Keeping control of the machines

The growth of automated trading since the introduction of MiFID has led to a greater focus on the risks associated with algorithms in the directive's forthcoming review, but how will the proposed changes impact market participants?
By None

How is MiFID II going to limit the systemic risks to Europe's equities markets posed by high-frequency trading (HFT)?

Bureaucracy might turn out to be the best one-word answer. In the consultation document issued on 8 December, the European Commission (EC) proposed that all firms that conduct ”automated trading' be regulated under MiFID II and that all their algorithms be registered with the relevant authorities. The document identified high-frequency traders as an important sub-set, but gave such a broad definition of automated trading – i.e. all forms of trading that use “computer algorithms to determine any aspects of a trade” – as to potentially require regulated firms to provide information on the workings of even the most basic trade automation tools, such as a broker-supplied VWAP algorithm. Quite how much detail the regulators will require is yet to be determined, but it's possible that the competent authorities would need information on the design, purpose and function of every algorithm deployed in the European equities market.

While the national regulators would have the main responsibility for monitoring algorithms, the European Securities and Markets Authority (ESMA) could have a role to play for any algorithmic strategies that work across borders, given its clear remit to regulate supranational businesses, such as credit rating agencies.

It seems a bit unnecessary to register every single algo, but how much harm can some extra form-filling do?

True, the current MiFID II proposals do not imply that the regulators would have the power to prohibit the use of an algorithm or mandate changes to the way it operates, but a logical next step might be for regulators to introduce mandatory stress testing of strategies to ensure that algos will continue to function properly in extremely volatile markets. This is something some large buy-side firms already do when evaluating broker algos. However, if regulators become too involved in authorising algos for use on European equities markets, restrictions resulting from concerns about systemic risk could lead to less choice and less scope for customisation to suit the individual buy-side trader's needs.

Let's row back a bit. What threats do algorithms pose to Europe's equities markets?

In its consultation, the EC emphasised the risks it thought algorithmic trading could pose to the functioning of an orderly market. In particular, it mentioned the risk posed by rogue algorithms that overreact to market events and the pressure exerted on trading venues by a large number of algorithmic orders generated by automated trading.

This suggests that the blanket approach outlined in the MiFID II consultation document stems from two separate market developments. First, the growing proportion of European equities trading volumes accounted for by highly-automated HFT strategies means that the firms that execute them can no longer sit outside MiFID. ESMA's forerunner, the Committee of European Securities Regulators, conducted an investigation into tech-led trading trends in the run-up to delivering technical advice to the EC on MiFID's impact. Clearly, CESR's report gave the EC sufficient concern about the existing market structure's ability to handle new types of trading as to warrant closer inspection.

Second, the 6 May ”flash crash' in the US seems likely to have provided the impetus for the EC to widen the scope of its registration drive to potentially include all algorithms. A joint report by the Securities and Exchange Commission and the Commodity Futures Trading Commission blamed a rogue algorithm that was set by a trader at a long-only firm to sell US$4.1 billion worth of S&P E-mini futures contracts without price or time constraints. If execution algorithms in the hands of buy-side traders can cause such ructions, the EC may have decided that it's best to have them listed with the regulators too.

There's a third possible reason for the EC's apparent decision to require all algorithms to be registered. A handful of domestic exchanges in Europe currently require high-level descriptions of algorithms from their members, but this is not mandated. However, as domestic exchanges continue to lose their grip on equity trading volumes, regulators may feel as though its time to step in.

What are the potential implications of this for market participants?

It rather depends how far the EC wants to go. It might decide that only sell-side suppliers of execution algorithms and proprietary trading houses – i.e. the primary developers of algorithms – need provide information. This might mean that the only change for buy-side traders is that their algos now come pre-approved with a ”safe to trade' stamp from the regulators. As noted above, it may also mean their algos are less customisable and hence less effective.

For the brokers and the prop shops, the prospect of algo registration is far more worrying. Both types of firms regard their algos as a source of competitive advantage and neither will want to explain the nature and design of proprietary strategies for fear of revealing commercially sensitive information.

Will algo registration make the market safer?

Unlikely, in the view of many. Instead of trying to understand how good the engine of a car is – i.e. looking at the design of each algo – market participants argue the EC would be better off setting a speed limit or imposing restrictions on the outputs of algo strategies, such as limiting the order-to-cancel ratio.

In addition, there is doubt that regulators currently have the knowledge and expertise to be able to provide useful insight and analysis on how risky an algo might be. For example, the methodology used by VWAP algorithms for mid-cap stocks may be broadly similar across multiple providers, but differences may arise from technicalities such as how market signals are interpreted, risk thresholds and tolerance to price movements. Even if the regulators invest in the expertise required to understand different algorithms, is this really the most cost-effective way to protect investors?

Unquestionably, the more regulators understand about the trading strategies of different types of market participants, the more able they are to anticipate threats to market stability. But this still feels like a sledgehammer to crack a nut.

Click here to vote in this month's poll.