Enron Social Network

An enjoyable project from my Social Network Analysis course involved creating and analyzing the social network formed through the publicly released Enron emails.

Duration: 2 weeks

HCI Methods Employed: Social Network Analysis, Data Visualization (Gephi)

For the course in Social Network Analysis, we had a project analyzing the social network derived from the Enron emails.

Within the released network, there were 1102 nodes, which in this case are email addresses, 1802 directed edges, out of 1,213,302 possible edges, with a density of 0.001485.

The Enron network appears to be very hierarchical, most information flows from the top down, and there were not many cases where information went from the edges to a more central location. The majority of the centrally located nodes, using the weighted Eigenvector centrality scores, were executives with Dasovich (1), Shapiro (0.829), Steffes (0.729), Mara (0.722), Kean (0.518) and Richard Sanders (0.339). Most of the non Enron emails were very low in centrality other than gfergus@brobeck.com, mday@gmssr.com, rcarroll@bracepatt.com, and dwatkiss@bracepatt.com. Of those Gary Fergus is the most central with an Eigenvector score of 0.51813 and a weighted degree of 24. While he is now the Attorney General, at the time of the Enron scandal he was a 21 year veteran at the lawfirm: Brobeck, Phleger and Harrison in San Francisco. His only connections outside of the pink clique are Kean and a few Enron Executives. This lead me to believe that the pink clique is likely the in house legal team. From the content of the emails, it seems that he was involved with bankruptcy conference calls, handling the government’s subpoena on Enron (with Richard Sanders), and the CPUC investigation. Most of the emails were about conference calls or providing phone numbers, so not much of a record of his interactions with Enron can be determined.

Most of the information flows from the central part of the network, to what I am assuming are middle management, to workers. Very rarely did workers in a clique (likely a department) have emailed exchanged between them. Most of what connects them electronically is emails from superiors. The middle management seems to have written to some higher ranking/ more central managers, but the bottom level or two receives email but does not send it to anyone in the network. One of the biggest exceptions noted on the network graph is Kaminski, who emails several people less central in the network, but also receives email from some of them. He has the 4^th highest in-degree score with 75 connected nodes. Kaminski appears to have been a risk analyst at Enron. While he did not talk to members outside his clique often (forest green), he communicated bilaterally with many members within his clique. Only three of his connections are more central to the graph than him, which include Greg Whalley (president after Mr. Skilling), Wade Cline (legal & CEO of Enron India) and Billy Lemmons (Vice President). Given Cline and Lemmon’s importance to the company, it is really surprising the only email exchanges in network are to Kaminski and Kean. Whalley had some edges connecting him to the olive group, but Kaminski is his only green connection, and Kean his only blue. With such a dynamic field as risk management, in a company that so actively needed it, Kaminski’s emails covered a broad range of topics. He communicated a bit with the research and development group, as well as external calls to Power Risk Management firms, data security, and correspondence with potential hires for Enron. He dealt so much with hiring, he had a very full folder on his email just for keeping resumes. Several emails he sent were regarding internal Enron politics, for the research department in particular. I think the involvement in internal politics is one of the reasons for the high in-degree score without an equally high Eigenvector Centrality score (0.172). He is in a position of power, and does have an array of responsibilities, but deals more with management and politics within his own department of Risk Management. He also had several emails corresponding with people outside of Enron about his work presented at conferences. His role at Enron was primarily technical, with some managerial roles of technical individuals. Given the breath of his knowledge/ command, the connections to universities, and the number of people with a PhD he was to interview, I’d also expect that Kaminiski is highly educated, either formally or informally. He dealt with data mining, information security, and data bases before there were many classes or widely distributed information on such subjects.

The node with the highest Eigenvector Centrality Score (1) and 6^th highest in-degree score (69) is Enron’s Government Relations Executive Jeff Dasovich. He emailed several key characters who are not all highly represented on their main emails, on alias emails. These emails to Kean, Shapiro and Steffes to non enron emails shows a closer relationship than those who only emailed their company emails. These “Skytel” emails also were very brief with very little formality or proper grammar. Most of these were sent for coordinating logistics for meetings. Dasovich also penned many jovial personal correspondences to government employees, thanking them for great interpersonal reactions outside of work. His email indicates that he kept up to date with relative legislation dealing with energy, with many topics related to those stored in specific folders pertaining to their subject matter. One item in particular he kept tabs on was the California energy crisis, both in respective folders and in the trash, so might have played an advising or negociating roll pertaining to that. He seems to not have been close to Alan though, because there were many emails to Kaufman and Shapiro when Alan left, trying to understand the circumstances and implications. The ability to ask such things shows a decently close connection to those two. Another note worthy thing from his email is that he appears to be enrolled in an MBA program at the same time he worked at Enron. He has several folders labeled MBA with class titles following. While on the Gephi network chart, his importance is very clear, measures other than Eigenvector centrality seem to indicate he might not be quite as central as he appears to be. Using Creamer et al (2009)’s social score method, which weighs a combination of the number of emails, the response time, how often the respond, as well as factoring in the cliques, degrees and centrality he only had a social score of 37.52 out of 100. Which was still higher than the very connected Steven Kean at 33.02 but less than a few others. Whereas people like Lay, who have only 6 connections on our data set, has a score of 53.41. This makes me question whether the dataset shown on Gephi is the same as they used, while cited to be the same one that the government released. Regardless, I do think their method for tie strength using the information easily obtained by email has a lot of merit. I would fathom that Shapiro and Steffes Eigenvector centrality scores were also heavily affected by their ties to Dasovich, because Shapiro only has 19 weighted degrees (18 in, 1 out), and Steffes 32 (15 in and 17 out).

