InfoVis 2005 Contest
Geographical Visualization of Technology Data in the US

Contest webpage: http://tara.slis.indiana.edu/outgoing/infovis05/submission/standardform2005.htm

This webpage complements our two-page paper submission, video and complementary web page.

Authors and Affiliations:

• Colin Murray, Indiana University, Bloomington; University of Sydney, Australia; National ICT Australia cmurray@it.usyd.edu.au
• Weimao Ke, Indiana University, Bloomington wke@indiana.edu
• Hana Milanov, Indiana University, Bloomington hmilanov@indiana.edu
• Mark Meiss, Indiana University, Bloomington mmeiss@steinbeck.ucs.indiana.edu
• Sharavan Rajagopal, Indiana University, Bloomington shrajago@indiana.edu
• Katy Börner, Indiana University, Bloomington katy@indiana.edu

Tool(s):

Chizu was extended to generate the geographic visualisation with the US map as the background.

PostgreSQL database was used to analyze the dataset by creating queries to obtain different views of the dataset.

Microsoft Excel was used to generate the charts.

Adobe Illustrator was used to improve the look of our charts and to add legends, etc to the geographic visualisations.

Microsoft powerpoint was used to create our "video".

TASK 1: GeoVis and Chart of Company Movements

• Process:
  For the first image we calculated the number of times each company moved based on a different zip code from one year to the next. We then extracted the companies that moved the most. For each company we also calculated the distance of each move. We then plotted the sales and employees on a line chart and added bars to indicate when moves occured and the extent of the move. We also used Chizu to plot the trajectory of the company and added this to the chart.
  For the third image we used Chizu to draw the trajectories of company moves for all companies over the 15 years.

• Image 1.1 :
  
• Insight:
  From this image we can see the relationship between the movements of the company and its sales and number of employees. For this particular company we can see that along with a move in 1997 came a big burst in both sales and in the number of employees.
  From the inset GeoVis we can get a better understanding of the movements made by the company. We can see if the company moved long distances and can also see if the company returned near a location it was previously at.
• Caption for exhibit:
  Individual company movements and it's relationship to sales and number of employees.
• Image 1.2 :
  
• Insight:
  This image is similar to Image 1.2 but for a different company. The interesting insight found from this visualisation is that the company moved from the west coast to the east coast and back again during the 15 years. Despite this large change in geographical location the company was able to maintain steady growth in both sales and number of employees.
• Caption for exhibit:
  Individual company movements and it's relationship to sales and number of employees.
• Image 1.3 :
  
• Insight:
  Plotting trajectories of all companies over the map of US results in a rather cluttered visualization. A closer examination shows that there is a small set of persistent patterns of relocation strategies. Many companies move toward large cities and there is a great deal of movement between the major business centers of the East and West coasts.
  There are plenty of questions that this visualization invites. Unfortunately, the contest data does not provide insight for the reasons of relocations. Those would naturally depend on which part of the company changed the zip code. Here we speculated it is the headquarters of the company. Given richer data on R&D centers and production facilities locations, it would be fascinating to see potential emerging patterns in relocations of companies’ laboratories into certain locations, potentially forming a geographically distinct cluster of knowledge (e.g., caused by collaboration with strong research university centers). The results of these analyses could then be linked to differences in innovation rates between companies that relocated R&D facilities and those that did not. Intensive R&D is expensive, and some industries (e.g., biotechnology) heavily rely on cooperation (which is facilitated by geographical proximity) seeing such patterns evolving on geospatial maps would be really interesting, especially given potential consequences for firms’ performance and innovation.
  Direction of movement is a fruitful area for visualization, as it reveals possible patterns which then invite different questions. Reasons for companies’ relocations can be multiple, each of them interesting in its own way. In addition to the aforementioned formation of knowledge clusters, other reasons for relocations could be a result of specific developmental policies creating motives for relocation of companies in search for fruitful opportunities. On the other hand, some policies (e.g., placing strict ecological standards) might increase operating costs and drive companies away.
• Caption for exhibit:
  Trajectories of all company movements over 15 years.

TASK 2: GeoVis of Each Industry

• Process:
  We use a map of the US to generate density plots of all 18 industries. The year is a node for each company each year placed at the company's location that year. The recent years are colored lightly and the early years are colored darker. Nodes for recent years are drawn on top of nodes from earlier years. The node size is once again mapped to total sales.
  For the second image we show two industries (biotechnology and pharmaceuticals) on the one map in different color shades.
• Image 2.1:
  
• Insight 2.1:
  This map of the telecommincations & internet industry shows that it has experienced extensive growth in recent years. We can see this from the high percentage of large brightly colored nodes.

