Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

EUROSPHERE

WORKING PAPER SERIES

Online Working Paper No. 13, 2008

Analysing Interview Data


Possibilities and challenges
Bjarte Folkestad

This paper can be downloaded without charge from: http://eurosphere.uib.no/knowledgebase/workingpapers.htm

ISSN 1890-59886

EUROSPHERE ONLINE WORKING PAPER SERIES Title: Analysing interview data: Possibilities and challenges Author(s): Bjarte Folkestad Working Paper No.13 This version: December 2008 Webpage: http://www.eurosphere.uib.no/knowledgebase/workingpapers.htm

EUROSPHERE, 2008 http://www.eurosphere.uib.no

2008 by Bjarte Folkestad All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including notice, is given to the source. The views expressed in this paper do not necessarily reflect those of the EUROSPHERE Project. The statement of purpose for the EUROSPHERE Online Working Paper Series is available from the EUROSPHERE working papers website, http://www.eurosphere.uib.no/knowledgebase/workingpapers.htm

Bjarte Folkestad Unifob Global / University of Bergen bjarte.folkestad@global.uib.no

ISSN 1890-5986 (online)

Analysing Interview Data Possibilities and challenges


Bjarte FOLKESTAD
This paper examines the possibilities and challenges in analysing qualitative interview data. The main inspiration of this paper draws upon the work of Anne Ryen (2002)1 and the seminal work of Matthew B. Miles and A. Michael Huberman (1994). Based on their contributions I will present the major epistemological perspectives in qualitative research. Furthermore the paper presents different analysis techniques. To enhance the understanding of the various analysis techniques this paper will include examples from my own research project on Euroscepticism among political parties.2 The paper will also discuss issues concerning comparability of the data and analysis. Since my project aim at exploring Euroscepticism among various political parties in Nordic countries and Central Eastern Europe it is important to consider various strategies for comparing and analysing interview data.

Why do interviews?
When a research project is conducting qualitative interviews there is (and presumably must be) a reason for it. One of these reasons might be because it provides a new insight into a social phenomenon. So when I want to explore the content of Euroscepticism, conducting qualitative interviews with elites in political parties might therefore been seen as one option within many that gives insight into this phenomenon. This, of course, will require reflections on the reasons for selecting elites and parties, which respondents should be selected, and how. As the main focus here is on the data analysis, this paper will not consider these issues thoroughly, but as it will be shown below, a distinction between data analysis and data collection is sometimes hard to draw, and it is therefore important to have this in mind when evaluating how interview data can be analysed. Interviews allow the respondents to reflect and reason on a variety of subjects in a different way than say opinion polls or party manifestos. Thus

Det kvalitative intervjuet: Fra vitenskapsteori til feltarbeid (The Qualitative Interview: from philosophy of science to fieldwork) is a methodological book in Norwegian. All quotes from Ryen 2002 are my own translation. 2 It should be noted that the data collection for my project are a part of the Eurosphere project (www.eurosphere.uib.no) this means that I will not personally collect all the data for my own project. Furthermore the data collection will also consist of data that I will not use for exploring Euroscepticism.

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

when elites are asked about issues such as the EU, migration and citizenship we can get a deeper insight in how they think and reflect. It goes without saying that such data are hard to obtain. A thorough interview guide must be developed; appointments must be scheduled; travelling from place to place; and finally the interview must be realised faceto-face. Although there are several issues that might affect the research in these phases, the analysis part is perhaps the most contested part of qualitative method:
The most serious and central difficulty in the use of qualitative data is that methods of analysis are not well formulated. For quantitative data, there are clear conventions the researcher can use. But the analyst faced with a bank of qualitative data has very few guidelines for protection against self-delusion, let alone the presentation of unreliable or invalid conclusion to scientific or policy-making audiences. How can we be sure that an earthy, undeniable, serendipitous finding is not, in fact, wrong? (Miles, 1979:591)

This paragraph neatly sums up much of my own thoughts when I was surveying the literature for guidelines. This is not to say that there are no guidelines in the literature, quite on the contrary. But in order to do surgical reading and to find alternatives/suggestions for how to analyse qualitative interview data at the same time as comparing across countries/ organisations I had to spend quite some time reading through the literature, which again made my strategy of surgical reading quite flawed. In the following I will try to depict some of the major epistemological perspectives in qualitative research as well as exemplify some analysis techniques. What are our preferences? Miles and Huberman (1994) argue that researchers must lay out their epistemological starting points: It is good medicine, we think, for researchers to make their preferences clear (ibid:4). A crucial point in making our preferences clear is therefore to assess how we view the data that is collected. This involves finding answers to questions like: how is the role of the interviewer treated? Do the answers of the respondents represent the reality or is the reality produced through the interaction between respondent and interviewer? While it is beyond the scope of this paper to give a full presentation of the main paradigms that exist within qualitative research today, a short summary/ review could be useful in order to get a better insight into how the analysis is performed. Based on Ryen (2002) and Silverman (2001) the table below present some major paradigm within qualitative research.

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

Table 1: Major paradigms in qualitative research3


Paradigm Characteristics the social reality is real important that the researcher does not affect/ influence the data quotes are marked Focus on Whatquestions search for production of meaning reality is produced through interaction focus on howquestions purpose: to get on the inside focus on the subjects own feelings both the researchers and the subjects feeling is empirical material worried about selfconsciousness research constructs the reality by producing description on it. Status of Data Methodology Random samples Standard questions Tabulations

Naturalistic Positivism

Facts about behaviour and attitudes

Etnometodology Constructionism

Mutually constructed

Unstructured Open-ended interviews

Emotionalism

Authentic experience

Any interview treated as topic.

Post-modernism

Deconstructing texts

Of these, the naturalistic paradigm is held to be the most dominant one. I believe one of the reasons for this is the fact that this paradigm offers some concrete solutions in how to analyse qualitative data. This paradigm calls for a standardised procedure, with structured or semi-structured interviews, that usually are pre-tested. This result in data ...that give us access to 'facts' about the world (Silverman 2001:86). While being dominant it has been subject to criticism from a range of researchers, which explains why there are several alternatives in how to collect and analyse qualitative data. In particular the role of the interviewer and how the interview/conversation are interpreted are crucial points of departure when criticising the positivist paradigm. So while the positivist interviewer is objective the interviewer in the emotionalist tradition is subjective. As Denzin put it:
I wish to treat the interview as an observational encounter. An encounter... represents the coming together of two or more persons for the purpose of focused interaction (Denzin, 1970:133 quoted in Silverman 2001:95).

The interviewees own feelings through the interview are thereby also empirical material for analysis. Ethnometodologist/ constructionist are on the
3

This table is based upon Ryen (2002) chapter 4 and Silverman (2001) chapter 4. It should be noted that there are differences in the labelling of the paradigms. This is shown in the table for the two first rows (Ryens labels are mentioned first).

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

other hand are ...searching for produced meaning or conversation on how the world is constituted by its members (Ryen 2002:64). This means that the interview and the interaction between the respondent and interviewer are important to consider. Within the Post-modernist tradition (which is not mentioned by Silverman), the issue of representation is crucial. Post-modernist de-constructs texts in order to re-present the reality. The research-procedure constructs the reality by producing images of it. The reality is put in parenthesis and what is left, is a world of images and representation (Ryen 2002:70). The purpose of this short review is to pave way for the argument that both single research projects (conducted by a single or small group of researchers) and more multiple/ large projects (conducted by a large group) should make some assumptions on how the data is treated. I believe that there is much to the fact that the data is influenced by the very interview situation itself. The quality of the data will vary not only from country to country but also from interview to interview. For example factors like where the interview is conducted (at the respondents office or in a public space like a restaurant or caf) might affect the data response. Also the difference in elite structure (Bygnes, 2008), formality of the interview situation, nervousness (both interviewer and interviewee) is factors that should be accounted for. At the same time it would probably be a too time-consuming process if factors like these should be built into the final data-report, particularly in larger research projects and when comparing across cases/ countries is important. A suggestion would therefore be to use the naturalistic/ positivist perspective as a starting point for analysing the data. There are two reasons for this: first of all because this perspective offers some clear guidelines for analysing the data. Secondly these guidelines could be useful in enhancing the comparability of the data, because it gives some standardisation of data collection, reduction and analysis.

