Discovering hidden patterns in technology management phenomena: three essays on using data analytics for exploration of causal inferences

The availability of copious amounts of data, increased computational power to analyze them, and the readiness of various techniques to extract information and find patterns among them, are having a significant impact on the way companies are managed. However, the ability of these methods to generate causal inferences is still one of the fundamental questions that must be resolved before its mainstream incorporation into quantitative research in management. To answer the question -- of what represents the current data deluge and data analytics methods for management research from both a philosophical and methodological perspective -- is the main purpose of this dissertation. In order to address that question, from a philosophical perspective, we adopt a realist view of causation. From a methodological perspective, we argue that the use of data analytics methods should start with a phase strongly grounded in theory and end with the abduction of new theories from the patterns. In the three essays of this dissertation, we exemplify the use of data analytics methods in the study of technology management phenomena. The first essay focuses on the phenomenon of open source software development. Despite the tremendous popularity of certain open source software applications, one major challenge for Free/Libre Open Source Software (FLOSS) online communities has been the high mortality rate of initiatives, as observed in the decline and sometimes fading away of development activity for the clear majority of projects over time. Another challenge is that the sheer number of simultaneously coexisting projects in these communities highlights the importance of visibility as a way for a project to distinguish itself from the collection. In this study, we examine the relationship between endurance and visibility connected to keyword characteristics. Our results suggest that higher interest of users is positively associated with the endurance of the projects in the communities. Further, we find that the selection of keywords, reflecting the functionality of the software and the operating system, is strongly associated with user attention. Per our study, increasing the visibility of the project is an important mechanism to sustain its activity over time. The second essay centers on exclusivity in technology licensing. While prior research has significantly advanced our understanding about exclusivity in licensing, there are still significant gaps in our knowledge about how licensing exclusivity is impacted by the interplay between different contextual and intrinsic attributes of the license. Exclusivity in licensing can be highly complex and contingent, potentially reflecting the interactions between different theoretical explanations, and the boundary conditions that apply to each theory. The exploration of such contingencies and complexities is hampered in conventional econometric analyses, which we seek to overcome by employing a novel empirical technique called decision tree induction, a powerful machine learning tool for uncovering nested “multiple theoretical viewpoints.” Implications for the empirical and theoretical literature on licensing, and for abductive theory development by leveraging “big data” are discussed. The third and final essay addresses alliance formation in the computer services industry. Strategic alliances have steadily increased over the last three decades as popular instruments for interfirm cooperation. While there are competing and complementary theoretical bodies of work that have addressed this phenomenon, the question of who allies with whom is still relevant due to the complexities surrounding the phenomenon and the challenging nature of the prediction task for alliance formation. Social network approaches have substantively contributed to our understanding of the partner choice in alliances; however, they also bring forth some limitations. We extend that previous work by addressing some of them through the introduction of the concept of heterogeneous networks and the application of a novel machine learning intensive technique to predict alliance formation. Our results suggest a high predictive accuracy of the technique. Implications for the path dependence of alliance formation processes are also discussed.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Business Administration

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2017

Author dc:creator

• Fernandez Corrales, Carla Beatriz

Contributors dc:contributor

• Subramanyam, Ramanath
• Somaya, Deepak
• Larson, Eric C.
• Mahoney, Joseph T.

• Data analytics
• Machine learning
• Causal inferences
• Abduction of theories
• Technology management
• Alliance formation
• Technology licencing
• Open source software

Statement dc:rights

• Copyright 2017 Carla Beatriz Fernández Corrales

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/98173

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/98173