Qualitative Coding
Created for the USF Muma College of Business DBA Program
Joann Farrell Quinn, PhD, MBA
Qualitative Content Analysis through Coding Module (12-15 hours total for all sub-modules)
Coding is an analytical process in which data, in both quantitative form (such as questionnaires results) or qualitative form (such as interview transcripts) are categorized to facilitate analysis. The data is transformed into usable information through labeling and organizing the data. This then allows the researcher to tell a story about the data, making it relevant to the receiver, which is particularly useful for practitioner-scholars who are tasked with finding actionable solutions to real world problems.
In this module, you will be provided an overview of different coding methods, as well as different methodologies that you may choose to explore your research question(s).
Qualitative research often involves an iterative process. This module will not instruct you on data collection methods, but rather qualitative analysis of data through coding.
You can't learn how to use qualitative methods by just watching videos, so this module will include collecting data through observation or interviewing and on analyzing and interpreting the collected data in other assignments. We will also touch on best practices, ethics, writing some methods of analysis, and mixing methods.
Learning Objectives:
By the end of this short course, students will be able to:
Describe various styles of interpretation of qualitative data.
Recognize which approach is most appropriate for different types of studies.
Apply one or more analytic approaches to data.
Activities:
Reading and watching videos about coding on various research methods, reading about methodologies of coding; applying coding techniques to selected research methods.
Sub-Module 1: Introduction to Coding
Coding is a form of qualitative analysis which can be used on a variety of data sources, including social media feeds, texts, transcripts, films/videos.
Social scientists use both qualitative and quantitative research methods to investigate research questions.
There are times when a qualitative approach is a better fit for the research needs and questions, or used in addition to quantitative methods to gain a better understanding.
Exploration: When something isn’t well defined, qualitative methods are helpful. For example, you can explore in a qualitative study the problems customers encounter, the needs users have and can’t articulate, or misunderstandings customers have in finding information or using a product.
Complexity: While complicated problems can be quantified, when you need to describe the complexity and subtlety of how users interact with a product or accomplish goals, qualitative research can distill the complexity into more manageable parts.
Context: Understanding the context and environment a user is in provides for better product direction. What are the products, places, people, and challenges customers deal with when accomplishing their goals? Some of the richest qualitative data isn’t collected in a contrived lab; it comes from observing and collecting data in person.
Explanation: When you need to explain linkages or mechanisms that cause things, a qualitative method can be fruitful. For example, when you want to know why people aren’t paying their bills via the mobile app or calling customer support because of an error, hearing customers’ own words help form theories and establish a testable hypothesis (Links to an external site.).
Measures don’t fit the problem well: While there are good ways of measuring usability (Links to an external site.), many interactions can be hard to quantify. Observing users as they struggle to accomplish a goal and probing on the source of the problems helps define what ultimately needs to be measured. It’s not very helpful to precisely measure the wrong thing. Qualitative data helps uncover the right things to measure.
(Reference for reasons to use qualitative research-- https://measuringu.com/qualitative-study/)
Why do we need to code? Coding makes text quantifiable and ready for analysis. Codes are a credible way to organize and analyze data. Without a formal process of coding and analysis, you aren't doing research-- you are extrapolating and guessing based on your preconceived ideas. This is why it is important to code qualitative data.
Watch the following short video on why we need to code.
Now, watch the following video to hear an introduction to coding and then keep reading to explore the different techniques for coding.
For the purpose of this module, we will stick to two basic techniques: 1) deductive coding and 2) inductive coding.
Deductive Coding – A type of coding of qualitative data in which you start your analysis with codes already in mind, based on previous research, a theoretical framework, or your own experience. This type of coding starts with a code book.
Inductive research involves the conversion of raw, qualitative data into more useful quantitative data. Unlike deductive analysis, inductive research does not involve the testing of pre-conceived hypotheses, instead allowing the theory to emerge from the content of the raw data.
Methodologies employed within inductive: phenomenology and grounded theory*. This type of coding is started without a code book or map in mind, and all seemingly relevant bits of data are coded and selectively reduced through an iterative process of reduction to themes through:
1) Open coding:
Adding content labels
Creating groups/categories of labels
2) Axial coding:
Merging and dividing groups
Finding code frequencies and relationships
Creating memos, reviewing them, relating them to codes, definitions, comments
3) Selective coding:
Identifying central theoretical phenomenon
Building a storyline that connects the categories of codes
Iteratively confirming/validating the theory
Coding is not an exact science, as we discern codes based on our perspective or lens, our experience, our biases, etc. Coding is interpretive and therefore it is important to maintain rigor and process to ensure your findings are representative of the data.
Coding is the initial step toward analysis and interpretation of data.
As coding proceeds, data is grouped, segregated, linked, regrouped-- this is an iterative process.
For an understanding the different types of qualitative approaches, please see this paper by Starks and Trinidad (2007) (Links to an external site.), which gives a fairly concise explanation.
Starks, H., & Brown Trinidad, S. (2007). Choose your method: A comparison of phenomenology, discourse analysis, and grounded theory. Qualitative health research, 17(10), 1372-1380.
