How to Build Strong Theories Using Case Studies: A Practical Guide Inspired by Eisenhardt”

Okay, let’s assume you are a researcher wanting to use Kathleen Eisenhardt’s approach to multiple case studies. Based on the sources, here is a detailed step-by-step walk-through of how you would typically proceed:

1. Start with a Research Question and Concepts:
   • Like any research, you begin with a research question.
   • A key characteristic of the Eisenhardt approach, compared to some other qualitative methods like grounded theory, is that you usually start with some concepts already in mind that you want to study.
   • For example, if you were studying power and politics, you would decide beforehand that these are the concepts you are interested in exploring.
2. Select Your Cases Using Theoretical Sampling:
   • You need to study multiple cases, typically four to 10 different cases. The number often ends up being between eight and 10 when this approach is followed.
   • You select your cases using theoretical sampling. This means you choose cases specifically because they are likely to provide insights into your concepts and, importantly, because they show variation in those concepts.
   • If you are studying power and politics, you should try to get cases where there is variation in both. You might intentionally seek out companies you know have centralized power and others where power is not centralized.
   • Case selection is an ongoing process. If you start with four cases and find that three have centralized power but only one does not, you would then focus on finding more cases where power is not centralized for the remaining cases you plan to study. This ensures you have sufficient variation for comparison.
3. Collect Data Using Multiple Methods (and Overlap with Analysis):
   • You should collect data using multiple methods. While interviews are common, you can also use survey forms (like giving an informant a paper survey before an interview) or data from databases, such as financial data.
   • Crucially, data collection and analysis need to overlap in qualitative research, including this method. You start analyzing early data while you are still collecting more. This helps you understand processes and allows you to focus later data collection efforts on the most important aspects that emerge from your early analysis.
4. Analyze the Data (Initial Steps):
   • You need to be very transparent about how you analyze your data because qualitative research offers flexibility and doesn’t have rigid, standard analysis techniques like quantitative methods. You must explain exactly how you did it.
   • Analysis often starts by organizing and sometimes quantifying your data and seeking patterns.
   • You will code your data to find evidence of the concepts you chose at the beginning (like power and politics) and identify different levels or variations of those concepts.
   • You can also develop profiles or short descriptions for each case or key units within the case (like executives or decisions).
   • Timelines are often created to map key events and help construct descriptions of processes.
5. Analyze the Data (Core Iterative Process):
   • Analysis always starts with within-case analysis. You analyze and understand each case thoroughly on its own first.
   • Then you move to cross-case analysis, which involves comparing the cases. You might compare them pair-wise (e.g., a high-performing company vs. a low-performing one) or compare groups of cases. The goal is to find patterns that explain how cases differ or are similar.
   • You seek associations between the different variables or concepts you are studying.
   • Once you find patterns or associations, you seek evidence for causality. You need to explain how, why, and when the causal processes work, not just state that an association exists.
   • This process of seeking patterns, looking for evidence of causality, and comparing across cases is iterated many, many times. If a pattern only appears in a few cases, it might not be the best basis for your theory. You refine your understanding through repeated comparison and analysis.
6. Handle Conflicts in Data:
   • You will likely encounter conflicts in the data, where different people describe the same event differently or give conflicting accounts. People’s accounts are influenced by their own interpretations, which can introduce bias.
   • In this approach, you highlight these conflicts and investigate why they exist. By understanding the reasons for conflicting accounts, you try to infer what the actual reality was. The goal is to eliminate or manage the influence of a person’s interpretation from your analysis.
7. Understand Patterns are Not Laws:
   • Remember that the patterns and associations you find are not seen as absolute laws. Just because most cases with centralized power showed a lot of politics doesn’t mean centralized power always leads to politics.
   • You are looking for some kind of association, which may not be perfect or even always strong; they can be weak as well.
8. Develop Your Theory and Propositions:
   • Through the iterative analysis and comparison process, a theory emerges.
   • The goal is to develop propositions, which are statements about potential causal relationships between two concepts. You must provide explanations of the causal process, not just claims.
   • This method is generally used to find associations and causal relationships between concepts that you chose before the study began, rather than developing entirely new concepts.
9. Compare with Prior Research:
   • Once your theory starts taking shape, you compare it with existing research (prior literature). You look for things in your theory that are similar to or conflict with what has already been found. This helps position your contribution.
10. Know When to Stop: Reach Theoretical Saturation:
    • You continue adding cases and collecting data until you reach theoretical saturation.
    • You know you’ve reached saturation when adding a new case no longer gives you more information to guide or refine your theorizing. If the latest case doesn’t really add much value to your developing theory, you can decide to stop adding more cases.
11. Report Your Findings:
    • The final output often includes tables showing associations between concepts, a written narrative or description explaining the relationships and processes, and quotes from your data to support your findings and provide evidence for the processes described. The reporting often reflects the quantitative focus by classifying cases and seeking associations.

By following these steps, you systematically move from initial concepts and case selection through rigorous data analysis and comparison to build a theory grounded in the evidence from your multiple cases.