Section 6.4: Data analysis
6.4.1 Approaches to qualitative data analysis
6.4.1.1 Content analysis
Content analysis, as Bauer (2000) stated, is “systematic classification and counting of text units [to] distill a large amount of material into a short description of some of its features” (cited in Marvasti, 2004, p. 90). It helps the researcher to translate the content of many pages into organized segments. Content analysis is a process of analyzing text which refers to “the recorded information about social life in the form of visual images, published written materials or transcribed interviews” (Marvasti, 2004, p. 90). Under content analysis, text could be seen as the reflection of public opinion, so it demonstrates how public opinion has been shaped, or could be seen as a cause of public opinion in which content analysis help to searching for the “themes that could connected with certain cultural practices or attitude” (Marvasti, 2004, p. 93).
Content analysis was seen as a quantitative approach in its early stages, because it originally counted the frequency of word appearances, or how much time was devoted to themes (Payne and Payne, 2004). However, content analysis began to emphasize attitudes, values and motivations behind the words, and gradually was used as a qualitative approach, in which qualitative researchers could bring their own cultural meaning to interpretation of the texts (ibid.). Qualitative content
analysis, which focuses on the meaning of text (unit) itself, offers a more “objective” approach for social research since large amounts of descriptive data could be translated into quantified categories or standard codes. However, researchers have pointed out that content analysis always breaks the respondent’s statement into “quantifiable chunks”, which will miss some important information when regarding qualitative data in its full social context (Riessman, 2002; Marvasti, 2004).
6.4.1.2 Thematic analysis
Thematic analysis is an approach which shares some principles and procedures such as “themes” and “codes” with content analysis. As suggested by Given (2008), thematic analysis is “a data reduction and analysis strategy by which qualitative data are segmented, categorized, summarized, and reconstructed in a way that captures the important concepts within the data set” (p. 867). It “moves beyond counting explicit words or phrases and focus on identifying and describing both implicit and explicit ideas within the data, that is, themes” (Guest, et al., 2012, p. 10). In thematic analysis, a theme or coding category is always drawn deductively or inductively. It could be derived from existing theoretical idea which the researcher brings to the data, or from the raw data (Joffe and Yardley, 2004, pp.57). Practically the methodology framework of thematic analysis comprises several approaches such as grounded theory, positivism, interpretivism, and phenomenology (Guest, et al., 2012). It selects from a range of techniques from different methodological camps to “identify and examine themes from textual data in a way that is transparent and credible” (ibid., p. 15). However, thematic analysis can be challenged if a qualitative researcher draws the richness of the themes from the raw information without reducing the insights to a trivial level for the sake of consistency of judgement (Boyatzis, 1998). In other words, it could be problematic if researchers simply pick text chunks to support the argument one wants to make (Silverman, 1993). Meanwhile, as “themes” or meanings behind symbolic texts are emphasized by thematic analysis, the number of times that a theme or category
appears does not always necessarily indicate the extent of importance or relevance; on the contrary, one could be of high relevance and conceptual importance even if it just appears once, or is mentioned by only one interviewee (Joffe and Yardley, 2004). Therefore, it is mainly an interpretive approach.
6.4.1.3 Narrative analysis
Since stories or storytelling is a common form of sharing information, thus, understanding what stories convey and how they convey them becomes an important part of qualitative analysis (Marvasti, 2004). Narrative analysis, as Riessman pointed out, is analyzing data as the order in which story is told (Riessman, 1993). Or as Cortazzi noted, analyzing following a particular pattern of telling, such as the substance of the story, the structure, the purpose that story serves and the context of story (Cortazzi, 2001; Marvasti, 2004). Because narratives are made meaningful under a certain set of practices and context, thus narrative analysis could help to simultaneously examine what respondents say and how they make it meaningful (Marvasti, 2004). Furthermore, as the purpose (of storytelling) and context are emphasized in narrative analysis, this approach goes beyond the asking of whether the story is true or not.
Narrative analysis is good at analyzing text in rich data sources such as in-depth interviews because it always tries to fully understand how various pieces of data relate to one another through the way they are articulated and the context (Riessman, 2002; Marvasti, 2004). It intends to put discrete elements together into big picture in order to display individuals, cultures and societies as wholes (Riessman, 1993).
6.4.1.4 Grounded theory
Grounded theory was developed by Glaser and Strauss in 1964 (Glaser and Strauss, 1967). This approach intends to inductively build up a systematic theory based on observation (Strauss and Corbin, 1994). The grounded theorists begin with
observation and an ongoing conceptual categorizing process upon observation, and then test these categories over time with more observation. A theory will evolve after carefully refining and linking conceptual categories (Glaser and Strauss, 1967). In other words, using grounded theory is a process of suggesting plausible relations between concepts by generating concepts from qualitative data, and then carefully reviewing the data and developing (refining) the concepts. Grounded theory has experienced several changes on its original conceptualization over time (Strauss and Corbin, 1990; Strauss and Corbin, 1994), although many ideas appeared, they do share some components. As Charmaz (2002, cited in Marvasti, 2004, p. 85) listed: Simultaneous data collection and analysis;
Pursuit of emergent themes through early data analysis Discovery of basic social processes within the data
Inductive construction of abstract categories that explain and synthesize these processes
Sampling to refine the categories through comparative processes
Integration of categories into a theoretical framework that specifies causes, conditions, and consequences of the studied processes.
The understanding of grounded theory, as Strauss and Corbin stated, has been influenced by “contemporary intellectual development such as feminism, political economy, and varieties of postmodernism” (Strauss and Corbin, 1994, p. 276). Furthermore, the grounded theory approach highlights the importance of the role of researcher. Existing theories can offer a guideline or framework to analyze data; however, it could become a serious restriction for data analysis because researchers often intend to analyzing data in order to “fit” some theory or framework. In the process of grounded theory, a theory evolves by continuous interaction, such as refining and reviewing, between researcher and data. Meanwhile, the researcher’s knowledge and experience play a crucial role in understanding the data. Thus, the researcher plays a more influential role in grounded theory than in other approaches of qualitative data analysis.
6.4.1.5 Discourse analysis
Discourse analysis is hard to define as a unified pattern of research since it has various meanings and application across many disciplines such as sociology, psychology and linguistics (Marvasti, 2004). Nonetheless, there are several central themes shared by researchers across different disciplines:
Viewing discourse and language as being productive of social reality (discourse doesn't simply describe reality but it creates it as well)
Treating discourse as a type of social action in its own right (discourse is not just a description but it does things)
Emphasizing the rhetorical functions of discourse (discourse is to promote one side of a conflict) (Gill, 2000, cited in Marvasti, 2004, p. 107).
There are different kinds of discourse analysis according to the topic of interest. For example, one is interactional discourse analysis. According to Austin (1962), interactional discourse analysis is based on the assumption that language is akin to social action and examines how the spoken word make us accountable to others in order to illustrate “how discourses accomplish reality in everyday talk” (Silverman, 2001, p. 178; Austin, 1962; Marvasti, 2004). Other topics such as the relationships between text and context, or between discourse and power also leads to different kinds of discourse analysis.
6.4.1.6 Conversation analysis
Conversation analysis, according to Silverman (2001), is an approach to investigate everyday conversation in order to show how social reality is produced by verbal exchange. The very focus of conversation analysis is the term “talk”, which could be categorized as ‘naturally occurring’ or ‘researcher-provoked’ (Silverman, 2001, p. 159). Naturally occurring talk refers to what people say in everyday common situations, such as at home or at the bus stop, while researcher-provoked talk refers
to what people say in response to specific questions, such as responses to interview questions (Marvasti, 2004). In conversation analysis, the conversation is always transcribed in as much detail as possible in order to show the “artful use of talk and capture every aspect of a talk such as pauses and intonations” (Baker, 1997, p. 131). Unlike discourse analysis, conversation analysis neither focuses on written text nor socio-cultural phenomena; it places emphasizes on social interaction such as everyday conversation and interpretations. Additionally, these interactions are often analyzed moment-by-moment.
6.4.2 Implications for this research
As mentioned in Section 3.1, one of the main goals of this research is trying to find out whether Pierson’s three institutional factors play as important a role in China as in Western federalist countries. Therefore, Pierson’s theory of three institutional factors offers a clear framework to analyze relevant data. Based on such an objective, thematic analysis could be the appropriate approach to achieve the expected goal.
Firstly, the analyzing process of Pierson’s three factors has clear existing concept and theory as boundaries. To some extent, it is more likely to be a theory-testing analysis rather than a theory-building analysis. Therefore, grounded theory is not appropriate. Secondly, among all other approaches of qualitative data analysis, thematic analysis has an advantage to achieve the research goal. Analyzing Pierson’s three factors aims at finding out how these factors work in practice. Answers could be achieved through exploring respondents’ ideas or experiences about relevant institutional mechanisms or social phenomena. That means the meaning of the answers given by respondents is very important to this research. However, conversation analysis focuses on the detail of conversation and sequential organization. Rather than knowing what it is, conversation analysis is more likely to examine how it constructed. Narrative analysis focuses more on purpose and order as well as the “whole image” of the respondent’s response rather than content and
meaning. Discourse analysis focuses on the interaction or relationship between different genres of discourse or between discourse and other elements such as context or power. All three approaches listed above emphasize other perspectives of respondents’ answer rather than the meaning, and are less appropriate for understanding the policy process. Thematic analysis allows the researcher to cut interview data into “discrete chunks” according to the analytical framework, which makes it possible to understand clearly about respondents’ idea or experiences of Pierson’s three factors.
However, the research does not look to only “test” Pierson’s three factors. This research was started with an open mind to the possibility of other factors that may have significant influence on the central-local relationship or the healthcare policy process. Based on this goal, there is space to employ an inductive thematic analysis approach in order to explore those “potential” factors. Codes and themes will be selected and the definition of the problem will be refined after careful observation of the interview data. Both frequency and distribution of indicators will be checked as well as exploring the specific relationship between the indicator and other social phenomena. After careful review and attempts to discover negative evidence, inductive thematic analysis could help to suggested plausible relations between the indicator and the central-local relationship or healthcare policy.
It is often claimed that inductive approaches are always related to qualitative research, whilst deductive approaches are often related with quantitative research. Deductive approaches often start with a hypothesis which is derived from existing theory, then test the hypothesis by collecting data and exploring the empirical world (O’Reilly,2009). Therefore, a deductive approach suits testing existing theories. On the contrary, an inductive approach starts with as few preconceptions as possible and keeps an open mind to allow theory to emerge from the data (O’Reilly, 2009). This makes an inductive approach suitable for challenging existing ideas and developing new theories. However, the deductive approach is often criticized by qualitative researchers that a theory-testing social researcher’s focus is often restricted by the framework imposed from the outset, and it is hard for them to think
outside the framework because when data are “collected with theories in mind it has already been formed into working hypotheses” (O'Reilly, 2009, p. 105).
After a series of interviews done in China, interviews were transcribed so that they could be analyzed. As mentioned in Section 2 of this chapter, this thesis is conducted based on the framework of Pierson’s (1995) three institutional factors. Interview transcripts were mainly coded in a “deductive way”, in which the main purpose of the data analysis is to explore how the three factors work in practice. Interview transcripts will be coded and analyzed in a structure of Pierson’s three factors. To some extent, this thesis is trying to find out whether these three factors play a role as important as that seen in Western federalist countries. Meanwhile, how these factors affect the central-local relationship in China will also be explored. Based on this, the data were mainly analyzed in a deductive way. However, it will not be a purely deductive analysis. Interview transcripts were also analyzed line-by-line in order to discover other factors (apart from Pierson’s three factors) that significantly influence the central-local relationship. There would be no answer to whether these factors exist or how these factors affect the central-local relationship until the data analysis has been finished. From this perspective, inductive analysis was also conducted in this research. Therefore, this thesis will employ a combination of inductive and deductive analysis, in which deductive analysis is the mainstream.