Submitted in partial fulfilment of the requirements for the degree of

Transcription

1 In what way is there a problem in the Principal-agent relationship between the home minister and the Europeanised national civil servant and what factors explain higher or lower degrees of loyalty of the agent to the principal? Submitted in partial fulfilment of the requirements for the degree of Master of Science in Public Administration (Specialisation in International and European Governance) 2015/2016, Faculty of Governance and Global Affairs, Leiden University, The Hague Ulf Schönert Lich 2016

2 Thesis supervisor: Dr. Casper F. van den Berg Faculty of Governance and Global Affairs, Leiden University, The Hague 1

3 Content 1 List of figures List of tables Introduction and research question Theory and literature review Scoping study Discussion of the literature Principal-agent relationship Europeanisation of national civil servants Influence of the Europeanisation of the national civil servant on the Principal-agent relationship Theoretical model and causal mechanism Research design Key concepts and their operationalization into variables Research approach and design Measurement and data source Defining the relevant populations, unit of analysis and case selection Common threats to inference Data collection and findings Measure of central tendency Measure of variation Statistical correlations Linear regression with dummy variables Multiple linear regression with dummy variables and beta Multiple linear regression with interaction effects Non-linear influences of the independent variables + test for missing variables Regression diagnostics Data collection and findings interviews Brief summary of the chapter Analysis Hypothesis Hypothesis Hypothesis Hypothesis Hypothesis Discussion analysis

4 6. Conclusion List of references Annex Stata do file Survey explainaire Survey questions English Survey questions Dutch Interview questions Invitation interviews Transcription Guidelines Interview Consent form Interviews Declaration of authorship

5 1 List of figures Figure 1 Principal-agent problem Figure 2 Principal-agent problem Figure 3 Chain of delegation and accountability in the Netherlands Figure 4 Chain of delegation and accountability in the European Union Figure 5 ACPR plot for the feeling of loyalty of the Dutch national civil servant towards his/her own policy views Figure 6 ACPR plot for the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative system of the EU Figure 7 ACPR plot for the feeling of a clear mandate for negotiations on EU level from the minister List of tables Table 1 List of variables Table 2 Descriptive statistics of all relevant variables Table 3 Measurement of central tendency and measure of variation Table 4 Statistical correlation Table 5 Linear regression with dummy variables Table 6 Multiple linear regression Table 7 Multiple linear regression with control variables Table 8 Standardised multiple linear regression with control variables Table 9 Regression diagnostic VIF test Table 10 Regression diagnostics heteroscedasticity test Table 11 Data collection and findings interviews

6 1. Introduction and research question The majority of the European Union (EU) member states are parliamentary democracies. In most parliamentary democracies, the legislature and executive can be aligned in a singular chain of delegation and accountability. Depending on the state s system, these chains differ but all start with the voter, contain a parliament and ministers and lead to the national civil servants (Ennser-Jedenastik 2013: pp ) (Van Thiel/ Yesilkagit 2008: pp ). The chain of delegation and accountability guarantees (in theory) that all power comes from the voters and that through the chain, the voters can hold everybody in the chain accountable (Ennser-Jedenastik 2013: pp ). Directly arising from there, one of the most common anxieties of our time and the topic of comprehensive research is the fear of a runaway bureaucracy. In other words, the fear that the chain of delegation and accountability is broken because the national civil servants are unresponsive towards instructions from their political superiors and will not be held accountable for their actions (Van Thiel/ Yesilkagit 2008: pp ). The person directly delegating the national civil servant and holding him/her directly accountable is the home minister. The connection between the minister and the national civil servants is one of the most important connections in every EU member state because it connects policy making and policy implementation (Beblav y 2009: pp ). While the problem of a runaway bureaucracy within EU member states has been researched extensively in the past, over the years, the EU has become an ever evolving factor which undoubtedly has influenced, at least to some extent, the complete national administration. Hence, it is quite astonishing that only a very limited number of studies focus[es] on the impact of the EU on individual [national] civil servants (Schmidt/ Van den Berg (in print): n.p.). Therefore, this master thesis will combine the Principal-agent relationship between the minister and the national civil servant and the Europeanisation of the national civil servant. The principal-agent theory was chosen because it can elucidate central problems in interaction between principals and agents in both policy implementation and public policymaking concerning performance and remuneration (Lane 2013: pp. 85). In the most straightforward structure of the Principal-agent relationship, two actors are in a cooperative relationship with each other and exchange resources. The two actors are the principal and the agent (Braun/ Guston 2003: p ) or, in this case, the minister and the national civil 5

7 servant. Whenever a Principal-agent problem arises between the minister and the national civil servant, the national civil servant acts disloyally and in his/her own interest that are contrary to those of the minister. The minister will try to control the national civil servant in order to prevent Principal-agent problems from developing but this is only possible to a certain degree because the control of the national civil servant is costly (time and money). The minister is therefore to some extent dependent on the loyalty of the national civil servant (Braun/ Guston 2003: p ) (Gailmard 2014: pp ). The impact of European integration occurs in almost all areas of administration throughout but in each area at different intensities and in different forms (Mastenbroek/ Princen 2010: pp. 154) (Schmidt/ Van den Berg (in print): n.p.) (Meyer-Sahling/ Van Stolk 2015: pp ). Whereas a small group of Dutch national civil servants are deeply affected by Europeanisation, the majority is shallowly affected (Mastenbroek/ Princen 2010: pp. 154) (Schmidt/ Van den Berg (in print): n.p.) (Meyer-Sahling/ Van Stolk 2015: pp ). But what must be understood is that Europeanisation is the subject of many different disciplines causing a huge number of different definitions. Most examinations of the influence of Europeanisation on the national civil servant concentrate only on one definition. One very good example: Europeanization as an analytical concept generally refers to the reorientation or reshaping of political and administration in the domestic arena in ways that reflect, structures, polices and practice advance through the EU system of governance. When we talk in particular about the Europeanization of civil servants and their work, what is meant is the degree to which national civil servants are involved in EU-related activities, what this involvement entails, and the impact of this involvement on their outlook, role-conception, career paths, discretion, and relationships vis-à-vis other administrative, political and societal actors (Schmidt/ Van den Berg (in print): n.p.). But this author learned that the different conceptions complement rather than exclude each other (Olsen 2002: ) and are furthermore necessary in the case of this master thesis to be able to completely analyse the influence of the Europeanisation of the civil servant as to the Principal-agent relationship between the minister and the national civil servant. To the knowledge of this author, such an in depth analysis with this particular focus on the Principal-agent relationship between the minister and the national civil servant and with such a broad understanding of Europeanisation has never been done. The literature gap this 6

8 master thesis will start to fill, is the influence the Europeanisation of the national civil servant has on the Principal-agent relationship and, based on this examination, the factors influencing the feeling of loyalty of the national civil servant towards his/her own minister. The goal of this research is to theoretically examine whether there is problem in the principalagent relationship between the home minister and the Europeanised national civil servant and to examine the factors influencing the feeling of loyalty of the national civil servant towards his/her own minister. The social relevance is obvious: if the Europeanisation of the national civil servant were to cause a problem in the Principal-agent relationship between the minister and the national civil servant, this could cause the system of delegation and accountability between the minister and the national civil servant to fail. In other words, the connection between policy making and implementation would be broken or at least be unacceptably skewed. The Europeanisation of the national civil servant could enable the national civil servant to do whatever he/she pleases without any oversight. Society would have to ask itself whether such a system would be bearable or would need some serious reform. The type of research by which this examination will be conducted is a mixed methods approach. By combining different designs and approaches can [ ] the relative weaknesses of each individual one [be alleviated] and [ ] their respective strengths [be amplified] (Toshkov 2016: p. 310). In this case, a survey and interviews will collect different types of data (mixed data collection) (Toshkov 2016: p. 312). The quantitative analysis of the survey is a secondary analysis. The survey had originally been conducted with a similar topic in mind (Gehring/ Weins 2009: pp ). The large number of cases collected in the survey suggest a Large-N design. To compare the large number of cases (national civil servants) a cross sectional design will be used at one point in time (Toshkov 2016: pp ). The qualitative analysis of the interviews are guided interviews. To be more precise, the type of interviews used in this master thesis are a derived form of a guided interview: the problem centred interview (Mayring 2002: p.67). The problem centred interview is centred on a particular (social) problem, which has previously been objectively illuminated by the researchers. The problem-centred interviews are particularly suitable in the more theory-driven research, which deals with specific issues (Mayring 2002: p.71). 7

9 For this analysis, a particular country was chosen: The Netherlands. Without revealing too much at this point, it can be said that the Netherlands are a constitutional monarchy with a parliamentary democracy system. Besides the fact of being a parliamentary democracy and an EU member state it was simply chosen because many Dutch national civil servants volunteered to provide information about themselves in interviews as well as in a survey to contribute to the examination at hand. The question this master thesis will try to answer is: In what way is there a problem in the principal-agent relationship between the home minister and the Europeanised national civil servant and what factors explain higher or lower degrees of loyalty of the agent to the principal? The second chapter following this introduction will present the literature review. The literature review will briefly introduce the performed scoping study and discuss the theory. The focus will be on the Principal-agent theory and, following that, different perspectives of Europeanisation and its influence on the national civil servant. The theory discussion will end with one of the core parts of this master thesis, the reconciliation of both theories, and examine the influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between the minister and the Dutch national civil servant. The second chapter will be concluded with the theoretical model and the causal mechanism. The third chapter will present the research design. Its first part will demonstrate the key concepts and their operationalization into variables; afterwards the research approach and design will be introduced and further elaborated in the part dealing with the measurement and data sources. The fourth part of the research design will define the relevant population, the unit of analysis and the case selection. The third chapter will end with the discussion of the common threats to inference. In the fourth chapter, the data collection and findings will be presented. An increasing approach will be used which means that this chapter starts with the basic measure of central tendency and the measure of variation, moves on to the statistical correlations and linear regression models including dummy variables and ends with a complete tested multiple linear regression model including dummy variables and standardized regression coefficients. In the following two parts of chapter four, the multiple linear regression model is then tested for interaction effects as well as non-linear influences of the independent variables and/or missing variables. Afterwards, the validity of the multiple 8

10 linear regression model will be established in the regression diagnostic. Then, the data collection and findings of the interviews will be presented in a separate part of chapter four. Chapter four will be concluded with a brief summary of the chapter. In chapter five, the analysis will be presented. The analysis will test the 5 hypotheses, which were elaborated in the last part of chapter two, the theoretical model and causal mechanism. The author will work through all 5 hypotheses and terminate this chapter with a discussion of the analysis. Finally, this master thesis ends in a conclusion. 2. Theory and literature review The following chapter will present the theory. The first part will be about the performed scoping study. Afterwards will in the second part the literature be discussed. This will include the Principal-agent theory, following that, different perspectives of Europeanisation and its influence on the national civil servant and ending with the influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between the minister and the Dutch national civil servant. The chapter will be concluded with the theoretical model and the causal mechanism Scoping study This scoping study will be used to examine the extent, range and nature of research activity, to determine the value of undertaking a full systematic review, to summarise and disseminate research findings, to identify research gaps in the existing literature (Arksey/ O'Malley 2005: pp ). The first methodological framework to carry out a scoping study was published by Arksey and O'Malley in Their developed framework consists of 5 stages, which can be extended by a sixth stage ( identifying the research question, identifying relevant studies, study selection, charting the data, collating, summarising and reporting the results and consultation exercise ) (Arksey/ O'Malley 2005: pp ). Arksey and O'Malley s framework provides, on the one hand, an excellent foundation for a scoping study but, on the other hand, no universal scoping study definition or purpose exists. Arksey and O'Malley themselves encourage other authors to adapt and refine their framework to enhance the methodology (Arksey/ O'Malley 2005: pp ). The scoping 9

11 study in this master thesis will use the Arksey and O'Malley framework (first 5 stages) and aim to build on it. The research question for the scoping study and the research question of the master thesis concur. But there is more to the first stage than identifying the research question; it is also important to decide the aspects of the research question which will be the focus for the next stages of the scoping study. The scoping study in this case will focus on three aspects: Principal-agent relationship, Europeanisation of national civil servants, Principal-agent relationship influenced by the Europeanisation of national civil servants (Arksey/ O'Malley 2005: pp ). The second stage serves the identification of key references, if possible for all three aspects. The idea here is to enhance this author s background knowledge regarding the topics, to make him an expert in the field and herewith improve the quality of the scoping study (Arksey/ O'Malley 2005: pp ). The key references consist primarily of introductions to the topics, textbooks, more general, or theoretical works. They create the basis for this scoping study. In the third stage, search terms are determined. As already indicated, the whole point of scoping the field is to be as comprehensive as possible in identifying primary studies [ ] and reviews suitable for answering the central research question (Arksey/ O'Malley 2005: pp ). The adopted strategy is an internet search of the terms: Principal-agent relationship, Europeanisation of national civil servants, Principal-agent relationship Europeanisation national civil servants. The findings of the search term Europeanisation of national civil servants have turned out to be too broad, which is why it was adopted into Europeanisation national bureaucracy. The internet search was completed via google (Google Deutschland 2016: n.p.), google books (Google books 2016: n.p.), google scholar (Google Scholar 2016: n.p.). Stage four is about identifying the key journals for the research question. Journals of International relations (Foreign Affairs/ Foreign Policy/ International Organization/ International Security/ International Studies Quarterly/ European Journal of International Relations), Public administration/ Public policy (Journal of Public Administration Research & Theory/ Journal of European Public Policy/ Journal of Public Administration Research & Theory/ Journal of Public Policy/ Policy Review/ Policy Studies Journal/ Public Administration 10

