Link to

March 15, 2011

Followers and Foes: Defining Industry Groups with Twitter

Delineating industry groups of related firms and identifying strategic peers is important for both financial practitioners and scholars. We have explored the relationship of S&P 500 stocks in terms of joint mentions in stock microblogs (for example, the tweet “Big banks up or down with Bernanke's re-nomination? $C $BAC" jointly mentions Citigroup and Bank of America and thus associates these companies with each other). The figure below shows the relationship of S&P 500 stocks in terms of joint mentions in 439,960 English-language, stock-related microblogging messages. For a rough comparison with the classic SIC industry classification, the shape and color of each stock symbol represents the one-digit SIC industry group. We can interpret proximity of a group of stocks roughly as the degree to which they are related.

We find that many subgroups of stocks are consistent with our intuition of classic industry delineations. Financial firms, such as Goldman Sachs (GS), JP Morgen (JPM), and Bank of America (BAC) are closely interconnected. There are also tight-knit smaller groups of stocks, such as the media companies Disney (DIS), CBS Broadcasting (CBS) and Time Warner (TWX). In addition, some subgroups exist that are not connected to the rest of the network, for example the insurance companies WellPoint (WLP) and United Health (UNH), the logistics firms United Parcel Service (UPS) and Fedex (FDX), and the hardware stores Home Depot (HD) and Lowe’s (LOW). Their isolated position suggests that these companies form micro industries that are often subject to the same news items, but not frequently associated with other firms.
The network graph not only confirms classical industry groupings, but also reveals interesting connections between these industries. For example, while Exxon Mobile (XOM) is obviously closely related to other major energy firms such as Chevron (CVX) and ConocoPhillips (COP), it is also linked to subcontractors such as the exploration firms Anadarko Petroleum (APC) and Halliburton (HAL), which in turn are associated with their equipment suppliers Schlumberger (SLB) and Cameron (CAM). Note that these relationships cut across traditional SIC categories. Online retailer Amazon (AMZN) is another interesting example in this respect, because the company appears to be a hub that is associated with traditional “brick and mortar” retailers, such as Walmart (WMT), Target (TGT) and Costco (COST), online retailers, such as Priceline (PCLN) and Ebay (EBAY), computer soft- and hardware firms, such as Apple (AAPL), Microsoft (MSFT) and Intel (INTC), and communication providers, such as AT&T (T). The network also highlights particularly competitive relationships, such as those between Coke (KO) and Pepsi (PEP) or Visa (V) and Mastercard (MA), as well as issues related to corporate control and ownership, such as the strong link between Disney (DIS) and its parent company General Electric (GE).
We conclude, that joint mentions of stocks on Twitter offers unique and rich insights into the relationship between companies.

The size of the stock symbol represents the total number of mentions, the thickness of the lines is indicative of the relative frequency of joint mentions (i.e. relatedness of two stocks). For better readability the figure was limited to stocks that were mentioned together at least 50 times.