High Frequency Trading & Job Displacement
Foreword: I wrote this paper my senior year - Spring 2019 - and found its content relevant to today as pundits claim AI will replace an increasing amount of white collar work. Hopefully through reading this paper, you are able to think about whether those claims are valid or should be challenged.
High Frequency Trading & Job Displacement
Wall Street investors engage in high frequency trading, or, more acutely, have automated trading algorithms that transact a large number of contracts at fractions of a second, in contrast to traditional trades. Although there are proponents that contend high frequency trading is more cost-efficient, effective, and can replace all financial analysts, there are numerous shortcomings with incorporating high frequency trading. I’ll denote problems associated with high frequency trading, as well as defend the role of financial analysts; further, I’ll advocate that investment banks adopt an approach favoring human-computer interaction (HCI). In this assignment, I’ll be critiquing the following simplified argument by Martin Ford in Rise of the Robots:
If the utilization of artificial intelligence is more profitable than the utilization of financial analysts, then corporations will replace all financial analysts with artificial intelligence.
The utilization of artificial intelligence is more profitable than the utilization of financial analysts.
_________________________________________________________________________________________
Corporations will replace all financial analysts with artificial intelligence.
Now, I’ll provide insight into the displacement of financial analysts in-relation to high frequency trading and later in the paper provide a possible response to Martin Ford, as well as a solution to this displacement.
Here, I’ll provide more insight into high frequency trading (HFT). HFT incorporates smart computers and, by extension, complicated algorithms to go beyond simply executing trades, as such tools effectively spot opportunities for arbitrage, or, more broadly, exploit opportunities for profit; further, they, often, deceive other algorithms through entering decoy bids and withdrawing within fractions of seconds. There are a variety of different approaches to designing algorithms, however, one of the most powerful involves the application of an artificial neural network, or, in function, systems that model the human brain. The supercomputer effectively incorporates self-learning systems and can adapt to unpredictable scenarios; moreover, the efficiency of these systems determines the ultimate success or failure of an investment bank. To provide insight into the efficiency and profitability of HFT, I’ve included this quote from Martin Ford:
In the realm inhabited by these continuously battling algorithms, the action unfolds at a pace that would be incomprehensible to the fastest human trader. Indeed, speed-in some cases measured in millionths or even billionths of a second-is so critical to algorithmic trading success that Wall Street firms have collectively invested billions of dollars to build computing facilities and communications paths designed to produce tiny speed advantages. (Ford, 93).
From this quote we can effectively conclude that HFT is more efficient than financial analysts, as these smart computers are able to process information at a rate incomprehensible to financial analysts and, correspondingly, profit from this information in microseconds; however, there are diminishing returns on the profitability of HFT, as it’s unlikely for a trader to consistently profit from access to public information.
Most economists agree with the claim that, without inside information, very few investors can consistently make accurate predictions about whether the price of an individual stock will rise or fall on a day. (IGM). Further, 55% of economists strongly agreed, 40% agreed, and no economists had disagreed with this claim.1 Though, having access to public information before most investors permits some to profit, as algorithms are able to process public information before it becomes reflected in the stock’s price and arbitrage profits.
Now, firms investing billions in this strategy suggests it’s getting harder to have early access to information.
Increasing the number of firms engaging in HFT will cause expected returns to diminish, which is why small speed advantages are so valuable. Though, it’s possible that with more and more money going into research on HFT, arbitrage from the strategy will cease once these speed advantages are no longer possible. This will happen when HFT becomes standardized, as companies are able to profit from having early access to public information and will lose the advantage once more firms can process information at the same rate.
