Qualitative analysis: Deductive and inductive approaches
How you analyze qualitative data depends largely on your methodology, your personal organizational and analytic preferences, and what kind of data you have. That being said, all qualitative data analysis processes are going to fall into one of two categories: deductive or inductive. In this post, I’m going outline what each of those terms mean for qualitative analysis, and discuss some general deductive and inductive analytic strategies. Some of what appears here comes from a book chapter I wrote with Dr. Patty Witkowsky on inductive and deductive approaches to qualitative data analysis. The chapter appears in Vanover, Mihas, and Saldaña’s book Analyzing and Interpreting Qualitative Research: After the Interview.
I have also recently published another article that expands on the ideas presented here. This article will help you further develop your qualitative data analysis skills.
The material presented here can be cited as:
Bingham, A.J. (2023). From data management to actionable findings: A five-phase process of qualitative data analysis. International Journal of Qualitative Methods. https://doi.org/10.1177/16094069231183620
Bingham, A.J., & Witkowsky, P. (2022). Deductive and inductive approaches to qualitative data analysis. In C. Vanover, P. Mihas, & J. Saldaña (Eds.), Analyzing and interpreting qualitative data: After the interview (pp. 133-146). SAGE Publications.
Deductive Analysis
Deductive, or a priori, analysis generally means applying theory to the data to test the theory. It’s a kind of “top-down” approach to data analysis. In qualitative analysis, this often means applying predetermined codes to the data. The codes can be developed as strictly organizational tools, or they can be created from concepts drawn from the literature, from theory, or from propositions that the researcher has developed.
Deductive analysis can help you to:
Sort data into organizational categories, such as data type, participant, or time period. To do this, you can develop attribute codes (Miles et al., 2020) to organize your data. These attribute codes are applied to categorize your data by data type (e.g., “interview”), location (e.g., “Binary High School”), participant (e.g., “Mr. Perez”), time period ( e.g., “Fall 2017”), etc. I generally do this as a first cycle of coding to create an organizational schema.
Organize data into categories to maintain alignment with research questions. Deductive analysis can help you maintain focus on the purpose of your research. Qualitative data can get…unwieldy. So, it’s helpful to have a plan to stay focused on your research questions. To do this, during my first read-through of my data, I create broad topical categories of interest based on my research questions and then sort the data into those categories. For example, if I’m doing study of how teachers’ implement technology-facilitated competency-based learning practices in their classrooms, I might develop the codes “Classroom practice,” “Assessment,” “Technology,” and “Data use” to reflect the broad topics of my research. I then sort the data into those categories, which allows me to focus on relevant data in subsequent rounds of analysis.
Apply theoretical or conceptual frameworks. Deductive practices can also help researchers to apply theory or conceptual frameworks. For me, this process generally comes after I do some inductive analysis of my data to identify themes, but you could draw on these strategies at really any point in the analysis process. For my process, I create codes based on the components of my theoretical framework, and then sort my data into those predetermined theory-based categories. Theory-based codes from institutional theory, for example, might include “Isomorphism” or “Decoupling.”
Inductive Analysis
Inductive analysis, on the other hand, is a more emergent strategy, where the researcher reads through the data and allows codes to emerge/names concepts as they emerge. It’s more of a “bottom-up” analytic strategy. There are many forms of inductive analysis, but some common practices are open coding (sometimes called initial coding), in vivo coding (codes developed from participants’ own words), and constant comparative analysis. Memoing plays a key role for me as well. I memo to keep track of my analysis process and the decisions I make and to make sense of the data I’m reading. I also keep a running memo of the themes and findings I am starting to see, and I have a memo where I keep interesting or generative participant quotes or excerpts from field notes, and any evidence relevant for my themes and findings as they develop.
