The development of HFT, which involves mainly shares but also foreign currencies and derivatives, is an evolution in, and not a revolution of, trading practices.  The main explanations behind it are:

• technological advances which make it possible to place orders in time periods that are increasingly shorter with, on some markets, orders being placed in a few micro seconds (a micro second is one millionth of a second). As the regulator Andrew Haldane said, not without humour, if supermarkets were managed like HFT programmes, you could do a lifetime's shopping in less than a second. When can we expect such trades to take place in a nano second?

• the fragmentation of organised exchanges in Europe by the regulator, through the MIFID directive, with the aim of breaking up exchange monopolies and of creating new ones in order to bring down trading costs through the introduction of competition and to avoid the creation of unjustified profits. A similar evolution took place at the same time in the USA but for different reasons.

HFT is not in itself a trading strategy. It is a collection of technical resources using the latest technological advances in terms of access to markets, routing and execution of orders, implemented to carry out traditional trading strategies.

HFT is characterised by the combination of the following factors:

• the issue, without human intervention, by algorithms running on computers, of a very large number of orders, many of which are cancelled practically immediately (60% in the following second, and finally, between 95 and 99%);

• in order to make a profit between a buyer and a seller;

• since at the end of the day, the net position is zero or insignificant;

• and most of the time, the securities are only held for very short periods of time, 7 seconds on average;

• which explains the very tiny margins on each trade;

• requiring great speed in the execution of orders;

• and, accordingly, to set up a firm’s computers as close as possible to those of the market platform in order to reduce the latent period;

• and all of this working as a proprietary trader, and not on behalf of clients;

• and focusing on very liquid shares.

* * *

The main strategies that rely on HFT are:

• market making:this involves offering a price bracket for a share, for example €15 to €15.01, indicating by this that the trader is prepared to buy at €15 and to sell at €15.01. Market makers who have this status, are obliged at all times to display a price bracket at which they are prepared to buy (the lower end of the bracket displayed) to a seller in the hope, a few seconds later, of being able to sell to a buyer at a higher price (the higher end of the bracket displayed, in as far as it hasn’t changed). Without market makers, markets don't work as well (fewer transactions, more volatile prices), as nothing guarantees that at any time, buyers and sellers will spontaneously come into line – on the contrary.  Other traders are engaged in market making without having the status and the restrictions of market maker. The profit is made on the difference between the price at which the trader buys and the price at which he/she sells, and even on discounts offered by stock exchanges seeking to encourage the presence of market makers on their markets rather than those of their competitors, in order to offer the liquidity that is likely, in turn, to attract other investors.  It is estimated that 50% of HFT is the result of market making activities, whether contractual or free;

• statistical arbitrage, which is in fact a form of speculation. This involves for example, because the Coca-Cola share price has risen, but that of Pepsi Cola hasn’t, quickly buying Pepsi Cola shares, the price of which historically is highly correlated statistically to that of Coca-Cola, and then sell them later. But, if the Pepsi Cola share price doesn’t rise, for some or other reason, a loss is recorded. So this is not an arbitrage in the true sense of the word. One quarter of HFT is the result of statistical arbitrage;

• pure arbitrage: by buying and selling at the same time, the same share that is trading at two slightly different prices on two different markets (Euronext Paris and Chi-X for example), or an asset and its derivative the value of which is not exactly what it should be given the value at the same time of the underlying asset, or between an ETF (2) and its components;

• order anticipation: this involves detecting large orders on the market, that are on the order books, but that have not yet been executed. Accordingly, sell orders could be submitted for small quantities, and, on the basis of the speed of the response by buyers, the presence or not of a large buy order can be detected. The technique then is to buy at a slightly higher price than this order.  If the market rises, a profit is made as it would have been possible to sell on at this higher price. If the market declines, the loss will be very limited, since the large buy order, while it is still in place, provides protection in this case. But because all of this happens within a very short space of time, the risk is limited;

• very short-term speculation: this is not based on a reflection of the value of the share, but rather on the fact that over very short periods, price differences may not follow the random walk of share prices (3), but a momentum, the result of which is that a fall is more likely than a rise, immediately following three already recorded price falls, for example. Alternatively, it is possible to bet on a return to the average. This is obviously speculation, since nothing guarantees that it will turn out this way. Finally, there is a difference between following a trend and hoping that it will last for a few more fractions of a second, making it possible to make a profit, and creating a trend to artificially push the price of a share up or down and thus make a profit. This would be market manipulation and is sanctioned accordingly.

In order to implement HFT techniques, considerable IT resources are required, costing tens of millions of euros.  It is an unending race, with the speed of light the only limit, to such a degree that one regulator refers to a “race to zero” (4). In other words, only the very largest investment banks, a few brokers alongside their market making activities (such as Getco for example), quantitative hedge funds (Citadel) or highly specialised firms (Tradebot, Wolverine Trading, Renaissance Technologies) have the financial, technical (and human) resources necessary to implement HFT techniques.  

The difference between the high-end and the low-end of the bracket displayed enables the market maker to make a profit on the risk that he/she takes on changes in the opposite direction in the price of a share held, while waiting to be able to sell it on to a buyer. The shorter the period for which the share is held, the smaller the price difference between the high-end and the low-end of the bracket will be, while still providing the market maker with an acceptable profit as the risk is taken over a very short period. Of course, the single trade is repeated again and again throughout the day.

A narrow bracket allows the seller to sell at a higher price or the buyer to buy at a lower price than a wide bracket. Accordingly, they benefit from this very fast race, which is not a question of passion but of rationality.

But why place so many orders and immediately cancel such a large number of them?

