A network graph of Corporate America
Having loaded a bunch of SEC Data in Freebase, I started exploring what sort of cool visualizations one might do with it that can show a lot of different data at once. I created this chart, which shows companies that are connected by shared board members. It got a lot of my business-type friends very excited:
Update: Not all of the data about when someone left a board is available. Therefore some links may represent people who have since stepped down.
(click on the preview to see the full size version)
I tried to put as much data on here as I could — genders of the people, CEO indicators, market cap, revenue, location, etc. What’s fun about data visualization is that rather than just making a point, it sometimes makes us ask questions that we might not have otherwise considered.
Connectedness
I generated a graph from the 400 largest companies by market capitalization. What’s shown here is the largest strongly connected component, which has 212 nodes. So 212 of these companies are run by people who all work together through some number of steps.
Gender balance
Upon looking at the graph, I noticed a surprisingly high amount of red (female board members). The reason it’s surprising is that only 12% of corporate board members from the sample were female — this graph, however, shows only people who are on more than one board. A quick calculation showed that about 21% of the lines are women. This difference is statistically significant to <0.001%. It supports anecdotes I've heard about women who manage to reach that level being asked to join more boards.
Tech clustering
Also striking was how closely tied all the West Coast technology companies are. At the bottom of the graph, you find Oracle, Cisco, Google, Yahoo and Sun, all tightly clustered. The rest of the graph does not seem to exhibit such tight clustering within an industry. Digging into the data, we find that a lot of successful West Coast VCs end up serving on the boards of many of the tech companies in which they’ve invested.
Energy industry
The deep green on the graph immediately draws out eyes to the companies with the most revenue. Looking quickly, besides Wal-mart, we see that the energy industry has been doing very well recently.
The board members were generated from SEC filings forms 3 and 4 for 2007, so the data may be outdated or incomplete. Again, I’m just trying to get you thinking!
May 1st, 2008 at 4:39 pm
Great chart! Microsoft seems kinda lonely for its size.
Would be interesting to see a variation where you focus on board members, e.g. by calculating the weight of a person based on the market cap of companies he or she is involved with.
May 1st, 2008 at 4:40 pm
Interesting post. What would be interesting is visualising this kind of information dynamically in a tool such as Centruflow. This would allow for the data to be constantly as up to date as the information in the database, whilst still giving a graph with just as much semantics as your picture above has.
A bit of a disclaimer: I’m the developer of Centruflow. I’ve never come across freebase before, and am always on the lookout for interesting datasources to visualise. If you think freebase is a possible datasource for a tool like Centruflow, I’d love to hear your thoughts.
Cheers,
Jonathan Giles
May 1st, 2008 at 6:16 pm
Really interesting would be adding data for felony convictions, out of court settlements, and those secretive wrist slappy deals so popular under the current regime.
In the past the FBI has refused to publish white collar crime statistics, probably because things like Enron’s ponzi scheme weighed more in monetary damage than YEARS of highly reported human crimes.
May 1st, 2008 at 7:36 pm
Can you explain the statistic you used to work out that the females were over-represented in this graph compared to the whole example - is it a Chi2 test, or something different?
May 1st, 2008 at 7:38 pm
kd –
It’s a cumulative hypergeometric test. It calculates the probability of getting a given number of successes in a sequence of n draws from a finite population without replacement.
May 1st, 2008 at 7:49 pm
This is interesting. I was wondering if you could explain how to access the same data from Freebase (I could only find an incomplete and error-prone list that looks like it does not come from SEC data). Could you also share what other kind of SEC data (if any) you have put into Freebase?
May 1st, 2008 at 8:15 pm
Wondering what visualisation package/software you used to generate this ?
May 1st, 2008 at 8:26 pm
Great stuff Toby, as always. Really intriguing.
May 1st, 2008 at 10:49 pm
awesome graph! do you think I could get a copy of the data you used?
May 2nd, 2008 at 1:12 pm
Do you draw any connection between the relationships you’ve discovered here with the notion of ’social’ networks? I’m thinking that the networks you’re describing would be more akin to corporate incest based on business need. I’d be interested in your perspective.
May 2nd, 2008 at 2:32 pm
[…] kiwitobes.com » Blog Archive » A network graph of Corporate America Interesting data visualization. (tags: visualization statistics business network interesting corporate data) […]
May 2nd, 2008 at 4:02 pm
Same as monkeyhelper I’m interested what kind of software you used to make the visual?
Thx
Dennis
May 2nd, 2008 at 4:11 pm
Sorry if this is a duplicate; I think my last comment got filtered for too many links. The data is at least partially wrong: Google & Yahoo do not share a board member, but Apple and Disney do.
May 2nd, 2008 at 4:19 pm
Tim — you are correct, Google and Yahoo do not currently share a board member. The data is not entirely up to date, and in some cases there are links which were not true at the same time, since Freebase does not always contain the date that the person stepped down. For example, Michael Moritz has served on the board of both companies.
The reason that the Apple/Disney link is missing is that Jobs only joined the board of Disney in 2006 and doesn’t seem to appear as a director in any of the filings for 2007
May 2nd, 2008 at 4:20 pm
For those interested, the graph was generated by taking a GML file generated in Python and laying it out using yFiles/yEd
May 2nd, 2008 at 5:12 pm
[…] analysis and figure from my friend Toby of the social networks among the public company boardroom set. Click for a larger […]
May 3rd, 2008 at 9:31 am
[…] 3, 2008 in history and current events by eric The interlocking directorates, today: Toby Segaran makes a map of corporate board membership showing the 212 most connected of the top 400 companies by market […]
May 3rd, 2008 at 9:47 am
Besides Microsoft, Pfizer is also lonely (or autonomous) with links only to Exxon and to two small companies.
I wonder if density of linkage might not lead to inertia. It doesn’t necessarily seem like a good thing for the linked company, unless the links are all to dominated smaller companies.
May 3rd, 2008 at 12:19 pm
What do the diamonds represent, couldn’t find a key or symbol description.
May 3rd, 2008 at 1:16 pm
Great concept– however, it is difficult to read or decipher. I suspect it is too complex for all but the data junkie (like me!)
Thanks for a great idea!
May 3rd, 2008 at 1:58 pm
[…] wait, here’s something! A network map of shared board directors published today. Not dynamic, but interesting […]
May 9th, 2008 at 12:11 am
[…] over at kiwitobes posted a really neat visualization of the relationships between board members in corporate America. […]
May 23rd, 2008 at 4:09 pm
Toby-
Looks great!
Can you please share the code? I could not find this kind of example in your book.
Thanks!
June 13th, 2008 at 1:36 am
[…] in San Francisco. They’ll be presenting Freebase’s new features as well as discuss some interesting mashups and visualizations. So if you’re in San Francisco, RSVP now, and go get some free pizza, beer, and […]
December 4th, 2008 at 5:46 am
super, ya tozhe tak hoshu
January 12th, 2009 at 10:37 pm
Great chart!
May 9th, 2009 at 1:43 pm
Fun stuff!
You might be interested in They Rule http://www.theyrule.net/ which takes an interactive look at the same data, although it doesn’t really do any analysis. It was made by Josh On of Futurefarmers.
August 22nd, 2009 at 10:31 pm
Hi, Toby,
Do you mind to explain if the distance between two nodes means anything? For instance, are companies that share more directors closer than the others? If not, how is the distance between the nodes determined?
Thanks.
September 17th, 2009 at 4:10 am
Hi, Toby,
Do you mind to explain if the distance between two nodes means anything? For instance, are companies that share more directors closer than the others? If not, how is the distance between the nodes determined?
September 17th, 2009 at 4:11 am
Edit AVCHD is a pratical and professional
AVCHD Editing software for HD Camcorders.Easily archivable and vesatile functions make AVCHD
Editor stand out among others.
September 17th, 2009 at 4:12 am
Blu Ray DVD Ripperis currently the
best software which can help users to ripper Blu Ray DVD to SD/HD video formats, such as
AVI, DivX, XviD, WMV, MPEG4, MOV, MKV, FLV, HD AVI, HD MPEG, HD MOV, H264/AVC, HD WMV, etc.
easily and fluently.
January 13th, 2010 at 6:54 pm
What do the diamonds represent, couldn’t find a key or symbol description.