Mary Hain seems to be a harbor of information summarization on some issues to Dasovich, Mara, Kaufman and Comnes. She also served to communicate Enron employee/ logistical information to individuals outside the company. At the end of her time at Enron she was Director of Federal Regulatory Affairs and resigned in April 2001 before the scandal heated up, and shortly after the Subpoena. Since many previous academic works show that the Enron network grew more interconnected during the crisis, it would explain why she has only a weighted degree of 70 and Eigenvector centrality of 0.574, despite her position and connections to many people central to the network. Upon her exit from Enron, she had a few people inquire if they could communicate open positions to them, both from within Enron and its connected companies. Additionally she received many fond emails wishing her the best of luck in her new ventures. She seems to have been fairly personable/ approachable, despite the normally brief emails she wrote as part of her roll.

Steven Kean, Enron Chief of Staff, who as previously discussed has high centrality scores as well as degree weights, has an email inbox that reflects the breadth of departments he is in contact with. Due to the nature of his position as Chief of Staff he has ties across most of the cliques. Unsurprisingly he has a much higher out degree (350), and than in degree score (17), given the previously speculated on hierarchical nature of Enron. He sends out general emails to most departments, but only the executive and higher ranking managers address him directly. A lot of the emails he does receive are forwarded, giving him a general idea of what each department is up to. Surprisingly a lot of the emails addressed to him aren’t overly formal, but aren’t really very personal either. They are short and to the point, including a lot of information on current statuses of projects, many of these are communicated to him by Kelly Kimberly for the California crisis. From research she appears to be Ken Lay’s spokes person at Enron with a focus on marketing and public relations. Other projects seem to have a person in a similar position to Kimberly, whom for the most part also address Kean informationally and without formalities. There were a few personal connections noted, such as Wyoming official S. Reyno, and Enron’s Elizabeth Linnell, whom sent condolences at the loss of his father. Even these were rather formal and not overly personal though. The impression of his connections is that while he was widely connected with many groups of people at Enron, he was not particularly close to most of them socially, and likely was a busy man mostly interested in facts. He does not seem to be as political or charismatic as many of Enron’s executives. But he does seem well organized and to have positioned his network so that he can stay on top of things without having emails from too many sources.

The degree distribution of the Enron network appears to be almost exponential in nature. There are a lot of employees who aren’t very well connected, and a few individuals who are well connected to the network. The most connected individual was Kean, but even he only had about 32% of the network as first degree email connections. The next most connected was Alan Comnes with 7.6% of the network as a first degree connection. The average degree in the network was 3.27, with the average weighted degree being 1.635. For this particular instance, it might measure the hierarchy of the company from CEO to the bottom employee. I was not able to find a reliable measure of what an average company’s degree within network was, additionally the Enron network contains emails from people who don’t work at Enron, but with that information could make more sense of the average degree. From examining the chart it appears that many of the cliques have 2 or 3 levels of managers/ communicators to reach the edges of the graph. Most of them go up in one direction, the exception is the teal group, which I am assuming are the research groups given their organization and ties to Kaminski, which have cliques within the module. The pink, likely legal or regulatory team, has sub-departments as well. The forest green group seems to be the further removed, and some of the individuals are 6 or 7 degrees removed from the network’s center, this group appears to be involved in compliance.  The Navy group is unique in its structure, that they are all directly directed by Kean, but don’t have much structure within their module group. About half of those navy nodes have non Enron email addresses, so are comprised of outsiders and Enron employees who were not listed on the index/ able to view their email boxes. The hot pink and greenish teal modules are unique that they are connected, and almost entirely comprised of non Enron emails. The nearest Enron point they seem to converge on is jdasovic, which would appear to be an alternative email node to the previously explored Jeff Dasovick, and is indeed a first degree connection to the official Jeff Dasovick. This email only has an in-degree of 4 and no out degree, and also is unfortunately not listed in the index for the data base. Upon searching the database, the emails that are accessible to that address appear to be a lot more casual in nature than the other address, with subject’s such as “hey whassup? you still mad at moi?”.  With many of the emails being to a csilva@redsky.com and planning outings and talk of Burning Man, as well as 2740741@skytel.com that seem more platonic. Perhaps that was Dasovick’s personal email, despite it being associated with Enron.

With more data mining tools (such as network changes over time) and standards with which to compare Enron’s measures with other companies, more analysis could be completed.