• In order to detect, as we saw above, the presence of large orders that are likely to impact heavily on the market, but also to identify the limit prices of traders, to give the impression of a large volume of activity so that certain traders modify their price behaviour, to slow down the speed of analysis and that of the reaction of buyers or sellers that are potentially competing.  This is the dark side of HFT.

• Some of the order cancellations are not really cancellations, but rather modifications of the market makers’ price brackets which are permanently adjusted in line with market developments, i.e., the arrival of new orders.  Although this adjustment, which is inherent to the function of market makers, is made once a day at foreign exchange counters, it is made here every fraction of a second.

* * *

There is an increasing amount of academic work being done on the social usefulness of HFT (let’s not forget that this is all recent and that more time is needed before any in depth analyses on the topic can be published).  Conclusions are, on the whole, positive, but reservations have also been expressed:

• HFT, by arbitraging prices on different stock exchanges, contributes to the technical efficiency of markets;

• HFT has probably helped the downward trend of differences between the offer and the asking price, thus reducing transaction costs for investors;

• the reduction in the size of the average order on the stock market, by a factor of 5 to 10 since 2000, increases the total cost of trades, which are now, for the biggest orders, split into several orders, resulting in higher post-trading costs.  The average trade on the NYSE has thus been reduced from 1,600 to 200 shares, in 15 years.  Moreover, this development could encourage traders, seeking to avoid this order splitting, to leave these platforms and go and trade in dark pools (6), which are less regulated. 

• any positive or negative impact of HFT on the volatility of share prices has not been clearly established;

• HFT has resulted in a costly race to invest in technology, with all players being more or less obliged to align their execution costs or to leave the market.

• HFT brings liquidity to the market, which is good for the running of the market. However, the amount of this liquidity is often very limited and it dries up when it is most needed, i.e., in crisis periods (miracles don’t exist in finance!).

* * *

Nevertheless, it is very worrying that, two years after the “flash crash” of May 6, 2010, we still don’t understand why this formidably loud clap of thunder occurred in a bright blue sky, or  the exact role played by HFT in setting it off.   The consensus (soft) is that HFT did not set it off, but it did accelerate it.  What happened was that the major HFT players withdrew from the market to protect themselves.  Let’s remember that on that day, in 20 minutes, the S&P 500 index fell by around 9% before climbing back up by as much, and that the Accenture share went from $40 to $0.1 and the Sotheby’s share from $34 to $99,999.99! Computers worked all by themselves, illustrating, if this was necessary, that if speed and automation are pushed very far, they will clearly bring systemic risk with them. 

Since then, other similar forms of dysfunctionning, although less spectacular, have been observed, including very recently with the Apple or BATS shares.

If going faster was always systematically the right answer, speed limits on roads would have been abolished a long time ago, and neither Air France nor British Airways would have put Concorde in a museum.

* * *

The reader will not be surprised to see us sorting through the HFT techniques and separating those that in our view are socially useful (arbitrage, market making and speculation), and others that seem to us to be a bit more doubtful, such as order anticipation or speculative techniques that are tantamount to market manipulation. 

Similarly, it isn't normal that HFT players should be regulated (banks by Basle 2 and 3: prudential ratios, inspection by regulatory authorities, external testing of procedures, etc.) while others (players other than banks) escape most or even all of such regulation.

The bombarding with orders that are practically all cancelled almost immediately and the race to set up computers adjacent to those of stock exchanges, could give an initial negative impression of HFT, as if it were based on cheating, manipulating and unfair competition, to the detriment of other market players who represent, although no longer the majority of orders placed, the vast majority of traders on the market.

At a time when the greed, selfishness and even the dishonesty of a handful of people has unfairly discredited the vast majority of people working in the field of finance, we can no longer allow the development, without reacting, of techniques, the collective utility of which is uncertain, which the regulators are only capable of overseeing from a distance, as a result of a lack of resources, and which lead to risks which are not easy to measure but which are potentially very high.

Fortunately, the market authorities seem to have measured up the problem.  In Europe, as part of the revision of the MIFID Directive, all substantial players on the HFT market must be regulated and must be under restrictions in terms of equity and risk management.  Those who play a de facto role of market makers should have rules imposed upon them.  The co-location of computer servers should be offered to all players who submit a demand, on a non-discriminatory basis, although there are physical limits to the number of computers that can be installed.  The notion of market abuse should be better defined and thus easier to sanction.  Automatic market shut-downs, in the case of excessive changes in share prices or in the case of volatility, could be imposed.

Combined with plans to tax financial transactions (7), THF should be better managed, making the market more certain, at the probable cost of less efficiency, but this would certainly not cripple it. After all, who complains about not being able to fly from Paris to New York in 3.5 hours as in the days of Concorde?

And if we may, we’d like to end with a bit of advice for the HFT industry.  Communicate! At the moment, the industry has no form of external communication. This sort of opacity, and the cult of secrecy for secrecy’s sake, is no longer fashionable and can only do a disservice to the industry.

Just like we compare the levels of CDS(1) for countries in order to achieve an overview of the way investors assess their solvency, the same exercise can be carried out on economic sectors on the basis of the individual CDS of the companies making up those sectors.

 

Over the last 3 months, the gaps have grown considerably, given the heightened risk aversion in a deteriorating economic climate.  So, you now have to pay 15 times more to protect yourself against default in the airline sector than in the pharmaceutical sector, compared with "only" 12 times three months ago.  But since the 3 major European airline companies are currently operating at a loss, there are no surprises here.  After all, didn’t Richard Branson, who founded Virgin Airways once say:  “How do you become a millionaire?  You start out as a billionaire and then you buy an airline company!”