Authors:

Matthew Corritore – Desautels Faculty of Management, McGill University 

Amir Goldberg – Stanford Graduate School of Business

Sameer B. Srivastava – Haas School of Business, University of California, Berkeley

Interviewers:

Nur Ahmed – Ivey Business School, Western University

Ying Li – Gies College of Business, University of Illinois

Article link: https://doi.org/10.1177/0001839219844175

1. Where did the idea for this project – to study cultural heterogeneity – come from? How did you start the project? Were you inspired more by data, theory, or computational linguistics?

The project started with our interest in using Glassdoor data to measure organizational culture in novel ways, with the hope that doing so could help us answer important questions about the complex relationship between organizational culture and performance. We had the sense that organizational researchers hadn’t yet taken full advantage of the crowdsourced, anonymous data that Glassdoor collects from employees working across many companies in different industries. We saw similar types of online data being used by others in management and sociology, such as Yelp data to study how audiences evaluate organizations with different identities. And we were excited about the potential of applying some natural language processing techniques in order to analyze the free-text comments that employees write about the culture in their organizations. We knew using the Glassdoor data would also bring challenges, such as concerns about which employees self-select into writing a Glassdoor review, but we hoped the advantage of being able to track corporate culture over time for many companies would outweigh any limitations.   

Much of our inspiration to focus on cultural heterogeneity was pioneering work on organizational culture by Jennifer Chatman, Charles O’Reilly, and their collaborators. One aspect of culture that they write about is cultural agreement, or the extent to which organizational members characterize the organization using the same norms and values, and they developed a measure of interpersonal agreement using their popular survey tool, the Organizational Culture Profile. However, one of the limitations of culture surveys is that it’s difficult to collect data from employees for a large, diverse set of organizations. As such, prior studies on the relationship between organizational culture and performance were quite limited. 

So it’s difficult to say definitively whether data, theory, or method was the main inspiration. What we can say is that our focus on intrapersonal cultural heterogeneity is an example of a methodological innovation driving a conceptual one. We vividly remember the moment where the proverbial light bulb went off. After we first trained a topic model to identify cultural topics in each employer review, we realized that the method differed in an important way from traditional surveys: it allowed us to detect when an employee characterizes the firm with reference to a broad versus a narrow set of cultural norms and beliefs. We initially introduced intrapersonal heterogeneity as a control variable, but quickly started to hypothesize about how this type of diversity was distinct from its interpersonal counterpart, and how it might affect an organization’s capacity to grow and innovate.

2. What was the process like for you to decide on using the terms “interpersonal heterogeneity” and “intrapersonal heterogeneity” to capture the meaning of the concepts you hoped to examine? In addition, about “interpersonal heterogeneity,” how important was it to distinguish it from organizational subculture, both theoretically and empirically? About “intrapersonal heterogeneity,” did you consider using alternative terms, such as “cultural plurality” mentioned in your paper? What’s your general advice for doctoral students, who aspire to introduce a novel concept but are frequently advised to demonstrate the necessity of doing so and the difference of the new concept from existing but related concepts? 

There was a lot of trial and error in deciding how best to label these constructs. For example, at the earliest stages we were reverse coding interpersonal heterogeneity and calling it “cultural agreement,” in accordance with work on strong culture. Once we shifted to focus on heterogeneity, we at one point labeled it “compositional” heterogeneity, in reference to an organization’s culture being composed of individual members’ beliefs about the organization. But, with the help of some great suggestions by the reviewers, we ultimately decided that interpersonal heterogeneity best emphasized that this form of cultural diversity manifests between individuals. 

It was similarly difficult to label intrapersonal heterogeneity. We originally thought of it in terms of cultural “breadth” versus “focus.” At one point we labeled it “content” heterogeneity in reference to members describing the organization using many norms and values. But again, we ultimately thought that intrapersonal heterogeneity emphasized the key idea that this is cultural diversity that manifests within-persons. We never considered “cultural plurality” as far as we remember. Giving constructs labels that are descriptive and evocative is hard. Even now, we’ve come to refer to inter- and intrapersonal cultural diversity rather than heterogeneity, especially when talking about the paper in the popular media. 

It was important to us to conceptually distinguish interpersonal heterogeneity from subcultures, or diversity stemming from distinct groups of employees with homogeneous beliefs. Based on existing theory and our close reading of the Glassdoor comments, we were confident that these differences in organizational culture were not adequately explained by subcultures. Nonetheless, we examined this empirically during the review process. We used a modularity statistic from network research to examine whether individuals’ cultural beliefs clustered into discrete cliques, and found no supporting evidence. 

Our advice for doctoral students working to introduce and defend a new concept is to be charitable towards the existing ideas that the concept builds upon. All conceptual innovations build on related ideas. Our notion of intrapersonal cultural heterogeneity is novel given our focus on members’ perceptions of the organizational culture and the organizational level of analysis. But the concept builds on important ideas from cultural sociology, as well as some work on diversity in teams. Notably, we were inspired by Klaus Weber’s work on corporate cultural repertoires, which itself builds upon Ann Swidler’s influential theorizing about how actors use elements of their cultural toolkit as means to action. Additionally, one of the reviewers turned us on to a nice paper by Bunderson and Sutcliffe (2002), which teases apart different types of functional diversity in teams, one being intrapersonal functional diversity. While they weren’t studying cultural diversity and were focusing on the team-level, those researchers did highlight the importance of separating between versus within individual variation, which was also a key part of our paper.  