Analysing qualitative interviews


In the following part I will present some of the analysis techniques that are applied in the literature. This part builds upon the examples and methods described in Ryen (2002), Miles and Huberman (1994) and Erlandson et al. (1993) Before introducing the possible analysing techniques it should be noted that the analysis phase in itself is a continuous process, and that we can not easily distinguish the collection, reduction and analysis phases from each other: data analysis does not occur in a vacuum (Erlandson et al. 1993:113). This is because researchers are continuously interacting between the respondents and the research tools. For each interview that is conducted more knowledge is possessed: not only about the phenomenon that is studied in it self but also about the interview guide as well. As the researcher becomes more experienced he or she will find several buttons to push in order to get the information that we are searching for. A technique that might not demand particular systematising than a transcription is quote-research. Here we use quotes from interview as illustrative or confirming examples.

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

For instance if we have the assumption that says that: Euroscepticism is a position that views the EU as a market-liberal project. A quote like: I think that EU is too market-liberal and I am therefore against the whole project would then be taken as an example that would confirm this assumption. The challenge here is to distinguish between journalism and science. While using quotes to describe the data is not a problem in itself it becomes a problem when this is the only way of analysing the data (Ryen 2002:169). Let us therefore consider the naturalistic/ positivist paradigm where a more sophisticated technique has been sought developed. Lincoln and Guba argues that the data Analysis involves taking constructions gathered from the context and reconstructing them into meaningful wholes(Lincoln and Guba, 1985:333 quoted in Erlandson et al. 1993:116). This process has according to Erlandson et al. (1993) four elements: 1)unitizing data, 2)emergent category designation, 3)negative case analysis and 4)bridging, extending and surfacing data. Here we consider the two first steps: unitizing and categorizing which follows a step-by-step procedure: 1. 2. 3. 4. Read the first unit of data Read the second unit Proceed in this fashion until all units have been assigned to categories Develop category titles or descriptive sentences or both that distinguish each category from the others 5. Start over

The data reduction here starts with the raw data which can be a section or the entire answer to one question. This must then be split into entities which in turn can be ordered into categories. In the figure 1 below I show this procedure is done in practice. The example is taken from an interview with an MP from the Norwegian Progress Party (Fremskrittspartiet). The quote is a word-forword transcription of what the respondent said when I asked about his own vision of the future of EU. The quote has some unfinished sentences which might show that the respondent was reflecting on the question (or talking while thinking). It should also be noted that this is also a translation from Norwegian to English. The text beginning with I: and in italics represent the interviewer, whereas the text beginning with R: represent the respondent.

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

Figure 1: An example of data reduction I: In which direction should the EU develop itself in the future in your opinion, specifically concerning the relationship between the central EU-institutions, member states, and regions R: I am a kind of EU-minimalist. For the EU is about making life easier, a little more flexible and to realise the four freedoms that they have been dealing with for a long time now. For me the EU should create a dynamic that make an economic and [pause] yes, make sure so happens. The EU that have moved in other directions as well, and in my opinion this has made things worse, made it more difficult and at least at certain areas made it more troublesome. If I were to decide I would have lead the EU back to basic. More of what used to be, in particular the thoughts that were dominating at the end of the 80s and beginning of the 90s. The chase in deepening the cooperation on other areas, I believe has created more pain than gain for EU as an organisation. I: So if the trend is more political power to Brussels you are against this? R: Absolutely, Absolutely. I mean that all experience shows that the nation state, also in Europe has come to stay.

I am a kind of EUminimalist

EU is about making life easier

Realise the four freedoms

The EU that have moved in other directions as well, and in my opinion this has made things worse, made it more difficult and at least at certain areas made it more troublesome.

If I were to decide I would have lead the EU back to basic. More of what used to be, in particular the thoughts that were dominating at the end of the 80s and beginning of the 90s

The chase in deepening the cooperation on other areas, I believe has created more pain than gain for EU

I mean that all experience shows that the nation state, also in Europe has come to stay.

Right trajectory of EU

Wrong trajectory of EU

Nation-state

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

This figure is an example of how data reduction can be done in practice. I have not reduced all the data from the raw data because of space limitations. For example one quote that I would have added is the fact that this respondent thinks of the EU as an organisation. This I believe fit into the pattern of how the respondent perceives the EU: namely as an economic organisation that benefits the nation-states of Europe. I have reduced the quote to three categories: right and wrong trajectory of the EU and a more ontological category, Nation-state. Whereas the two first categories shows the answer the respondent is giving to the question (right and wrong directions of the EU), the last category tells us something about what kind of political organisation/ polity he views as the most proper. The question is whether it is possible to extract more information from the data than what is displayed here? As I noted above this quote is from a MP representing the Progressive party. This is a party known for having a market-liberal/ libertarian economic profile. Combining this background information with the answer, it becomes clear that the respondent does not want the EU to deal with social policies. After all if you are against heavy taxation and official redistribution of wealth in the first place (such as the national government), why would you want an additional body to do the same thing (such as the EU)? As the figure also shows there is a dotted line from the I am an EU minimalist entity to the Nation-state category. I have included it because of uncertainty of where to place this entity. On the one hand this might be a statement where the respondent is arguing for the right trajectory of the EU, but on the other hand this could also be a part of his argument for favouring the nation-state. This shows one of the dilemmas the researcher is faced with when analysing the data. The same entity can be placed in several categories but if the researcher does this too often, it will undermine the value of categorising. It does not make analytical sense if many entities exist in most categories (Ryen 2002:153). This is one of the reasons why we have to repeat this process several times, both within and across cases. We will therefore be involved in a time-consuming yet creative process where new categories will be introduced and old ones rejected. Iterative analyses While Miles and Huberman (1994) have some similarities with Erlandson et al (1993) there are some differences in their focus. They distinguish three processes in the analysis4: 1. Data reduction this starts at the very initial research phase when concepts and methods are developed and subjects/ phenomenon are selected. 2. Data display seeking meaning on a limited part of the data (summaries, diagrams and text-matrices) 3. Conclusion comparing, contrasting, searching for patterns, triangulation etc.

These points are based on Ryen 2002:155.

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

Miles and Huberman also spend time on discussing the inductive and deductive approach in interview research design. According to them the iterative analyses combine both:
After one has inductively identified a theme, one goes on to try verifying or confirming the finding (deductive), which again gives an inductive loop. Huberman and Miles sees it as legitimate and useful to both start with conceptual analytical categories, that is deductive, or to gradually develop them, that is inductive (Ryen 2002:157)

This shows that there is a lot of potential in the data that we have collected. For one I have at the initial phase an inductive approach to my own thesis. By asking What is Euroscepticism? I seek to explore the content of Euroscepticism among the respondents that are interviewed. At the same time I travel into the field with a conceptual baggage. The categories (concerning trajectory of the EU) that I suggested in the example above (figure 1), bear some resemblances to how Kopecky and Mudde (2002) conceptualise support and opposition to European integration. According to Kopecky and Mudde it is useful to separate between scepticism towards EU on the one hand and scepticism towards the general ideas of European integration on the other. Building on an Eastonian (Easton, 1965:124ff.) system the purpose of introducing these two kinds of scepticism is to distinguish between diffuse and specific support. This in turn leads to a table consisting of four categories:
Table 2 Kopeck and Muddes conceptualisation of Euroscepticism Euroenthusiasts (pro-integration, pro-trajectory) Eurosceptics (pro-integration, anti-trajectory Europragmatists (anti-integration, pro-trajectory Eurorejects (Anti-integration, anti-trajectory)

Source: Kopeck and Mudde (2002:303). If this conception of Euroscepticism is our starting point, we must ask ourselves whether this the data reduction that is done in figure 1 fits this table or not. In other words can we place the respondent within one of the four cells in the table? If we assume that the trajectory of the EU is more than economic cooperation, then the respondent is quite clearly anti-trajectory. At the same time the respondent does not seem to reject the idea of integration as a whole (that is economic integration). It seems therefore plausible to categorise the respondent as a Eurosceptic. What we must then consider, and this would thus be a part an inductive loop, is if the concept Eurosceptics lacks precision. We should for instance expect answers that are similar to the one I have presented in figure 1, but almost turned upside down. That is, a respondent that rejects the economic cooperation, but welcomes the social Europe. This person would not be against integration, but we could not categorise the respondent as pro-trajectory either (that is Euroenthusiasts in the table). In