Now, I’ll examine the motives behind firms replacing financial analysts with artificial intelligence. In The Social Responsibility of Business is to Increase its Profits, Milton Friedman highlights shareholder theory, which denotes the only duty of a corporation is to maximize the profits accruing to its shareholders. This responsibility derives from a contractual obligation between investment banks and the shareholders they represent. In the context of investment advisors, they have a fiduciary responsibility to act in the best interest of their clients, as they’re acting on the behalf of an investment bank. There are a few limitations to fiduciary responsibility. One, investment banks ought not to contract to do anything immoral or illegal: moral reasons can cancel or override these promises. Two, if all shareholders want to sacrifice profits for social responsibility, then managers are required to comply. The motive behind replacing financial analysts with artificial intelligence stems from the contractual obligation investment banks have with their shareholders, as it, presently, maximizes shareholder value to adopt HFT strategies. Not adopting HFT and keeping financial analysts would, according to Friedman, impose a tax on shareholder value, as he outlines:
[…] the corporate executive would be spending someone else’s money for a general social interest. Insofar as his actions in accord with his “social responsibility” reduce returns to stockholders, he is spending their money. Insofar as his actions raise the price to customers, he is spending the customers’ money. Insofar as his actions lower the wages of some employees, he is spending their money. (Friedman, 2).
Here, it should be noted an employee isn’t a shareholder and they shouldn’t be working on their own behalf; instead, they are representing the interest of anyone that has invested their money with the investment bank. Although some analysts invest money where they work, employee interests aren’t also shareholder interests. The social interest in this context would be the employment of financial analysts, as some contend that the role of corporations is to create jobs; however, replacing financial analysts with AI is eliminating positions. In order to uphold an investment bank’s fiduciary responsibility, artificial intelligence, seemingly, needs to replace all financial analysts, as this decision maximizes shareholder profit.
In shareholder theory, if all shareholders prefer to fund the social interest, then the firm must fund the social interest. Here, some assert there’s a concurrent civic obligation in contrast to profit maximization; moreover, they assert companies can impose costs in-line with their corporate social responsibility (CSR). This doctrine maintains, in part, that companies are responsible for the creation of jobs, not displacing jobs. Proponents of CSR appeal two-fold: welfare matters, and everyone’s welfare assessment should be equal. Now, there have been several protests where stakeholders felt corporations had violated their responsibility. For example, Occupy Wall Street was a protest against an array of social issues, such as job displacement. Although the momentum for the protest had dissipated before any real, corporate change had been realized, some executives such as Howard Schultz, chief executive of Starbucks, do still support stakeholder interests. In a debate with Milton Friedman, Schultz stated:
Companies that hold on to the old-school singular view of limiting their responsibilities to making a profit will not only discover it is a shallow goal but an unsustainable one, […] Values increasingly drive consumer and employee loyalties. Money and talent will follow those companies whose values are compatible. (Bussey, 2).
Now, it’s important to note that there can be overlap between the interests of shareholders and stakeholders; further, promoting policies or practices that service the public interest can be profit maximizing and ethical. Howard Schultz suggests that efforts to achieve profit maximization may, ultimately, hurt a firm’s profits, as this restricting corporate culture may influence employees to work elsewhere and demotivate investors. Value and profit are symmetric, as money follows firms with compatible values to those of the shareholders; moreover, money, in general, is a means in which consumers can attach value to a firm’s service or product.
If investors value the services of financial analysts, then firms will continue to employ financial analysts. Though, the motivation behind investing is, often, maximizing one’s expected return for one’s risk aversion,
not pursuing one’s social interests. Thus, we’d expect a decrease in the number of financial analysts overall.
In this paragraph, I will provide insight into the recent decrease in financial analysts employed on Wall Street, as well as in other industrialized areas. Now, disruption is nowhere more evident than at the New York Stock Exchange (NYSE). The NYSE is famous for having a noisy trading floor dependent on having direct communication with brokers and dealers, but algorithms now handle the majority of trades. To provide reference into the scope of high frequency trading, I have highlighted a quote from Martin Ford:
… automated trading algorithms are now responsible for nearly two-thirds of stock market trades, and Wall Street firms have built huge computing centers in close physical proximity to exchanges in order to gain trading advantages measured in tiny fractions of a second. (Ford, 56).
Now, this passage is important for a few reasons. One, it denotes a majority of trades are done using high frequency trading, which means artificial intelligence can and have massively impacted the world economy.
Two, there are now fewer brokers and dealers, as most of the trading is, instead, done in computing centers.
