How fast should trades be settled on financial markets? If I sell a stock now, how long does it take until I can use the cash proceeds?
Until recently, it could be several days.
In September 2016, the Securities and Exchanges Commission (SEC) proposed to shorten the U.S. markets' settlement cycle from three to two days. In a market where high frequency traders update quotes every nanosecond, talking about a post-trade infrastructure where delays are measured in days feels like a different era altogether. Trades need to follow a long route, through a central counterparty, custodian banks, and a central securities depository.
Blockchain (or, generally, Distributed Ledger Technology or DLT) is poised to change the status quo and bring post-trade infrastructure up to speed. DLT uses a distributed messaging protocol (a "ledger") to create consensus across all counterparties and maintain a unique, authoritative shared record.
Regulators agree that with Blockchain, long settlement chains are obsolete (see, for example, some recent reports from European Central Bank, ESMA, or the Fed Board). The decentralized nature of Blockchain allows trades to be validated through market consenus almost instantaneously: that is, immediate settlement.
Does this mean we should settle trades immediately? Not necessarily. Fredrik Voss (VP of Blockchain innovation at Nasdaq) envisions a world where
we [..] allow participants to select the pace at which they want to settle, which has been challenging to do in the market today.
More anectodal evidence that immediate settlement is not always best comes from Moscow, where a reform to implement same-day settlement has been reversed in March 2013: trades now settle in two days, like in most places around the world.
In a new study with Mariana Khapko (University of Toronto), we take a fresh look at the economics of settlement times. The main market imperfection, in our view, is not the length of time-to-settlement per se as much as its rigidity (so far, due to institutional and technical requirements). What Blockchain and DLT bring to the table is the option of shorter settlement times, rather than simply faster settlement.
There are two main channels we focus on. First, a long time-to-settlement means that a trader is exposed to counterparty risk for an extended period of time -- and might not actually receive the trade proceeds on the settlement day. However, at the same time longer times-to-settlement could help liquidity. Forcing traders, especially market-makers, to settle immediately is equivalent to asking them to hold (risky) inventories to settle against. If settlement times are longer, market-makers can act as intermediaries and match buyers and sellers without having to hold large inventories. Exactly how large settlement time should be depends on -- potentially time-varying -- (i) counterparty risk and (ii) search costs in a given market. Having the option to adjust time-to-settlement in time and across assets is therefore valuable.
Who should exercise the settlement time "option" introduced by Blockchain? One way would be that the exchange sets a fixed time-to-settlement for all trades (in a given security, or on a given day, or both). Another way is to build a "3-D" order book, where traders specify prices, quantity, and preferred times-to-settlement.
We find the first option yields the largest welfare. In our model, if market-makers are allowed to choose times-to-settlement, they choose to specialize. Some market-makers offer expensive, but fast settlement, whereas other market makers offer cheap contracts, but with long settlement times and high counterparty risk exposure.
Having the option to propose trades with different levels of counterparty risk allows market-makers to relax price competition and earn excess profits. Moreover, such rent-seeking behaviour is more pronounced when counterparty risk is already high, as that widens the scope for specialization. Conversely, if the exchange sets a unique time-to-settlement, it stimulates price competition between market-makers.
Whether and how large an impact Blockchain technology will make on financial markets remains an open question. The overarching message of our study is that Blockchain's main value for trade settlement is its flexibility. Market design choices influence which trading stakeholders are best positioned to take advantage of such flexibility -- with important consequences for liquidity and market quality.
P.S. The paper is available [here].
A famous Monet, worth at least $1 million, is being auctioned off. Imagine first a naïve owner, in the following sense: He sells the painting to the first bidder offering more than $1 million. There is no compelling reason to bid more. Rather, potential bidders would compete to be "fast:" that is, buy the most powerful car and be the first one to knock on the sellers' door. It is a perfect world for car sellers, not so much so for art collectors with a need for cash.
