You are short a put struck near the initial level (100%) whose knock-in sits at the protection barrier (60%). While the underlying stays above 60%, the put is dormant and harmless. Breach 60% and it knocks in, and now you’re short an in-the-money put — meaning your loss tracks the underlying 1:1 all the way down.

So the autocall feature is really a knock-out that can end the whole package early; the coupons are digitals; and the gut-punch is a sold barrier put. Net-net: you are short optionality.

Run the loss case from the island in full. The underlying ends at 52% of its start, below the 60% protection barrier. The put has knocked in and is deep in the money against you:

You collected one 8% coupon and ate a 48-point drop in principal. That’s the short put paying off — against you. Notice the grotesque asymmetry: your upside was capped at the coupon and short-lived (the note autocalls away your best income years), while your downside ran the full 1:1 below the barrier.

So when you hold an autocallable you are, in Greek terms: short vega (you sold options — rising implied vol hurts you), short skew (you’re short a deep out-of-the-money put, the most skew-sensitive thing there is), and — for the basket versions we’re about to meet — long correlation. Plus a subtler curse: the note autocalls away exactly when you’d want to keep the income — markets are calm, the underlying is up, life is good, and poof, your cash comes back to be reinvested at lower yields. That’s reinvestment / duration risk baked into the product.

Worst-of baskets (the dial that juices the coupon)

Single-stock autocallables exist, but the species that fills brochures is the worst-of basket. Instead of one underlying, the note references several — say three indices or stocks — and every barrier test uses the worst-performing one. Coupon paid only if the weakest name is above 70%. Autocall only if the weakest is above 100%. Protection breached if the weakest finishes below 60%.

This is why worst-of coupons are so juicy, and it’s the opposite of diversification. With one underlying, you need that name to crater. With three, you need only any one of them to crater — three independent ways to hit the trapdoor. More names = more breach paths = more risk = bigger coupon.

Quick intuition with three names. Suppose each of three underlyings independently has, say, a 10% chance of finishing below the protection barrier. For a single-name note the breach chance is 10%. For a worst-of on all three, a breach happens if any of them falls through — roughly 1 − (0.9)³ ≈ 27% if they moved independently. Nearly triple the tail risk, which is exactly what funds the fatter coupon.

And there’s a hidden Greek here: that breach probability depends on correlation. If the three names are highly correlated, they tend to crash together, so “worst-of” isn’t much worse than “any-of” — the breach probability stays closer to the single-name 10%. If they’re uncorrelated, the worst-of has many more independent chances to dive, pushing the breach probability up toward that 27%. So the holder of a worst-of note is long correlation: you want the names to move together. You’re effectively short dispersion — the same dispersion risk from the correlation-trading lesson, smuggled into a retail note.

Cliquets (ratchets): resetting the strike

Now meet the autocallable’s stranger cousin. A cliquet (from the French for “ratchet”; also called a ratchet option) is a series of forward-starting options chained end to end, where the strike resets to the prevailing spot at the start of each period. Picture a ladder of consecutive at-the-money options: each period you start fresh from wherever the underlying now sits, measure that period’s return, and the gains ratchet — once locked in, a later fall can’t claw them back (that’s the ratchet teeth catching).

Most cliquets clip each period’s return with a local cap and floor (a per-period maximum and minimum, often a 0% floor so a bad period contributes nothing rather than a loss), and sometimes a global cap on the lifetime sum. The classic “X% per year, gains locked in annually, never goes negative” structured note is a floored cliquet.

Worked example — a 3-period cliquet, local cap 5%, local floor 0%. Each period we measure the return from that period’s reset strike, clip it to the [0%, 5%] band, and add it to a running total that ratchets up:

Two things to notice. The floor in Period 2 means the down-move contributes a clean 0% instead of dragging the total down — the ratchet holds. And the strike reset in Period 3 is the magic: even though the underlying (110) is barely above where Period 1 ended (107), Period 3 still earns a fresh +5% because its strike reset to 103, not the original 100. You get paid on each leg’s local move, not on the cumulative distance from the original start.

Why issuers love them, and why traders dread hedging them. Issuers love cliquets because the “locks in your gains every year, can’t go down” pitch is irresistible to retail. But a cliquet’s value depends overwhelmingly on the volatility of future, not-yet-started options — the forward volatility and how bumpy it is, the vol-of-vol — and on the whole term structure of volatility, because each forward-starting leg is priced off vol for a period that hasn’t begun. A vanilla cares about today’s implied vol; a cliquet cares about vol between future dates you can’t yet observe. That makes it exquisitely sensitive to things that are hard to mark and hard to hedge, which is why cliquets have a long history of blowing up the desks that mispriced their forward-vol exposure.

When (if ever) to buy one

Time for the honest take. An autocallable isn’t a scam — it’s a legitimate way to sell volatility that can pay you well if the deal is fair and you know what you own. It’s a reasonable trade when all of these hold:

• You have a genuinely mildly bullish-to-sideways view — you think the underlying will hover, drift up, or at worst sag modestly, not crater.
• You accept the tail — you’ve looked the 1:1 downside in the eye and you can live with it.
• You understand you’re selling volatility (and skew, and on a worst-of, correlation) — you’re the insurer, eyes open.
• You’re paid a fair premium — the coupon genuinely compensates you for the risk, after the issuer’s costs (which is the next lesson’s whole subject).

It’s dangerous when it’s sold — and bought — as a bond substitute to someone reaching for yield who never sees the short put hiding under the coupon. That buyer thinks they own protection; they actually sold it. Here’s the trade-off at a glance:

Putting it together

An autocallable is the world’s best-selling structured note: a high-coupon wrapper checked against three levels on a schedule — an autocall barrier (clear it, redeem early with a coupon), a coupon barrier (clear it, get the coupon — often with a memory feature that recovers missed coupons later), and a protection barrier (breach it at maturity and lose principal 1:1). Decompose it and the truth appears: a bond-like funding leg + a strip of digital coupons + a short deep down-and-in put. The fat coupon is option premium, not yield — you’re a vol seller harvesting the volatility risk premium, short vega and short skew, with a capped upside that autocalls away early and a downside that runs the full distance below the barrier. Worst-of baskets turn the coupon dial by adding underlyings (more breach paths) and make you long correlation. The cliquet is the cousin that resets its strike each period, ratcheting in gains with local caps/floors, with value buried in forward vol and vol-of-vol — a hedging nightmare. Buy one only when you’re mildly bullish-to-sideways, know you’re selling vol, and are paid fairly; never as a “safe bond.” Next we’ll decompose structures like this piece by piece and price each part — the only way to tell whether the coupon is a gift or a trap.

Next — decomposing structured products — we take a note like this apart into its exact building blocks (zero-coupon bond, digitals, barrier options) and value each piece, so you can finally judge whether any structure’s headline coupon is a fair deal or a dressed-up fee.