Trade offs and potential impact on market stability - Frequent Batch Auctions, Part III
Mark-to-market
Let’s take a checkpoint. In trading, you must “always know your position”.
In my previous posts in this series, Hold on… this doesn’t make sense - Frequent Batch Auctions, Part I and The $5 billion heist that doesn’t exist - Frequent Batch Auctions, Part II, we’ve examined the fuzzy math and illogic of recent academic research suggesting $5bn annual investor harm in the form of “latency arbitrage tax” and other anti-social maladies [FCA Insight “Big bucks from small change,” and Occasional Paper No. 50 Quantifying the High Frequency Trading ‘Arms Race’: A new methodology and estimates (Aquilina, O’Neill and Budish, Jan 2020) and related papers (collectively, the “FBA Papers”)].
We established that FBA Papers have yet to prove this sloppily constructed figure, that this problem exists nor its true victims. In attempting to solve for “computation simplicity advantages,” a problem for seemingly nobody, the FBA Papers proposed a partially-baked technical feature as a wholesale substitute to our robustly functioning, interconnected, multi-asset global market structure. Despite acknowledging nearly 5 years ago that “an analysis of the effect of frequent batch auctions on market stability… is an important topic for future research,” the FBA Papers’ authors have made no attempt to contribute to that research, nor even envision how this might work for practitioners across interconnected financial instruments and market centers.
Instead, they’ve coined a term, “latency arbitrage tax,” and published a headline grabbing $5b cumulative global figure for it to recapture our attention. They extrapolated that figure using:
5-year old data from one venue with 50% of UK equities’ volume,
a lot of really complex-looking equations, and
a lot of AWS computing power
Continuing on, we have established, using the FBA Authors’ observations, that this hypothetical $5b figure is likely netted closer to zero, with pounds and pence traded amongst the firms competing to maintain razor thin profitability at top of book. Third-party research estimating total HFT profitability (inclusive of all trading strategies) is much lower than the FBA Papers estimate of “latency arbitrage sniping tax,” thus calling into doubt Occasional Paper 50’s methods and conclusions. The FBA Papers’ authors chose not to sanity check their estimates against other industry research in order to explain their hypotheses and conclusions. But they continue to publish and proselytize at universities and on YouTube.
That’s where we’re at.
So if you’re as unconvinced as I am, you can choose to read on or cut your losses. If you choose to read on, I’ll focus the much delayed conversation on “the effect of frequent batch auctions on market stability… an important topic for future research.” (emphasis added)
What would the FBA/anti-CLOB future look like?
Now seems like and appropriate time to question the impact of eliminating continuous trading in CLOBs and replacing that market structure with FBAs:
Could a community of market makers make money without a bid/ask spread?
Could market makers keep tight markets across instruments and market centers if matching were queued and pulsed like a lighthouse?
Who takes the other side of the trade in a rapidly adjusting market with no spreads and delayed execution?
Would volatility be dampened or amplified, particularly while the market absorbs new information and firms and algos wait for it to be reflected in the tape or the fills to come back?
Would market centers consolidate either by mandate or as a result of this structure?
What else could coexist alongside FBAs, if continuous matching via CLOBs is replaced?
What’s the trade? What are we giving up to resolve this one problem? How does it all work for everyone? Is it worth it?
None of these questions are addressed by the authors. That’s because there is no free beer.
Discrete time makes it technologically possible to disseminate public information symmetrically.
- Budish, Cramton Shim, November 2015
This is true, but it is only part of the story. Discrete time market structure does not guarantee that symmetrically disseminated information will be received, processed, and acted upon without unintended side-effects to exchange platforms and liquidity-providing intermediaries. It assumes primacy for computational simplicity via symmetrical matching and dissemination. While we aren’t convinced that technologically-driven asymmetry is a material problem, common-sense suggests that discrete time FBAs are only one potential solution. However if we are to assume FBAs to be the solution, we are asked to assume a lot, and to potentially give up even more.
Practitioner feedback on FBA’s impact on market stability
Industry experts identified the weaknesses of this position as early as Oct 2014, shortly after Eric Budish, associate professor of economics at the University of Chicago Booth School of Business, presented his idea at the European Capital Markets Institute’s Annual Conference in Brussels.
Comments in a Markets Media article by Terry Flanagan included:
Mark Hemsley (former Cboe Europe CEO): said it is unclear how a discrete batch auction exchange would function alongside continuous time exchanges and that it is unclear how a frequent batch auction equity exchange would interact with a derivatives exchange.