Imagine now the Monet owner is somewhat savvier. He schedules a one-day exhibition: potential buyers may come, see the painting, make a bid to the seller, and leave. At the end of the day, the seller compares all bids and chooses the best one. The outcome is a more competitive world for buyers: In all likelihood, the highest bid for the painting will exceed $1 million. At the same time, the demand for powerful cars goes down: no need to be the first one to submit your bid, just to bid the best price.
If you had a Monet to sell, which of the two worlds would you like to live in?
The first setup is a metaphor for incumbent continuous-time limit order markets. The art buyers stand in for high-frequency traders (HFT) and the Monet is an arbitrage opportunity: say, a stale quote. On limit order markets, HFT arbitrageurs earn high profits as they do not compete in prices. What does matter for them is to be the fastest: Hence, a costly arms' race emerges to own the most advanced trading technology (Budish, 2015, QJE).
The alternative is discrete-time trading, or batch auction markets. Such markets clear at fixed intervals (say, 100 milliseconds) via an auction. HFT arbitrageurs who observe a profitable opportunity may submit their bids before the market clears. As Monet buyers in the previous example, they will outbid each other and as a consequence earn lower profits. Lower profits for HFT arbitrageurs implies lower losses for the (HFT) liquidity provider with the outstanding stale quote. As discrete-time trading depresses the cost of providing liquidity, spreads go down and liquidity improves.
The argument is compelling. So compelling in fact that in October 2015 the Securities and Exchanges Commission granted approval for the Chicago Stock Exchange to launch a batch-auction platform, CHX SNAP. The London Stock Exchange is also preparing to test-pilot a midday auction in March 2016. Regulators and exchanges are taking active steps to improve the market design.
In a joint paper Marlene Haas, we aim to pin down both the advantages and the "hidden costs" of discrete-time markets in a world of HFTs.
We argue that three factors influence, in a non-trivial manner, the intensity of HFT competition and therefore the magnitude of the liquidity improvement: the interval between consecutive auctions (how long does the Monet exhibition last?), the HFT speed to process new information (how fast are the bidders' cars?), and the number of HFTs (how many potential bidders?)
A first conclusion of our study is that on a batch auction market, the HFT arms' race improves market quality. If our potential Monnet buyers have faster cars, more of them make it to the exhibition. More bids stimulate price competition, pushing down profits for arbitrageurs and therefore the opportunity cost of providing liquidity. The result is, at a first glance, surprising: Previous research found that such "race to the bottom" among HFTs does not improve the quality of limit order markets. We bring a new insight: If the arms' race cannot be curbed by regulators, discrete time markets could be a clever way to channel it towards price competition and better liquidity.
So far, batch auctions sound like an ideal market design. There is, however, a caveat.
We argue that, despite the enhanced price competition, batch auction markets do not improve liquidity under all circumstances. This is due to trade rationing, a consequence of dispensing with time priority and the queue from limit order markets.
To understand the mechanism, consider three HFTs competing to provide liquidity to a trader who, in expectation, only needs two units of the asset.
In a limit order market, two of the three HFTs would have posted quotes first and captured the liquidity demand. The third one "lost the race," so he might as well quit the market. The two winners trade one unit each.
In the auction market, however, arrival order becomes irrelevant. All HFTs post a quote, and consequently trade two-thirds of the asset in expectation. Excessive liquidity supply leads to trade rationing: profits from providing liquidity are lower than on the limit order market and HFTs compensate by setting a higher bid-ask spread. Is that enough to cancel the advantages of the discrete-time markets? Our study shows it can indeed be the case, if HFT competition is not strong enough.
The financial industry is starting to take batch auction markets seriously, with several such exchanges scheduled to draw their curtains soon. Their optimal design is still very much an open question. To promote liquidity, regulators should stimulate HFT competition, and not discourage technological investments. In discrete-time markets, competition in speed promotes competition in prices -- and better liquidity for investors.