40-Times-Leveraged Short Volatility Trade Goes Horrifically Wrong For Everyone Involved

…or, Managing and Mitigating the 2014 CNH FX TARF blowup

(Author’s note: I wrote this a year or so back for a few friends on Finance Twitter who wanted to know what FX TARFs were and what effect the huge vol overhang from USDCNH TARFs would have on the spot market. I figured that the shenanigans in USDCNY yesterday made it an appropriate time to dig this up and republish it for a wider audience. Onward…)

In 1985, it was CHF-denominated loans to Aussie farmers. In 1998, it was USDIDR cross-currency swaps with Indonesian noodle makers. In 2008, it was USDKRW KIKO options to Korean corporates and PRDCs to Japanese investors. In 2014, the FX blowup du jour is USDCNH European-knock-in target redemption forwards (EKI TARFs): SME and wealth management clients across Asia made gargantuan, leveraged, unfunded bets that the PBOC would continue to manage the CNY for a slow appreciation, and were caught long-and-wrong when the trend turned in February of this year.

Your instinctive reaction might be “if the customers are losing, the exotics books on the other side must be having a good time” – but you’d be mostly wrong. When TARFs go wrong, they can go wrong in all sorts of unexpected ways: imperfect hedging; counterparty risk; and collateral risk among them1; and most of those are bad news for both the bank’s exotics book and any smaller banks that might have back-to-backed the trade along the way.

The trades in question are strips of leveraged knock-in forwards with an autocall feature2. The typical pitch runs something like this:

“So here you go, USDCNH spot’s at 6.0350, and you want to sell it because you think the PBOC’s going to keep guiding it lower. Here’s the deal: with this trade, you sell a million USD a month, every month for a year, at 6.15 – not 6.0350 – that’s an extra hundred thousand yuan a month in your pocket. If spot fixes between 6.15 and 6.20, you don’t do anything; if spot fixes above 6.20, you have to sell 2 million USD a month at 6.15, but that won’t happen, will it now? And if you accrue more than 300,000 yuan of positive PnL, the trade will knock out: your PnL’s locked in and you’ve got no more risk.”

The client says “you’re done”, the salesperson clicks “Save”, and the risk hits the trader’s book.

The risk profile of a brand-new TARF, once you’ve hedged out the delta, looks (if you squint) like an economy-size long position in short-dated low-delta wings. A one-year TARF isn’t like a vanilla, and doesn’t have the peak of its risk in the one-year ATM-ish bucket: the peak of the vega risk on the time axis is around the expected knockout date of the trade, and the peak on the strike axis is slightly inside the strike of the TARF (or, in this case, the EKI level).

So the default hedge is to sell vanillas with expiry roughly equal to the expected life of the trade (or, equivalently, the middle of the bucketed vega profile) and strike at the peak of the vega, in enough size to roughly hedge out the vega of the trade.

The vega of our example trade is bimodal in the 2mth and 6mth buckets, so the natural hedge is to split the difference and sell 3mth 6.15 USDCNH calls in about CNH 250k of vega. The vega hedge notional ends up close to USD 40 million, which – when you compare it to the “$1mio a month” notional that the client sees on their confirmation – should tell you something about the leverage embedded in these structures.

The problem for the exotics book arises when spot reaches the knock-in level of the TARF. At that level, the TARF’s expected maturity extends from 2-3 months to a full year, and the book gets rapidly less long gamma from the TARF (because it’s behaving like a longer-dated option) at the same time as it’s getting rapidly shorter gamma from the vanilla hedge as it becomes an ATM strike3. There’s also sizable digital risk around the EKI strike on the fixing dates, just in case your exotics traders’ lives aren’t exciting enough already.

The upshot, then, is that as the TARFs fall out of the money, the exotics book needs to turn around and buy spot and gamma – right when their customers and the other exotics desks on the street are panicking and doing the same. And that’s what happened: when USDCNH spot spiked from 6.05 to 6.25 between February and April, short-dated option volatilities more than doubled (from 1.6 to 3.8 in the 1mth) because of relentless buying by exotics desks and unwinding by customers.

The estimates that did the rounds in early April, as these trades were blowing up, were that eighty to a hundred dollars of these trades had been printed just in USDCNH, and just in the first three months of 20144.

“So if the exotics desks are losing, and the customers are losing – who wins?” you may ask yourself5. The real winners are probably the EM FXO flow books that the exotics books did their hedges with in the first place; the market dynamics are a lot like the pre-crash EM FXO markets in 2007 when the street kept getting stuffed with gamma from customer flow and having to turn around and sell it. This turned into a game of pass-the-parcel between flow desks, as traders kept trying to catch a falling knife in EM vol markets and kept getting stopped out6. The end result was front-end KRW vol trading on a 3 handle in 2007… and triple digits in 2008 when the music stopped and the happy vol sellers became distressed vol buyers.

And if that weren’t enough: there are other, non-market, risks on these trades as well.

The business model for a structured FX business is often pure back-to-back: a local bank (henceforth Tinybank) will print a TARF trade with its client, then turn around and back the TARF out with one of the big global banks (henceforth Megabank London), pocketing an upfront premium for its trouble. In a world of CSAs and potential counterparty defaults, though, there’s still risk on that “riskless” trade.

If Tinybank has a collateral agreement with Megabank London but not with its customer, they can end up in an awkward situation where Tinybank has to post collateral to Megabank with no corresponding collateral coming in from customers on the other side. When all of a bank’s customers are the same way round in a trade, this can add up to a significant amount of collateral calls all falling due at the same time7. If those collateral calls trigger a ratings downgrade, which triggers more collateral calls, which trigger further ratings downgrades, the consequences can be disastrous.

Customers have been selling volatility in search of a yield pickup since options were invented, and the complexity of the trades that they print is only limited by the imagination of the structuring desks at the largest FX-center banks. These complex trades only stop if regulators intervene to block them, or if markets intervene to blow them up – and even then the trades still have to be risk-managed to maturity.

Banks need to manage these trades responsibly, whether they choose to take the risk on or back it out8. The days of managing the risk on spreadsheets and relying on external valuations are over. Banks need to be able to adequately price and risk-manage these trades themselves. They need capital to support the trades: capital to support the funding and collateral demands, capital to pay the recurring costs of dynamic hedging, and capital to fund the hefty decay bills on hedges for the second-order greeks. And they need human capital: traders who understand the embedded risks in structured FX trades and the metastable market dynamics created by these crowded one-way trades.

FX structured product blowups have happened before and will happen again. Banks that participate in the FX structured product market, whether they take the trades into their own books or back them out, have an obligation – to their counterparties, to their employees, to their customers, and to their shareholders – to implement robust risk management systems before the next crisis comes knocking.


  1. Also there’s the risk that the regulators will storm in and fine you and shut your lucrative TARF business down, which has happened to four banks in Taiwan as I write this. [return]
  2. They’re also not a million miles from the “5/30 swap” 17×-leveraged-shortdate-low-delta-spread-option trades that blew up P&G back in 1995. If everything else from the 90s is coming back, I’m thinking it’s time for some opportunistic distressed-asset trades in flannel shirts and X-Files VHS tapes. [return]
  3. A more sophisticated hedge would be to buy low-delta USDCNH call strikes – 6.35s or thereabout – as well as selling the higher-delta 6.15 call strikes. On one hand, this makes the short topside vanna and volga positions much less painful on a move higher in spot. On the other hand, the steep risk-reversal premium for low-delta calls makes the hedge cripplingly expensive to hold: at the vols that prevailed in late January 2014 when these trades were printing, a vega neutral sell 6.15s/buy 6.35s spread leaves you simultaneously short gamma and paying theta. [return]
  4. And the greeks on these trades are titanic in both the small-T and capital-T senses. In late April, Morgan Stanley’s research team estimated that the Taiwanese market alone was short $32 billion of USDCNH delta and short $76 million vega from its TARF positions. Closing those trades out would be the equivalent of one to two weeks’ volume in the USDCNH spot and options markets hitting the street all at once. [return]
  5. And you may ask yourself: “where is that beautiful house?”. And you may ask yourself: “where does that highway go?”. And you may tell yourself “my god, what have I done?”, especially if you’re rapidly getting shorter spot and vol as they both scream higher. David Byrne: FX exotics specialist. [return]
  6. Myself included. Good times; good times. [return]
  7. The collateral issue is especially awkward for TARFs. Because the TARF early-terminates when its PV becomes positive, Tinybank with its one-sided CSAs will never find itself in the happy situation of receiving collateral from Megabank and not having to pass it on to customers. The bank’s funding desk effectively ends up long the TARF and short FX convexity alongside its customers, with even more painful results when the trade goes wrong. [return]
  8. And most banks would prefer to take the trades onto their own books if they can. The margins on TARFs make them a pleasingly cheap source of gamma, and it’s hard to imagine an FXO trader who’d say “no, I don’t want that cheap gamma, I’m going to back the trade out; hello Megabank London can I give you a lump of 3-month 10-delta USDCNH topside half a vol below your bid?” [return]