Research Article: What can we learn from corporate sustainability reporting? Deriving propositions for research and practice from over 9,500 corporate sustainability reports published between 1999 and 2015 using topic modelling technique

Date Published: April 12, 2017

Publisher: Public Library of Science

Author(s): Nadine Székely, Jan vom Brocke, Rajagopalan Srinivasan.


Organizations are increasingly using sustainability reports to inform their stakeholders and the public about their sustainability practices. We apply topic modelling to 9,514 sustainability reports published between 1999 and 2015 in order to identify common topics and, thus, the most common practices described in these reports. In particular, we identify forty-two topics that reflect sustainability and focus on the coverage and trends of economic, environmental, and social sustainability topics. Among the first to analyse such a large amount of data on organizations’ sustainability reporting, the paper serves as an example of how to apply natural language processing as a strategy of inquiry in sustainability research. The paper also derives from the data analysis ten propositions for future research and practice that are of immediate value for organizations and researchers.

Partial Text

Growing legislative pressure and increasing public concern about the global climate and the carrying capacity of the earth have led to increasing demands for organizations to act in sustainable ways [1]. Consequently, the number of organizations that publish information on their sustainability practices has grown steadily [2]. One way in which organizations communicate these practices to stakeholders is through sustainability reports—usually published annually with financial reports [3]—that report on the organization’s “economic, environmental and social impacts caused by its everyday activities” [4].

We employ a semi-automated text-mining technique on publicly available sustainability reports to determine the topics they address. These techniques usually represent documents as vectors. In the easiest form, such a vector includes for each term in the document the number of appearance. However, such a vector has a high number of dimensions (each one reflecting one term). Thus, we need to reduce the dimensionality of the resulting vector [29] in order to be able to handle these huge amount of data.

We analyse 9,514 sustainability reports published between 1999 and 2015 by 3,906 different organizations. The most common industries were financial services, followed by the energy sector, the mining sector, and food and beverage products.

Our analysis applies topic modelling to more than 9,000 sustainability reports in order to identify sustainability practices. We identify forty-two topics that are related to sustainability from which we make ten observations. In the following, we discuss these observations and develop ten related recommendations for organizations and researchers.

Increasing numbers of organizations are publishing sustainability reports about their sustainability practices [2]. The present work used topic-modelling techniques to analyse 9,514 sustainability reports published by organizations between 1999 and 2015 and derives ten specific propositions to guide future research and practice.

Table 7 provides on overview about all topics, including the label (if related to sustainability), the most probable terms and the sustainability dimension.




0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments