Algorithmic Trading Could Create Risks To How Markets Function

i. Introduction

Economics has struggled to make sense out of the price movements that occur in financial markets, or on a larger scale, economies as a whole. For a long time the ideas of equilibrium were very important to the study of economics, but when these ideas are looked at critically, some problems, and in fact, paradoxes, begin to emerge.

Eugene Fama's ideas built on earlier economists' work that took the view that the capital market not only had a goal, but also was somehow designed to be the most efficient allocator of a society's investment capital. There are a few assumptions inherent in this view, but from a more practical perspective, it relies heavily on the idea of information efficiency.

One of the largest flaws in an idea like Fama's original theory is that information isn't seen as having any sort of unique properties, nor is it seen as being a direct influence on the ability of a market participant to act in their own best interest. Because a market is assumed to be efficient, all participants are seen as having access to a completely accurate representation of value.

At times the results from the application of these ideas has been shown to represent market action in the real world. But overall, it isn't difficult to see how far removed from reality a theory that posits an efficient market based on perfect information is, at its core. There is also the paradox of how the information came to be manifested in the prices in the first place, because this model assumes, in essence, an eternally existing perfect market.

In the real world, access to information isn't uniform, and markets may or may not move due to fundamental factors. Human traders are capable of making mistakes, and they are also driven by emotions. Fear and loss, and greed for gain are powerful psychological drivers, and the information transmitted in a price isn't a fair representation of these forces at all.

Information's Value

Because market prices are exogenous, it is easy to understand why Fama would imply that a market is a perfect allocator of value. But many sophisticated investors are able to obtain information that gives them the ability to outperform the broad market, because their decisions are based on a view of reality that is larger in scope than a price in a market.

The companies' shares that are bought and sold in a share market have a human value, and their operations are susceptible to effects from both the economy, and environment. These factors are what successful investors seek to better understand, and none of them are ever obvious from the price of a share. Today quantitative trading makes a big impact on most of the major markets as well, and this takes information gathering to a whole new level.

All of these influences add a layer of reality to economic theories, and help to explain why some investors, or financial entities are able to outperform the broad markets on a consistent basis. Quantitative investing relies heavily on data driven decision making, and algorithmic trading strips human frailty out of the buying and selling of financial instruments. A realistic view of the financial markets has to take all these ideas into account, and while they are simple enough to outline in an abstract, the reality of the marketplace is far more nuanced.

ii. A Method for the Madness

The Efficient Market Hypothesis isn't really helpful in the real world, and of course there are many other ideas out there. The Random Walk theory has the advantage of being applicable to empirical data, but much like the Efficient Market Hypothesis, it fails to take into account the advantages that unique information gives to investors that can afford to acquire it.

Lasse H. Pedersen is a Danish economist, and he recently published a book that describes the financial markets as efficiently inefficient. He posits that markets are in the end efficient, as they seem to arrive at correct prices eventually, but they are ineffective as they go to extremes in their search for those correct prices.

One of this theory's most effective assertions is that information is what makes the difference between success and failure to an investor, and thus creates a way to assign value to financial information. According to this theory, an investor who has better access to information will prosper, while smaller investors who can't afford the information will underperform.

Unfortunately for this theory, it creates a world where “valuable” information will inherently lead to outperformance for the investors that can afford it, and this isn't realistic in the least. Despite this, Mr. Pedersen's concept of an efficiently inefficient market warrants a much deeper look, because it creates a level of reality in economic theory that is very useful from a practical perspective.

His theory also describes the formative function that price movements occupy in an economy, and this is becoming more relevant as the major markets' price movements are increasingly a result of algorithmic trading. When algorithmic trading is used to any great extent, the market no longer relies on the value of information, in the way Mr. Pedersen defines it. Instead it functions as a means of gaining from complex math, and when used on scale, can actually create the economic reality outside of the exchange it is operating in.

