Saturday, August 6, 2016

SEBI’s Discussion Paper on Algorithmic Trading

Background to Algorithmic Trading

Similar to most other fields, the use of technology is being optimized in trading in the stock markets. Stock trading is getting increasingly automated through use of sophisticated computer systems that operate through algorithms, which minimize human involvement and decision-making. Not only does this lead to the extensive use of technology by stock traders and investors, but it may also create imbalances in the stock markets whereby certain players can use technology to favour their own commercial interests at the cost of other similar players in the markets.

On the one hand, such algorithmic trading (or algo trading) is beneficial as it operates instantaneously based on information available in the markets, and hence makes the markets more efficient. On the other hand, it has been criticized on the ground that it creates distorted incentives in various market players that could lead to imbalances and consequently significant risks to the stock markets as well as the economy as a whole. Instances such as the flash crash that occurred in the US markets in 2010 due to erroneous order entry into the computer systems only highlight the risks of automated trading.

Interestingly, the debate surrounding algo trading achieved high intensity a couple of years ago, and was even the subject matter of a book Flash Boys: A Wall Street Revolt by Michael Lewis. In the meanwhile, regulators the world over have been consider the merits of restricting such algorithmic trading so as to mitigate its risks.

Moving to India, the scope and extent of algo trading has not been clear. Recognizing the need to address the issues pertaining to algo trading, particularly from an operational standpoint, the Securities and Exchange Board of India (SEBI) has issued a series of guidelines to the stock exchanges (here, here and here). However, in order to consider the regulatory aspects of algo trading on a comprehensive basis, SEBI yesterday issued a Discussion Paper on ‘Strengthening of the Regulatory Framework for Algorithmic Trading & Co-Location’.

SEBI’s Discussion Paper

At the outset, the Discussion Paper seeks to clarify the various technical concepts relating to the topic. Algo trading is the broader concept (as discussed above), which provides greater speed to stock trading and also offers anonymity. A sub-set of algo trading is “high frequency trading (HFT)”, which uses high-speed networks and locational advantages to create trading opportunities within miniscule fractions of a second. HFT usually relies upon co-location, which in essence means that certain traders place their computer servers within close proximity of the stock exchange system such that they have access to trade information and prices a split second faster that others, which they can work to their advantage.

After introducing the concept, the Discussion Paper goes on to state:

2.5. Adoption of such advancements in technology in our market in recent years have resulted in vast majority of orders being generated through trading algorithms. Currently, more than 80% of the orders placed on most of the exchange traded products are generated by algorithms and such orders contribute to approximately 40% of the trades on the exchanges.

The use of algo trading seems quite pervasive, although it is not clear whether the Discussion Paper refers to these statistics with references to global markets in general, or whether it relates to India. If the numbers pertain to India, they look much higher that I would have thought.

The Discussion Paper then goes on to make a series of proposals in order to combat the distortive effects of algo trading. Some of them are set out below:

- Minimum Resting Time for Orders: This would ensure that there would be a minimum time between when an order is received and when it is allowed to be amended or cancelled. This is to avoid a situation of “fleeting” orders that can be prevalent in algo trading where large orders are made and then cancelled or amended in a split second, thereby distorting the market. SEBI has proposed a minimum time of 500 milliseconds before an order can be amended or cancelled. SEBI has also indicated that its approach might be unique in that such a minimum resting mechanism is not mandated by any regulator yet.

- Frequent Batch Auctions: Under this mechanism, buy and sell orders would be accumulated for a length of time, such as 100 milliseconds, at the end of which orders received during the time interval will be matched. This will obliterate any advantages that HFT players may have through co-location.

- Random Speed Bumps: This involves “introduction of randomized order processing delay of few milliseconds to orders”. This is again to nullify co-locational advantages that some players might have.

- Randomization of Orders: Under this proposal, all orders received during a predefined time period (such as 1 to 2 seconds) would be randomized in order to perform the order matching routine. Here again, any advantage of speed or location would be of no use.

A few other similar, and specific, proposals have been made in the Discussion Paper.


It is not clear what form the final set of SEBI’s regulations relating to algo trading will take, if at all. However, from nature of the proposals and the tone of the Discussion Paper, it is clear that the approach is towards imposing greater regulation on algo trading so as to bring about a level playing field in the market. In other words, the use of technology ought not to create disparities in the system (although proponents of algo trading may argue that there is nothing unfair about it). To that extent, SEBI seems to be adopting a rather strict approach towards regulating algo trading. Although I have not undertaken detailed research on the issue, it appears from the Discussion Paper and the surrounding debate that SEBI’s efforts in reining in algo trading may have gone farther than most other regulators around the world. More broadly, the idea seems to be that speed is not always beneficial, and sometimes delayed responses may be necessary and may yield better results (as Professor Frank Partnoy has argued in his book Wait: The Art and Science of Delay).

Comments on the Discussion Paper are due from the public on August 31, 2016.

No comments: