PART III ANALYSING AND WRITING

UP YOUR RESEARCH
9 HOW TO DO DATA ANALYSIS : THE BEGINNER’S
GUIDE TO CODING

LEARNING OBJECTIVES
By the end of this chapter you will have the tools to:

Prepare your data for coding

Develop a codebook

Understand the basic structure of coding, including codes, categories,

and themes

Conduct pre-coding, first-cycle, and second-cycle coding

Chapter summary

We often start a project seeking that perfect ah-ha moment. Yet, as Richards
(2009) notes, these insights do not arrive out of thin air. Instead, most
discoveries are the product of good research design and ongoing analysis. In
this chapter, we present one of the main ways researchers bring order to
qualitative data: coding. We take you from the early stages of preparing your
data to first- and second-cycle coding.

INTRODUCTION

The majority of projects arrive at a good conclusion through analysis

processes rather than a grand moment of discovery. Arrival will be confirmed
by growing confidence that you really know what is going on. It happens, in
other words, over time, through thinking and working with the data.
(Richards, 2009, p. 143)

We have written this chapter for qualitative researchers who are relatively new to the
process of data analysis. Qualitative data analysis can be daunting and confusing, so
we have avoided the ‘everything and the kitchen sink’ approach. Our goal is to offer
concrete strategies for bringing order to qualitative data by coding. We recognize
other ways to analyse qualitative data, and coding is inappropriate for some projects
and approaches to qualitative methods (Saldana, 2013). Our discussion is also
primarily focused on analysing texts such as transcripts, field notes, documents and
online materials. Once you have a good understanding of at least one of the primary
methods for analysing qualitative data, you can expand your toolkit.

In addition to understanding the mechanics of coding, Matt Rafalow (featured in this

chapter) reminds us of two important lessons. First, Rafalow’s example illustrates the
role of ongoing reflection throughout the research process. Good researchers do not
just ‘decide’ and ’move on’; they continuously question whether a particular decision
or approach honours the data. They ask: ‘Am I getting it? Does my analysis reflect
what is going on?’ Second, Rafalow provides a poignant example of how our
personal feelings can be an important data source. By carefully ‘separating out’
observed data from our reflections, we can evaluate both others and ourselves in the
pursuit of knowledge.

Below we detail the generic timeline of qualitative data analysis discussed in this
chapter. Some of these steps, such as developing a codebook, start early and
continue throughout the project’s data collection and analysis phases. Others, such
as deciding on the tool you will use to analyse your data, occur only once. More
experienced researchers will likely come into a project knowing that they intend to
use a particular computer-assisted qualitative data analysis software (CAQDAS)
program, for example.

This chapter outlines the steps to prepare your data, create a codebook, write
memos, and choose the appropriate coding tools. We provide you with the tools to
develop a codebook, engage in various coding approaches, and the steps of pre-
coding, first-cycle coding and second-cycle coding.

1. Step One: Getting Prepared: We will outline how you should prepare your data,
including developing a codebook and selecting the tool(s) you will use to analyse
your data.
2. Step Two: Pre-Coding, First-Cycle and Second-Cycle Coding: We will explain
how you should approach data collection, including pre-coding your data while
collecting it and are still in the early phases of data analysis. Next, we will show
you how to conduct first-cycle and second-cycle coding.

STEP ONE: GETTING PREPARED

Key takeaways

Preparing, labelling and organizing data early will help you keep the
project organized
Develop themes and codes early on and continue to develop them as you
engage in the coding process
There are many types of coding, from manual coding to using computer-
assisted software

Data analysis is an ongoing process

In this section, we discuss preparing your data, creating a codebook, writing memos
and selecting the tools you will use to code (e.g. CAQDAS). You should not leave
these tasks until the end of data collection. Treating your data analysis as an ongoing
process will allow you to gain a deeper familiarity with your materials and likely
generate more meaningful insights.

Preparing your data: early considerations and tasks

As you collect your data, you must prepare your data for analysis.
 Labelling: Organize and label all transcripts, field notes, pictures, or other
qualitative materials. A label should include all the information you need to
readily retrieve and identify the data, such as the name, location, contact
information and date of data collection. You may think that you will remember,
but as you collect more data or engage in larger projects, labelling, and
organization are crucial to success.
Making decisions about what to code: In all likelihood, you will not be able to
code all the data you have collected. In the process of selecting a community
group, for example, you may have collected information about many other
community groups. Depending on your research question, these data may be
relegated to a paragraph in your methods section about how you ended up
selecting group ‘A’ over groups ‘B’ or ‘C’. It is also not uncommon for researchers
to decide whether they want, at least initially, to code the entire transcript or
sections of the transcript. Your interview schedule, for example, may have been
divided into several sections to capture a reasonably wide range of topics, with
only one or two being pertinent to the research question you plan on answering
in your thesis or a particular paper. In collaborative projects, decisions may have
been made early about which part of the interview each person ‘owns’.
Preparing documents: If you have transcripts or field notes, Saldana (2013)
recommends double spacing text or creating ‘stanzas’ (Gee et al., 1992). Stanza
is a term used to describe grouping texts that represents a particular topic or line
of discussion. Line spaces are used to separate stanzas to define a new topic or
direction. In the example below, we divided the text into three separate stanzas
to differentiate the discussion about parent engagement, worries about friends
and friends as bad influences. Each stanza is separated by a researcher
question.

Example: Preparing Your Documents – the Role of Stanzas

QUESTION: What does the term parent engagement mean to you?

PARTICIPANT 1: Well it’s like being there, you know. Being supportive.