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

that sense we might end up categorising two respondents with quite different view of what is the best way of organising Europe into the same category. These reflections I believe show one of the advantages of doing qualitative interviews. Analysing the data in the way shown above, is useful particularly when doing single-case analysis or an analysis of a certain phenomenon such as political party elites perception of European integration. Also it helps to familiarise oneself with the data. However in order to do cross-country comparison this might not be sufficient enough. One reason for that is that the data reduction might vary from country to country and that we will not end up with the same categories. While this finding would be interesting in itself it would complicate the comparability of the data. This means that for larger research projects, and when comparison is an important goal one must try to find other alternatives of reducing and analysing the interview data. On the other hand I do believe that the analysis-technique suggested might be useful in a larger comparative project. For instance this could be an interesting explorative process that will strengthen the analysis later in the project. In particular this can be helpful when for instance comparing the data reports to ideal types (more on this below). Also as I have shown with the iterative analysis process, the interaction between the data and concepts could give way for improving concepts and theories. Phenomenology As stated above this paper is not aiming of given an exhaustive description of the various paradigms that exist in qualitative research. I have deliberately taking the naturalist/ positivist paradigm as a starting point for illustrating how we can go about when analysing interview data. This was justified because we here have some clear guidelines for analysing data. However it should be noted that there are other paradigms that also offer more structured ways of analysing data, and phenomenological research is one of them. Kvale and Brinkman (2009) gives their presentation of this perspective:
This kind of interview seeks to obtain descriptions of the interviewees lived world with respect to interpretation of the meaning of the described phenomena. It comes close to an everyday conversation, but as a professional interview it has a purpose and involves a specific approach and technique; it is semi-structured it is neither an open everyday conversation nor a closed questionnaire (Kvale and Brinkmann, 2009:27).

For the examples I have used above (e.g. Figure 1), it makes sense to claim that we are approaching the lived world of politicians. So when asking a politician about what direction the EU should move I am engaging in a conversation that for most politicians is an important theme in their life and work. Using examples from Giorgi (1975), Kvale and Brinkman (2009) suggest five steps for analysing interview in a phenomenological tradition: 1. 2. 3. 4. Read through the whole interview to get a sense of the whole The researcher determines the natural meaning units The natural meaning unit is restated as simply as possible Interrogating the meaning units in terms of the specific purpose of the study 9

ISSN 1890-5986

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

5. Essential non-redundant themes of the entire interview are tied together into descriptive statement I will not go into detail by using examples to show how analysing interviews in this way can be done. But as we can see from this short review of phenomenology is that there are some similarities to both the naturalistic and ethnomethodological tradition. For one we see that data reduction is an important key in analysing the data. Also we see that the focus meaning is central is many of the paradigms presented above.

Comparing interview data


In my own research project comparative research is one of the core goals. Interestingly enough Ryen does not acknowledge this as the main intention of qualitative interview:
The main intention of qualitative interview is not to compare cases/ units but to get access to actions and events that are viewed as relevant for the research/ study. Access to the single respondent and the way he or she views the world is central (Ryen 2002:85).

Miles and Huberman (1994) on the other hand claim that cross-case analysis could be of value. This is because it enhances generalisability as well as deepening the understanding and explanation of a phenomenon. That being said they also recognise that the goal of generalisability is disputed among researchers (doing qualitative analysis). One factor that could enhance the comparability of the data is when the collected interviews are, at least initially, rather standardised. This involves that the interviewer has a fixed questionnaire that is adhered to in a rather strict fashion, and that there are clear guidelines for selection organisations and respondents within these. It could also involve pre-testing of the interview-guide/ questionnaire to enhance the validity of data. In this respect Miles and Huberman calls for a clarification of whether to use a variable- or case-approach. Ragins (1987) summary of these approaches or strategies is worth quoting at length:
The two strategies are surprisingly complementary. The Case-oriented strategy is best suited for identifying invariant patterns common to relatively small sets of cases; the variable-oriented strategy is best suited for assessing probabilistic relationships between features of social structures, conceived as variables, over the widest possible population of observations. The main weakness of the case-oriented strategy is its tendency toward particularizing (often while pretending to great generality - for example, a theory of ethnic political mobilization based on one case); the main weakness of the variableoriented strategy is its tendency towards abstract and sometimes vacuous, generalizations (Ragin, 1987:69).