Automated trading began to be used with the assumption that the price exists beyond the control of the algorithm that is trading. Today with the scale of algos in some of the markets, this is no longer a realistic view. It also can create a situation where the price of the stock or commodity that is being traded is no longer a representation of the underlying asset.

These ideas challenge the concept of an efficiently inefficient market as a realistic way to view markets, and in some cases it does so to an extreme degree. This is a result of the formulas that are at the heart of an algorithmic trading program being applied to a market en masse, and when this occurs, a host of new ideas emerge.

iii. No Longer a Random Walk

The concept of a successful investor is something of a conundrum for academic economists, especially from the perspective of either an Efficient Market, or Random Walk Hypothesis. If viewed from the efficient market perspective, access to information is meaningless, as all participants have the same position.

The Random Walk view is equally dismissive of the value of information, because in a truly random market, no amount of information could help an investor succeed. Today markets are moving toward a state that eclipses both these ideas, and also challenges some long held ideas about price discovery.

Andy Hall made a fortune in the oil futures market. His secret was large amounts of data gathering, and analysis of the fundamentals. This is an example of the power that information has to benefit a dedicated investor, but recently, he decided to close down his flagship fund. The fund had lost more than 30% of its value this year, and he cites automated trading for his exit from the market.

In a letter that was obtained by Bloomberg, he was quoted as writing, “Algorithmic trading systems have increasingly come to dominate," and he went on to state that, "Investing in oil under current market conditions using an approach based primarily on fundamentals has therefore become increasingly challenging. It seems quite likely this will continue to be the case for some time to come.”

The oil futures market is dominated by algorithmic trading (60%+ of the trades are made by algos), and as this example shows, the way in which information is used in a changing trading environment can make a big difference to the outcome of its application. Hall's nickname of “god” was given to him after he cleared a $100 million dollar bonus as a oil trader at Citi, but today, it looks like “god” is being run out of town by the machines.

The Meaning of Information

The concept that Mr. Pedersen puts forth is worthwhile because it not only describes a market that has information that is valuable, but it supposes that information can be used to knowable ends. When fundamental factors were widely held as being valuable vis a vis price discovery, people like Andy Hall could use their ability to gather information to make legendary gains.

The efficiently inefficient market hypothesis would support the value of information to a trader like Andy Hall, so we are left wondering how price discovery will happen in a market that is dominated by formulas that are traded on by computers.

Quantitative trading can be implemented in a number of ways, but today, there are many hedge funds that rely chiefly on fully automated systems that more or less remove humans from the market to a degree that has never existed before. While quantitative analysis isn't new to the world of trading, this level of fully automatic trading is changing the way markets function.

When Mr. Pedersen uses information as a property in his theory of an efficiently inefficient market, it is assumed that access to information will make the difference between a successful investor, and one that is at the mercy of the overall market direction. But with algorithmic trading, and the rise of artificial intelligence, this assumption may be a tenuous one.

iv. The Correctness of Price in an Algorithmic Marketplace

Investing legend Benjamin Graham famously said, “In the short run, the market is a voting machine but in the long run, it is a weighing machine.”

With these words Mr. Graham was describing a market's tendency to move into extreme conditions over short periods of time, due to human emotion, or other oversights. Over the long term however, he posited that the true value of an item or company would be found, as its fundamental qualities would outlast irrational market conditions.

Mr. Pedersen's idea of an efficiently inefficient market seems to be very much in line with this old adage, but there may be something more at work today when it comes to how information is used to determine price, regardless of time frame.

An algorithmic trading system is a curious creature. While they are very complex, and one could consider their potential forms to be endless, they can be thought of as existing in two modes. When an algorithm is used in a market, its mode can either be adapting, or forming.

When quantitative trading began, it operated solely in the adapting mode. That is to say; the quantitative equation looked for data that would help it to adapt to the price action in a market, and profit from it. Today, as in the oil market, algorithms are working in forming mode, where they are creating market direction to a level that minimizes the influence of fundamentals, as the term has been used up until today.

