Table of Contents
In the wake of FTX‘s collapse a lot of people are struggling to understand what happened. FTX’s former CEO has been on an unprecedented media tour following the collapse of the recently-well-respected business he co-founded. And while coverage is voluminous not a lot of useful new information has come out.
One thing is however clear: many folks in the crypto markets do not understand in detail what an exchange is, or how an exchange for levered products is supposed to work.
Following the collapse of an exchange which offered leverage it became painfully clear many users had no understanding of the consequences when a levered product exchange fails. This may even extend to the folks who ran FTX as their odd entanglements with Alameda Research dramatically pushed up the exchange’s risk level.
Margin trading, futures and even puts and calls are nowhere near new. This 17th century book describes a lot of the products offered on FTX. Here we are going to try to break this down a bit and explain what is going on.
A Cash Exchange
The simplest sort of exchange offers one product: cash trading. People enter orders to buy or sell and the exchange matches them up. Everything settles on some agreed convention (i.e t+2 for 2 days post-trade, or t+1 or whatever number) as time marches on. Generally the exchange will not hold anybody’s assets — it merely facilitates settlements among members. Some other party — known as a custodian — holds all the assets and takes instructions to settle.
There might be some insurance fund or margining process among the custodians in case someone defaults on a payment. Those details are not terribly important. What does matter is that baring any defaults nobody’s assets ever go missing. And even if someone fails to deliver the worst that happens is that trade breaks. This is not great — but it is not fatal.
By limiting the size of individual trader’s outstanding positions this risk can be contained. And losses associated with this type of unlevered exchange are exceptionally rare in financial history.
Crypto tends to blend the custodian into the exchange. And this introduces some risk. But, generally, as long as there is no leverage nobody will experience catastrophic losses. The phrase “fully reserved” when applied to this sort of system refers to the custodians. And it simply means the custodians have all the assets they claim to. This is normal.
Brokers Offering Margin Trading
If you have a cash exchange and want some leverage the simplest solution is to arrange it yourself. This is precisely what happens when a broker arranges a margin loan. The exchange still gets full payment — but the broker has some kind of lending relationship with a client.
Here there is more risk. For example the client can run out of money. And there is no foolproof solution to that problem! In practice what happens is the brokers are regulated in two specific ways. First they must maintain a sufficiently large reserve fund (their own capital) relative to the amount of credit they extend. And second their lending book must be sufficiently diversified.
Say a broker has 10 units of capital and is allowed under the rules to make up to 100 units of loans. Under a modern regulatory regime they are not allowed to loan all 100 units to a single fund. And, regulations aside, no well-run business is likely to lend even 10 units to a single client here. Why would you risk your entire capital base on a single client?
In a political scandal related to the 2016 US election some loans were extended amounting to over 20% of a bank’s capital base. And that drew a lot of scrutiny. The bank CEO ended up in jail when the loans were ruled to be essentially bribes.
Here if a client defaults on a margin loan nobody knows about it except their broker. Losses only propagate to the wider system if the broker goes under too. And that is why the regulators require brokers to carry capital and limit concentration in their loan books.
For this setup we can still, somewhat, apply the term “fully reserved” to mean both that the custodians have everything they are supposed to and that the brokers are sufficiently capitalized. Losses are still possible but they should be limited to a failed broker and that broker’s clients.
Finally, the exchange itself can offer leverage. This is a common enough structure in traditional finance. Exchanges such as the CME, LME and EUREX operate almost entirely on this structure. And parts of many other large exchange groups do as well. Here the exchange itself asks members to post margin — and there is the possibility of catastrophic losses.
In practice what happens is the exchange has margin rules for their members. And the membership includes both large individual traders and brokers with their own clients. As these are leveraged positions the phrase “fully reserved” does not really apply. If the exchange was holding enough money to pay out every position in every possible scenario there wouldn’t be any leverage!
Were the price of something to gap a long way there is a good chance some exchange members would be unable to make their required margin payments. This happened in 1987 to the Hong Kong Futures Exchange. It went bankrupt and even the regular Hong Kong stock exchange was closed for several days.
The defense against these losses is threefold. First, the margin requirements themselves are designed to include some buffer. Second, the exchange generally holds some combination of reserve funds and a facility to call up capital from something like an insurance policy (sometimes with members or an insurance company). Finally the exchange can haircut winners to fund a large losers’ losses.
It is this last mechanism that transmits risk throughout the system. Think how a futures exchange works: every buyer has an matching seller. So no matter how large someone’s losses are they must fit inside the gains of all the winners. The key thing to notice is that, functionally, this is the same as the exchange going bankrupt and a court distributing assets to creditors. A winner is owed 100 units. But someone suffered a loss over the margin amount such that the exchange can only pay 70. Whether the winner is haircut and paid now, or collects their 70 in bankruptcy court…they are still getting only 70.
This is a built-in feature of an exchange that offers leverage. FTX offered leverage and, effectively, operated out of a single legal entity. (What FTX really did was worse — it operated out of a web of related companies with no clear policies or procedures.) Consequently when it blew up there was no meaningful way in which it could have been “fully reserved” — and creditor losses are expected.
FTX gave Alameda special non-liquidation privileges. We’ve seen this movie before with the Roger Ver – CoinFlex dispute where certain traders are allowed to run negative balances on the exchange. The tacit assumption is that they will make good before other clients try to withdraw those funds.
It is this feature in particular that caused the epic losses for non-levered traders on FTX. The sort of liquidation waterfalls described above only have a chance of working if they are run quickly. Someone’s balance hits 0? Seize their margin, liquidate their positions, use whatever backstops are required and so on down the line until the issue is closed.
But you can only haircut winners to covers losers losses when those losses come from open positions. If you persistently hide losses it is easy to end up in a situation where even seizing all open position profits does not bring in enough money. Then you need to claw back profits from prior trading. Where does that money sit? It sits in trader’s cash and token balances. This is where FTX’s client funds went.
Note this is almost exactly what happened in Hong Kong in 1987.
Reducing the chance of these sorts of losses is a key function of a regulator. But that does not mean regulations will eliminate them. NASDAQ’s Nordic power exchange suffered massive losses following a member default in 2018. In that case the parent company had sufficient funds to plug the hole. But the risk of a collapse and cascading losses was real. And there is zero chance NASDAQ would have made the exchange whole if a persistent months-long fraud was responsible for systematically hiding losses.
FTX, knowing what we know now, was not a well-run exchange. And better procedures would have led to a better outcome for the exchange’s members. But there is no magic “fully reserved” formula that fixes this problem once and for all.
This is part of the reason many jurisdictions impose extra requirements for people to trade with leverage. Not only might their own trading lead to large losses — the rest of the market infrastructure will necessarily expose them to new risks.
So long as the crypto markets demand exchanges provide leverage directly to end users everyone’s on-exchange funds will be at risk. This may not be the way some traditional exchanges work — but it is a standard, expected, outcome for derivatives exchanges.
We are watching an age-old fight between demands for leverage and demands safety. Watch this space.