There are many different strategies for case-oriented strategies, but those suggested in Miles and Huberman seems to leaning towards testing the cases against a theoretical framework or groups/ families. Many researchers approach cross-case comparison by forming types or families. You inspect ISSN 1890-5986 10

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

cases in a set to see whether they fall into clusters or groups that share certain patterns or configuration (Miles and Huberman, 1994:174). This bears resemblances to the method involving ideal types which I shall return to below. In a variable-oriented strategy the inner-dynamic of the case are replaced with a search for patterns and themes that cut across the cases. This means that the data reduction is taken a step further and (perhaps) more pressure is put on the researcher in terms of interpreting the answers so that they can be reduced into variables. Before moving on to the ideal type method we need to ask ourselves: What is a case? Gerring (2004) defines a case study as: an intensive study of a single unit for the purpose of understanding a larger class of (similar) units (ibid:342). He then makes the following clarification:
A population is comprised of a sample (studied cases) as well as unstudied cases. A sample is comprised of several units, and each unit is observed at discrete points in time, comprising cases. A case is comprised of several relevant dimensions (variables), each of which is built upon an observation or observations.

Thus using my project as an example it is possible to set up the following table:


Table 3: Population, units and cases Population European party elites Units Cases Sample of different European organisations (e.g. political parties) Respondents from the political parties

On the basis of this table we could however identify a number of cases. In the first place we have the individual interviews which can represent a single case. Secondly we have the organisations that each of these individuals represent. Thirdly the respondent can represent larger communities than organisations be it party-families (such as social democrats), regions (Central Eastern Europe) or nation-states. It is therefore important to have this in mind when finalising the data report. Should for instance the data report cover a summary of findings/ answers from each respondent or should it be a summary of the entire organisation as a whole? If the latter is preferred: should one have a variable oriented approach where the focus on the main pattern and themes that exist within this organisation? Another question that arises is what to do with answers and respondents that are not fitting the pattern? In my own data collection I experienced several times when the respondents claimed to be at odds with what the party they represented meant about an issue. This is of course anticipated, after all representing a party does not entail that one agrees with every issue the party claims position. The challenge is how we can include such findings in the data report, when/ if we are expected to summarise the patterns for the party/ organisation as a whole? ISSN 1890-5986 11

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

Perhaps a more fruitful way of clarifying these considerations is to distinguish between units of observation and units of analysis. In the first case the units of observation is the single respondents themselves (i.e. the interviews). However in order to enhance the data collection we should also include data about the organisations/ countries that the respondents represent. Background and institutional data collection would also contribute to the units of observation. Considering units of analysis we proceed in the same way as with the units under observation (individuals, organisations, regions and countries). What is important to notice in this respect is that the researcher must be aware of on what level the observation (data collection) and analysis is situated. That being said, it is not always easy in an interview to distinguish between an organisation and individual. Often when interviewing spokespersons, employee or MPs representing a party on will often find answers like: We in the Conservatives believe that, and My personally opinion is that. So if we do not explicitly ask for personal or normative opinions of the respondent then we must be aware of such distinctions. In perhaps more interesting cases one can also find statements like: Well on this issue I disagree with my own party, in my opinion.... Depending on the research such answers could be compared to other data sources such as party programmes, election manifestoes and so on. This would be a part of the data triangulation, an issue which I have not elaborated on so far. When triangulating: the researcher seeks out several different types of sources that can provide insights about the same events relationships (Erlandson et al., 1993:115). For a comparative project additional data sources are important in increasing the validity of the findings. As mentioned above in figure 1 it is important to know the background/ ideology of the party that the respondent is representing. So if a research project about the EU includes Norway as one of the countries, the saliency of the EU-issue should be included as important data. An example would be whether the party representatives are unleashed by the party on this issue or if the representatives are forced by the party whips to take a position that they themselves do not adhere to. Ideal Types and deductive approaches As the example I have presented above has a certain inductive character we should also consider more deductive oriented strategies. An option that might be more appropriate is the use of ideal types:
In this method, a phenomenon under study is first defined based on an existing theory. The features of this phenomenon, which this theory presupposes to exist, are delineated through deductive reasoning. Then, empirical cases proximity (similarity) to or distance (difference) from this theoretical ideal type is measured or described. (Sicakkan, 2006: 90)

This way of analysing social phenomena is well-known in social science with Max Weber standing out as the main contributor (Ritzer, 2000). Perhaps the most recognised example is his ideal-type of the bureaucracy where he among other things shows the rationalisation of the world (which was an effect of bureaucratisation). This study is an example of how an ideal-typical model of a certain phenomena can be used to study the effects of large societal ISSN 1890-5986 12