The question that emerges now isn't the value of information, but one of how price is viewed as a reflection of reality, over any period of time. Human traders know that a major company has value, or that the commodity that is underlying the price on a mercantile exchange has some sort of intrinsic utility. These ideas are completely foreign to a trading algorithm, and because of this, the price of a share or commodity can be pushed to an extreme that is wildly disconnected from reality.

Making Markets Move

Market fundamentals can be equated with the information that Mr. Petersen cites as making the difference between outperforming a market, and regular returns. For a share in a company, the price can be representative of the people who are employed, the plant and stock and any other commercially valuable items that the company has.

The algorithms that are employed by a quant fund may assess the value of a company to some degree, but when they are traded on, they are unlikely to step in to buy a share when the market is crashing. Most programs will be designed to wait out the volatility, and come back in when things settle down. This can be contrasted with human traders and investors, who have a better idea of the underling value that a company has.

When an algorithmic trading system is adapting to a market where prices are being discovered by humans, the damage they can do is minimal. Ultimately humans will know that a massive steel conglomerate has a value above zero, and when madness prevails in the exchange, some cool heads will stand to make a fortune. Algos don't have this ability, and as the Flash Crash of 2010 shows us, the movements that can take place when the machines are in control are striking.

This sort of market dynamic takes the inefficiency in Mr. Pedersen's theory to an extreme, and also creates the risk that well informed traders, like Andy Hall, will simply refuse to deploy capital in such an inhuman marketplace. In the absence of the well informed, cool headed investors that step into the breach when lunacy breaks out, there really is no telling where prices will land.

v. Arriving at Now

In 2013, algorithmic trading accounted for 13% of the shares traded in the major exchanges in the United States, and today that number is north of 20%. That is a big rise in automated trading over a short period of time, and while less a less extreme percentage than the oil futures market, more than a fifth of the largest equity exchanges in the world is a significant amount.

The idea of an efficiently inefficient market that relies on transitory price extremes to eventually find a fair value was probably a fairly accurate way to view the market mechanism a decade or two ago. Bot now, with algorithmic trading on the rise in the biggest markets on earth, it is reasonable to ask how advanced mathematics will affect the value of the markets themselves.

What is often overlooked in economics is the humans that are at the core of any economic entity, a stock market, or commodity exchange included. People like Andy Hall spent years learning everything there was to know about the oil market, so that he could buy and sell as efficiency as possible. That was the magic of the market, and it was how he made a fortune trading them.

It remains to be seen if algos will do the same job, but given the growth that algorithmic quant funds are enjoying, we may find out before too long. As it stands today, algorithmic quant funds are becoming far more popular, due to their ability to consistently outperform other types of investment vehicles, and the broad indexes themselves.

In the first quarter of 2017, automated quant based hedge funds gained around $4.6 Billion U.S. Dollars in assets, and now they hold around 30% of the $3.1 Trillion U.S. Dollars of total hedge fund assets. With those kinds of numbers, it may only be a matter of time before mathematical models are assigning value to the biggest companies' equity via the exchanges, and to a much greater degree than human traders. These appraisals of value will by necessity be determined not by assets, but by market performance. This is algorithmic modeling in forming mode on a scale never before imagined.

It is perhaps heartening to see the evolution of the marketplace in the light of Mr. Pedersen's view of efficiently inefficient development. Only in this case, it isn't the value of the transactional goods that is being discovered, but the value of the market itself. Because equity and commodity exchanges are the heart of the modern financial system, the rise of algorithmic trading could be seen as a new means of assigning value within the global marketplace.

When viewed from this perspective, the level of responsibility of the mathematical traders becomes far more serious than just making a profit. There is no way to know what will come of this mega trend towards automated quantitative trading, but it is difficult to imagine a situation with higher stakes, or more people held in the balance of mathematical formulas that were designed to generate profits, and nothing else.