• Image 2.2:
  
• Insight 2.2:
  This map shows both biotechnology and pharmaceuticals on one map. Observing interdependence between industries is a fascinating area for business research. One such example is offered by the increasing interdependence between pharmaceutical and biotechnology industry. These interdependencies can be related to geographical proximity, which can be seen on the map. We can see biotechnology companies around large pharmaceutical companies. The importance of observing such linkages is emphasized when we think about some of the numbers involved:
  - Global sales of prescription (including both branded and generic drugs) and over-the-counter (OTC) remedies top $300 billion annually.
  - One factor driving the industry is the world's increasing elderly population. The over-65 set, which consumes three times as many drugs as younger populations, is expected to reach 690 million by 2025, and people are living longer thanks to drugs.
  - While the buyers may be living longer, monopoly profits from patents don't last forever. Different profit and noticed demand pressures force companies to come up with new products, and R&D costs are sky-rocketing.
  All these factors shaped the collaboration between profit-hungry pharmaceutical companies, and growing biotechnology companies. The interesting part is that the young biotechnology industry, with many start-ups needing money for research, knowledge and contacts for registration and distribution of their drugs, today boasts nearly 1,500 biotech-focused companies in the US, more than 300 of which are public. Fascinating history of Genentech serves as one of many examples of collaboration of a biotech company with a pharmaceutical sector. Founded in 1976, it marketed its first drug through a license agreement with a pharmaceutical giant, Eli Lilly. Indeed, developing alliances has been one of Genentech's key strategies for success from its inception. This is one of the numerous examples of an exchange relationship between pharmaceutical companies and young biotechnology ventures, where the pharmaceutical company offers financial support for a project, requiring in exchange the rights to discoveries of the biotechnology firm.
  These are just some of the numerous reasons for mapping these two industries in search for insights on their interdependence. Besides the sole geographical distribution of industries (e.g. visible Boston and San Francisco Bay area points in biotech), which is closely related, such mapping, particularly over time can give some interesting insights. To an observer, such maps can show evolution of the relatively young biotechnology industry. Scholars found that in biotechnology, companies which form many partnerships experience higher growth rates. Are these patterns location bound or spanning different states? Also, latest trends of industry consolidation could be interesting to observe, given the richer data. Who is buying whom? How many of the mergers/acquisitions are inter vs. intra-industry? Are companies seeking new knowledge buying biotechs or going for economies of scale and scope merging horizontally? Finally, are new pharmaceuticals being founded? Do they aim to be in geographical clusters close to knowledge domains, or instead going to new spaces in search for unsaturated markets?
  Observing, comparing and contrasting patterns of sales growth, companies being founded, their collaboration and M&A activity is a fruitful area for visualizations. Richness in data, and more detailed zoom-ins can give insights into interesting evolutionary patterns and interdependencies between these (and other examples of interdependent) industries, and serve as valuable sources of information on these dynamic and complex industries.
  Another insight we get from this visualisation is that there are no biotechnology or pharmaceutical companies in the state of Wyoming as this state is coloured white.
  
  Sources:
  
• Hoovers online, industry reports. www.hoovers.com, July 15, 2005
• Genentech corporate website, www.genentech.com
• Stuart, T. (2000): Interorganizational Alliances and the Performance of Firms: A Study of Growth and Innovation Rates in a High-Technology Industry. Strategic Management Journal, 21: 791-811.
• Powell, W. W., K.W. Koput & L. Smith-Doerr. (1996) Interorganizational Collaboration and the Locus of Innovation: Networks of Learning in Biotechnology. Administrative Science Quarterly, 41: 116-145.

TASK 3: Charting Bursts of Sales and Employees

• Process:
  For each year, we calculated the number of companies in each industry whose sales doubled from the previous year. We then plotted this information on a line chart. We used the same method for the number of employees.
• Image 3.1:
  
• Insight 3.1:
  What industries experience a sudden increase of the number of companies? Which ones had a sudden increase in their sales or employees? Bursts in the number of employees and sales might indicate that a company finally developed a marketable product while prior to that the company was mostly involved in research and development. Hence a burst would define the year when their product hit the market and turned first sales. From these charts we can see when the different industries experienced major growth.

• Image 3.2:
  
• Insight 3.2:
  Similar to image 3.1 but for employees instead of sales.

TASK 4: GeoVis of the Growth of Companies

• Process:
  A 15 frame animation was created representing the years 1989 to 2003. You may also view this in our video. Each time frame uses the very same base map: a map of the US in which states are color coded by the total sales of all their companies. Richer states are given a darker shade. Overlaid over this map are nodes representing the companies that are in existence in this particular year. The area of nodes reflects company sales in millions of dollars. Node color represents the age of companies - younger companies are given a lighter color while well established companies appear darker. Companies that are in their first year of existence are colored yellow and companies in their last year of existence are colored red. The maps were generated by extending the Chizu system implemented by Mark Meiss, online at http://steinbeck.ucs.indiana.edu/~mmeiss/L579/project4/final.html. The time frame 2002 of the 15 year animation is shown here. We labelled the headquarters of some major US companies as a frame of reference.
• Image 4.1:
  
• Insight 4.1:
  A closer examination of the 2002 map reveals that California and New York have the most companies with the largest sales. We also see clusters around major cities. By zooming into a local region we can access details.
  Unfortunately, the contest data provides no information on why companies emerged, strived, or died. It would be interesting to know if a red node is a consequence of a company's bankruptcy or an acquisition or merger with another firm. One could speculate that 'smaller' red nodes were either acquired or went bankrupt due to the liability of being a small or new company. We could also speculate that two large red nodes in one year could have formed an even larger yellow node (new companies) in the subsequent year suggesting big mergers, like the one that formed Verizon in 2000. Ideally, the maps could show the merger waves characterizing many industries, reflecting the general trend of consolidation in some industries.

COMMENTS: Acknowledgements

We appreciate the effort by G. Grinstein, U. Cvek, M. Derthick, M. Trutschl in providing the Technology Data in US contest data set. This work is supported by a NSF CAREER grant under IIS-0238261, NSF grant IIS-0513650, and an equipment grant by UITS, Indiana University.