Sviatoslav Rosov, analyst in the capital markets policy group at CFA Institute: “To get around the asynchronicity of price discovery between exchanges, all exchanges would have to become frequent batch auctions and have the auctions perfectly synchronised for the discrete-time auction model to work, in its strictest interpretation”
The article continued, “According to Rosov, Budish argued that any frequent batch auction exchange would only be responsible for eliminating latency arbitrage on its own venue.”
- “Batch Auctions Could Combat HFT”, Markets Media, Nov 2014
If we cannot solve the problems summarized by Hemsley and Rosov, then the solutions aren’t rooted in reality. To solve latency arbitrage on a single venue, or to mandate that all venues resolve latency arbitrage themselves, is to ignore how markets work. Nearly all major stock indices have equity index futures and options that track the indices, all of which are traded on venues separate from the underlying stocks and index tracking ETFs. Within the USA, equity derivatives and underyling securities fall under different regulators. In most developed markets including the G7/EU countries, underlying stocks and ETFs are traded on several competitive venues, in multiple national jurisdictions, in different currencies (also on fragmented venues), and under different regulatory regimes. The chances of regulatory symmetry working alongside discrete time symmetry are not strong and hardly imaginable.
What we have to assume in order for FBAs to be feasible
To this date, the FBA authors make no attempt to provide a plausible hypothetical future-state of the “FBA/anti-CLOB,” thus we are left to assume the following are feasible and/or desirable:
We are to go back to a pre-MiFID I (2007) concentration rule and abolish competition
We can ignore the effects between underlyings, their derivatives and other fungible instruments traded at the same location or elsewhere
Equities, ETFs, DRs
Index Futures and Options
Commodities and ETPs
Currency Pairs (e.g. GBP/USD - EUR/USD - EUR/GBP - EUR/JPY - USD/JPY…)
Somehow market centers will all operate coordinated FBAs and compete, differentiate with net-positive effects to the trade lifecycle across interconnected asset classes
Beyond these implausible assumptions, we are told with new FBA/anti-CLOB Exchange spreads at zero, immunized from the spread-widening effects of stale-quote sniping, that the market will somehow also be deeper (Section VII, Budish, Cramton Shim, November 2015).
Following on, we are commanded to believe that with the arms race resolved via FBAs, a breadth of liquidity providers can function in such an environment. That they can operate in an enhanced way without spreads, dutifully delivering their social usefulness without differentiation or a source of profit, willing to bear the fixed and variable costs of doing so.
We emphasize that our results do not imply that on net HFT has been negative for liquidity or social welfare. Our results say that sniping is negative for liquidity and that the speed race is socially wasteful. Frequent batch auctions preserve (in a sense, enhance) the useful function served by HFTs—liquidity provision and price discovery—while eliminating sniping and the speed race.
- Budish, Cramton Shim, November 2015
“Enhanced?” Nope. As we discussed in Part II of this series, the LP doesn’t earn a living without a positive bid-ask spread and a trade to capture that spread. The more frequently they can round-trip their trades, whether to capture the spread or reduce losses, the more they can provide “socially useful” liquidity. FBAs limit the frequency that LPs can trade, the choice of timing in which LPs can do it, and eliminate their spread.
Don’t believe me, listen to the words of Johannah Ladd, former Secretary General of the FIA EPTA trade group, who represents European Liquidity providers.
“Under such a market design, liquidity providers would be forced to delay hedging until the next batch auction occurs,” Ladd added. “This longer holding period, even if measured in milliseconds, increases inventory risk as market participants cannot immediately hedge new inventory.”
- Source: “Batch Auctions Could Combat HFT”, Markets Media, Nov 2014
What Ladd describes is similar to trying to keep a multi-car caravan going in a congested city versus on a highway with no traffic lights. Even the most skilled drivers fail to "hold it together" as traffic lights cause part of the group to pause while others make it through the cycle of the batch. It's much easier to keep things together with continuous driving on a motorway. When the group breaks up, it's hard to know if everyone is continuing on with the plan, or if someone has pulled off the road for whatever reason. You don’t “know your position”.
The same is true for the multi-asset LPs who rely upon hedging in continuous markets to keep traffic flowing; in the FBA, the strategy breaks down should they have to wait to hedge two or three instruments and/or currencies, and those marks move while they wait to execute the transaction. In such scenarios, overall liquidity and interplay would diminish across the spectrum of instruments and marketplaces.
We’ll once again revisit a conclusion from the FBA Papers:
“The research indicates that ordinary households are not significantly impacted by the costs of this activity in their retirement and savings decisions.... flawed market design significantly increases the trading costs of large investors, and generates billions of dollars a year in profits for a small number of HFT firms, who then have significant incentive to preserve the status quo.”
- Aquilina and O’Neill, FCA Insight, “Big bucks from small change”, Jan 2020
These guys are literally all over the place.
But I would have to agree that status quo beats assured mayhem.
