Regulate, don't stifle

Sebi's proposal to treat all APIs as algo trade is unwarranted

In a consultation paper last week, the Securities and Exchange Board of India (Sebi) proposed regulating retail or third-party algorithmic trading (algos). The regulator believes these modes of trading are risky and there is little understanding of how they function. Sebi says these “can be misused for systematic market manipulation as well as to lure the retail investors by guaranteeing them higher returns”. The potential loss from a failed algo strategy may be huge.

Sebi wants brokers to give inputs to help formulate a policy framework on third-party algo providers. It has proposed to treat all orders based on the Application Programming Interface (API) as algo-driven and said such orders should be tagged with an ID unique to the brokerage. It wants brokers to perform a sequence of stringent checks on any API-based trades to ascertain if these are algos. It has also stated third-party algo providers could be treated as investment advisors and that two-factor authentication (which implies human intervention) be put in place.

This may be an over-reach in some respects. It would impact retail traders and brokerages in terms of the cost of compliance. It would retard the use of API-based technology, which smoothens trading processes for all investors.

Most brokerages offer APIs, which allow the investor to directly connect some sort of trading programme to the brokers’ server. Brokerages also offer their own algos for traders interested in using these. An API may also be used to hook up an algo, coded by the client, or a third-party.

Algos are programmes that automatically monitor price-volume action and make trades without human intervention, buying and selling when pre-set instructions are triggered by price moves. They react very quickly, especially when hooked directly to the exchange’s own servers (this is co-location). Some algos trade arbitrages — small price differentials arising for brief periods in different markets. Many algos are complicated “black boxes” — they may trade combinations of stocks, commodities, currencies, and derivatives.

A brokerage has to carry out extra checks on “own-account” algos it offers, get these cleared by Sebi, and tag them with a unique ID. But it is not very easy to identify clients using self-designed or third-party algos. There are telltale signs — human beings don’t respond with the same speed. But it would require a lot of extra number-crunching power and data analysis, to create detection systems.

Algos can place the user at greater risk. This is partly due to the lack of human intervention, and partly because they can be programmed to make simultaneous trades of different markets. In the early days, this could spiral into a huge market-wide risk owing to lack of circuit filters. The famous “Black Monday” crash of Wall Street on October 19, 1987, occurred because algos sold heavily without human intervention.

However, Sebi has many robust checks in place to ensure adequate margins are collected. It has circuit filters to halt trading if there is an extreme price move. While “flash crashes” — instantaneous drops that hit the circuit — do occur, the damage is limited. By the National Stock Exchange’s estimate, about 14 per cent of the trading volume (and around 45 per cent of the trading value) is algo-driven. Major retail brokerages estimate around one in 2,000 clients uses algos. This can be tackled by adequate margining. Imposing high costs of compliance under the assumption that every API user is an algo trader would punish every investor.