12 Review/ Public Policy and Administration) and Political science (British Journal of Political Science/ Comparative Political Studies/ European Journal of Political Research/ Journal of Politics/ Governance/ Policy Studies Journal/ Political Science Quarterly/ Political Studies) were explicitly searched in view of the terms developed in stage three. All other texts in any other than the above mentioned journals were included as well as long as they had a strong connection with the master thesis and met the inclusion criteria (Please see stage five). In stage five, inclusion criteria based on the research question were created to limit the number of studies to those that could enhance the output of the scoping study. All texts had to be scientifically researched at least as to the level of a master thesis but preferable the level of a dissertation. To be eligible, articles had to be published in one of the key journals for political science, international relations or public administration (please see stage four) or be of comparable scientific quality. All texts had to be published in either English or German. Other languages were excluded for practical reasons (cost and time to translate other texts). The author of this thesis had to accept that due to this unnecessary limitation potential relevant texts might have been missed. The texts needed to be in a scientifically quotable format. The texts had to be in a storable format (e.g. pdf, doc, etc.). The texts had to be accessible via the internet or in a library. The texts could not be older than from before 2000 as the focus of this scoping study is to present the current level of the research and its gaps in the literature. Furthermore, each text needed to have a strong connection with the master thesis (Arksey/ O'Malley 2005: pp ). The last stage of the scoping study was to chart the data in an Excel file. Charting describes a technique for synthesising and interpreting qualitative data by sifting, charting and sorting material (Arksey/ O'Malley 2005: pp ). The goal was to obtain key information from the findings, bring them into an orderly fashion and achieve some sort of comparability of the findings. The key information are the author(s) names; the title of the text; the year it was written in; the number of citations; definitions used in the text; the qualitative, quantitative or mixed method used as the research design; the country (s) examined; the dependent variable(s); the independent variable(s); the key findings in the text; the identified research gaps and/or future directions. The established data will be used as a basis for the next part (Arksey/ O'Malley 2005: pp ). It is important to keep in mind that the scoping study and the following discussion of the literature was done for and ultimately with the resources (time 11

13 and space) of a master thesis. The goal of the scoping study and the discussion of the literature is not to give a complete picture but to provide the reader with all necessary information in order to understand the background of the thesis. The next part of Chapter 2 will summarize and report on the results of the scoping study but will also identify any research gaps in the existing literature. The structure of this part is based on the three aspects the scoping study is based on: Principal-agent relationship, Europeanisation of national civil servants, Principal-agent relationship influenced by the Europeanisation of national civil servants Discussion of the literature Principal-agent relationship The principal agent theory has been developed in the context of rational choice and transaction cost theory (Braun/ Guston 2003: p.302) (Scott 2000: pp ). It is in fact a highly flexible family of models, rather than an overarching set of assumptions and results (Gailmard 2014: pp ). In the most straightforward structure of the Principal-agent relationship, two actors are in a cooperative relationship and exchange resources. The two actors are the principal and the agent. Nevertheless, a Principal-agent relationship can have multiple principals and agents. The principal needs the agent to realize his/her interests and delegates tasks from his/her area of responsibility to the agent. The reasons for using the agent can be manifold, for example, the agent can possess a certain skill which the principal needs to fulfil to meet his/her interest or simply because the principal wants to put his own resources to better use and save time by delegating minor tasks or simply because it is too much work, etc. The agent works on behalf of the principal ( extension of self ) in exchange for some benefit (e.g. payment, social recognition, etc.). The limited information on the agent s abilities make it very difficult for the principal to find the perfect fit for the job. Both actors are assumed to be rational trying to maximise their benefits according to their preferences (Braun/ Guston 2003: p ). The agent has two advantages over the principal. First, the agent can more easily obtain better quantitative and qualitative information (hidden knowledge) than the principal for several reasons. Under normal circumstances, the agent is an expert in his/her field, the 12