QUESTION: Can you elaborate on what you mean by ‘being there’ and
‘supportive’?
PARTICIPANT 1: Being engaged means that I take responsibility for raising
my kid, not the school or anyone else. I make it a point to know what is
happening at school, you know in terms of the curriculum, deadlines,
homework, what teachers expect, what Sara should be working on to
succeed. We work with Sara every night on her homework. We check it. But
beyond school stuff, being engaged also means knowing your kid’s friends,
and what your child is doing. I guess for me it’s hard to articulate because it’s
so pervasive. There’s no one thing, it’s really everything. School, after-school,
what they’re watching on TV, who they’re texting, what’s on their phone …
monitoring their friends. Wow, that’s a big one.

QUESTION: How so?

PARTICIPANT 1: The friend thing is so huge now. All this texting and hanging
out. Kids coming over to the house. As a parent you’re always wondering, who
is this kid, do I want them hanging around my house? Who are their parents?
What are they up to? You wouldn’t believe the stuff that I’ve heard.
QUESTION: Like what?

PARTICIPANT 1: Well, there are so many kids that are just bad influences.
One kid, I thought I knew her, seemed so sweet. Well, she was selling drugs
out of her mom’s car no less. And I heard about another one, I can’t recall his
 name. But anyway, another bad egg. He stole a case of wine from his job, he
was bussing tables. Of course, he got caught. These kids, they think they’re
so clever sometimes. But anyway, you just have to be so careful. Here’s this
kid, who looks sweet, has good grades and is all ‘yes, sir, please and thank
you’, when they come to your house, and next thing you know they’re busted
for selling drugs. Really scary stuff.

Formatting: If you import your data into CAQDAS, format it to the software
program specifications (e.g. .docx, pdf). The variation of the programs and
ongoing program upgrades makes it impractical to list all the formats. Consult the
specifications of the software program.
To hard copy or not: Even if you plan on importing materials into CAQDAS, some
researchers find it helpful to have a hard copy handy as a reference guide. If you
have accretion measures (e.g. photos), you can lay out your materials on the
floor or use a corkboard to display them. Some researchers find it helpful to
physically see, touch, and arrange their materials before and during data
analysis.

The codebook
A codebook ‘is a set of Codes, definitions, and examples used as a guide to help
analyze data’ (DeCuir-Gunby et al., 2011, p. 138). Starting a codebook should be part
of the planning process, but it is an ongoing task that gets modified along the way as
you refine your analysis. The point of the codebook is to formally operationalize your
codes, and maintain consistency across coding and, in some cases, across coders.
A very elaborate and multi-person project may require a lot more detail, while a
researcher working alone may require short and simple descriptions to stay on track.
Specify each code, define what each code means, and the limits or exclusions of
each code. You should also include a representative quote to remind you of the
essence of a particular code. Modifying MacQueen et al. (1998) and others (Saldana,
2013), a codebook usually includes the following information:

Code name: The label that you have assigned to the code

Code definition: A short description of the code

Inclusion and exclusion criteria: The criteria or central characteristic that justifies
the material’s inclusion or exclusion from a particular code

Examples: One or two examples (e.g. interview quote) that best represent the
code

There are other categories that you may include in your codebook. Saldana (2013)
suggests including a ‘close, but no’ category to specify material that could be
mistaken for a particular code (in other words, the ‘close but no cigar’ example).
Some researchers also suggest separating the ‘inclusion’ and ‘exclusion’ criteria into
separate categories. If a particular code is related closely to one of your interview
questions, you may also want to include an ‘Interview Question’ column.
While it sounds very technical, a codebook can be developed using a simple Word
document or Excel file. Some CAQDAS will populate a codebook for you after you
create codes with the descriptions and information. Rather than be too prescriptive
about the codebook format, we have provided you with a simple template that is
easily modified to suit a range of topics and organizational preferences (Table 9.1).
Some researchers may want to organize codes alphabetically; others may prefer to
organize their codebook by topic, concept, or theory.
Table 9.1 Codebook template

Code Description Inclusion/exclusion Example

Data analysis tools: from manual to CAQDAS options

Finally, you will need to decide on the tools you will use for data analysis. There are
options: a) Manual; b) Word or Excel; and c) CAQDAS.

Manual options
The most low-tech option is to code manually. There are two basic strategies for
manual coding (see Table 9.2 for a summary). The first strategy is to use a hard copy
of the data – transcript, field notes, and so forth. Some researchers like to colour-
code text with a highlighter or marker, using different colours to signify a particular
code or category (e.g. green for ‘hate school’ and yellow for ‘love school’). You can
use sticky notes to add notes or memos on the side.
Other researchers prefer to write the main text on recipe cards, with each recipe card
containing one passage of text. Similarly, some researchers prefer to cut text directly
out of the transcript, with each strip of paper containing one passage of text.
Researchers arrange and rearrange the cards or strips of paper into different piles
that correspond to a particular code or category (e.g. pile 1 = hate school; pile 2 =
love school). Regardless of the method, researchers eventually arrange codes and
categories into more prominent categories and themes.

Manual coding is not without its drawbacks. If you think this is an easier or quicker
option – think again. Manual coding is very labour intensive and can become
unmanageable if the project involves a lot of qualitative data or multiple researchers.
However, for a (very) small pilot project, manual coding is a reasonable option. For
newer and less experienced researchers, manual coding may help them feel closer to
the data and gain a deeper understanding of the process of coding. And some
researchers prefer to code manually for personal or methodological reasons.
Table 9.2 Coding manually

Possible tools

Researchers use a colour-coding system to differentiate Printed transcripts,

codes, patterns, or themes memos, or
unobtrusive data
Researchers often write in the margins
Pens, highlighters,
Some researchers write chunks of text on recipe cards or recipe cards, Post-it
cut transcripts up by text passages. Each card or strip of notes, scissors
paper serves as one code, and researchers organize the
cards or strips into broader categories and themes