Finally, these centers neighbor exchanges such as the NYSE because location provides speed advantages. These details underscore the scope of job displacement, which Ford highlights in the following quotation:
The impact of all this automation is clear: even as the stock market continued on its upward trajectory in 2012 and 2013, large Wall Street banks announced massive layoffs, often resulting in the elimination of tens of thousands of jobs. At the turn of the twenty-first century, Wall Street firms employed nearly 150,000 financial workers in New York City; by 2013, the number was barely more than 100,000-even as both the volume of transactions and the industry’s profits soared. (Ford, 115).
Corporations have placed more focus on short-term profitability and, correspondingly, shareholder theory, as institutions have incorporated the usage of high frequency trading and have displaced tens of thousands. Moreover, firms have managed to execute a greater volume of transactions and have received higher profits. It’s important to note that similar displacement is occurring in other industrialized parts of the world and it seems more corporations attempting to maximize profits will replace employees with artificial intelligence.
Now, I’ll examine the decline in labor’s share of national income in-relation to financialization, or, more broadly, the growth of finance activity. Previously, it was noted that investment banks have seen an increase in transactions and profit, despite continued lay-offs; further, an increase in high frequency trading directly parallels the decline in labor’s share of national income, as fewer workers in finance are employed. Advancements in information technology are highly correlated with growth in the finance sector because most finance innovations such as collateralized debt obligations (CDOs) necessitate the utilization of super- computers; nevertheless, these developments have allowed some significant catastrophic, economic fallout. For example, subprime mortgage backed CDOs were heavily implicated in the 07 – 09 financial recession, as they hid short-sell positions from AAA seeking investors through the volume of mortgages within CDOs. Moreover, these mortgages were massively profitable for big banks, as they earned more commission on each bond sold; however, banks began to run low on credit worthy investors and issued riskier mortgages. These mortgages are ranked in a hierarchy based on their level of risk, and AAA bonds have the least risk. Although this is how it’s supposed to work in theory, this differs from the actual practice of creating CDOs; instead, banks, around the time of the financial recession, had put high risk bonds together with low risk. Credit agencies, such as Moody’s, then apply a higher ranking to these portfolios based on diversification. This financial instrument had collapsed from massive defaults on mortgages and left millions unemployed. I’ve included a quote from the Organization for Economic Cooperation and Development to highlight this:
The OECD area economy has entered recession and labour market conditions are rapidly deteriorating in many countries, […] projections indicate that the average unemployment rate in the OECD area may reach 6.3% in the last quarter of 2008, from 5.5% a year earlier. […] Overall, these projects suggest an increase in the number of unemployed persons in the OECD area from 34 million in 2008 to 42.1 million in 2010… (OECD)
This serves as a cautionary note, as financial innovations, such as high frequency trading, can cause fallout.
In this paragraph, I’ll outline some initial warning signs surrounding high frequency trading (HFT). Although the earliest computerized trading can be traced to the 1980s, HFT popularized in the 21st century; further, on May 6, 2010, The DOW Jones Industrial Average had plunged from one investor utilizing HFT. To provide more reference into this incident, I have included the following quote from Martin Ford below:
The average trade dropped from 10 seconds to just 0.0008 seconds, and robotic, high speed trading was heavily implicated in the May 2010 “Flash Crash” in which the Dow Jones Industrial Average plunged nearly a thousand points and then recovered for a net gain, all within the same space of just a few minutes. (Ford, 24).
Navinder Singh Sarao, the trader, had implemented a spoofing algorithm, which caused the market to crash. Navinder had placed massive sell orders, which had brought down the price and motivated others to sell. Once the price was brought down significantly, Navinder turned off the algorithm, bought, and then waited.
Though, when Navinder managed to bring down the price of the futures contract, there was a flash crash. This strategy is illegal because Navinder committed fraud and benefited directly at the expense of others. Navinder wasn’t arrested for the “Flash Crash” and was warned to stop spoofing by the futures exchange; however, Navinder didn’t stop spoofing and, instead, he had responded to the FCM “Kiss my ass.” (Levine). He continued to manipulate the market for five years and then finally was arrested for numerous incidents of manipulating markets. Matt Levine provides insight in, Guy Trading at Home Caused the Flash Crash:
The Commission has presented evidence that, from at least April 2010 to January 2012; July 2012 to June 2014; and September, 2014 to present (“Relevant Period”), Defendants have manipulated, attempted to manipulate and/or spoofed the near month of the Chicago Mercantile Exchange […] Similarly, the CFTC presented evidence that Defendants have profited over $40 million from E-mini S&P trading during the relevant period. (Matt Levine).