14 reason why the principal hired him in the first place. Furthermore, the agent has more time to collect all the necessary information; if this were different the principal would not have had to hire the agent in the first place. The agent s second advantage is that he/she knows all the possible courses of action that can be taken in one situation (hidden action) (Lane 2013: pp ). But a problem for the principal out of these agent s key advantages arises only when the agent has different interests compared to the principal and is therefore prone to act against the principal s interests. The different interests of the agent can be manifold, for example, he/she can have different political views or the agent tries to shirk his/her responsibilities. The benefits of the principal, on the other hand, depend on the condition that the agent performs the tasks in accordance with the wishes of the principal. The principal is in the dilemma that he/she does not know whether the agent is motivated to act upon his/her own interest, contrary to those of the principal (Braun/ Guston 2003: p ). The agent is able to act in his/her own interest because it can be very costly for the principal to control the agent s actions; furthermore, it is very unlikely for the agent to be caught. In order to completely control the agent, the principal would have to collect enough information to close the information gap and would have to be aware of all the courses of action possible in every specific situation. This is contrary to the reason of the principal deciding to delegate certain tasks to the agent, namely to save his/her own resources. The principal can therefore not always ensure that the agent is acting in the principal s interest. The principal has to rely on the loyalty of agents towards him to some extent. The result is that the agent possesses some leeway to exploit in his/her own interest. When a task becomes so costly for the agent that there a Principal-agent problem is incurred, the agent acts in his/her own interest and not in the interest of the principal and exploits the principal. The costs for the principal which can be the result of the agent acting in his/her own interest instead of those of the principal are called agency costs (Braun/ Guston 2003: p ). The rent seeking ambitions of agents can even lead to illegal activities like corruption, tax evasion, embezzlement, patronage, etc. (Lane 2013: pp ). Both actors are assumed to be rational and try to maximise their benefits according to their preferences, meaning when there is a Principal-agent problem, the agent will act in his/her own interest instead of those of the principal. The agent will be disloyal to the principal. When there is no Principal-agent problem, the agent will be loyal 13

15 (Braun/ Guston 2003: p ). It is important to note that a problem in a Principal-agent relationship is not always a necessity. In many cases, hidden knowledge or hidden action may be non-relevant, non-existent, or even be in favour of the principal. In others, the principal and the agent may share preferences and be allies (Van Thiel/ Yesilkagit 2011: pp ). Figure 1 Principal-agent problem hidden action = Principal-agent problem or/and * hidden knowledge different interest between the principal and the agent Source: Author s figure (1) based on Lane 2013: pp Despite the fact that complete control of the principal over an agent can never be achieved nor is it desirable, the literature examines different mechanisms enabling the principal to make sure that the agent acts in the principal s interest to achieve as little agency loss as possible (e.g. multiple principals controlling one agent, multiple agents controlling each other, the responsibility of the agent to make reports about all the collected information and the possible courses of action, to build trust between the principal and the agent, etc.) (Lane 2013: pp ) (Ennser-Jedenastik 2013: pp ) (Van Thiel/ Yesilkagit 2011: pp ). The principal(s) has to decide whether the additional costs which arise through the different mechanisms warrant preventing the expected agency loss, in other words the loss the principal would suffer if the agent acted against the principal preferences. Principals typically would find it prohibitively costly to eliminate agency loss completely, so they choose to tolerate some (Gailmard 2014: pp ). If the principal controls the agent too closely, it can have the effect that the agent loses all flexibility and is not able to use his/her expertise in the principal s benefit. Furthermore, the principal will act against the principal s interest every time his/her interests are supressed. If the principal left the agent some more freedom 14

16 to act, the possibility that the agent s interests overlap with those of the principal would be much greater with smaller agency loss as a consequence (Gailmard 2014: pp ). During the course of this master thesis, the principal-agent model will be used to elucidate central problems in interaction between principals and agents in both policy implementation and public policy-making concerning performance and remuneration (Lane 2013: pp. 85). When the Principal-agent relationship is applied to politics, a great number of Principal-agent relationships can be examined. This includes not only formal government actors being categorized as principals or agents but also interest groups, citizens, the media, etc. (Van Thiel/ Yesilkagit 2008: pp ). Depending on the different systems, there can be more or less Principal-agent relationships with one or several different principals and agents. In most parliamentary democracies, for example, the legislature and executive can be lined up in a singular chain of delegation and accountability. Each chain includes several Principalagent relationships. The Principal-agent relationships received most attention in the literature in this chain are between voter towards parliament, minister towards civil servant and minister towards the management of state owned enterprises (SOE) (Ennser-Jedenastik 2013: pp ) (Van Thiel/ Yesilkagit 2008: pp ). In the first case, the voters (as principals) select and instruct members of parliament (as agents) by voting their favoured candidate into office. The politicians present the policies they want to carry out in the election campaign. But as explained more generally above, in every Principal-agent relationship hidden information and hidden action is a distinct possibility. The voter is not able to retrace all the possible courses of action that a politician can take in a given situation. Furthermore, the politician can to obtain better quantitative and qualitative information more easily than the average voter. This is a problem for the voter when the politician acts in his own interest rather than the interest of the voter. Politicians can have their own agenda, which is made public after the election or kept hidden. They claim that they have a general mandate from the population to search for the policies that are in the national interest and it makes it easier to adapt their politics due to unforeseen circumstances (Lane 2013: pp ). Complicating things is the issue that election contracts tend to be very ambiguous or general. The politician will try to maximise his/her resources and will use them to remunerate or give favours to the people who helped them win elections, using massive propaganda to convince the principal about their suitability as 15