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

structures. Sicakkan discusses the problems of using singular ideal-types with showing that this might lead to to a lack of significant information about the phenomenon under study (Sicakkan, 2006: 91). He suggests two solutions to overcome this challenge: comprehensive inductive conceptualization comprehensive deductive conceptualization

In the first case all empirical cases are treated equal in order to conceptualise a given phenomenon. In the second we deploy multiple theoretical ideal-types in which empirical cases are measured against these. To specify this method of multiple ideal-types we can use an example from a quantitative study, where Sicakkan (2005) classified responses to immigration, in a Norwegian survey: 1. Identification of several relevant contesting normative theories 2. construction separate attitude models based on each selected theory 3. comparing cases with these attitude models and classifying them into groups with respect to their differences from or similarities to these models 4. constructing a single scale or multiple scales accounting for both the differences between the groups found in 3 and for the differences between the cases within each group Here we can argue that point 1 and 2 in this procedure is activated prior to the data collection. Thus we identify and construct the models we intend to use in the analysis, and then develop the appropriate research tools that will be used in point 3 and 4. What I believe can be one of the benefits in comparing the data towards theoretical models is that answers that might seem inconsistent might contain a sound logic in these models. As Sicakkan notes A subject can bear both positive and negative attitudes to immigrants in different dimensions(Sicakkan, 2005:56). As I have experienced in my own data collection, it is not unusual that respondents can speak furiously against European integration because they do not accept the loss of sovereignty for their own country, and then later in the same interview speak warmly about EUs eastward enlargement in 2004. While the method involving ideal types has a different starting point, one of the aims in the analysis remains the same compared to analysis by Erlandson et al. (1993) and Miles and Huberman (1994) that is to categorise and make sense of the collected data. To exemplify this way of analysing interview we could therefore first start with an expected answer scheme/ table:
Table 4: Ideal types and expected findings Ideal type A Question/ theme 1 A.1 Question / theme 2 A.2 Question/ theme 3 A.3 Ideal type B B.1 B.2 B.3 Ideal type C C.1 C.2 C.3

ISSN 1890-5986

13

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

The cells (A.1, A.2 etc) would consequently include expected findings that are expected according to the ideal type. It goes without saying that the expected findings are not to be found empirical, since the ideal types are not empirical. This means that while some responses from the interviews might fit into the cells the interview/ respondent/ organisation as a whole will not. We will not find organisations that are truly representing Ideal type A or B, but we will most likely find organisations that have similarities and differences to the various ideal types. For a more concrete example of this table we can view a shortened version of Sicakkans five ideal-typical models of response to immigration5:
Table 5: Five Ideal-Typical Models of Response to Immigration (shortened) Libertarian Immigrants right to vote in local elections after 3years residence Segregation as conflict resolution Liberalist Republican Communitarian Tribal

Agree

Partly agree

Partly disagree

Disagree

Disagree

Disagree

Partly disagree

Disagree

Agree

Disagree

Developing ideal-types for the phenomenon of Euroscepticism per se has provided some difficulties. One reason for this is that we are dealing with attitudes that can be related to stances on rather specific issues (e.g. rejecting European integration as a whole because of certain fisheries policies). In other words we are lacking some sense of generalness and as Ritzer notes: Ideal types should be neither too general nor too specific (Ritzer, 2000:116). In this case Euroscepticism is too specific. The strategy will therefore be to lower the level of abstraction and create some expected types of Euroscepticism, that are based on a) theoretical concepts and b) initial findings from the interviews (finding key characteristics inductively).

Summary
This paper helps to avoid what Kvale calls the 1,000-page question: How shall I find a method to analyze the 1,000 pages of interview transcripts I have collected(Kvale, 1996:176). The question should be avoided at least after the data collection is finished. If you have collected 1,000 pages without considering how to analyse them, then you have probably made quite a few epistemological and methodological errors along the way. As this paper has shown giving both general and specific recommendations for choosing analysis techniques is not easy nor is it always desirable. This is largely because we have already chosen some paths at the outset of our research. In that way no research is truly inductive in the sense
5

This is a shortened version of Table I in Sicakkan 2005:54. A review of the ideal-typical models is given in the same article.