3. How do you see intrapersonal heterogeneity contributing to future research on organizational culture? What are the potential areas this construct can contribute?  

We think intrapersonal cultural heterogeneity is an important construct that we hope will influence how academics and practitioners think about organizational culture. As we argue in the paper, management research often views cultural heterogeneity as necessarily a detriment to efficiency but a boon to creativity and innovation, largely because it focuses mostly on interpersonal heterogeneity. By highlighting that intrapersonal heterogeneity is a distinct dimension, and showing that it is associated with growth and innovation, we suggest that organizations can escape the efficiency-innovation trade-off, at least with regards to cultural heterogeneity. 

We think that a natural next step is to research the determinants of intrapersonal heterogeneity. It’s not clear why we observe variation in intrapersonal heterogeneity across otherwise similar organizations. One possibility is that some leaders and managers are socializing employees to work in accordance with many different cultural ideas. But another possibility is that some organizations are selecting employees, whether deliberately or not, who have worked in many different cultural environments and draw from those experiences to shape their current organization’s culture. Many models still treat new employees as blank slates who are programmed through socialization to hold management’s values. However, the reality is that employees can bring into organizations a diverse array of cultural perspectives, drawn from not only their experiences in other formal organizations, but other domains of life as well. 

We hope that the paper also inspires and supports work at the individual level of analysis. For example, Yoonjin Choi and Paul Ingram have been concurrently working on research linking cultural brokerage at the individual level to the ability to be creative. There is still much to learn about how individuals can use culture as a toolkit to construct action and spark innovation, and we think that some natural language processing tools could be useful in measuring how individuals engage with varied cultural domains. 

4. We loved the introduction of the paper, especially how you presented the tension between two constructs and how your paper contributes to the existing literature! How did you arrive at your current framing? Did you change your framing during the review process? It seems that you presented the project at number of conferences and seminars before submission. To what extent did you integrate the feedbacks from participants from these venues, and later from reviewers, into the final product that we see today?

Thanks, we’re glad that the introduction resonated with you. The basic framing was largely in place when we entered the review process, but our AE Marc-David Seidel and the reviewers were instrumental in pushing us to more precisely represent and draw upon prior work. Most notably, it was a long process to refine how best to engage with the expansive literature on demographic diversity and performance, which doesn’t directly speak to the type of cultural diversity we study, but nonetheless exhibits this tension about diversity’s performance implications.   

We were fortunate to get feedback on the paper at many different conferences and seminars, ranging from the more sociological (e.g., the ASA Annual Meeting) to more mainline strategy (e.g., the Wharton Corporate Strategy Conference). The paper of course benefited from these diverse perspectives, but our biggest takeaway was that the topic and the results were of fairly broad interest to different intellectual communities studying organizations. It increased our confidence that a general interest journal like ASQ was a viable outlet for the work. 

 Marc-David and the reviewers at ASQ were wonderful — the paper improved substantially in the review process. Most notably, the reviewers encouraged us to expand our firm-level outcomes to include measures of firm patenting that were better aligned with our theory, as well as helped us convince readers that using a computational approach to measure culture was reasonable. 

5. Computational linguistics is still an emerging research method within management discipline. How was the review process like for this method? How did you deal with reviewers’ concerns around the method? What are the potential areas do you see for management scholars to apply similar computational methods? Do you have any suggestions for doctoral students who are interested in using computational linguistics?

We’re encouraged that there is now more management work taking advantage of computational linguistics/natural language processing/text analysis, much more so than when we first started this paper. Nonetheless, these methods can still pose challenges in the review process.  

The main concern that we worked to address in the review process was related to the construct validity of our measures, specifically whether the topic modeling was producing topics that were germane to organizational culture. Establishing whether a topic is “cultural” is tricky — we were trying to measure employees’ cultural beliefs, and what is cultural is often in the eye of the beholder. We’re proud that the paper proposes a method that guides the detection of cultural topics using a training set, i.e. a subsample of phrases that make explicit reference to a culture synonym. Nonetheless, we were prompted by reviewers to do some manual validation of the cultural topics, an exercise which we had some mixed feelings about. The reality is that we’re still learning as a field about the best practices for validating these new approaches.  

Despite the challenges, we see huge potential for management scholars to apply similar computational methods. Beyond topic modeling, deep-learning methods like word embeddings and BERT can help researchers model fine-grained variation in how actors use language in organizational settings. One opportunity that we see is for researchers to use such approaches to understand how certain actors draw upon and fuse together variegated cultural logics to motivate action, which may be detectable through language.

For doctoral students interested in using computational linguistics, the barriers to learning and applying these methods are lower than ever. However, we advise using such techniques for the right reasons. As with many novel methods, there’s a strong temptation to wield these tools like a hammer, and view every research question or data source as a nail. However, the onus is on the researcher to demonstrate a strong rationale for their use, such as allowing for a new theoretical insight to be tested or overcoming an empirical limitation of prior methods.    

References cited:

Bunderson, J. Stuart, and Kathleen M. Sutcliffe. “Comparing alternative conceptualizations of functional diversity in management teams: Process and performance effects.” Academy of management journal 45, no. 5 (2002): 875-893.

Nur’s Bio: Nur Ahmed is a Ph.D. Candidate at Ivey Business School, Western University. He’s interested in innovation and industry evolution and computational social science. 

Ying’s Bio: Ying Li is a Ph.D. Candidate at Gies College of Business, UIUC. She’s interested in emergence and evolution of industries, authenticity, social evaluation of novelty, as well as cultural and creative industries.