Word or Excel
The middle-ground option is to code data using a Word or Excel file (see Table 9.3 for
a summary). However, like the manual option, using Word or Excel can become
unmanageable very quickly. In a Word document, you can colour-code text, copy and
paste sections of text into different codes, categories or themes, or do both. Like
manual coding, each section of text or colour represents another code, category or
theme. Additionally, you can copy and paste into new Word files, with each file being
a different code, category or theme. You can also use tools such as textboxes to
make notes or comments, such as through track-changes, to act like sticky notes
along the side of the document.
In Excel, you can organize columns and rows in various ways to separate your data
into different codes, categories or themes. The most straightforward option is to place
each interviewee or set of field notes into a separate row and then create columns
that correspond to a particular code, category, or theme.
Table 9.3 Coding using Word or Excel

Possible tools

In a Word document, researchers can do one or more of the Word:

following:
• Highlighting
• Highlight text and use a colour-coding system to
differentiate codes, categories or themes • Copy and
paste function
• Cut and paste sections of text, grouping text by codes,
categories, or themes • New Comment
function
• In the ‘Review’ toolbar, use the ‘New Comment’ function
to add memos or comments on the side Excel:
• In the ‘Insert’ toolbar, use the ‘Text Box’ function to add • Copy and
memos or comments within the body of the document paste function
In an Excel document, researchers can:
Create columns relating to codes, categories, or themes.
Create rows that relate to each individual piece of data or
person

CAQDAS options
CAQDAS programs are excellent organizational tools for storing, organizing, and
coding qualitative data (see Table 9.4 for a summary). Qualitative research projects
tend to generate mountains of data that can become unmanageable, even for the
most experienced researcher. It is not surprising that many share Saldana’s
sentiment about CAQDAS programs: when one considers the ability to quickly move
back and forth between analytical tasks and ‘recode, code, uncode, rename, delete,
move, merge, group, and assign different codes to shorter and longer passages of
text with a few mouse clicks and keystrokes … the advantages of CAQDAS over
paper and pencil soon become apparent’ (Saldana, 2013, pp. 33–34).
While CAQDAS programs include various advanced features, you can learn
everything you need to know to perform basic coding in about three hours. Your
university or college may offer workshops, and there are many videos, tutorials, and
other resources online (e.g. manuals).
Table 9.4 Coding using CAQDAS
 Possible
tools

Once the materials are imported into the selected program (see ATLAS.ti
software specifications for formatting), most CAQDAS software
allows researchers to: Dedoose

• Organize and store a large amount of data. Researchers can MAXQDA

import a variety of data including documents, jpegs, videos, mp3
files, and pdfs NVivo
• Assign data to codes and develop broader umbrella categories
and themes
• Organize codes separately or into a family-tree-like structure to
signify a pattern, relationship, or hierarchy
• Assign a passage of text, picture, or other to more than one code
• Create, add, delete, merge, or modify codes and their content as
the project develops

• Review materials line by line, picture by picture, and so forth

• Search for key words or phrases

• Link one piece of data with another

• Write memos
• Conduct a content analysis

• Create various kinds of displays (e.g. matrix) or data maps

Programs will vary, but many of the most popular brands allow researchers to import
a variety of data, including documents, jpegs, videos, mp3 or other audio files and
pdfs. Once the data are imported, CAQDAS programs will allow you to create codes.
The coding structure may vary slightly, but most will allow you to develop standalone
codes (e.g. Free Nodes) or codes that are structured like a family tree (e.g. Tree
Nodes).
Family-tree-like codes are usually organized hierarchically, with the master or parent
code first, followed by ‘child’ codes and even ‘grandchild’ and ‘great-grandchild’ codes
following it. NVivo, for example, offers researchers a lot of flexibility. Standalone Free
Nodes can be left as is or developed into more elaborate Tree Nodes as the project
develops; conversely, a Tree Node may be broken apart into many different Free
Nodes. And you may decide that a ‘child’ node should become a ‘parent’ node at
some point. The programs allow you to move around, arrange, and rearrange your
coding as you move from first-cycle to second-cycle coding.

Figure 9.1 is an example of Janice Aurini’s project on parent engagement using

NVivo. You can assign data to more than one code at a time and make changes to
the codes, assignments or the data itself at any point. The beauty of these programs
is that while only segments of data are assigned to a code (e.g. a passage of text),
you will be able to see where the data came from readily. So, if you are working with
transcripts, each small coded passage of text will include the label you have assigned
to it (e.g. Mary Smith, Parent, East End School, 1 July 2014). You will also be able to
readily access the entire transcript with a click of a button if you want to re-read the
whole transcript again or see the text just before or after the passage (Figure 9.2).
Figure 9.1 Coding example using NVivo (NVivo 10–2012)

Figure 9.2 Access to interview transcripts using NVivo (NVivo 10–2012)

What is the ’best’ qualitative software? Many programs will get the job done. If you
are new to qualitative data analysis, speak to colleagues who have some familiarity
with CAQDAS about your project and anticipated coding needs.

It is also not practical to provide you with instructions for using CAQDAS software as
the specifications change with each update. Once you have selected your software,
most brands provide detailed and easy-to-follow instructions (e.g. how to import
documents, how to create codes). Your post-secondary institution may even sponsor
information or training workshops.

There is a lingering misconception that CAQDAS programs code for you. Fortunately,
the field has matured. Most qualitative researchers appreciate that while many
programs allow you to search for words (e.g. auto code), it is ultimately your
responsibility to verify the accuracy and authenticity of the code by manually going
through each coded passage to ensure that it belongs. In short, coding that is sloppy,
selective, or distanced from the data can occur whether one codes manually or with a
CAQDAS program. Responsible and ethical coding that captures the essence of your
data results from careful and thoughtful data analysis, not the specific tool or program
you use to code.

STEP TWO: PRE-CODING, FIRST-CYCLE AND

SECOND-CYCLE CODING

Key takeaways

Coding evolves throughout the data analysis, from very tentative pre-
coding to more descriptive first-cycle coding, to more thematic second-
cycle coding

When engaging in data analysis, return to your codebook and the data.
New themes, and therefore new codes, may emerge throughout the
process

The terms pre-coding, first-cycle, and second-cycle coding generate an image of

a neat and orderly process that occurs in distinct and compartmentalized stages.
More experienced researchers know that coding is an iterative process that evolves
as the data collection and analysis progresses. Importantly as we and others (e.g.
Miles et al., 2014; Saldana, 2013) recommend, data analysis and coding should
occur throughout the data collection phase.

