The Swiss Franc Shock of 2015: How the SNB Floor Removal Created Negative Balances and Broke Brokers

On January 15, 2015 the Swiss National Bank (SNB) abandoned the minimum exchange rate of 1.20 Swiss francs per euro that it had enforced since September 2011. In the minutes after the 10:30 announcement in Zurich, EUR/CHF fell to an all-time low of about 0.85, a move of roughly 29%, and thousands of leveraged traders were closed out at prices far below their stop losses. Many ended the day owing their brokers more than their accounts had ever held, and some brokers did not survive.

In short: a currency floor looks like a free option to a leveraged trader, because the central bank appears to promise that the price cannot go lower. When the promise is withdrawn without warning, there are no buyers at the old level and a stop loss becomes a market order into an empty book. FXCM’s clients lost their $80 million of collateral and generated about $275 million of negative balances, which FXCM had to cover with a $300 million emergency loan at 10% interest rising to 20.5%. Alpari (UK) entered special administration four days later. The episode is the clearest public case study of gap risk, and of how a retail loss travels up the chain to brokers, prime brokers and banks.

This article sits alongside leverage and margin in forex, which covers the margin mechanics, and how retail forex brokers hedge, which covers the business models that determined who absorbed the losses.

The primary sources

Almost everything below comes from documents written by the people involved:

The SNB: its press release of January 15, 2015, Chairman Thomas Jordan’s remarks at that day’s press conference, Vice Chairman Fritz Zurbrügg’s speech of March 26, 2015, and the Annual Reports for 2014 and 2015.
FXCM’s SEC filings: the 8-Ks and press releases of January 2015 and the 10-Ks for 2014 and 2015.
Alpari (UK)‘s administration records at Companies House: KPMG’s Statement of Proposals (March 2015) and the final progress report (May 2018).
Regulators and the BIS: the FCA’s notice to Alpari customers, the CFTC and NFA actions of February 6, 2017, NFA Notice I-15-04, ESMA Decision (EU) 2018/796, the BIS Quarterly Review of March 2015, and “Downsized FX markets: causes and implications” by Michael Moore, Andreas Schrimpf and Vladyslav Sushko (BIS Quarterly Review, December 2016).

How the floor worked, 2011 to 2015

In August 2011 the franc was surging as investors fled the euro area crisis. The SNB first flooded the money market with francs, expanding banks’ sight deposits from CHF 30 billion to CHF 200 billion in a series of steps that August. When that did not hold the currency, it announced on September 6, 2011 that it would “no longer tolerate” a EUR/CHF rate below 1.20, and enforced the floor with foreign currency purchases.

Defending a floor is mechanically simple and financially expensive. Every time sellers pushed EUR/CHF towards 1.20, the SNB bought the euros they were selling, paying in newly created francs. Zurbrügg’s March 2015 speech gives the scale: CHF 188 billion of foreign currency purchases in 2012 alone, taking reserves from CHF 50.5 billion in 2007 to CHF 432.2 billion at the end of 2012. By December 2014, according to the BIS Quarterly Review of March 2015, the SNB’s balance sheet had grown to 87% of Swiss GDP.

Pressure built again in late 2014 as markets anticipated quantitative easing from the European Central Bank. On December 18, 2014 the SNB introduced a negative interest rate of -0.25% on large sight deposits to discourage franc holdings. In its 2014 Annual Report the SNB says that pressure “increased very significantly during the first half of January 2015” and that “the amounts spent on interventions were extremely high.” It concluded that keeping the floor would have required “an uncontrollable expansion of the balance sheet, to a level that might even have been several times Swiss GDP.”

Why a floor attracts leveraged traders

From a trader’s point of view, a credible floor seems to cut off the downside. Buying EUR/CHF at 1.2010 with a stop loss at 1.1980 looks like risking 30 pips to make whatever the euro might recover. That is an easy trade to pitch, and it invites high leverage: if the price cannot fall below 1.20, why not size up?

The flaw is that the floor was a policy, not a law of nature. The SNB itself called it an “exceptional and temporary measure.” A stop just under a policy floor bets that the policy will hold or end gradually. Zurbrügg explained in March 2015 why it could not end gradually: “An announcement, or even the slightest hint of an imminent decision to dispense with the minimum exchange rate, would have triggered massive speculation against the SNB.” The exit had to be a surprise, which meant that every stop under 1.20 was exposed to a gap.