Here, there’s repeated evidence of fraud and attempts to commit fraud that were permitted through HFT. Moreover, this evidence provides support towards enforcing regulations on HFT and increasing penalties on criminals engaged in this sort of fraudulent activity.
Having an economy in which spoofing is permissible would cause mass speculation and, in short, suggest a firm’s stock price follows an unpredictable path, as algorithms could simply manipulate its value. This differs from how the financial market currently functions, as public information detailing firm success is reflected in the stock price and not artificially inflated or deflated at the whim of a high frequency trader. Now, the contemporary market, at minimum, provides superficial comfort needed for traditional investors. When this comfort is eroded, investors will pull their money. To provide an example of investors losing faith in financial institutions, we should look at the Great Depression, as a record of 16 million shares3 were sold between October 24th and October 29th of 1929, or, more famously, what’s known as “Black Tuesday”. Consumer confidence vanished from the stock market crashing; further, this crisis was accompanied by a decline in spending and, correspondingly, led businesses to layoff numerous amounts of their employees. Linda Levine, labor economics specialist, provides insight into unemployment during the Great Depression:
The unemployment rate rose from 3.2% in 1929 to 24.9% in 1933 during the Great Depression’s more severe first downturn. While almost 1.6 million persons were unemployed in 1929, more than 12.8 million individuals lacked jobs in 1933. […] Although the Great Depression came to an end in June 1938, the unemployment rate averaged 19.0% for the year. (Linda Levine, 8).
In 1929, a decline in consumer spending had led to an increase in unemployment both domestic and abroad. Therefore, ensuring consumers have faith in the financial markets is crucial for ensuring a healthy economy; however, this faith in financial markets is violated when high frequency traders utilize spoofing algorithms. To prevent unemployment rates from increasing, there must be some regulations on high frequency trading. Now, the question that remains to be answered is the following: what regulations, if any, must be enforced? This isn’t a simple question to answer, however, in the following section, I’ll outline one possible approach. Finally, I’ll advocate in favor of firms adopting an internal policy to train their employees to use technology.
The first provision prohibiting spoofing appears in the Dodd-Frank Wall Street Reform and Consumer Protection Act following the 2008 financial crisis, as a measure to improve market transparency; however, the government has historically struggled to combat high frequency trading, as well as spoofing. To provide more insight into this provision, I outlined Capital Expenditure Authorization Section 4c(a)(5):
It shall be unlawful for any person to engage in any trading, practice, or conduct on or subject to the rules of a registered entity that- is, is of the character, or is commonly known to the trade as, “spoofing” (bidding or offering with the intent to cancel the bid or offer before execution. (Cornell).
In accordance with the statutory language of this provision, the term “spoofing” seems largely unspecified. To the trade, spoofing is a trading strategy wherein a large order is placed to long or short an asset, a stock, and the converse is placed on the opposite side; though, the trader has no intention to trade the larger order. Instead, the intent is to trade the smaller order and cancel the larger order, which creates market speculation.
Previously, we’ve discussed one instance wherein Navinder Sarao had utilized various spoofing techniques. Specifically, Navinder used an advanced layering algorithm and a traditional, manual spoofing technique. Manual spoofing, again, was found illegal in Section 4c(a)(5), however, there wasn’t consensus on layering. To provide more background into this layering algorithm, I included the following passage from the CFTC:
Sarao created a special algorithm referred to as “the Layering Algorithm,” which automatically and
simultaneously layered multiple exceptionally large price offers that shifted to ensure a gap of three to four levels away from the best asking price. (Miller, 4).
