Data analysis

For data analysis, the methods of “process tracing” (George & Bennett, 2005) and “qualitative content analysis” (Mayring, 2000, 2014) were employed, while the whole process of analysing documents and interviews was assisted by the software MAXQDA. The following sections explain the study’s data analysis methods.

3.4.1. Process tracing

Process tracing is a research method that is used by social scientists conducting case studies to analytically access the descriptive dimension of the case study and detect causal processes which do not necessarily appear in a linear way (George & Bennett, 2005; Maggi, 2016;

Vennesson, 2008). It provides a common middle ground for those interested in historical explanation and the complexities of historical events through researching individual and embedded cases (George & Bennett, 2005). In their work on case study research and theory development, George and Bennett (2005) argue that process tracing is an “invaluable method that should be included in every researcher’s repertoire” (p. 224), because it contributes to social sciences in ways that statistical methods cannot do in terms of both theory testing and heuristic development of new hypotheses. As such, process tracing has been a powerful method that has partly contributed to a “historical turn” in social sciences and renewed interest in path-dependent historical processes (ibid.).

Process tracing shares some basic features of historical explanation and, therefore, the difference between extensive historical description and process tracing can be blurry (Maggi, 2016). In general, process tracing differs from historical explanation because of the emphasis on an analytical explanation based on a theoretical framework that has been already identified when designing the research (George & Bennett, 2005). In other words, process tracing envisages to contribute to theory testing and/or theory development by identifying causal mechanisms within a single case (ibid.). Although there is still little consensus on how process tracing should be carried out, Maggi (2016, p. 60) argues that process tracing employs “the complete spectrum of qualitative data, such as histories, archival documents and especially interview transcripts – which are useful for a very detailed description of the studied case”.

Tracing the process that led to an outcome means narrowing down the list of potential causes, but even then it is rather challenging to eliminate all potential different explanations but one, particularly when human actors are involved (George & Bennett, 2005). However, even small

58 and unexpected empirical evidence revealed through process tracing can illuminate new aspects of a phenomenon that could be easily overlooked by statistical analysis.

Moreover, process tracing fits the interpretivist epistemology of the present study. In an interpretivist perspective, process tracing allows the researcher to examine how different factors are interlinked and the context in which this happens (Vennesson, 2008). This means that the focus of process tracing is not only on what happened, but also on how it happened.

As such, process tracing is an adequate method for empirical case study research allowing us to examine the reasons that actors give for their actions and to understand empirically their preferences and perceptions (ibid.). Process tracing can help to identify connections that appear as only plausible by treating actors’ preferences and perceptions as empirical questions that require careful empirical investigation. In contrast to positivism, an interpretivist perspective of process tracing explores how certain processes came about and how specific factors interact, although it faces difficulties in weighting the relative importance of the different factors (ibid.).

In the present study, process tracing has been employed to explore the relationship between policy changes in three countries, namely Austria, Greece and Hungary, and European developments with regard to teacher education. Among the varieties of process tracing, the so-called “detailed narrative” is considered the most appropriate for the present study. Detailed narrative is defined as the simplest variety of process tracing that takes the form of a detailed story presented in the form of a chronicle with the purpose to illuminate how an event came about (George & Bennett, 2005, p. 210). This type of process tracing makes no explicit use of theory but rather provides a detailed description of a sequence of events that may reveal the possible causal processes in a case (ibid.).

In our case, this sequence of events refers to the development of teacher education systems over time and includes three steps. First, a brief historical overview from when teacher education was institutionalised in each country and until the late 1990s illustrates some national priorities and challenges before the launch of the Lisbon Agenda. The narrative continues with developments after the year 2000, focusing on some major reforms with significant influence on the whole spectrum of teacher professional development with the purpose of identifying manifestations that could imply an explicit or implicit change of policy and practice connected to Europeanisation. The third step narrows down the scope of process tracing to the study’s three analytic categories (i.e. continuum, teacher competences, teacher educators), in order to filter the potential explanations of policy change. To frame this whole detailed narrative according to reliable manifestations, process tracing is combined with qualitative content analysis. George and Bennett (2005) argue that process tracing often complements other research methods and in the present study qualitative content analysis helps to cluster collected data as will be discussed in the following section.

3.4.2. Qualitative content analysis

Empirical material, including documents and interview transcripts, were analysed according to the research method of qualitative content analysis proposed by Mayring (2000, 2014).

Qualitative content analysis is “an approach of empirical, methodological controlled analysis of texts within their context of communication, following content analytical rules and step by step models, without rash quantification” (Mayring, 2000, §5). The specific method scans the collected empirical material with categories guided by theory and research questions (ibid.).

Some basic ideas of content analysis include the following: (a) the material should always be interpreted within its context; (b) rules of analysis are laid out in advance; (c) categories are in the centre of the analysis; and (d) criteria of reliability and validity are established (ibid, §7).

59 These essential components of qualitative content analysis can be processed in two ways, namely through inductive category development and deductive category applications (ibid.).

The main goal of content analysis in the present study is to assist process tracing by identifying the connections between national and European policy developments regarding teacher education. To do so, categories are developed in two phases according to both inductive and deductive approaches, considering that categories need to be carefully established and revised within the process of analysis (feedback loops) (ibid., §7). The first phase of inductive category formation has to do with empirical material regarding the European context, while the second phase of deductive category application relates to the national case studies. In both approaches, categories consist of coding units and context units. Coding units are the smallest component of material that can be assessed and can be a minimum portion of text falling within one category, while context units determine the largest text component falling within one category (Mayring, 2014, p. 51).

The analysis in Chapter 4 dealing with the European context of teacher education involved breaking down the empirical material, namely European policy documents and interview transcripts with European policy experts, into units through theory-guided and inductively developed categories. Mayring (2014) suggests that in the process of inductive category formation it is useful to keep content-analytical units very open-ended. Therefore, guided by the first research question, to explore how teacher education is consolidated in the EU policy process, categories were formulated regarding the mechanisms and content of Europeanisation at the European teacher education level. After thirty percent of the material analysed, the categories were revised and reduced to some main categories according to which data in Chapter 4 are presented.

Section 4.3.4 of Chapter 4 provides a summary of European policy thinking in teacher education, providing descriptors for each analytic category, namely the continuum of teacher education, teacher competences and the role of teacher educators. The specific analytic categories with their sub-categories and descriptors, as presented in Table 5 of Chapter 4, were deductively applied to empirical data collected for each country case study. The descriptors represent what Mayring (2000, 2014) terms “definition” for each category and help to cluster the empirical data. However, content analysis “is not a standardised instrument that always remains the same” (Mayring, 2014, p. 39), and therefore reading of the empirical material for each case study could also lead to the emergence of new categories inductively. Since each case study represents a unique teacher education system, it is natural that content analysis can lead to unique categories for each case study.

However, it was generally attempted that all three case studies in Chapters 5, 6 and 7 are analysed and presented in a symmetrical way by applying deductively the same overarching analytic categories in order to allow comparisons later in the discussion chapter. Figure 7 below illustrates the rules of the qualitative content analysis which help to analyse the material step by step by devising the material into content analytical units (Mayring, 2000), according to the present study’s research questions and theoretical framework.

60 Figure 7. Procedural model of qualitative content analysis

Due to the amount of documents and interview transcripts, the qualitative data analysis software MAXQDA was utilised to help with the coding process. The specific computer software allows the user to organise and analyse a diverse range of data in a flexible and quick way and is often used for qualitative content analyses. For the European context analysis and for each case study a different project was established in MAXQDA. Inductive analysis of the European context material provided categories and codes that were then applied to the case study material, and separately for each case study, so that new categories could also emerge inductively. It should be noted, however, that since the analysis was driven by the research questions and the theoretical framework, coding focused on some aspects of the data, rather than the data overall. Especially in the analysis of the case study materials, the predefined categories helped me to code and extract the relevant text segments, although I was generally open to the emergence of new categories.

Overall, I approached the process of formulating categories and clustering the interview data in a context-sensitive way that paid attention to the interpretations and meanings of participants rather than the direct use of a particular set of words. The perspectives of research participants are discussed according to their professional group (e.g. policy expert, teacher educator) and direct quotations are used throughout the empirical analysis chapters as a way of illustrating overarching categories and patterns of thoughts. Since participants could express

61 different viewpoints depending on their institutional affiliation, I envisaged to include an even selection of quotations from different interviewees regarding the development and enactment of specific policies. However, it could be that some participants are quoted more than others when it comes to certain issues, due to their particular expertise and responsibility over specific aspects of policymaking. The following section refers to the present study’s trustworthiness criteria, establishing this way the final criterion of Mayring (2000) regarding quality standards for assessing research.