January 15, 2015

The press release was short. The SNB was “discontinuing the minimum exchange rate of CHF 1.20 per euro.” At the same time it cut the rate on sight deposits above the exemption threshold by 0.5 percentage points to -0.75% and moved the target range for three-month Libor to between -1.25% and -0.25%. The reason given was that “divergences between the monetary policies of the major currency areas have increased significantly.” At the press conference Jordan added that the floor was discontinued “with immediate effect” and that the SNB would “cease foreign currency purchases associated with enforcing it.”

What happened next is described in the SNB’s 2015 Annual Report: “In the minutes after the announcement, the CHF/EUR exchange rate fell to an all-time low of CHF 0.85 per euro. However, the euro was rapidly able to recoup some of its losses and the pair settled close to parity in the afternoon of 15 January.” By the end of January it was back around 1.05. The BIS reported that the main Swiss stock index fell almost 9% on the day and that EUR/CHF trading volume on EBS hit levels not seen since May 2010.

The timing was deliberate. Zurbrügg said a European weekday morning let participants “start from a level playing field,” whereas a weekend announcement would have hit the thin Australian session first. He acknowledged “some bottlenecks immediately after the announcement.”

Why stop losses did not protect anyone

A stop loss is an instruction to sell at the best available price once a trigger level trades. It is not a guaranteed exit price. On January 15 the first prices below 1.20 were far below it, because the natural buyer at 1.20, the SNB, had left. In FXCM’s words, “liquidity became extremely scarce and shallow, which affected execution prices.” Stops triggered and were filled at whatever bids existed, often far below the stop level. Brokers that marked positions to their liquidity providers’ prices passed those fills straight through.

Worked example: one retail account through the gap

The figures below were computed in Python. The account is hypothetical; the price levels are the SNB’s.

A trader with CHF 20,000 of equity buys EUR 500,000 of EUR/CHF at 1.2010, with a stop loss at 1.1980. To keep the arithmetic in one currency, the account is denominated in francs, the quote currency of the pair.

Notional value: 500,000 x 1.2010 = CHF 600,500, or 30 times equity (the 30:1 limit that now applies to retail CFDs on major pairs in the EU and UK).
Planned risk: 500,000 x 0.0030 = CHF 1,500, or 7.5% of equity.
The fill that wipes out the account exactly: 1.2010 - 20,000 / 500,000 = 1.1610, a move of just 3.33%.

Any fill below 1.1610 leaves the trader owing money:

Fill price	Move from entry	Loss (CHF)	Account balance (CHF)	Loss as multiple of equity
1.1980 (the stop, as planned)	0.25%	1,500	18,500	0.1x
1.1500	4.25%	25,500	-5,500	1.3x
1.1000	8.41%	50,500	-30,500	2.5x
1.0500	12.57%	75,500	-55,500	3.8x
1.0000 (near the afternoon level)	16.74%	100,500	-80,500	5.0x
0.8500 (the intraday low)	29.23%	175,500	-155,500	8.8x

The stop changed the intended exit, not the size of the position. A trader who thought they were risking CHF 1,500 could owe several times their deposit: at 30:1 a 3.3% gap is fatal, and the actual gap was five to nine times larger.

The US rules of the time would have allowed even more leverage on this trade. NFA Financial Requirements Section 12 set the minimum security deposit for the Swiss franc, one of ten “major” currencies, at 2% of notional, or 50:1. On January 21, 2015, NFA Notice I-15-04 raised the Swiss franc requirement to 5% (20:1) with effect from the next day. On this position that is the difference between a CHF 12,010 and a CHF 30,025 minimum deposit.

FXCM: an agency broker that still lost $275 million

FXCM, listed on the New York Stock Exchange, built its retail business on what its 10-K called an agency model: “when a customer executes a trade with us, we act as a credit intermediary, or riskless principal, simultaneously entering into trades with the customer and the FX market maker.” The model is meant to remove market risk. January 15 showed that it does not remove credit risk.

In a press release dated January 15, filed with the SEC on Form 8-K the next day, FXCM announced that clients had “generated negative equity balances owed to FXCM of approximately $225 million,” and that the company “may be in breach of some regulatory capital requirements.” The 10-K for 2014, filed in March, gave the full picture:

“At the time of the SNB announcement over 3,000 of our clients held slightly over $1 billion in open positions on EUR/CHF. Those same clients held approximately $80.0 million of collateral in their accounts.”
The move “wiped out the account equity of those clients as well as generated debit balances owed to us of $276.0 million.”
The mechanism: “While clients could not cover their margin call with us, we still had to cover the same margin call with our banks.”