17 political agents (Lane 2013: pp ). Also on this point, it should be said that a problem in a Principal-agent relationship is not always a necessity but a possibility. The politician has to act in his own interest and not in that of the voters (Van Thiel/ Yesilkagit 2011: pp ). Although complete control by the principal over an agent can never be achieved and besides the above mentioned mechanisms for the principal to make sure that the agent acts in the principal s interest to achieve as little agency loss as possible, the voters have additional mechanisms at their disposal to achieve some control over the member of parliament (e.g. parliamentary opposition, referendum and recall, political decentralization or federalism, strong involved civil society, to statutorily regulate politics, re-election of members of parliament, oversight hearings, separation of power, etc.) (Lane 2013: pp ) (Van Thiel/ Yesilkagit 2008: pp ). The voter has to decide whether the additional costs arising due to the different mechanisms warrant preventing the expected agency loss. In the second case, a minister (as principal) uses civil servants (as agents) to implement policies. The minister needs the civil servant to realize his/her interests and delegates tasks from his/her area of responsibility to the civil servant. The civil servant works on behalf of the minister in exchange for some benefit from the minister (e.g. payment, social recognition, etc.) (Braun/ Guston 2003: p ). Also in this relationship, hidden information and hidden action may exist. The civil servant can more easily obtain better quantitative and qualitative information than the minister. Furthermore, the minister is not able to retrace all the possible courses of action that a civil servant can take in a given situation. This presents a problem for the minister when the civil servant acts in his own interest rather than the interest of the minister (Lane 2013: pp ). The different interests of the civil servant can be manifold, for example, he/she may have different political views or the civil servant tries to shirk his/her responsibilities or the civil servant tries to maximise his/her budget whereas the minister tries to cut costs. The minister s benefits, on the other hand, depend on the condition of the civil servant performing the tasks in accordance with the minister s brief. The minister is in the dilemma that he/she does not know whether the civil servant is motivated to act upon his/her own interest, an interest that might be contrary to those of the minister (Braun/ Guston 2003: p ) (Gailmard 2014: pp ). The civil servant is able to act in his/her own interest because it can turn out to be very costly for the minister to control the civil servant s actions. To some extent, the minister has to rely on the loyalty of the civil 16

18 servant towards him. When a task becomes so costly for the agent that a Principal-agent problem arises, the agent will act in his/her own interest and not in the interest of the principal and end up exploiting the principal (Braun/ Guston 2003: p ). The minister has some additional mechanisms at his/her disposal as well in order to achieve some control over the civil servants (e.g. different kinds of contracts for the agent with different freedoms of action or reservation prices and incentives, reorganization of the ministry structure, audits, reporting sanctions, budgetary powers, administrative procedures, to statutorily regulate processes, staffing of the civil servant posts with party affiliates who have similar policy preferences (ally principle), limit the policy decisions the agent is allowed to make and establish decision making criteria, control by judicial actors, clear negotiation mandates, to make his/her priorities clear to the national civil servant, to coordinate between negotiation and implementation, etc.) (Lane 2013: pp ) (Van Thiel/ Yesilkagit 2008: pp ) (Ennser-Jedenastik 2013: pp ) (Kassim/ Menon 2003: pp ) (Gailmard 2014: pp ). The minister has to decide whether the additional costs arising through the different mechanisms warrant preventing the expected agency loss. Many of the different measures mentioned in this text are highly disputed among scholars. Every control measure restricts the civil servant s flexibility (Kassim/ Menon 2003: pp ) and can render the civil servant unable to use his/her expertise to the minister s benefit. Furthermore, the principal will act against the principal s interest every time his/her interests are supressed. Whereas, if the minister gave the civil servant some more freedom to act, the possibility that the civil servant s interests overlapped with those of the minister would be much greater with smaller agency losses as a consequence (Gailmard 2014: pp ). One Principal-agent relationship catching a lot of scientific attention in the last few years was the relationship between a minister and the management of state owned enterprises (SOE) (Ennser-Jedenastik 2013: pp ) like independent administrative bodies, contract agencies, government foundations, legal-entities with statutory tasks, etc. The reason for this attention resulted from SOEs having been seen equally in the literature and by government as the ultimate tool to reduce government overload and public deficit and more importantly in this case to solve or at least improve the problems within the Principal-agent relationship between the minister and his/her civil servants which led to the excessive creation of new SOEs. Tendering/bidding under a regime of short-term contracts was supposed to 17