For further reference, here is a blog post on phenomenology and grounded theory:
What gets coded? Units of social organization, are what get coded, according to Lofland, Snow, Anderson and Lofland (2006):
1 cultural practices (daily routines, occupational tasks, microcultural activ-"
"ity, etc.);
2 episodes (unanticipated or irregular activities such as divorce, championship games, natural disasters, etc.);
3 encounters (a temporary interaction between two or more individuals such as sales transactions, panhandling, etc.);
4 roles (student, mother, customer, etc.) and social types (bully, tight-ass, geek, etc.);
5 social and personal relationships (husband and wife, party-goers, etc.);
6 groups and cliques (gangs, congregations, families, jocks, etc.);
7 organizations (schools, fast-food restaurants, prisons, corporations, etc.);
8 settlements and habitats (villages, neighborhoods, etc.); and
9 subcultures and lifestyles (the homeless, skinheads, gay leather bears, etc.)
And these units are combined with aspects such as the following for analysis:
1 cognitive aspects or meanings (e.g., ideologies, rules, self-concepts, identities);
2 emotional aspects or feelings (e.g., sympathy in health care, road rage, work- place satisfaction);
3 hierarchical aspects or inequalities (e.g., racial inequality, battered women,"
"high school cliques).
PLEASE READ THE FOLLOWING ARTICLE which reviews the two main approaches to coding-- through use of an existing or created coding structure (deductive) and the open coding technique (inductive).
Blair 2015 Download Blair 2015
Although I cannot provide the link here for copyright reasons, The Saldana book can be found online, or of course purchased online. Please read chapter 1.
Saldaña, J. (2015). The coding manual for qualitative researchers. Sage.
Here are some helpful definitions:
Code = a label for a piece of text
Theme = a pattern, a group of data or something that emerges from data
Category = categories are defined/explained by their properties
Inductive = little or no predetermined theory, structure or framework (guided by research question)
Deductive = use of structure, theory or predetermined framework (code book, existing codes, themes, etc.)
Thematic = identification, analysis and reporting of pattern
*A note on grounded theory... grounded theory is a methodology, not a theory, as the name would imply. And quite often grounded theory, while a good process for approaching data, does not result in theory creation.
Sub-Module 2: Inductive Coding
This sub-module is going to focus on inductive coding. Inductive coding is a more labor intensive and creative process than deductive. If you can master inductive coding, then deductive coding will come easily. Therefore, this sub-module will have the bulk of the learning and understanding.
Remember from the introductory module, inductive research involves the conversion of raw, qualitative data into more useful quantitative data. Unlike deductive analysis, inductive research does not involve the testing of pre-conceived hypotheses, instead allowing the theory to emerge from the content of the raw data.
Methodologies employed within inductive: phenomenology and grounded theory*. This type of coding is started without a code book or map in mind, and all seemingly relevant bits of data are coded and selectively reduced through an iterative process of reduction to themes through:
1) Open coding:
Adding content labels
Creating groups/categories of labels
2) Axial coding:
Merging and dividing groups
Finding code frequencies and relationships
Creating memos, reviewing them, relating them to codes, definitions, comments
3) Selective coding:
Identifying central theoretical phenomenon
Building a storyline that connects the categories of codes
Iteratively confirming/validating the theory
One specific inductive approach that is frequently referred to in research literature is grounded theory, pioneered by Glaser and Strauss.
This approach necessitates the researcher beginning with a completely open mind without any preconceived ideas of what will be found. The aim is to generate a new theory based on the data.
Once the data analysis has been completed the researcher must examine existing theories in order to position their new theory within the discipline.
I suggest that the first step (after simply reading through your data but before open coding) is pre-coding, which is identifying the sections of text that you wish to identify with codes. You review the data and highlight or underline the words, phrases or sentences that stand out to you and seem like they may be important. This is what Boyatzis calls 'codable moments' (Boyatzis, 1998).
It is important that your coding process begins while you are collecting your data, not after all of your data has been collected. This is important, as you may wish to modify your method of data collection.
While you are working through this initial phase of highlighting and then coding, you also want to ensure you are keeping your research question in front of you, and reminding yourself regularly what you are seeking to understand. This doesn't mean what you hope or expect to find, but anything related to answering the question that emerges from the data.
OPEN CODING
What is a code?
In the open coding step, your goal should be identifying concepts (e.g., actions, interactions, responses, events, etc.) that you find significant in the context of your study. This step will allow you to go from text to concepts, providing an abstract, conceptual representation of your data. The concepts you identify are called “codes” and the process of marking your data for these codes is called “coding.”
As you code your data, you will see that some of the concepts you identify reoccur and you start reusing the codes you created at the beginning of the coding process. At some point, the coding reaches a level of “maturity” and no new codes emerge. This is a natural
point to stop open coding and move to axial coding.
A coding pattern can be characterized by:
similarity (things happen the same way)
difference (they happen in predictably different ways)
frequency (they happen often or seldom)
sequence (they happen in a certain order)
correspondence (they happen in relation to other activities or events)
causation (one appears to cause another)
How many codes, etc should I have?