$1 billion of positions on $80 million of collateral is 12.5 times leverage, and collateral of 8% of notional. Losses of about $356 million (the $80 million of equity plus $276 million of debit balances) are 35.6% of notional, larger than the 29.2% peak-to-trough move in EUR/CHF itself. FXCM did not break the positions down, so the gap cannot be decomposed precisely, but one contributor is clear: the losses were in francs, which the SNB notes “initially appreciated almost as much against the US dollar as against the euro,” so each franc of loss cost more dollars. Across 3,000 clients the debit balances average roughly $90,000, though they were highly concentrated, as shown below.

The Leucadia rescue

Regulators required FXCM’s regulated subsidiaries to restore their net capital “on an expedited basis.” On January 16, 2015 FXCM signed a credit agreement with Leucadia National Corporation (the parent of Jefferies). The terms, from the January 19 press release and the 2015 10-K:

A $300 million, two-year term loan, with net proceeds of about $279 million.
Interest of 10% a year, rising by 1.5 percentage points every quarter while the loan was outstanding, capped at 20.5% in the final agreement (the initial release said 17%).
A $10 million deferred financing fee (a further $30 million was avoided by early paydown), and monthly prepayments out of any money collected from clients’ debit balances.
A side agreement giving Leucadia a share of future distributions and sale proceeds. After the loan and fees were repaid, Leucadia would receive 50% of the next $350 million, 90% of the next $500 million, and 60% of everything after that, with the right to force a sale of the business from January 2018.

The arithmetic shows who sets terms in a weekend rescue. Had the full $300 million stayed outstanding for two years, the rate path (10%, 11.5%, 13% and so on up to 20.5%) averages 15.25% and would have cost about $91.5 million in interest plus the $10 million fee. The equity kicker was larger still. On a hypothetical $1 billion of proceeds after repayment, Leucadia’s tiers give it $175 million plus $450 million plus $90 million, or $715 million (71.5%), against $285 million for FXCM’s shareholders. FXCM’s own 10-K said the Leucadia deal “was the only financing that we were able to arrange in the very short timeframe we were given by the regulators.”

Forgiveness and recovery

On January 28, 2015 FXCM announced that it would forgive about 90% of clients with debit balances in certain jurisdictions. It continued to pursue the rest: institutional, high-net-worth and experienced traders who made up about 10% of the clients but “over 60% of the total debit balances owed.” It later extended forgiveness to about 97% of clients. By December 31, 2015 it had recovered $9.8 million, or 3.6% of the $275.1 million debit balance figure in the 2015 10-K, and booked a net loss after recoveries of $265.4 million. The 10-K added: “We do not expect any further recoveries.”

FXCM also removed currency pairs it believed carried “significant risk due to over active manipulation by their respective governments either by a floor, ceiling, peg, or band,” naming among others USD/DKK, EUR/DKK, EUR/PLN, EUR/CZK, USD/SGD and HKD/JPY, and raised margin requirements.

The 2017 enforcement actions

FXCM’s US business ended on February 6, 2017, for reasons unrelated to the franc but central to its agency pitch. The CFTC found that from September 2009 to at least 2014 FXCM told retail customers its “No Dealing Desk” platform had no conflict of interest, while it “had an undisclosed interest in the market maker that consistently ‘won’ the largest share of FXCM’s trading volume,” a firm that shared most of its trading profits with FXCM. FXCM, FXCM Holdings and founders Dror Niv and William Ahdout were ordered to pay a $7 million civil penalty and to withdraw from CFTC registration. On the same day NFA barred FXCM, Niv, Ahdout and Ornit Niv from membership. The conflict between a broker and its clients is the subject of forex broker conflicts of interest.

Alpari (UK): from solvent to special administration in four days

Alpari (UK) Limited was an FCA-authorised provider of CFDs and spread bets on currencies, indices and commodities, with 166 employees. KPMG’s Statement of Proposals records the sequence:

Thursday, January 15: “many of the Company’s clients sustained losses on their CFD, foreign exchange and spread betting transactions with the Company.”
Friday, January 16: KPMG and the law firm Ashurst were engaged. The weekend was spent trying to sell the company. The efforts “were ultimately unsuccessful.”
Monday, January 19: the board resolved to appoint Richard Heis, Mark Firmin and Samantha Bewick of KPMG as joint special administrators under the Special Administration Regime for investment firms, with the consent of the Bank of England.

