Micro Level: Technological Capabilities and Organizations

Jupyter Notebook Link

On the third and final part of the main analysis, the results of the application of the methodology to the micro level will be presented. Here, the focus will rely on studying organizations as the units of technological capability. Moreover, the analysis will focus on the study of the Technical University of Denmark and of organizations based in Denmark.

1.Characterisation of organizations

The first part of the micro level analysis seeks to characterize an organization by its capability matrix, as previously done for years at the macro level, and countries at the meso level. By filtering all of the documents whose owner contained the “Technical University of Denmark”, the technological capability matrix of the University was achieved (LINK). The normalized version of this matrix, has the following properties:

Matrix Properties

DTU capability matrix

Dimensions

352x352

Mean

3.7e-04

Standard Deviation

7.8e-03

Maximum

0.82

Minimum

0.00

Symmetry

True

In its capability matrix, DTU has the maximum value in the position (200, 286) with the value of 103.8. This position refers to the terms biogas and anaerobic digestion

2.Organization correlation matrix and organization profiles

Following the methodology applied in sections 4.1.2 and 4.2.2, we turn our attention to organizations. A comparison was made between the capability matrices of 112 organizations, by the means of the calculation of the Pearson correlation index between them. The database contains a total of 10638 organizations; therefore, in order to maintain the quality of visualizations, a triage of the organizations was made. By only retrieving organizations with 7 assets or more and for reference purposes added a number of Danish organizations. As a result the list was reduced to a total of 112 organizations. The Danish organizations added manually were: DTU, Novozymes, Aalborg University, Aarhus University, University of Copenhagen, University of Southern Denmark, DTU Riso, and Dong Energy.

By comparing every organization’s capability matrix and calculating the correlation between each of them, the capability matrix of organizations was created. In the figure below is the clustered version of this matrix. In this clustering, Danish organizations do not appear as particularly related. DTU for example, appears in a cluster with a majority of Chinese organizations (Tsinghua University, Shanghai University and Beijing University). On another note, Dong, Riso, Aalborg and the University of Copenhagen are highly connected in this clustering. Novozymes appears rather isolated, while the University of Southern Denmark and Aarhus appear in the same cluster. The organizations in focus are highlighted in red in the graph below.

When cutting this correlation matrix in the rows corresponding to DTU and Novozymes, we are able to access the profiles of both of these organizations. In the case of DTU, its top 3 correlations are with Shanghai Jiao Tong University (70%), Tsinghua University (70%) and the Portuguese University of Minho (69%). The organization that is not a university with the highest correlation with DTU is CSIR (African Research Organization). In the case of Novozymes, the correlations are on average lower. The Indian Institute of Chemical Technology (57%) and the University of Copenhagen (55%) make for the most related organizations.

Organization correlation matrix: Clustered Version. (Danish organizations in red.)

Country correlation profile: Technical University of Denmark.

Country correlation profile: Novozymes.

3.Comparing organizations

In order to prove the concept of comparing the capabilities of two organizations, the Technical University of Denmark and the University of Tsinghua were selected. These organizations are similar by a factor of about 70%, and for this reason the differences in the usage of term pairs across documentation is not necessarily important.

Comparing the usage of all term pairs and then ordering the absolute relative difference, resulted in the table below. Here, the top 20 term pairs that differed mostly in usage between these organizations are presented. The first observation to be made is the fact that several of these term pairs are simply not used by one organization. “Biodiesel-hydrolysis” and “sorghum-fermentation” for example, are simply not used by DTU; however, they both appear in 1.15% of all documents owned by Tsinghua University. Another interesting note is the difference in terms of feedstocks used. Tsinghua University for example, uses “sewage” and “waste” and “gasoline” in a more important way, while on the other side, “algae-anaerobic digestion” is not used at all.