19 strengthen the principal s position because the agent s only interest would be to receive the next tendering/ acceptance of a bid. Unfortunately, this kind of Principal-agent relationship had its own downside as it invited unreliable and untrustworthy agents to bid. Furthermore, there is the expenditure to coordinate the different SOEs and manage auctions and tournaments comparable to monitoring civil servants. The transaction costs for the principal were considerable (Lane 2013: pp ). The principals had even less political control in the SOEs because the latter were even more autonomous and specialized, which increased the problems of hidden knowledge and hidden action (Van Thiel/ Yesilkagit 2008: pp ) (Braun/ Guston 2003: p ). All of the Principal-agent relationships mentioned above are worth studying. However, the data collected for this master thesis concentrates on the Principal-agent relationships between a minister (principal) and his national civil servant (agent), which in this author s opinion is one of the most important Principal-agent relationships as it connects policy making and policy implementation (Beblav y 2009: pp ) within a state. Whereas the intention in this part about the Principal-agent relationship was to explain the relationship between the principal (minister) and the agent (civil servant) more general, will from this point onwards the focus be on the Principal-agent relationship between the minister and the national civil servant. Critics reproach the Principal-agent relationship taking only the principal and the agent into consideration and abstracting third parties. To some extent, this is correct. Although the Principal-agent relationship can theoretically be enhanced by several principals and agents, it still gives only a black box model which simplifies assumptions about the principal(s) and the agent(s). Therefore, it has to be mentioned openly that the Principal-agent relationship cannot offer the complete picture. Nevertheless, it offers the Principal-agent relationship a most appropriate starting point for theorizing and analysing the relation between a principal and his/her agent (Pollack 2007: pp. 3-21). In the next part, the author will examine the way in which a national civil servant can be Europeanised. 18

20 Europeanisation of national civil servants Europeanisation is the subject of many different disciplines resulting in a huge number of different definitions. The different conceptions complement rather than exclude each other (Olsen 2002: ). In this chapter, the author will briefly present the three most important ones and afterwards concentrate on those that allow this thesis to advance. The historical perspective refers to Europeanisation as the export of European political institutions, political practise and way of life beyond the European continent mainly through the means of colonisation and thus through coercion (Sittermann 2006: p. 3). The cultural perspective refers to the reshaping of identities in contemporary Europe in a manner which relativizes (without necessarily supplanting) national identities" (Sittermann 2006: p. 3). The political perspective, at least the part that concentrates on the European Union, must for a better understanding first be divided into three sub concepts. The first sub concept means by EU enlargement the term Europeanisation (Sittermann 2006: pp. 1-23) (Olsen 2002: ). The second one is the emergence and development at the European level of distinct structures of governance, that is, of political, legal, and social institutions associated with political problem solving that formalize interactions among actors and of policy networks specializing in the creation of authoritative European rules" (Sittermann 2006: pp. 4-5) (Olsen 2002: ). The third understands Europeanisation as the penetration of European rules, directives and norms into the otherwise differentiated domestic spheres (Sittermann 2006: pp. 4-5) (Olsen 2002: ). The influence on the domestic sphere to which the last sub concept refers happens on two levels. The first institutional adaptation examines effects of EU membership on institutions and actors. The second EU influence examines effects of the EU on national policies (Sittermann 2006: pp. 1-23) (Olsen 2002: ) (Bovens/ Christensen/ Jenny/ Müller/ Yesilkagit 2010: pp ). To advance this thesis, two of the above mentioned perspectives need to be elaborated. The cultural perspectives view speaks of reshaping of identities through Europeanisation (Sittermann 2006: p. 3). According to social identity theory, this can happen in one of two ways. Either, once somebody is exposed to European integration their national identity and their newly established European identity reinforce each other. The citizen then has multiple 19