The CFTC had released this in a Proposed Interpretive Order pertaining to spoofing, which held a market participant must act with some degree of intent to engage in trading practices prohibited in Section 4c(a)(5). Now, there has been action taken in the past against traders engaging in manual trades; however, how do we attribute intentionality in the context of high frequency trading, as these transactions are all automated? Intentionality, in this case, should hold that trading programs are an extension of the trader’s own purposes. Navinder Sarao was, ultimately, found guilty for both manual spoofing, as well as this layering technique.
Spoofing is an extension of traders and, some might advocate in favor of removing human traders; however, it’s not clear how this would be possible, or, more acutely, whether this would even be desirable. Presently, traders engaging in automated trading establish constraints for these trading programs to adhere. These programs learn through training protocols where they trade between themselves, similar to self-play.
When the program is suitable for real markets, the training protocol is turned off and it engages in real trade. It’s possible that these constraints are diminishing the expected return on an algorithm, however, these rules are needed to prevent market manipulation, such as in the context of layering or other spoofing techniques. With trades becoming increasingly more automated, the fallout caused by one computer has exacerbated; further, one oversight can turn into a series of errors and, ultimately, cause a firm to lose millions of dollars. To provide one example, I’ve outlined the incident response of Knight Capital, an investment firm that lost $460m from high frequency trading due to a software bug that had effectively bankrupted the corporation:
On August 1, Knight did not have supervisory procedures concerning incident response. More specifically, Knight did not have supervisory procedures to guide its relevant personnel when significant issues developed. […] Knight’s system continued to send millions of child orders while its personnel attempted to identify the source of the problem. (Knight Capital Americas LLC).
Here, it’s important to note that one error early in the development of a software patch led to several errors; further, this firm lacked the required supervision, quality assurance, needed before updating the algorithm. In response to this case and others, the Securities Exchange Commission (SEC) has mandated rule 15c3-5. This rule necessitates brokers, or brokerage firms, appropriately control risks associated with market access so as to not jeopardize their financial condition, that of other market participants, or the market’s stability.5 However, it seems that in the context of unsupervised high frequency trading there’s a lot of unhedged risk. This is because errors that aren’t caught early in the development process can lead to catastrophic outcomes. The potential for one glitch to cost a firm hundreds of millions suggests the need for more quality assurance.
To hedge risk on the behalf of brokerage firms and maintain market stability, financial analysts are needed. There may come a time when analysts aren’t needed but for now, financial analysts can be a valuable asset.
Now, I will establish the value of financial analysts in high frequency trading and risk management. Previously, I’ve stated that algorithms adhere to rules contained in their software that developers’ program. These constraints, often, derive from legislation and prohibit algorithms from engaging in explicit activities. This necessitates updating trading software and these updates should be supervised, as there’s a lot of risk. It would, seemingly, be impossible for financial analysts to review every transaction before it’s processed; however, there should be another program that reviews transactions and notifies analysts of suspect actions. To expand on this proposed safeguard, I included a quotation from the CFTC featured in RIN 3038-AD52:
[…] a pre-trade risk control such as a message throttle will prevent submission of orders that exceed a predetermined frequency per unit time. Such a control could be operated by the market participant generating orders, the clearing firm guaranteeing its trades, or the trading platform on which orders would be executed, and would limit the impact of an algorithmic trading system not operating as intended. (RIN 3038-AD52, 54).
Here, there’s an array of areas for a financial analyst to supervene when an automated system malfunctions, or appears to be attempting to engage in spoofing; however, these financial analysts will need more training. There ought to be a clear process at each station, otherwise there’ll be an increase in human error and time. Time, again, is essential in the context of high frequency trading but this time delay shouldn’t be a problem, as delaying all investors equally wouldn’t alter the order in which the contracts were received and executed. The former point on experience, however, is a legitimate concern, as errors can result in catastrophic fallout.
Firms should train financial analysts to work with a diagnostic software that spots potential errors or fraud. To expand on this idea, I’ve provided background on a similar human-computer hybrid approach at PayPal:
The fraudsters’ adaptive evasions fooled our automatic detection algorithms, but we found that they didn’t fool our human analysts as easily. So Max and his engineers rewrote the software to take a hybrid approach: the computer would flag the most suspicious transaction on a well-designed user interface, and human operators would make the final judgement as to their legitimacy. Thanks to this hybrid system – we named it “Igor” […] we turned our first quarterly profit in the first quarter of 2002 […] The FBI asked us if we’d let them use Igor to help detect financial crime. (Thiel, 108).