ISSN 1890-5986

14

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

that we have absolutely no idea of what to expect before starting our scientific endeavour. At the very beginning when we select or research phenomenon we have made some assumption and expectation of what we could possibly expect to find. Still I believe there is much to the fact that there is an interaction between inductive and deductive approaches when analysing qualitative interview data. And, depending on the phenomenon under study, it would therefore be very interesting to seek to combine the two analysis techniques as suggested above (naturalistic/positivistic and ideal types). By making our preferences clear throughout the research we will not only save time (and avoid the 1,000-page question) but also making the analysis and results more understandable to both the reader and the researcher him-/ herself. Since the comparative approach that I have selected for my own project is important, I also believe that a standardisation of the interview data both in the analysis and collection is equally important. This is creates several challenges that can not only be met with standardisation, but also with triangulation and close collaboration if the research is conducted by several researchers. Yet at the same time the standardisation should not be based on a deductive approach alone. As this paper has shown, there is an interaction between the different research phases and the researcher. This in turn creates many opportunities for development a creative and innovative research design that was not anticipated prior to the data-tool development and data-collection. Based on the considerations in this paper I believe that a combination of the standardised analysis/ collection and measurement against Ideal types/ theoretical models on the one hand and more inductive analysis on the other are fruitful for exploring Euroscepticism among political parties on the outskirts of Europe.

ISSN 1890-5986

15

EUROSPHERE WORKING PAPER No.13

FOLKESTAD

References
Bygnes, S. (2008) Interviewing People-Oriented Elites. The Eurosphere Online Working Paper Series, Working Paper No. 10 / 2008. Denzin, N. K. (1970) The Research Act in Sociology: a Theoretical Introduction to Sociological Methods, London, Butterworths. Easton, D. (1965) A Framework for Political Analysis, Englewood Cliffs, N.J., Prentice-Hall. Erlandson, D. A., Harris, E. L., Skipper, B. L. & Allen, S. D. (1993) Doing Naturalistic Inquiry: a Guide to Methods, Newbury Park, Calif., Sage. Gerring, J. (2004) What is a Case Study and What is it good for? American Political Science Review, 98, 341-354. Giorgi, A. (1975) An Application of Phenomenlogical Method in Psychology. in Giorgi, A., Fischer, C. & Murray, E. (Eds.) Duquesne Studies in Phenomenological Psychology. Pittsburg, PA, Duquesne University Press. Kopecky, P. & Mudde, C. (2002) The Two Sides of Euroscepticism: Party Positions on European Integration in East Central Europe. European Union Politics 3, 297-326 Kvale, S. (1996) Interviews: An Introduction to Qualitative Research Interviewing, Thousand Oaks, Calif., Sage. Kvale, S. & Brinkmann, S. (2009) Interviews: Learning the Craft of Qualitative Research Interviewing, Los Angeles, Calif., Sage. Lincoln, Y. S. & GUBA, E. G. (1985) Naturalistic Inquiry, Beverly Hills, Calif., Sage. Miles, M. B. (1979) Qualitative Data as an Attractive Nuisance Problem of Analysis Administrative Science Quarterly, 24, 590-601. Miles, M. B. & Huberman, A. M. (1994) Qualitative Data Analysis: an Expanded Sourcebook, Thousand Oaks, Calif., Sage. Ragin, C. C. (1987) The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies, Berkeley, Calif., University of California Press. Ritzer, G. (2000) Sociological Theory, New York, McGraw-Hill. Ryen, A. (2002) Det Kvalitative Intervjuet: fra Vitenskapsteori til Feltarbeid, Bergen, Fagbokforl. Sicakkan, H. G. (2005) Senses That Make Noise & Noises That Make Sense. Three Techniques for Scaling Attitudes to Immigration. Norwegian Journal of Migration Research, 5, 42-73. Sicakkan, H. G. (2006) The European politics of citizenship and asylum: a comparative analysis of the relationship betweeen citizenship models and the legal-institutional frames of asylum determination in West Europe. Bergen, Department of Comparative Politics, University of Bergen. Silverman, D. (2001) Interpreting Qualitative Data: Methods for Analysing Talk, Text and Interaction, London, Sage.

ISSN 1890-5986

16

You might also like