Discussion points: Which is it? Are HFTs good or bad? Costly profiteers or not? Where’s the profit and what did you do to confirm it? Do a small number of trading firms really want to preserve a negative sum arms race? Or do the promoters of this idea just have an agenda, and struggle to unify on a message while they fail to do the analysis to support their conclusion? And do you really believe, as Budish argued, that any frequent batch auction exchange would only be responsible for eliminating latency arbitrage on its own venue is workable?
I need a beer.
Would extreme volatility be worse in FBA-driven markets?
Looking at March 2020 COVID-19 + Super Tuesday + oil shock volatility, we can imagine that FBA/anti-CLOB lighthouse batched auctions would have frustrated market making and hedging in equities, options, futures and options.
Discussion points: On a week like this one below, who would have taken the other side of the trade in these fast moving markets with FBA/anti-CLOBs? How would liquidity providers hold themselves out or hedge across markets and instruments if unable to control when they are able to do so? What’s going on with spreads in Options, Futures, the 500 underlying stocks, similarly-composed ETFs, etc?
Are we expecting LPs to hold our beer while the market moves against them?
While this was a rough period, we can all agree that the markets were resilient and orderly. Limits and trading halts curbed further volatility swings, and while the markets were active and volatile, market structure is not a complaint being alleged. Not in the equities markets, at least.
Given what I’ve seen, and what Hemsley, Rosov and Ladd observed in 2014, I struggle to envision how price discovery would work, and how market makers operating exclusively on FBA markets could deliver two-sided liquidity within and across asset classes. Others way smarter than me should chime in on this, or perhaps the FBA authors themselves owe us that.
What would we be giving up in this future?
Let’s examine one assumed hypothetical future wherein one market center conducts FBAs. This is the most “equivalent” to the FBA Papers’ proposal and Occasional Paper’s analysis. In order to understand the potential implications, we need to appreciate what we’ve gained since MiFID I in 2007 when competition among marketplaces was first introduced in Europe.
Since that time the London Stock Exchange, like many primary exchanges across Europe, responded to competition in several meaningful ways, including:
Better, more resilient trading platforms (e.g. TradeElect replaced by MilleniumIT)
While Millennium IT and other global systems are not immune to disruption, such occurrences are far less frequent and shorter in duration than before competition. While the press coverage of the industry typically sensationalizes ever decreasing median response time (milli > micro > nano-seconds), the true benefits enjoyed by the industry are improvements in (measured by):
Throughput (more orders per second)
Functionality (more order types, order books, cross-order book interaction, market structure matching logic variations)
Scalability and consistency (smaller standard deviation of response time)
Stability (fewer outages)
Ease of implementation (common specifications, such as FIX, ITCH and OUCH)
By all measures, we are better off than before. We have more choices, which include market structures that do not deploy price-time priority. More anon.
Lower transaction costs throughout the trade lifecycle (execution and clearing)
The table below compares the Execution + Clearing Costs in basis points for 2007 (pre-MiFID I) and today in a hypothetical £10,000 notional trade::
FTSE 100 or 250 equity
With one execution via IOC order type in lit order book
Assumptions:
The highest cost pricing tier available to the exchange/clearing member (so as to not ignore smaller participants)
LSE vs lowest cost alternative venue and CCP available at the time, at the best tariff available not requiring a subscription or volume tier break (so as to not ignore smaller participants)
Sources:
London Stock Exchange 2007/2020, Chi-X Europe 2007, Cboe 2020, Turquoise 2020
LCH 2007/2020, EMCF 2007, EuroCCP 2020, SIX X-Clear 2020
Highbury Associates, LLC analysis
In 2007 as MiFID I eliminated the concentration rule, an IOC order executing on the LSE would cost 0.96 bps. The same order would cost 0.40 bps to trade and clear on Chi-X Europe. Today, those executions would cost 0.39 bps on the LSE, and 0.20 bps on Cboe/BXE, respectively. Competition has dramatically reduced the cost of execution and clearing, a whole 0.56 bps on the LSE, and 0.76 bps between LSE 2007 and BXE today. More beer for everyone!
Execution venues were the catalyst for clearing house (Central Counterparty, or “CCP”) competition. In 2007, the LSE offered a choice of two CCPs for clearing equities, LCH.Clearnet and SIX X-Clear. This was unique at the time, but didn’t catalyze competition as no other CCPs were allowed to participate. The alternative trading venues at the time, including Chi-X Europe, BATS Europe and Turquoise started with captive CCPs that were increasingly price-competitive. They soon architected a choice of multiple CCPs on their platforms that would facilitate the clearing of trades made between members using different CCPs. This was known as interoperability. Interoperability allowed clearing firms to net positions and more effectively manage margin at the CCPs, drove down pricing and simplified tariff structures. While the LSE allowed full interoperability by mid-decade, some national primary equity exchanges held out on open access and thus interoperability until forced to do so by regulatory mandate (EMIR). Other national primary exchanges, most often also owners of their legacy captive clearing houses, have implemented this to varying degrees.
