Chapter 2: Methodology
2.1 METHODOLOGICAL CONSIDERATIONS
2.1.3 Data Analysis and Coding Scheme
Given the congruence method, co-occurance of norms is the variable for analysis.
Behavioral changes of individual respondents are not the target of this research. Rather, the representation of norms in the selected policy documents and the responses from elite respondents are used as indicators to make statements about the implementation of norms.
The focus is not on the change of behavior on the Vietnamese side, or to measure the EU’s normative impact on the Vietnamese government, but on the translation of norms from commitments on paper to individual practice on the EU’s side.
6 A list of policy documents can be found in appendix 1. Transcripts are not published in this study due to confidentiality commitments. More information can be provided upon request to the author provided the respondents written consent.
63 Table 2: Levels of Policy Analysis
Level Primary Documents for Analysis* Empirical Data
Policy guideline TEU
*this is an exemplary list. The full list of policy documents can be found in appendix 1.
NPCD can be identified by the presence of norms on all levels of implementation. The degree of coherence can vary from level to level. To structure the measurement of NPCD, four levels of analysis have been defined reflecting the stages to be traced in the process-tracing
‘light’ approach (see table 2). These four levels are defined by the level of precision they represent in the relationship between the EU and third countries. While this classification per se can be applied to various relationships between the EU and its partner countries, the selection of country specific documents and interview partners reflects the choice of Vietnam as a partner country in this study.
The first level corresponds to the highest institutional level within the EU represented by the Treaty on the European Union and the Treaty on the Functioning of the European Union. As indicated above, this stage of the policy process reflects a rather rhetorical commitment regarding normative policy coherence for development. The TFEU lays out technical processes of policy-making such as voting rules and decision-making procedures, which could be understood as implementing mechanisms, but as there is no reference about how to implement coherence or how to promote norms, this will be taken solely as a guideline. These guidelines apply to all EU policies.
The second level represents the major agreements between the EU and Vietnam. This level, called policy formulation, is characterized by the position between the guideline and the implementation. In other words, the agreements are more precise that the guidelines on the one hand and tailored to the contracting parties on the other hand. Nevertheless, these agreements leave room for interpretation and, therefore, for the exact implementation.
64 Multiannual Indicative Programs (MIPs) represent the first policy implementation stage, i.e.
policy implementation on a program level. The MIPs apply to an entire country and cover a time span of five years on average.
The fourth level is the project level. This stage is the final implementation stage in this study, because projects are the smallest institutionally organized unit of implementation.
Projects often tend to be technical in nature, which reflects the highest level of precision regarding the objectives to be achieved. In contrast to the multiannual indicative programs, projects have a local focus.
Within each level, the EU norms need to be identified, that is to say that a norm either occurs or does not occur in a given document or interview. Therefore, the issue of how norms will be identified when they are not mentioned explicitly remains, which is why the various coding techniques were applied in the data analysis. Norms can be named either directly through the concrete mention of a norm or indirectly by describing a situation without using the terminology. A coding scheme supports the analysis. This scheme evolved a number of coding techniques that were applied during the first cycle of coding. All coding was done using the software Atlas.ti, which was also used for the interview transcription.
The coding scheme was split by four cycle-one coding methods: grammatical methods, elemental methods, affective methods, and exploratory methods (Saldaña, 2009). Cycle-two coding methods were not applied, because their purpose is to develop broader thematic categories, which are mainly used for theory building, which in turn is not aimed at in this study.
One type of the first-cycle methods are grammatical methods whose names derive from the coding technique. Grammatical methods are attribute coding, which relates to the participant’s information such as age, gender, or job experience, simultaneous coding, which allows for several codes attributed to one string of words, and magnitude coding, which reflects an attitude towards something (Saldaña, 2009). For example, the statement “I came in summer 2014”
represents a string that was coded under the attribute coding method and was given the code
“work experience less than 2 years”, which means that at the time of the interview the respondent had worked in the position the respondent held at that point for less than two years.
This information is background information about the respondent, which might be relevant when a respondent seems uncertain of their responses. A respondent who has achieved their
65 position only recently tends to have less insight than someone who has been working in that position for several years. Simultaneous coding implies that one string, either a full sentence or a set of words, can have more than one code attributed to it. One piece of information can be used for different categories and codes are not mutually exclusive. When asking the question
“Are your relations formal or informal?” and the respondent answers “both”, “both” would be coded as “formal” and “informal”. Magnitude coding is used to detect a country’s relationship with the EU and, in particular, whether that country would appreciate more coordination or collaboration with the EU or less. “Doing things jointly would be an advantage” is consequently coded as the country wanting more cooperation. Structural coding, one of the elemental coding methods that are used as an initial approach to qualitative data analysis, was applied to detect phrases which relate to the topic of inquiry (Saldaña, 2009). This method is a method which helps to identify passages that relate to normative policy coherence in EU development and trade policies broadly. Most of the codes were used to provide the context in which the EU operates, but some information is also more directly related to the topic as the example about policy coherence in table 3 below shows. In the example, the respondent directly addresses policy coherence, which is the topic of research.