This section outlines a condensed version of what Saldana (2013) refers to as

‘generic coding’ (p. 64). This approach spans several qualitative methods and is quite
amenable to grounded theory, one of the most widely used methods of analysing
qualitative data. As Miles et al. (2014) observe, most approaches to qualitative coding
share the same basic steps. Below, we present a condensed and slightly modified
version of their list that we will expand on in the remainder of the chapter.

Pre-coding: Assign preliminary codes to collected text (e.g. transcripts, field

notes, websites) as you collect your data. You can do this manually, in a Word
document, or start a project in your selected CAQDAS program.
First-cycle coding: Review pre-coding and make changes as needed. Continue
to assign codes to collected text (e.g. transcripts, field notes, websites). First-
cycle coding tends to be more descriptive and captures the central
characteristics embedded in the data. You should begin to develop tentative
propositions about what you think is going on, patterns, and even some possible
categories and themes.

Second-cycle coding:

Early stages: Review first-cycle codes and start to reorganize and subsume,
where appropriate, first-cycle codes into broader categories. You should
start to isolate patterns. If required, you can use these insights to inform the
next wave of data collection.
Later stages: Again, review categories and revise or add categories as
needed. Compare and contrast emerging propositions with established
 concepts, theories and findings. Start to develop master themes that capture
the overarching essence of the data. Themes each subsume, where
appropriate, several categories, which comprise a collection of codes.

Even at the early stages of second-cycle coding, start to develop a set of propositions
about what you think is going on. Your propositions will later inform the themes that
develop at later stages of data analysis.

Getting started: pre-coding

Saldana (2013) and others (e.g. Layder, 1998) recommend pre-coding your data. As
you collect your data, review your materials, make notes, highlight key passages, and
start to craft preliminary codes and memos.
Preliminary codes that can be created in advance, even during the early stages of
your data collection and analysis, include (see also Miles et al., 2014, p. 81):

Description or attribute codes: Attribute codes capture the basic characteristics of

the people, places, or things in your study.

Deductive codes: Deductive codes are generated from your research questions,
key concepts, theories that you have drawn on to design your study, and your
literature review.

Interview schedule: You can also use the questions posed in your interview
schedule to create an initial list of codes.

‘Potpourri’: Given the infancy of data collection and analysis, you should feel free
to follow your gut instinct. These codes may or may not initially fit with your larger
framework but instead strike you as important for their potential empirical or
theoretical utility.

Write ‘memos’ – your thoughts, hunches, theoretical musing, questions about the
data, problems with the analysis, relationships with the participants, your own
emotions, questions about your coding decisions – throughout the pre-coding and
coding process.
In Chapter 5, we discussed four types of memos:

Summative memos are a basic description of the participants and a general

overview of what happened during data collection.

Theoretical memos are conceptual ideas that emerged during data collection, in
the field or while reviewing your data.

Methodological memos relate to any methodological or data collection issues

that emerged during data collection.

Personal memos are all your reflections about issues that may have affected
the quality of the data collection.

Memos can be simply written in the margins or created in a separate document. Most
CAQDAS programs have some function that allows you to add memos either
separately, or connected to a piece of data. Saldana (2013), on the other hand,
dislikes using the ‘memo’ function in CAQDAS programs, and instead prefers to write
analytical memos freely first, before determining how to label them or situating them
within the larger project. Only you can decide what works for you – a simple pad of
paper, notes in the margins of your documents or CAQDAS memo function.
Memos are an essential part of data analysis and should be written throughout the
data collection process. Rafalow reflects on his use of brackets to add to his field
notes. This technique is similar to creating memos; it is a handy way to remember
everything that you are thinking and feeling along the way.

How to Relive, and Relearn, Middle School

Matt Rafalow, Google1

Middle school was a horrific time for me. I was bullied and isolated in ways
that affected me well into early adulthood. When I started the fieldwork that led
to my book Digital Divisions (2020), I justified its focus on middle schools by
explaining that middle school is important to study because it is a period that
affects students’ later life chances and opportunities, and that it is sorely
under-researched. But this was not the full truth of what motivated my study.
My own story motivated it, too.
1 This chapter does not reflect the views or opinions of Google.

Part of why I wanted to ‘go back’ was to examine the unresolved. As an

undergraduate, I remember reading Pascoe’s Dude You’re a Fag (2007) and
feeling deeply seen. She unpacked how constructions of gender and sexuality,
rooted in peer dynamics in high school, shaped troubling forms of harassment.
But it wasn’t enough. I needed to go to earlier grades, to the ‘scene’ of what I
experienced, to understand why kids go through what they do – and why I
went through what I did. Early on, I wilfully ignored what felt like a cognitive
dissonance of sociological and psychological motivations.

It didn’t help that my training gave me mixed messages about the value and
risk of honouring and using my own experiences in pursuing my fieldwork and
analysing what I observed. How do you generate and code field observations
into science, and not a report of your own subjective interpretations? There’s
been quite a lot written about this topic by ethnographers already (Flaherty et
al., 2002). But what I’d like to offer is a template for how, through my process
of observation and analysis, I both found some peace and produced science.
I did, indeed, relive some of the worst parts of middle school as I pursued this
project. But fieldwork and ongoing analysis exposed me to a process of
relearning middle school – in ways that helped me to more fully document
school social structures. I realized that this process of relearning was not too
different from how ethnographers carry with them into the field their
assumptions from existing literatures and, through their iterations between
observation and analysis, build on what we know in science. Further, and
most importantly, if I had not been reflexive about my past, I would not have
been able to uncover what I did.
One of the tactics that I was taught by my ethnographer-mentor David Snow
was to take field notes in ways that differentiate what I actually observed from
every other kind of thought I had by using brackets [ ]. Here is an example
from my field notes:

Walked out of a seventh grade class where we had just watched a

video documentary that included a bit about the statue of Lincoln
sitting in a chair at the capital. Six students who had just walked out
of the class, too, gathered together by lockers in the hallway. They
were making fun of another young man in the group by suggesting
he would want to sit on Lincoln’s lap. ‘Oh, you’d love to sit on his lap.
You’d totally love it, Eric!’ Another student responds: ‘Eew, he’s gay I
knew it!’ Another young man in the group started to mimic sitting
 down, and made a facial expression that resembled sexual
enjoyment. Eric periodically says ‘shut up’ but does not otherwise
respond. Ms. Leary, a teacher who had been present during the
entire interaction, chuckled. She says, ‘Eric, maybe if you paid
attention in class you’d have a better comeback. You’re letting your
man Lincoln down!’ The whole group laughs, and Eric is quiet.
[I wonder what Eric is thinking and feeling as this happens. This
seems to be a clear example of what Pascoe observed in her work
on harassment.]

[I’m not feeling well after observing this. I feel anxious. I think I had a
flashback from my own experiences of being harassed like Eric and a
teacher only making it worse. Are all middle schools like this? Are
status hierarchies just a fact of life that will always lead to someone
getting hurt?]

Notice I used brackets to differentiate observations from questions I have to

myself. But I also used brackets to share the flashbacks and anxiety I felt,
even linking these woes to existing work on status hierarchies that share
similar conclusions I had at that time. This is a practice that is an attempt to
‘separate out’ the real from the subjective (Krieger, 2018). Bracketing made it
easier to identify specific vignettes from the field in my later writing. But I later
realized that my bracketed reflections, including my feelings, were valuable
data, too.
My coding process was modelled from Emerson, Fretz and Shaw’s (1995)
work on field notes and analysis. I analysed field notes and interviews
throughout my data collection process, generating codes as emergent themes
came up; I identified how themes connected to existing scholarship; and then
developed new questions I would then investigate back in the field. As the
project continued, I expected I would whittle these codes down even further
into more specific categories that might tell me something new about the
literatures they speak to.

Early on in my analysis, I considered coding only non-bracketed observations

in the interest of producing the most scientific results. But I couldn’t shake the
feeling that ignoring my bracketed anxiety was somehow wrong. So, I coded
my thoughts and feelings, too, ensuring that I knew which were bracketed
excerpts and which were not. In analysing these sections, I found myself
creating new codes, like ‘anxiety’ or ‘flashback’. In a particularly dark period
early on as I iteratively analysed this data, I wondered whether bullying truly is
a fact of life (I even created a code to this effect), and suspected that maybe
there will always be kids like me and Eric that will suffer from the brunt of
harassment at school.
After this coding work, a process that specifically included mapping out my
own feelings, I realized what I ended up with were more research questions
that I needed to pursue in the field. Was bullying a fact of life at each of the
three schools I studied? How did bullying feel to the students at the different
schools? Did it feel like how I remembered it? Did teachers similarly
participate in student harassment?

If I had not coded these feelings and flashbacks, I would not have known to
seek out answers in the field. I did, in fact, observe bullying happen at each of
the three schools I studied. I could have let both my memories (‘bullying is
always traumatizing’, ‘bullying is a fact of life’) and existing work (‘bullying can
lead to life altering trauma’, ‘bullying is a durable feature of status hierarchies’)
lead me to move on and focus on other phenomena at school. But by testing
these assumptions in the field I realized that they were quite wrong.

While bullying did indeed occur at the three schools I studied, I found that the
meaning and structure of harassment varied quite significantly by school. The
school from the field notes I shared earlier was quite similar to the school I
grew up attending: harassment was rampant, often targeted towards
marginalized students based on gender, sexuality or race, and teachers
participated in bullying. But because I analysed both what I observed and
what I felt and used this to identify new questions, I learned that harassment
worked differently at the other schools. Not only was the subject of bullying
different at other schools – focusing, instead, on romantic dramas – but its life
cycle was different, too. Students at the other schools told me how after
bullying happened they would try to defuse the underlying cause, offering
support to both the harasser and the bullied, to understand what was going on
and how they could help. Further, I found out that teachers played a much,
much more important role in student harassment than I realized. Had I only
focused on observed bullying within peer student groups – what I initially felt
compelled to focus on due to my own history – I would have missed that
teachers at the school where bullying was the worst were being bullied by
other teachers, too. They described to me a hostile workplace, where they
learned to expect hostility from one another, and saw how this informed how
they treated students with hostility as well. This didn’t exist at the other
schools.

For me, coding and analysing data is a process of evaluating both others and
yourself in the pursuit of building on what we know about human behaviour. To
ignore your past and current experiences, including how you feel as you
collect data, is to ignore the full range of the social structures you seek to
understand. Armed with perspectives from scholars past, I analysed my own
experiences in the field – and in doing so, I was able to gain a more complete
picture of the structure of school experience, too.

Questions for reflection

1. What are the key ‘takeaway’ lessons?
2. What does Rafalow mean by ‘separating out’ the ‘real from the
subjective’? How is this accomplished at the data collection phase? How
is it accomplished at the data analysis and write-up phase?
3. What are the benefits and challenges of mapping out and incorporating
personal feelings and experiences?

References
Emerson, R. M., Fretz, R. L., & Shaw, L. L. (1995). Writing ethnographic
fieldnotes. University of Chicago Press.
Flaherty, M. G., Denzin, N. K., Manning, P. K., & Snow, D. A., (2002). Review
symposium: Crisis in representation. Journal of Contemporary Ethnography,
31(4), 478–516. https://doi.org/10.1177/0891241602031004004
Krieger, S. (2018). Beyond subjectivity. In A. Lareau & J. Shultz (eds),
Journeys through ethnography: Realistic accounts of fieldwork (pp. 177–194).
Routledge.

Pascoe, C. J. (2007). Dude, you’re a fag: Masculinity and sexuality in high

school. University of California Press.
Rafalow, M. H. (2020). Digital divisions: How schools create inequality in the
tech era. University of Chicago Press.

Quick tip: How to start coding

 As we have repeated throughout the chapter, you should code throughout the
data collection phase.

If you are using CAQDAS, refer to your codebook and create your initial
codes in the program.

Start slowly and with one transcript, one day’s worth of field notes, one
internet post, and so forth.
First, read over the data in its entirety. Review any memos that you have
written related to that data during the pre-coding phase.
Start at the beginning of the document, and code small sections of text at
a time, while keeping in mind the text in its entirety.
When appropriate, assign pre-existing codes to sections of text.
Develop new codes as they emerge or make changes and update the
codebook.

While you should be thoughtful, do not get too bogged down with creating the
‘perfect’ code. Throughout the coding process, you will have ample
opportunity to modify or discard codes, add additional examples or
clarifications, merge multiple codes into larger ones, and so forth.

Developing codes: first-cycle codes

First-cycle coding is when codes are ‘initially assigned to data chunks’ (Miles et al.,
2014, p. 73). In the literature, this stage is also referred to as ‘open coding’ (Corbin &
Strauss, 2008; Saldana, 2013). A ‘code’ is the term used to describe the word or short
phrase that captures the main essence of one small dimension of your data. As
Charmaz (1983) describes, codes can ‘serve as shorthand devices to label, separate,
compile, and organize data’ (p. 186). It is not uncommon for parts of an interview,
pictures, videos, and so forth to be assigned to more than one code, something that
is particularly easy to do with CAQDAS.
Table 9.5 is an example of a parent describing her contact with her daughter’s school
and how she organizes her daughter’s after-school time. The first quote was
assigned to four codes, while the second quote was assigned to three other codes.
As the coding progresses, we further refined the analysis to differentiate the
responses (e.g. ‘poor’ and ‘good’ communication). For now, we will keep our example
simple.
Table 9.5 Coding example

Data Codes

‘As a parent it’s my job to figure out what is going on with my Code 1:
kid. I’m in constant contact with the school, you know, “How’s Communication
everything going, how was her day at school, what should I be
working on at home, you know in terms of homework?”. We Code 2:
expect Sara to do well at school, and our job is to make sure Intervention
she does well. And if I see a problem, it’s my job to intervene,
find out what is going on and to be part of the solution.’ Code 3: Parent
expectations
‘We do a lot of things after school. I think you can get from our
discussion that I’m one of “those” kinds of parents! Sara Code 4: Sports
certainly has free time, but we make it a point to enrol her in a
 y p

Data Codes

lot of after-school activities like soccer and music classes. Code 5: Music
Homework seems to eat up a lot of time nowadays, now that
Sara is in Grade 6.’ Code 6:
Homework

Since a code represents an individual segment of data, you may develop 100–200 or
more codes; however, there are no hard and fast rules about how many codes you
should have. The number of codes depends on the size of the project and your
approach to coding. Bernard (2011) explains the difference between a ‘splitter’ and
‘lumper’ approach. Splitters break down text into small segments to differentiate each
idea expressed in the text, while lumpers create more summative codes that capture
the essence of a segment of text. Splitters will end up with many more descriptive
codes, while lumpers will end up with fewer codes that are broader and contain more
contextual information. As Saldana (2013) observes, there are benefits and
drawbacks with each approach. Splitting may produce more superficial codes, while
lumping may gloss over important nuances in the data. Early in the data analysis, you
may want to try both approaches to see which one helps you understand your data.
While you are creating these codes, do not get too bogged down about whether they
are ‘right’ or not. You want your codes to be thoughtful and true to your data, but
remember that most codes will be revised as your data analysis progresses. Our list
of first-cycle coding options (Table 9.6) is not as exhaustive as Saldana’s (2013), but
it will give you a good overview of what and how you may code your data.
Table 9.6 First-cycle coding options

Type of Description Sample Quotes Sample Codes

Codes

Descriptive: Captures who, ‘Our customers are Code 1:

Nouns what, and where really interested in clean Customers
eating. We have
locations all over the Code 2:
country, but you tend to Products for
find us in more affluent Sale
neighbourhoods. We sell
a lot of organic products, Code 3: Affluent
vitamins … stuff like Neighbourhoods
that...’

Descriptive: Captures actions, ‘I have a pretty good Code 4: Helping

Action interactions, and relationship with most of Customers
processes as they the customers. I really
are described by enjoy helping Code 5:
participants or customers, you know Answering
observed by the selecting products. Questions
researcher People who are just
getting into clean eating
often have a lot of
questions.’

In Vivo In Vivo coding ‘The Market pays really Code 6: Fair

creates codes from fair wages too, Wages
participants’ own especially compared to
words or phrases the last place I worked.
Type of Description Sample Quotes Sample Codes
Codes

As an employee, I feel Code 7: Feel

valued …’ Valued

Interpretations Captures how ‘It was really wrong what Code 8:

participants happened to Steve. He Wrongful
interpret situations shouldn’t have been Dismissal
or events fired. It wasn’t his fault’

Feelings Captures ‘I was really angry that Code 9: Angry

participants’ Steve got fired. I was so
feelings and confused. It was so Code 10:
emotions unlike the management Confused
…’

Belief systems Questions that ‘I just really question the Code 11:
examine fairness of it all. I just Fairness
participants’ think it’s wrong to fire
values, morals, or someone without a full
standards investigation.’