Inductive analysis can help you to:
Make meaning from the data. Because I typically use some deductive coding strategies to organize my data first, and sort it into categories that are relevant to my study purpose, my inductive analysis process generally starts after I’ve already done a pass through my data. However, you could start with inductive analysis (though I still recommend some attribute coding first to help you keep organized). When I start my cycle of inductive analysis, I finally begin to make meaning of the data I have. For me, this generally starts with open coding - a pretty classic form of qualitative coding during which the researcher reads through the data and develops and applies codes as they go, to represent what’s happening in the data. I read through the data in each category I created in my first round of deductive coding, creating and applying codes and identifying emerging topics or concepts as I read. I then use what’s called “pattern coding” to identify patterns across and within data sources. This is a process of condensing the codes created during open coding to chunk the data into fewer analytic concepts. This helps me begin to summarize the data.
Develop themes and findings. The key purpose of inductive analysis is to really dig into what is happening in the data, to understand the themes present in the data and to produce findings to answer your research questions. In my analysis process, I identify themes from the pattern codes through memoing and further condensing the pattern codes where I can. I then try to capture the themes in short phrases like “Real time data use” or “technology-facilitated competency-based assessment.” Many qualitative researchers use the words “themes” and “findings” interchangeably, which works for them. For me, I like to differentiate. My themes are usually words or phrases, and I develop my findings from those themes, by condensing and rewording the themes into short phrases that clearly answer the research questions. For example, “Teachers relied on real time data use to implement competency-based learning” would be a finding for the research question “How do teachers implement competency-based learning in their classrooms?”
Identify representative data to support findings. Throughout my analysis process, I will often develop in vivo codes from participants’ own words to point to data that is representative of particular findings. I also keep a running memo of participant quotes and excerpts from field notes, etc., as well as a memo of my emerging findings, where I will note representative evidence and write deeper descriptions and explanations. This process helps me to keep track of important evidence, and gives me a place to free write about my findings. I am a person who thinks best through writing, so I find this memoing process especially critical.
Explain findings using theory and literature. During one of my deductive cycles of coding mentioned above, I apply codes based on my theoretical framework(s) and/or the existing literature. But that doesn’t go far enough in terms of really applying theory and using it to make sense of and explain the findings (and write the discussion section). So, after the deductive cycle of coding in which I sort the data into theory-based categories, I do some inductive analysis, where I memo to develop short phrases that connect my findings with theory and existing literature. Though coding forms the basis for most of my analysis, memoing in this cycle helps me understand what it all means and why anyone should care about it. Combining deductive and inductive coding and memoing supports the development of the discussion and implications sections. It also helps you to understand your findings in relation to existing research, examine how the theoretical framework explains the findings (and where it doesn’t, which allows for theoretical contributions), and can support you in providing actionable, meaningful implications and recommendations.
As I see it, both deductive and inductive strategies are important in qualitative analysis, so in my work, I draw on both. I use deductive strategies to organize and focus myself, and I use inductive strategies to understand what is happening in the data, without forcing the data into what I think I’ll see. In short, a data analysis process that draws on both deductive and inductive analysis supports a more organized, rigorous, and analytically sound qualitative study.
See below for an example of how I organize deductive and inductive analytic practices into cycles.
This figure, adapted from Bingham & Witkowsky (2022) and Bingham (2023), gives an overview of my analysis process. Five cycles, encompassing both deductive and inductive processes.
I’ve got a new book out with Routledge that focuses on how theoretical and conceptual frameworks can be applied in research. In this book, we explore a variety of theories and further discuss their application in data collection and analysis. Please take a look!
The material presented here can be cited as:
Bingham, A.J., & Witkowsky, P. (2022). Deductive and inductive approaches to qualitative data analysis. In C. Vanover, P. Mihas, & J. Saldaña (Eds.), Analyzing and interpreting qualitative data: After the interview (pp. 133-146). SAGE Publications.
Bingham, A.J. (2023). From data management to actionable findings: A five-phase process of qualitative data analysis. International Journal of Qualitative Methods. https://doi.org/10.1177/16094069231183620