The FCA told customers that Alpari had “more than 100,000 customers” and that its initial view was “that the client money is whole.” According to the Statement of Affairs cited by KPMG, about £5.7 million was owed to Alpari’s prime brokers in respect of losses: the same pass-through mechanism FXCM described, landing one level higher.

The final progress report shows how it ended. Agreed client balances totalled $95.8 million. Clients received two distributions from the client money pool totalling 82 cents in the dollar, the second (27 cents) paid in mid-2017. The shortfall ranked as an unsecured claim, and unsecured creditors received 43.54 pence in the pound in total, the last dividend paid in February 2018. The special administration closed in May 2018. The FCA had told customers that, depending on their circumstances, they might also have access to the Financial Services Compensation Scheme for client money losses. The point for a retail trader is that even “segregated” client money is only as good as the firm’s records and the insolvency process: the final client money distribution came more than two years after the collapse, and the administration took more than three.

How the loss travelled up the chain

Think of it as a chain of credit, each link exposed to the one below:

The client owed the broker the negative balance.
The broker owed its banks or liquidity providers the loss on its hedge whether or not the client paid, as FXCM’s 10-K described.
The prime broker that gave the broker access to those banks had to settle with the executing banks regardless of whether the broker paid. Alpari’s £5.7 million debt to prime brokers is the visible example.

Each link held capital or collateral sized for ordinary days. None was sized for a 29% move in minutes. That is why the BIS later wrote that the move “sent shockwaves through the prime brokerage industry, causing prime brokers to raise fees and cut clients,” and that some dealer banks “scaled down their business exposure to retail brokerage platforms,” pushing retail brokers into “prime-of-prime” arrangements, where a non-dealer firm stands between them and the bank. The mechanics are covered in FX prime brokerage explained.

What changed afterwards

US margin. NFA’s January 2015 notice raised the Swiss franc security deposit to 5%. NFA Notice I-16-27 (November 28, 2016) cut it to 3% from December 5, 2016, citing lower CME and ICE margins on franc futures. NFA’s Executive Committee adjusts these deposits over time, so check its current notices.
Broker product lists. FXCM dropped pegged and managed currencies and raised margins.
EU negative balance protection. ESMA Decision (EU) 2018/796 of May 22, 2018 restricted retail CFDs from August 1, 2018: leverage limits starting at 30:1 for major currency pairs, a margin close-out at 50% of required margin on an account basis, and negative balance protection on an account basis, which ESMA described as a backstop that provides “an overall guaranteed limit on retail client losses.” National regulators later made these measures permanent.
US retail clients remain in a different position: US rules set minimum security deposits but do not mandate negative balance protection of the kind ESMA imposed, so what happens to a deficit depends on the customer agreement. Offshore brokers marketed by signal groups may promise protection that no regulator enforces; see offshore forex brokers explained.

How professionals read the Swiss franc shock

A peg is a policy decision with a gap attached. Professionals size positions in managed currencies on the move that follows the policy’s end, not on day-to-day volatility. Measured by its recent day-to-day volatility, which the floor itself suppressed, EUR/CHF looked like a low-risk pair.
A stop loss caps the intended loss, not the actual one. Risk-per-trade rules that assume the stop is filled, which is how almost every signal service describes its risk, break down exactly when they are needed. The general problem is covered in forex signal risks explained.
Scenario size, not leverage ratio, is the useful number. The question is “what do I lose if this pair gaps 20%?” At 30:1 the answer is about six times the account; at 5:1 it is about the whole account.
Know who stands behind your account. The FXCM and Alpari outcomes depended on broker capital, insolvency regimes and client money rules more than on anything the client did.
Central banks surprise when the alternative is worse. The SNB judged a hint more expensive than a shock.

The short version

From September 2011 to January 2015 the Swiss National Bank held EUR/CHF at or above 1.20 by buying foreign currency on a scale that took its balance sheet to 87% of GDP. On January 15, 2015 it stopped, deliberately and without warning, and EUR/CHF fell to about 0.85 within minutes before settling near parity. Stops under 1.20 were filled far below their levels. FXCM’s clients, holding $1 billion of positions on $80 million of collateral, ended up owing about $275 million; FXCM survived on a $300 million Leucadia loan at 10% rising towards 20.5%, plus a large share of its future value, and recovered less than 4% of the debts. Alpari (UK) entered special administration on January 19, and its clients eventually received 82 cents in the dollar from client money. The lessons are about gap risk, the limits of stop losses, and the chain of credit behind every retail account. This is an account of what happened, not a trading recommendation.