From Saldana, pg 20:
Lichtman (2006) projects that most qualitative research studies in education will generate 80–100 codes that will be organized into 15–20 categories which evenyually synthesize into five to seven major concepts (pp. 164–5). Creswell (2007) begins his analyses with a short-list of five to six Provisional Codes to begin the process of “lean coding.” This expands to no more than 25–30 categories that then combine into five to six major themes (p. 152). Other disciplines and varying approaches to qualitative inquiry may prescribe different sets of numbers as general guidelines for analysis. The final number of major themes or concepts should be held to a minimum to keep the analysis coherent, but there is no stan- dardized or magic number to achieve. Unlike Lichtman’s five to seven central concepts and Creswell’s five to six major themes, anthropologist Harry F.Wolcott (1994, p. 10) generally advises throughout his writings that three of anything major seems an elegant quantity for reporting qualitative work.
Here is a box of index cards from a qualitative study where I interviewed 25 physicians. Each card has a 'codable moment' on it. Of course, many were deemed to be irrelevant, so you may end up with many more codable moments initially coded than you include in your axial and then selective coding process.
How do I know if I am coding it properly?
Watch this video:
A note about coding--- How do you actually label the text, etc?
I use a process of printing out the transcript(s), leaving a margin on the right hand side and after highlighting what I believe are codeable moments, reviewing those chunks of text and giving them a label (ie the code).
Here is an example of an initial highlighting (pre-coding) and open coding on a portion of an interview.
sample coding 1.pdf Download sample coding 1.pdf
I then write all of the codes with interview (#14) and line number (#13) references (so I can find them again) on individual index cards (note this does not correspond with the sample coding 1 document). This allows me to physically make piles of the codes that seem like they 'go together' in the next step of axial coding.
Here is an example of a bit of text (fascinating learning experience to deal with manipulative doctor) from an interview that I felt was important and the associated code (self-assessment/awareness) on an index card from my own coding. Note this card also has been categorized as 'ineffective,' as my research in this case was comparing those nominated as effective (focus of study) versus those deemed ineffective physician leaders (control group).
As I said previously, you may be looking for similarities in coding, but you also may be looking for differences between groups, which I was in this case.
AXIAL CODING
In the axial coding step, your goal will be to categorize the codes you created in the open coding step. This step is essential identify abstraction in the data. You want to maintain a history of the individual code(s) and then the categories, because as you categorize, you also may find that as you go, you may wish to change the categories as you realize a different schema of understanding. Remember, as you categorize codes, they do not become irrelevant. It will still be important to refer to when you are creating your story.
SELECTIVE CODING
The selective coding step will allow you to identify relationships between codes and categories that you created in the previous step. You can ask the question, “What consequences arise from what actions/interactions under what circumstances?” Once you start answering this question using your codes and categories, you will start finding relationships.
If you are familiar with mind-mapping (Links to an external site.), this might be a useful technique in your repertoire to consolidate your ideas into selective categories.
Identifying themes-- read this article Download this article :
Ryan, G. W., & Russell Bernard, H. (2003). Techniques to Identify Themes. Field Methods, 15(1), 85–109. https://doi.org/10.1177/1525822X02239569 (Links to an external site.)
Watch this video to review the process of coding, outlined above.
Sub-Module 3: Deductive Coding
Deductive coding is a type of coding of qualitative data in which you start your analysis with codes already in mind, based on previous research, a theoretical framework, or your own experience.
It is the same iterative process of inductive, with the exception that you are not exploring your data with a blank sheet, but rather searching for the codes that you have in mind from an existing framework.
READ this article (Links to an external site.) on developing and using a code book for analysis.
DeCuir-Gunby, J. T., Marshall, P. L., & McCulloch, A. W. (2011).
Developing and using a codebook for the analysis of interview data: An example from a professional development research project. Field methods, 23(2), 136-155.
My advice is that unless you FIND an existing code book you wish to use, you might want to use more of a grounded (inductive) approach to your data analysis. Creation of a code book, unless based upon a rigorous framework or schema, is very time consuming.
If you would like to explore a text to understand more about coding, check out this text by Boyatzis on qualitative analysis: Transforming Qualitative Information, Sage, 1998.
Here is a simple sample text for deductive coding:
“We think that sometimes parents, we don’t talk about sex to our daughters. Therefore when they start to have sexual relations, they don’t have as much knowledge on how to use a condom and that puts them at higher risk. They say that girls have a higher risk because there is less information about sex.”
Now use the following codes to explore the above text.
Topic codes:
•Parent-child communication
•Lack of knowledge
•Vulnerability of girls
How would you code the sample?
Perhaps like this?
Topic codes:
•Parent-child communication-- parents, we don’t talk about sex to our daughters
•Lack of knowledge-- when they start to have sexual relations, they don’t have as much knowledge on how to use a condom and that puts them at higher risk.
•Vulnerability of girls-- girls have a higher risk because there is less information about sex.