Why Public Dissentiment Works
Public Dissentiment, an online tool that helps protestors negatively impact the price of a publicly traded stock by using the same sentiment analysis tools used by many stock trading algorithms. In the contemporary stock markets, high frequency trading bots that continuously buy and sell stocks have replaced human market makers. These bots provide the market with liquidity by being ready to take the buy or sell end of a stock trade most of the time. But HFT bots are also risk adverse, and withdraw from the market in time of uncertainty. When this happens, the market experiences an almost instant decline in the price of a stock, or group of stocks, followed by a speedy recovery. This is known as a ‘flash crash’ and is now a common feature the electronic stock markets.
Trading bots are also analyzing news and social media posts to gain an informational edge. This provides a space for people to intervene and make their voices heard. In April of 2013, the Twitter account for the Associated Press was hacked and a deceptive tweet that said two explosions had gone off in the White House caused the S&P to drop $136 billion before rebounding. And the recirculation and renewed popularity of old news stories can also affect stock prices. In 2008, a six-year-old news story about United Airlines erroneously appeared in as a top link in Google’s search results, UAL’s stock plummeted 75% before the mistake was noticed.
Like financial trading bots, Public Dissentiment machine-reads news articles about targeted companies and performs financial sentiment analysis. Articles with a strong negative sentiment are saved. Protestors can then use Public Dissentiment to generate negative social media posts that link to an article with negative sentiment and use some of the negative words found in the article. When a swarm of protestors targeting one company is large enough to create uncertainty around that company’s stock, they can cause the high frequency trading bots to stop trading that stock creating a flash crash. Shareholders may ignore public protests, but they will notice swings in the company’s stock price.
In the not too distant past, if you traded on a stock exchange, you were certainly trading with a person. To ensure that there would always be someone willing to trade with major stock exchanges designated certain individuals or companies to be market makers.
A market maker is a trader who is ready to buy or sell a stock which provides ‘liquidity’ so that capital can always flow freely allowing the market to function continuously. When human traders were designated as market makers, they were required to buy or sell the stocks no matter what the current conditions of the market happened to be. Even if a stock was in freefall, they were obligated to give a seller a quote to buy that stock.
Although, in practice, this did not always happen. During the Black Monday market crash of 1987, the market makers at Nasdaq stopped answering their phones—leaving investors with no way to sell their stocks as the market fell. Some investors lost everything.
The backlash from this event prompted Nasdaq to make their computer network that automatically executes trades under 1,000 shares mandatory for market makers. The ‘Small Order Execution System’, or ‘SOES’, was not the world’s first automated trading platform, but it was the first intended for small-scale investors, and the place where network technology and algorithms became more important than social connections or an institution’s size.
The 1990s saw the emergence of a group of day traders who used software programs to consistently outperform institutional investors and market makers. Pejoratively dubbed ‘SOES Bandits’, these traders worked in brokerages that looked like arcades and played the stock market like a video game. Rather than trading directly on an exchange, these traders typically used a passive computer network called an ECN, or electronic communications network. Some ECNs actively worked to instill hacker values, such as information transparency and openness, into the electronic stock market.
Like SOES Bandits, high frequency traders (HFT) use structure of the network and special technology to perform in the markets. Many HFT trading strategies resemble digital activist strategies like DDoS attacks or spoofing, while other strategies involve building alternate fiber optic networks or data beamed through microwave radio or lasers. For better or worse, the continuous buying and selling of stocks by HFT bots has replaced a system of market makers.
Regulation National Market System (Reg NMS) is a series of initiatives by the US SEC intended to modernize trading regulations for the electronic market system—including HFTs. Reg NMS was established in 2005, but wasn’t fully implemented until 2007. Some of the rules intended to protect investors have unintentionally led to new HFT arbitrage strategies.
On May 6, 2010 at 2:45pm EST, the stock markets inexplicably collapsed by more than $1 trillion and rebounded in the course of 36 minutes. The exact cause of the crash is still debated, but in 2015, a London based man living with his parents and trading on his home computer was arrested for placing large sell orders that he would quickly cancel to drive down the prices of securities he wanted to buy. When HFTs and other algorithmic trading bots detected odd market behavior, they simply stopped trading, helping to perpetuate the panic. Flash crashes in individual markets are now an almost daily occurrence.
On April 1, 2013, the Twitter account for the AP was hacked and a fake tweet about two explosions in the White House was made. Almost instantly, trading algorithms reacted to the tweet by halting trading causing a flash crash. Interestingly, human traders knew the tweet was fake because it said “Barak Obama” was injured when the AP always refers to the president as “President Obama”.
The idea that markets have a sentiment can be traced all the way back to Adam Smith and the idea that market swings are motivated by either fear or greed, two irrational human emotions. Countless investment strategies have been grounded on the belief that these irrational tendencies can be detected and even measured through rational means. There are numerous products that compare market factors, such as the difference between three-month T-bills and the three-month LIBOR rate, the moving average of the S&P, and dozens of other such statistically produced measures.
More recently, sentiment analysis practices from artificial intelligence and computational linguistics research have been applied to stock trading. These techniques machine-read texts and use either a lexicon-based or machine learning approach. In a lexicon-based approach a corpus of words is labeled with positive, negative, or neutral sentiment is used to analyze the text by assigning the values in the corpus to the words in the text. In a machine learning approach, a corpus of texts is fed into an algorithm that finds statistical commonalities among the texts. The texts in the corpus may be pre-labeled as positive or negative (supervised machine learning) or unlabeled (unsupervised machine learning) along the algorithm to discern its own criteria for distribution.
Applied to finance, the fundamental difference between these different approaches to sentiment is that traders using computational linguistic approach are seeking an informational advantage by being able to read, process, and trade on information extracted from text-based data faster than their competitors. Traders comparing various market indicators are looking to discern market trends and predict what direction the price of a security is headed. Although, in practice, these two paradigms for sentiment influence each other and are sometimes even combined. Self-fulfilling prophecies are a common feature of financial speculation. For example, if a significant number of traders (algorithms or people) are using a similar formula to decide if a security should be bought or sold, they will all get the same result and their combined behavior will push the price of that security up or down, prompting other traders to take notice, and possibly follow suit. Public Dissentiment helps protestors communicate directly with computational linguistic-based sentiment analysis algorithms and potentially impact their behavior.
The creators of the following images used do not endorse Public Dissentiment.