PayPal was experiencing a problem that had caused the company to lose millions and needed a solution. The company incorporated a solution that involved using automatic detection algorithms, but it didn’t work. This problem can, also, be seen in the Flash Crash, as Navinder had incorporated a new layering algorithm; however, this layering algorithm was not detected until later and had allowed the market crash to continue. PayPal had dealt with this problem of fraud through the incorporation of a human-computer hybrid strategy. This strategy proved effective for the company and should be implemented to address high frequency fraud; further, this approach should work to stop automated trades from orders that would bankrupt corporations. Thus, multiple parties can benefit from the use of financial analysts in partnership with diagnostic software.
In this paper, we examined high frequency trading and discussed some benefits, as well as dangers. In the future, artificial intelligence may replace analysts, however, there’s currently too much risk involved. There’s an incentive for traders to spoof the market and human-computer hybrids provide the best defense; further, in the case of Knight Capital, there’s powerful reason to safeguard against malfunctioning software. These cases caution profit-centric firms from replacing all of their financial analysts with automated traders. Regulations on financial institutions prohibiting complete replacement of financial analysts must continue. Otherwise, there’s an increasing likelihood that high frequency traders will cause the market to crash again. We can see such catastrophe in the Flash Crash, 07–09 Financial Recession, as well as the Great Depression. Now, these financial analysts should receive more inhouse training to adapt to technological advancements. Financial institutions must be cautious with high frequency trading, as their actions impact millions of lives. Thus, there must be more supervision over firms engaging in high frequency trading and financial analysts, for now, provide the proper safeguard.
Works Cited
Ford, Martin. Rise of the Robots Technology and the Threat of a Jobless Future. Basic Books, 2016.
Friedman, Milton. “The Social Responsibility of Business Is to Increase Its Profits.” SpringerLink, Springer, Berlin, Heidelberg, 1 Jan. 1970, link.springer.com/chapter/10.1007/978-3-540-70818-6_14.
Bussey, John. “Are Companies Responsible for Creating Jobs?” The Wall Street Journal, Dow Jones & Company, 28 Oct. 2011, www.wsj.com/articles/SB10001424052970204505304577001930473006096.
“Guy Trading at Home Caused the Flash Crash.” Bloomberg.com, Bloomberg, www.bloomberg.com/opinion/articles/2015-04-21/guy-trading-at-home-caused-the-flash-crash.
“Impact of the Economic Crisis on Employment and Unemployment in the OECD Countries.” OECD, www.oecd.org/els/emp/impactoftheeconomiccrisisonemploymentandunemploymentintheoecdcountries.htm.
Levine, Linda. “Long-Term Unemployment and Recessions .” Fas.org, 22 Nov. 2010, fas.org/sgp/crs/misc/R41179.pdf.
“7 U.S. Code § 6c. Prohibited Transactions.” Www.law.cornell.edu, www.law.cornell.edu/uscode/text/7/6c.
Miller, Jen Paul, et al. “The Anti-Spoofing Provision of the Dodd –Frank Act: New White Collar Crime or ‘Spoof’ of a Law?” Www.thompsoncoburn.com, www.thompsoncoburn.com/docs/default-source/News-Documents/spoofing.pdf.
“ORDER INSTITUTING ADMINISTRATIVE AND CEASE-AND-DESIST PROCEEDINGS, PURSUANT TO SECTIONS 15(b) AND 21C OF THE SECURITIES EXCHANGE ACT OF 1934, MAKING FINDINGS, AND IMPOSING REMEDIAL SANCTIONS AND A CEASE-AND-DESIST ORDER.” Www.sec.gov, www.sec.gov/litigation/admin/2013/34-70694.pdf.
“RIN 3038-AD52.” Www.cftc.gov, www.cftc.gov/sites/default/files/idc/groups/public/@newsroom/documents/file/federalregister112415.pdf.
Thiel, Peter, and Blake Masters. Zero to One: Notes on Startups, or How to Build the Future. Virgin Books, 2015.