Assessment Captures ‘I think the impact was Code 12: Impact

participants’ huge. I mean the firing
assessments, just sent a chill through
estimations, or The Market’
valuations

Frequency Captures ‘It was so unlike the Code 13:

participants’ management … it was Unusual Firing
understandings really unusual. It just
about duration, doesn’t happen here’
regularity, or
commonality

Local Captures how ‘I think the firing was Code 14:

causation participants really personal. Fred, Personal
understand why the head guy, just never Reasons
something liked Steve. There was a
occurred lot of personal stuff. I bet
he was just looking for
any excuse to fire him.’

The emergence of categories: early and later stages of

second-cycle coding
Second-cycle coding uses the resulting first-cycle codes to condense, integrate, and
layer them into broader and more coherent categories and themes (Miles et al.,
2014). Second-cycle coding is the act of ‘pull[ing] together a lot of material into a
more meaningful and parsimonious unit of analysis. They are a sort of meta-code’
(Miles & Huberman, 1994, p. 69).
Second-cycle coding involves interpreting and making sense of the data by
identifying patterns, relationships, and explanations. Part of this process includes
structuring the codes hierarchically to identify categories. Your original codes may be
relabelled, subsumed by other codes, rearranged, or eliminated. Second-cycle coding
may also suggest that additional data collection is needed to develop an emerging
theory (referred to as ‘theoretical sampling’).
Developing categories occurs after you have created several codes; it is part of what
some researchers refer to as ‘pattern coding’ (Saldana, 2013) and ‘focused coding’
(Corbin & Strauss, 2008).

Categorization

Categorization is the process of grouping codes under more significant unifying

classifications. As you start to develop codes, you will find yourself arranging and
rearranging your data into broader classifications or typologies based on the patterns
that begin to emerge or for some analytic, practical, methodological, or theoretical
reason. The same codes may be used in more than one category. Since a category is
a grouping of codes, the average project can include numerous categories; there is
no magic number of categories a project ‘should’ include or number of codes each
category ‘should’ contain.
In Table 9.7, we demonstrate the creation of two categories: parent engagement and
after-school. Category 1 captures four codes that relate to several dimensions of
parent engagement. Category 2 captures the three codes that relate to how parents
construct their children’s after-school time. Note that both categories include the code
‘homework’ since it is related to parent engagement and after-school activities.
Table 9.7 Categorization example

Representative quote Categories

‘As a parent it’s my job to figure out what is going on with my Category 1:
kid. I’m in constant contact with the school, you know, “How’s Parent
everything going, how was her day at school, what should I be Engagement
working on at home, you know in terms of homework”. We
expect Sara to do well at school, and our job is to make sure Code 1:
she does well. And if I see a problem, it’s my job to intervene, Communication
find out what is going on and to be part of the solution.’
Code 2:
Intervention

Code 3: Parent
Expectations
Code 6:
Homework

‘We do a lot of things after school. I think you can get from our Category 2:
discussion that I’m one of “those” kinds of parents! Sara After-School
certainly has free time, but we make it a point to enrol her in a
lot of after-school activities like soccer and music classes. Code 4: Sports
Homework seems to eat up a lot of time nowadays, now that
Sara is in Grade 6.’ Code 5: Music
Code 6:
Homework
Themes

As Saldana (2013) observes, a theme is an ‘outcome of coding, categorization, and

analytic reflection, not something that is, in itself, coded’ (p. 14). Specific definitions
will vary, but for our purposes, think of a theme as the big ideas that emerge from
your data. We suspect that you will have anywhere from two to five themes, but there
is no magic number.

Whereas codes and even categories tend to be more descriptive, themes tend to be
the outcome of interpretive processes (Rossman and Rallis, 2003); they may also
relate to an established concept or theory. In short, themes emerge after some pretty
significant analysis and reflection, and certainly after the pre-coding and first-cycle
coding stages.

In Table 9.8 we have expanded our example to three major themes: ‘cultural capital’,
‘social capital’ and ‘economic capital’.
Table 9.8 Putting together codes, categories and themes

Sample Codes Sample Categories Sample Themes

100–200 Codes 15–25 Categories 2–5 Themes

Code 1: Communication PARENT CULTURAL

ENGAGEMENT CAPITAL
Code 2: Intervention
AFTER-SCHOOL
Code 3: Parent Expectations
Code 4: Sports

Code 5: Music
Code 6: Homework

Code 1: Communication INFORMATION SOCIAL CAPITAL

NETWORK
Code 3: Parent Expectations
CONNECTION TO
Code 7: Good Relationship with SCHOOL
Teacher

Code 8: Access to Resources

Code 9: Attend School Events

Code 10: Private Pre-school RESOURCES ECONOMIC

CAPITAL
Code 11: Saving for University
Code 12: Tutoring

First Stage Coding--------------Early Second Stage Coding-------------Later

Second Stage Coding
Moving from codes, to categories, to themes

You can think of the process of coding like a puzzle. First, start with the individual
pieces (codes); next, you put together groupings of pieces that constitute smaller
segments of the puzzle (categories); and after a lot of arranging and rearranging, you
put large sections of the puzzle together (themes). The completed puzzle is the story
you are eventually able to tell people about your study.
But how do you get there? We present four coding strategies for advancing your
analysis from the more descriptive first-cycle to more explanatory second-cycle
coding by developing the following: a) pattern coding; b) focused coding; c)
frequency coding; and d) comparison coding.

Pattern coding
As your analysis develops, you may start to see reoccurring patterns in your data.
Patterns emerge when you find that formally separate codes are connected, reoccur,
or develop in similar ways. Such patterns may represent organizational, social-
psychological or underlying processes about what is happening, how it is happening,
and the assumptions that participants hold to be true about the nature of reality
(Charmaz, 1983). Miles et al. (2014) observes that pattern coding can be very useful
when there are many cases or data. Once identified, a pattern code can serve as a
category or even a theme that subsumes several other codes and even other
categories.