Term pair comparison of DTU and Tsinghua University:

First Term

Second Term

DTU

TSINGHUA

Difference %

bioethanol

fermentation

0.656250

0.188679

0.467571

anaerobic digestion

biogas

0.828125

0.452830

0.375295

ethanol

fermentation

0.734375

0.490566

0.243809

anaerobic digestion

algae

0.218750

0.000000

0.218750

anaerobic digestion

biodiesel

0.203125

0.000000

0.203125

hydrolysis

biogas

0.203125

0.000000

0.203125

biodiesel

transesterification

0.171875

0.000000

0.171875

ethanol

catalysis

0.171875

0.000000

0.171875

biogas

manure

0.203125

0.037736

0.165389

straw

fermentation

0.203125

0.037736

0.165389

ethanol

bioethanol

0.218750

0.075472

0.143278

methanol

catalysis

0.140625

0.000000

0.140625

methanol

transesterification

0.140625

0.000000

0.140625

ethanol

transesterification

0.140625

0.000000

0.140625

hydrolysis

bioethanol

0.250000

0.113208

0.136792

fermentation

sorghum

0.000000

0.132075

0.132075

hydrolysis

biodiesel

0.000000

0.132075

0.132075

hydrolysis

cellulose

0.187500

0.056604

0.130896

ethanol

straw

0.203125

0.075472

0.127653

algae

biogas

0.125000

0.000000

0.125000

4.Collaborations

A particularity of the micro level analysis is the study of the types of collaboration made between organizations. By applying the same methodology as section 4.2.3.1 for each organization in the list, details about the nature of the collaborations that were put in place was obtained. By filtering these partners, and detecting the word ‘’Univ” in their names, the program designed was able to detect the nature of the collaboration, and categorize them into “University Partnerships” or “Organizational Partnerships”.

In the two bar charts below, the organizations in the database are ordered according to the percentage of partnerships of Universities.

The first chart, presents all of the organizations in the original list. The Agricultural Research Service (ARS) appears in the first position with 70% of partnerships made with universities and 30% with other organizations. Approximately a third of the organizations queried, have no university partnerships, and only collaborate with businesses or organizations. The Danish organizations and universities (in red), all collaborate with universities, the one that does so the least is “Riso DTU”, where university partnerships make up for approximately 48% of all collaborations. When looking at businesses, Novozymes’ collaborations are 60% with universities, and Dong Energy 57%.

The second chart is derived from the first, but on the x axis, only universities are presented. DTU, for example, is balanced in terms of university and organizational collaborations with ratios of 58% and 42% respectively. On the other side of the spectrum, two Brazilian universities (Universidade Federal do Parana and Universidade Estadual de Campinas) and one Canadian university (University of Alberta), only collaborate with businesses and research organizations. There are no organizations that collaborate exclusively with universities, but organizations that collaborate exclusively with non-universities do exist. Moreover, the Danish organizations in focus (red) all collaborate with universities in at least 50% of their shared assets. Finally over half of the non-universities queried, collaborate almost exclusively with other non-universities.

Collaboration by type of partnership: All organizations.
Collaboration by type of partnership: Universities.

5.Organizational spectrums

Similarly to the analysis and results carried out in section 4.2.5, the same approach was applied to organizations.

When observing the first representation of the organization spectrums for DTU, Novozymes, the University of California Berkeley and Tsinghua University, one can notice immediately some differences and similarities between the organizations and the source of their differences. Some term pairs such as “sugar/cellulose” appear in all of the 4 organizations. While others are exclusive to some organizations such as “willow/miscanthus” for DTU or “paper/grass” for the University of Berkeley. This representative visualization serves as a proof of concept to the following one, where all of the countries and term-pairs are presented.

In the full picture visualization, the countries were ordered according to the result of the hierarchical clustering of section 4.2.2. The first observation to be made is related to the long vertical lines along all of the organizational spectrums, these long bars are the representation of term pairs that are widely used by a large number of organizations, thus producing this effect. Moreover, some of the organizational spectrums seem more widely distributed than others. For instance, the organizational spectrum of Shell (5th to bottom) has a number of condensed black areas corresponding to many interrelated term pairs; on the other hand, the organizational spectrum of DTU appears to be more diverse and well distributed. Finally, one can note that the quality of clustering, this is because organizations and groups of organizations that use the same term pairs appear closer together.

Organizational spectrums: Limited to 45 term pairs and 4 organizations.

Organizational spectrums: Complete spectrum visualization.