All of this was brought upon by innovative, competitive execution venues.
Innovation of matching logic and order types
Competition promoted the emergence of creative execution destinations and techniques, including periodic auctions, speed bumps, dark trading with mid-point matching, swim lanes bifurcating LPs and natural liquidity at Aquis and SIX Swiss Exchange’s volume-priority EBBO, RFQ’s on Instinet Blockmatch, and Equiduct's volume-weighted market structure.
According to the ECB’s July 2017 Occasional Paper “Dark pools in European equity markets: emergence, competition and implications” by Petrescu and Wedow, in 2009 trading in dark pools was less than 1% of total turnover in Europe. By 2016, that number grew to 8%, with the UK exceeding 12% on the FTSE 100 and 15% of the FTSE 250 volume. There was clearly considerable demand for this market structure in some execution strategies, with benefits including lower price impact, hidden intent for large orders, and savings from crossing within the spread. Subsequently, Periodic Auctions were introduced across various venues, tallying nearly 4.5% of total FTSE 350 trading in January 2020 (source: Cboe). Continuous trading in the lit markets also conceded share to the closing auction. Dark trading is less prevalent today only due to the double-volume cap mandates introduced with MiFID II.
The point here is that the market values choice that only results from competition. Market participants are like beer drinkers. Some may prefer to drink a Bud like my dad, but many value the choice of different styles and craft brewers, too. As beer drinkers, big and small, don’t want to be told what to drink, nor do traders want to be told how to trade. Like craft brewers, venues have learned to listen to their constituents and build what consensus wants and will support.
We shouldn’t turn back.
The price we’ve paid for choice
As with any trade, there are some costs associated with competition and choice. Common grievances include:
Added complexity in market data integration and order routing
Added fixed costs to connect to more markets
Issues with fragmentation
Dispersal of liquidity
“Ephemeral” liquidity, oftentimes related to the same shares being bid/offered across different trading venues
Some degree of price dislocation related to race conditions between trading venues with fungible instruments
Service providers have entered the market to mitigate and manage this complexity, giving economies of scope and scale to participants. In doing so, the spread has narrowed between the technical elite and everyone else. As the FBA Papers’ authors point out, some market making participants have ceded some of their margins to fixed technology costs, only to maintain parity and a >51% weighted-average win rate.
Concluding thoughts
The benefits of well-regulated, diverse market microstructures, even net of “socially wasteful” telecoms for some, outweigh the costs of going backward. Backward is the implicit direction of travel proposed to move to FBA/Anti-CLOBs.
In the three editions of this commentary, we’ve debunked the assumptions, methods, analysis and conclusions of the FBA Papers, while at the same time addressed very real risks and trade-offs related to a move from current-state competitive, evolving market structure to a uniform Frequent Batch Auction-only future-state. It wasn’t pretty, and we all feel a little hungover from it.
I’m not equipped with a PhD, petabytes of data nor infinite AWS computing power. But I do have considerable industry experience at two electronic trading firms, a global bank and a leading exchange. I’ve spoken to colleagues on the buy-side and the sell-side, at liquidity providers, exchanges and market structure consultants; all share my skepticism and frustrations from the FBA Paper distractions. I thank them for their input. Still, I’m sure there are plenty of holes to poke in my analysis and commentary. Have at it. We can differ and I may have been off base in parts. Let’s have some creative abrasion.
For those of you who are legislators or regulators, or manage clients who may feel disadvantaged, marginalized or aggrieved, I hope these observations highlight the need to sanity check far reaching conclusions that should not be drawn from works with materially flawed assumptions, selectively chosen data sets, convenient extrapolations and theoretical solutions in search of a problem. Politicians should hesitate from exhibiting confirmation bias as they look for a villain in volatile or sinking markets.
Perhaps my distance allows me to be a bit more vocal than current practitioners. But as long as such headline grabbing works are sponsored and not sanity checked, the industry will live with the repercussions of its silence. How on Earth does work like this get funded?
While this is it for the moment with respect to Frequent Batch Auctions, I will have some future comments that frame this problem against other issues that deserve attention. Stay tuned. This isn’t about preserving status quo, but rather ensuring we continuously work to improve market structure and its ecosystem to facilitate efficient capital allocation, while protecting the interests of shareholders of all types, particularly those whose voices aren’t heard.
Because even if we have our preferences, we believe in innovation, competition and differentiation.