Focused coding
There are various terms that describe the process by which researchers start to
‘clump’ together ‘clusters’ of data. As Miles et al. (2014) explain:

it might be called ‘distilling, ‘synthesizing’, ‘abstracting’, ‘transforming’ and

even the abhorrent ‘reducing’ the data. Even though these are different
works and processes, they kind of mean the same thing … In all instances,
we’re trying to understand a phenomenon better by grouping and then
conceptualizing objects that have similar patterns and characteristics. (p.
279)

Unlike the initial stages of coding, focused coding is more selective and develops
more significant categories. Focused coding requires the researcher to select the
most salient or telling codes that best represent the data. While initial coding broadly
asks, ‘What do the data suggest?’ (Charmaz, 2006, p. 47), focused codes are more
iterative and represent more theoretically rich categories. When conducting focused
coding, the researcher takes a limited set of codes crafted during the first-cycle
coding phase. The goal is to develop more abstract and inclusive categories that
capture a larger amount of data.
Figure 9.3 is an example taken from Emily Milne’s PhD project. After carefully
creating dozens of codes, Milne developed three broad categories: ‘Interpreting
policy reform’, ‘schooling organization’ and ‘educators’. These codes were eventually
reorganized under the larger theme ‘dynamics of policy implementation’.
Figure 9.3 Focused coding example (NVivo 10-2012)

Frequency coding
Qualitative researchers tend not to think of themselves as ‘numbers’ people. After all,
our interest is primarily about the quality, not the quantity, of whatever dimension of
social life we are studying. However, as Miles et al. (2014) rightly point out, ‘when we
identify a theme or a pattern, we are isolating something that (a) happens a number
of times and (b) consistently happens in a specific way’ (p. 282). You can count a
variety of things, including words, phrases, and events. Counting can help you
identify a category and even a theme and it can verify your initial propositions about
what is going on in your data. Counting can also keep you ‘analytically honest’ (Miles
et al., 2014, p. 282) by forcing you to verify your hunches about what is happening ‘a
lot’.
In her research on Canadian school shootings, Stephanie Howells conducted a
framing analysis to determine how school shootings were discussed in the Canadian
news media. Each article was coded to determine the major frame, or theme, of the
article. As part of this research, she counted how many newspaper reports used each
of the major frames that she identified through her analysis. Through frequency
coding, she was able to determine that most (over 60 per cent) of the newspaper
articles focused primarily on three main frames: ‘perpetrator’, ‘victims’, and ‘the
school’ (Howells, 2012b).
Depending on the search, CAQDAS programs make counting particularly easy. Most
programs allow researchers to perform simple word searches, for example, that can
be displayed by raw number of counts (Figure 9.4), Word Clouds (Figure 9.5) and
Word Trees (Figure 9.6).
Figure 9.4 Word frequency result (NVivo 10–2012)

Figure 9.5 Word cloud

Figure 9.6 Word tree

Comparison coding
Comparison coding is a common analytical strategy employed by qualitative
researchers. Comparisons can often be anticipated well in advance, either because it
makes practical sense or based on prior knowledge. As you are creating initial codes,
you may have also divided the data into logical comparison groups (e.g. teachers and
students), allowing you to compare and contrast how each group responded. These
analyses may also point to new ways to understand the data.

As we discussed in Chapter 2, you can also look for internally driven comparisons.
Internally driven comparisons occur when two or more units of interest (e.g.
communities, organizations) are similar or different on the key attribute of interest.

CONCLUSION
This chapter has outlined concrete strategies for analysing qualitative data by coding.
We first described preparing your data for analysis and starting a codebook. We also
discussed three main data analysis tools: manual, Word or Excel, and CAQDAS
options. The main part of the chapter detailed specific steps to move from standalone
codes to the development of broader categories and themes.

The next chapter completes our journey by outlining how to write up qualitative data.
While qualitative research is more accepted, researchers still struggle to publish
qualitative data in peer-reviewed journals, books, and policy venues. Some granting
agencies are also less receptive to qualitative data. Learning how to communicate
qualitative research effectively is critical for overcoming these hurdles and
disseminating your research.

FURTHER SUGGESTED READING

For a more detailed discussion of qualitative data analysis, we recommend two
books. Each book provides an excellent in-depth examination of qualitative data
analysis and coding. We refer to these books many times throughout the chapter.
These books are suitable for novice and experienced qualitative researchers and are
part of our library of ‘must have’ books.
Miles, Matthew B., Huberman, A. M., & Saldana, J. (2019). Qualitative data analysis:
A methods sourcebook (4th edn). Sage.

Saldana, J. (2021). The coding manual for qualitative researchers (4th edn). Sage.

SAGE CASE STUDIES

DeWeese, A., Jennings, P., Brown, J., Doyle, S., Davis, R., Rasheed, D., Frank, J., &
Greenberg, M. (2017). Coding semi-structured interviews: Examining coaching calls
within the care for teachers program. In SAGE Research Methods Cases.
www.doi.org/10.4135/9781473958319
Heckemann, B., Thilo, F. J., & Wolf, A. (eds). (2020). Working via distance—using
computer-assisted qualitative data analysis software, email, and teleconferencing in
qualitative content analysis: Experiences, pitfalls, and practical tips. In SAGE
Research Methods Cases. www.doi.org/10.4135/9781529714197
Orphanidou, M., & Kadianaki, I. (2020). A guide for developing and applying a coding
scheme in qualitative media analysis: Representations of depression in the press. In
SAGE Research Methods Cases. www.doi.org/10.4135/9781529715217
Swain, J. (2018). A hybrid approach to thematic analysis in qualitative research:
Using a practical example. In SAGE Research Methods Cases.
www.doi.org/10.4135/9781526435477

KEY TERMS

CAQDAS Comparison Coding Personal Memos

Categories or Categorization First-Cycle Coding Pre-Coding

Codebook Focused Coding Second-Cycle Coding

Codes Frequency Coding Summative Memos

Coding Methodological Memos Themes

Pattern Coding Theoretical Memos