News and Blog

Social Media and the Energy Sector

18 Sept 2013
by Niels Rosenquist
C.E.O Janys Analytics

IHS is a global leader as a provider of insight to the energy sector and hosts one of the most important conferences related to the global energy market, IHS CERAWeek. When IHS asked us to take a look at how social media can be analyzed to look at issues of interest to attendees at CERA week, we came up with some materials that were presented today at the conference.


1) What can large-scale social media tell us about sentiment related to important topics in the U.S. energy market?

2) How do the connections between users of social media impact the spread of messaging related to energy topics?

3) How did the movie “Promised Land” impact sentiment on hydraulic fracturing (or “fraccing”)


Twitter data (supplied by GNIP) from December 2012 through February 2013. In particular, we looked at approximately 20,000 “tweets” related to fraccing, as well as the film Promised Land, which was released on January 3rd 2013.


Applying our proprietary sentiment and network technology to the data, we analyzed the general sentiment of the US population using Natural Language Processing algorithms. That supplied sentiment for locations (at the county level), time, date, and specific user. We then looked at a select number of followers for individuals we knew to represent important stakeholders in the industry. Finally, we applied our technology to determine what communities of users these individuals belonged to, as well as how the communities were related to each other globally.


We found a few interesting things. First, it appeared that sentiment regarding fraccing was lower than the national sentiment (without regard to topic) over the time period measured. Notably, this sentiment has been stable throughout the past year, and did not change upon the release of Promised Land, a film that carried a strong anti-fraccing message. This led us to ask some additional questions:

Where are people messaging about fraccing?

Figure 1 shows a subset of messages geo-located to the user’s home area. Much of the distribution can be seen along the highly populated coastal areas. However, one can also see some significant messaging levels in areas above the shale deposits where public sentiment is particularly relevant to public policy and drilling permits.

Figure 1

Who appear to be particularly influential individuals within the Twitterverse?

Here, we sought to identify a number of users who were well known stakeholders in the energy space. We began by analyzing the followers of users from a variety of sectors, including oil and gas executives, environmental activists, media members, and public officials.

Figure 2

From these people‚Äôs networks, we were able to generate “communities” of closely tied individuals. Using our technology, we were able to determine a number of communities who are highly interlinked as well as interlinked with other stakeholder groups.

It’s interesting to note that a couple of individuals appeared to be in communities that were not related to their actual occupation. Since we based community membership upon whom someone is linked to, and not their stated occupation, these situations can occur. This form of being “embedded” in another community can be important, especially if the community someone is in is part of their target audience.

For our assessment of influence, we did not rely upon the number of a person’s followers, which is a non-specific predictor of influence (i.e. getting messages out). Rather we looked at whether or not a person’s messages were likely to be re-tweeted out of their follower network. This likelihood is based in large part on the lattice-like structure of individual’s networks that goes far beyond their own followers and whom they follow. The figure below shows the final energy network, along with who appears to be influential within those communities.

Figure 3

This kind of analysis is not meant to minimize the importance of the message(s) being sent from users. Clearly the content of the messages is important. However, the linkages themselves can be just as important. To use an analogy from the power industry, predicting the impact of a power surge has as much to do with the layout of the power grid as it does the size of the surge itself.

What about Promised Land? What explained it’s apparent (lack of) impact?

Given the focus and message of Promised Land, there was a non-zero likelihood of it driving chatter and exposure to one particular point of view about the subject. What we found, however, was that while chatter about fraccing increased around the time of the release of the film, interest in the film itself was minimal, with chatter about Promised land an order of magnitude smaller than even low-budget Hollywood releases. While it cannot be proved conclusively with the data, it appears reasonable to assume that the anti-fraccing sentiment the movie was limited in its spread in part due to the fact that people did not show a lot of interest in the movie itself, something reflected as well in the eventually disappointing box office results for the film.

Figure 3

We have a white paper that contains these results as well as some descriptive statistics. If you would like a copy, please contact us at info@janysanalytics.com.