A Nonlinear Impact:
Evidences of Causal Effects of Social Media on Market Prices
We provide empirical evidence that suggests social media and stock markets have a nonlinear causal relationship. We take advantage of an extensive data set composed of social media messages related to DJIA index components. By using information-theoretic measures to cope for possible nonlinear causal coupling between social media and stock markets systems, we point out stunning differences in the results with respect to linear coupling. Two main conclusions are drawn: First, social media significant causality on stocks' returns are purely nonlinear in most cases; Second, social media dominates the directional coupling with stock market, an effect not observable within linear modeling. Results also serve as empirical guidance on model adequacy in the investigation of sociotechnical and financial systems.
Keywords: #Financialmarkets, #complexsystems, #socialmedia, #nonlinearcausality, #informationtheory
Reducing Uncertainty in Uncertain Markets:
Investigating When Directed Social Sentiment can Lead Individual Stock Volatility
Sentiment-loaded Twitter messages that reference an individual security/stock (e.g. $APPL, Apple Inc.) have been verified to reduce uncertainty about the corresponding security’s future volatility by calculating the Mutual Information Surplus between sentiment-loaded Tweets, S and the change in daily high and low prices, Fvolatility. In concordance with the paper’s results, such predictive behavior may require a critical mass of Tweet volume, which is characteristic of large-cap stocks. 1 From 01/14 to 12/15, 97 NYSE stocks exhibit significant leading information surplus when using sentiment as a leading indicator of Fvolatility with an average optimal lag at 5 days, suggesting uncertainty about an individual security can be reduced five days prior to a given trading day.
Keywords: #k-means, #mutualinformation, #sentimentanalysis, #volatility
Non-Linear Relationships Between Stock Sentiment and Volatility:
An Unsupervised Clustering Approach
We explore new methods in examining non-linear relationships between Twitter sentiment data and the stock market. We created the Clustered Social Index (CSI). The CSI is an index based on clustering of companies’ sentiment data. Its predictive ability over the volatility of the stock market is tested.
Trading on Social Media Mood
This paper will discover the influence of social media on DJIA components’ prices and the ways to utilize it. Firstly, causal relationship between social media and stocks are studied. Secondly, linear and non-linear models are built to predict direction of stock prices. Strategies to use aforementioned predictions are suggested and evaluated with return to volatility metric. The study showed existence of linear causal relationship between Social Media and volatility, whereas linear causality between SM and log returns occurred only in 2 stocks out of 30. The trading strategies suggested in paper outperformed the DJIA index and other strategies based on models without Social Media signals.
When Can Social Media Lead Financial Markets?
This paper explores the social signs of financial dynamics. Existing academic research shows social media does lead financial markets, we have built on this by performing a direct examination of “When?”.
To do so, we seek to identify an increase in Mutual Information (MI) and Pointwise Mutual Information (PMI) scores when social media data is offset in an ex-ante configuration to precede data from financial markets. Following this, we cluster output results, on a company-wide and daily-wide basis, to determine the most informative social and financial configurations. Applying our method to the NASDAQ Top 250 (by Market Cap) yields the result: 176 companies posses a statistically valid Net Information Surplus, maximised in the technology sector and strictly associated with periods of increased activity (social and financial), strong positive social sentiment intensity and good financial performance (log returns). This concludes that social media data can lead financial markets and proves an answer to “When?” can be given as a pattern of social and financial variables tailorable to each (group of) stock(s).
Research and implementation was completed as part of the UCL MSc Business Analytics (with specialisation in Computer Science) degree for the COMPG011: Data Analytics module.
Keywords—#SocialMedia, #FinancialMarkets, #Time-seriesanalysis, #Ex-anteconfiguration, #MutualInformation, #k-meansClustering.