576 Qualitative data analysis

Tips and skills

Coded text from the Disney project
Interviewer OK. What were your views or feelings about the presentation of different
cultures, as shown in, for example, Jungle Cruise or It’s a Small World at
the Magic Kingdom or in World Showcase at Epcot?
Wife Well, I thought the different countries at Epcot were wonderful, but I # uncritical 5
need to say more than that, don’t I? $ enthusiasm 4
4
Husband They were very good and some were better than others, but that was 5 4
down to the host countries themselves really, as I suppose each of the 4 not critical 4
6 4
countries represented would have been responsible for their own part, 4 of Disney 6 aesthetic
so that’s nothing to do with Disney, I wouldn’t have thought. I mean 7 4 critique
some of the landmarks were hard to recognize for what they were 5 4
4 4
supposed to be, but some were very well done. Britain was OK, but there 4
4
was only a pub and a Welsh shop there really, whereas some of the other content 4
6
pavilions, as I think they were called, were good ambassadors for the critique 7
4
countries they represented. China, for example, had an excellent 360 4
degree film showing parts of China and I found that very interesting. 7

Interviewer Did you think there was anything lacking about the content?
Husband Well I did notice that there weren’t many black people at World Showcase, 5
particularly the American Adventure. Now whether we were there on 4
6 visitors’ 5
an unusual day in that respect I don’t know, but we saw plenty of black 4 ethnicity 4
Americans in the Magic Kingdom and other places, but very few if any 7 4 ethnicity
6
in that World Showcase. And there was certainly little mention of black 5 4 critique
visitors’ 4
history in the American Adventure presentation, so maybe they felt 6
7
alienated by that, I don’t know, but they were noticeable by their absence. 7 ethnicity

Interviewer So did you think there were any special emphases?

Husband Well thinking about it now, because I hadn’t really given this any 5
consideration before you started asking about it, but thinking about it 4
4
now, it was only really representative of the developed world, you know, 4 nationality
Britain, America, Japan, world leaders many of them in technology, and 6
critique
4
there was nothing of the Third World there. Maybe that’s their own fault, 4
maybe they were asked to participate and didn’t, but now that I think 4
about it, that does come to me. What do you think, love? 7

Wife Well, like you, I hadn’t thought of it like that before, but I agree with you.

Steps and considerations in coding recording interviews, you begin transcription at a

relatively early stage.
The following steps and considerations need to be borne
• Read through your initial set of transcripts, field notes,
in mind in preparation for and during coding. documents, etc., without taking any notes or consider-
• Code as soon as possible. It is well worth coding as ing an interpretation; perhaps at the end jot down a
you go along, as grounded theory suggests. This may few general notes about what struck you as especially
sharpen your understanding of your data and help interesting, important, or significant.
with theoretical sampling. Also, it may help to allevi- • Do it again. Read through your data again, but this
ate the feeling of being swamped by your data, which time begin to make marginal notes about significant
may happen if you defer analysis entirely until the end remarks or observations. Make as many as possible.
of the data collection period. At the very least, you Initially, they will be very basic—perhaps key words
should ensure that, if your data collection involves used by your respondents, names that you give to
 Qualitative data analysis 577

themes in the data. When you do this you are coding— forging interconnections between codes, and reflect-
generating an index of terms that will help you to ing on the overall importance of your findings for the
interpret and theorize in relation to your data. research questions and the research literature that
have driven your data collection.
• Review your codes. Begin to review your codes, possibly
in relation to your transcripts. Are you using two or
more words or phrases to describe the same phenom- Turning data into fragments
enon? If so, remove one of them. Do some of your
codes relate to concepts and categories in the existing The coding of such materials as interview transcripts has
literature? If so, might it be sensible to use these typically entailed writing marginal notes on them and
instead? Can you see any connections between the gradually refining those notes into codes. In this way,
codes? Is there some evidence that respondents portions of transcripts become seen as belonging to
believe that one thing tends to be associated with or certain names or labels. In the past, this process was
caused by something else? If so, how do you charac- accompanied by cutting and pasting in the literal sense
terize and therefore code these connections? of using scissors and paste. It entailed cutting up one’s
transcripts into files of chunks of data, with each file
• Consider more general theoretical ideas in relation to
representing a code. The process of cutting and pasting is
codes and data. At this point, you should be beginning
useful for data retrieval, though it is always important to
to generate some general theoretical ideas about your
make sure that you have ways of identifying the origins
data. Try to outline connections between concepts and
of the chunk of text (for example, name, position, date).
categories you are developing. Consider in more de-
Word-processing programs allow this to be done in a way
tail how they relate to the existing literature. Develop
that does not rely on your DIY skills so much (see Research
hypotheses about the linkages you are making and go
in focus 22.4 for an account of this use of word-processing
back to your data to see if they can be confirmed.
software). Nowadays CAQDAS software is increasingly
• Remember that any one item or slice of data can and
being used to perform these tasks (see Chapter 25).
often should be coded in more than one way.
As Coffey and Atkinson (1996) observe, following
• Do not worry about generating what seem to be too many Strauss and Corbin’s account (1990) of grounded theory,
codes—at least in the early stages of your analysis; codes should not be thought of purely as mechanisms for
some will be fruitful and others will not—the import- the fragmentation and retrieval of text. In other words,
ant thing is to be as inventive and imaginative as they can do more than simply manage the data you have
possible; you can worry about tidying things up later. gathered. For example, if we ask about the properties
• Keep coding in perspective. Do not equate coding with and interconnections between codes, we may begin to
analysis. It is part of your analysis, albeit an important see that some of them may be dimensions of a broader
one. It is a mechanism for thinking about the mean- phenomenon. For example, as shown in the next chapter
ing of your data and for reducing the vast amount of (see especially Figure 25.1), ‘ethnicity critique’ came to
data that you are facing (Huberman and Miles 1994). be seen as a dimension of ‘ideology critique’, along with
You must still interpret your findings, which means ‘class critique’ and ‘gender critique’. In this way, we can
attending to issues like the significance of your coded begin to map the more general or formal properties of
material for the lives of the people you are studying, concepts that are being developed.

Tips and skills

Too many codes
The initial coding of a large corpus of data can generate an alarming number of codes. Charmaz (2004), for
example, recommends as a first stage in coding for grounded theory ‘line by line coding’, whereby virtually every
line in a transcript or other source of data will have a code attached to it. She argues that this process means that
the qualitative researcher does not lose contact with his or her data and the perspectives and interpretations of
those being studied. However, this process will almost certainly result in a proliferation of codes. This should not
be alarming. What the analyst of qualitative data needs to do is ask questions about what these codes have in
common so that they can be combined into higher-order and more abstract codes.