Value coding, a coding method that falls within the affective methods that address personal experiences and subjectivity, aims to detect the respondent’s values, beliefs or attitudes (Saldaña, 2009). This type of coding is particularly relevant for this study as it aims at normative representations across policy commitment and policy implementation. Even though this coding addresses subjectivity, a distinction between a respondent’s personal values or professional values is not made, because it is assumed that personal values translate into professional values.
Taking the example given in table 3, the respondent stated that gender equality and human rights are substantial for the respondent’s country. This is not their personal point of view, but the country’s position, which in the case of delegation staff might merge. What is important is not whether that is a personal or professional statement, but whether or not norms occur in different settings. If the respondent claims that gender equality is important, then this would have to reflect in other settings as well, such as in the monthly meetings or projects. The mere statement of a norm is, thus, not sufficient to claim that norms are actually relevant in development or trade. Indirect statements of norms need to be evaluated carefully. Promoting
“the idea of an open society” was coded as “freedom”, because of the freedoms a society were
66 to enjoy in a non-discriminatory environment. As is the case with coding, and with qualitative analysis in particular, all interpretations are subjective to the researcher’s understanding of what was said and what it is supposed to mean. Consequently, inferences and conclusions need to be expressed with great care.
The last type of coding methods are exploratory methods, among which is hypothesis coding.
Hypothesis coding is applying codes that are derived from theories to the data set to test given hypotheses (Saldaña, 2009). The analysis can be done either qualitatively or quantitatively but was done in a qualitative manner for this study. Sub-questions were used for this research instead of hypotheses to leave room to the data to reflect reality. The sub-questions, however, were phrased to reflect on the proxy factors (political commitment, policy networks, and regional interests) and to address the relationship between these proxy factors and NPCD. Thus, the questions were derived from theory and were guiding themes during the coding phase.
Regarding policy networks, for example, the question is how they influence normative coherence. One of the respondents gave an answer that gives some insight about the sectoral divisions and overlaps: “as less and less member state are involved in development, it is the commercial/economic attachés who take that place”. This string was coded as “knowledge overlap”.
Table 3 gives a summary of the coding methods with examples from the interviews.
For the policy documents, value coding has been used to determine the occurrence of norms in those documents. Search strings were simplified in order to provide more flexibility. For example, in order to identify whether the norm “democracy” occurred in a document, the search string did not cover the exact word. Instead, the abbreviated term “demo” was utilized to allow for variation between the words “democracy” and “democratic”, which both represent the norms “democracy”. Similarly, for “gender equality” both “gender” and “equality” were used as separate search strings. This allowed the recognition of “equality between men and women”, which would have been excluded by using solely the precise wording “gender equality”. Table 4 provides an overview of all twelve norms, the search string and positive and negative example, i.e. example that show commitment to norms and examples that do not.
67 Table 3: Applied Coding Methods and Examples from Interviews
Coding advantage. We also try to look for the coming year, the EU is planning a number of activities, where we could bring [our]
“To try to promote [our country] also with the values that we stand for: gender equality is very important for [us], human rights;
“What we see is that on the development counsellors meeting, as less and less member states are involved in
development, it is the commercial attachés or economic attachés who take that place.”
REL-DEV_knowledge overlap
Source: author’s compilation based on Saldaña’s (2009) typology
Creating the codes and revisiting the data set to ensure the rigorous application of those codes is already subject to the researcher’s precision and interpretation. All responses and search strings were evaluated based on the context. The sole appearance of a term did not suffice to be taken into account as a normative reference. Taking one example from table 4 “free from discrimination” could be a reference to trade rules, but in this case taken from the EU-Vietnam Partnership and Cooperation Agreement, it is under the labor, employment and social affairs heading making reference to non-discrimination at the work place based on nationality. Thus, for each normative reference in both the interviews and policy documents, the context in which the reference was made was taken into account.
68 Table 4: Norm Search Strings and Examples from Policy Documents
Norms Search strings Positive Example Negative Example Democracy “Democracy”
Liberty “Liberty” Liberty Liberal market economy
Peace “Peace” Peace Peace mission
Rule of law “Rule of law” Rule of law -
Drawing conclusions from the codes and making generalizable deductions adds another layer of difficulty, which is why the coding results needs to be evaluated cautiously. Normative policy coherence for development is qualitatively analyzed via the proxy factors indicated above. Taking the example mentioned above, involving economic attachés in development counsellor meetings would indicate that existing structural divisions between the development network and the trade network are starting to diffuse, which in turn makes knowledge transfer more fluid. The potential of normative spill-over from one network into the other is higher given the closer contact, but the direction of the spill-over remains unclear.