21 identities (inclusive national identity) and who one is depends on which groups one identifies with (Hooghe/ Marks 2005: p.422). Or, once somebody is exposed to European integration and opposes it, the national identity is reinforced and seen as a defence against the influence from Brussels. The citizen has an exclusive national identity (Hooghe / Marks 2005: pp ). These group loyalties can be extremely powerful in shaping views towards political objects (Hooghe/ Marks 2005: p.423). In other words, what Hooghe and Marks proved in their text Calculation, Community and Cues: Public Opinion on European Integration is that communal identities partially influence the citizens (political) attitudes (Hooghe/ Marks 2005: p.423). The national civil servant must in this regard be treated like a citizen. All of the above said therefore applies to him/her as well. The second perspective which needs to be further elaborated is the third sub concept within the political perspective. Europeanisation is seen as the penetration of European rules, directives and norms into the otherwise differentiated domestic spheres (Sittermann 2006: pp. 4-5). This happens on two levels, on the level of institutions and actors, and that of national policies (Sittermann 2006: pp. 1-23). The influence of EU membership on institutions and actors will be observed more closely (Laffan 2006: pp ) (Laegreid, Per/ Steinthorsson, Runolfur Smari/ Thorhallsson, Baldur 2004: pp ) (Harmsen, Robert 1999: pp ) (Ruffing, Eva 2015: pp. 1-21). In Time for EU Matters: The Europeanisation of Dutch Central Government, Mastenbroek and Princen explain the impact of the EU both on individual [national] civil servants and on organizations (Mastenbroek/ Princen 2010: pp. 154) in their analysis. Mastenbroek and Princen asses that a small group of [national] civil servant is deeply involved in EU matters but the majority is involved to a much lesser extent (Mastenbroek/ Princen 2010: pp ). The impact of European integration on government organizations remains limited to particular organizations [in particular policy sectors (Schmidt/ Van den Berg (in print): n.p.).] and particular [national] civil servants within central government (Mastenbroek/ Princen 2010: pp. 154) (Schmidt/ Van den Berg (in print): n.p.) (Meyer-Sahling/ Van Stolk 2015: pp ). The impact happens in three different dimensions (structure, culture and staff). In their structural dimension, they examined the personnel allocation which was distinguished by the EU involvement of the national civil servant, time spent on EU matters by the national civil 20

22 servant and dispersion of EU-related activities. The cultural dimension should not be confused with the cultural perspective explained at the beginning of this part. What happens as to the cultural perspective is that once somebody is exposed to European integration either his/her exclusive or inclusive national identity is reinforced. But it is still his/her identity. In the cultural dimension, which Mastenbroek and Princen divided into working practices (Training, Recruitment and Career development) and Outlook (Mastenbroek/ Princen 2010: pp ) ( the clarity of negotiation mandates, the priority given to EU matters, and the coordination between negotiation and implementation (Mastenbroek/ Princen 2010: pp. 162)) the national civil servant is taught to think European. His/her identity or believes do not change. In theory, these two things might be perfectly separable but the author of this text is aware that when the civil servant is taught to think European or is exposed to European integration in his work, it is very likely that his/her identity is influenced as well. But as Mastenbroek and Princen were not looking into this kind of influence on identity, these influences will be subsumed under the reshaping of identities in the cultural perspective in the next part. The third and last dimension Mastenbroek and Princen looked at was the staff which they divided into projection (Preparation, Council working groups, Commission working groups, Bilateral contacts) and reception (Transposition, Application and enforcement and Consideration of EU policies) (Mastenbroek/ Princen 2010: pp ). This part showed that the national civil servant can be Europeanised in two ways by reshaping his/her identity and by the penetration of European rules, directives and norms into the otherwise differentiated domestic spheres. The next part will examine how the Europeanisation of the Dutch national civil servant influences the Principal-agent relationship between the minister and the Dutch national civil servant Influence of the Europeanisation of the national civil servant on the Principalagent relationship Countless scholars examined the Principal-agent relationship including the relationship between the minister and the national civil servant. When it comes to the Europeanisation of the national civil servant, the knowledge is particularly sparse, which is surprising because of the centrality of these civil servants (Mastenbroek/ Princen/ Van den Berg (in print): n.p.). But thanks to some scholars, this literature gap is decreasing. This master thesis will combine 21

23 the Principal-agent relationship between the minister and the national civil servant and the Europeanisation of the national civil servant. To the knowledge of this author, such an in depth analysis with this particular focus on the Principal-agent relationship between the minister and the national civil servant and with such a broad understanding of Europeanisation has never been done. The literature gap this master thesis will start to fill, is the influence of the Dutch national civil servant s Europeanisation on the Principal-agent relationship between the minister and the Dutch national civil servant and based on this examination the factors influencing the feeling of loyalty of the Dutch national civil servant towards his/her own minister. As mentioned above, the focus of this master thesis is the Principal-agent Relationship between the minister (principal) and the national civil servant (agent). The agent has two advantages over the principal (hidden knowledge and hidden action) (Lane 2013: pp ). A problem for the principal may arise when the agent has also different interests (e.g. different political views, an interest to shirk his/her responsibilities) and does therefore have an interest to act against the principal s interests. The principal cannot always ensure that the agent is acting in his/her interest. The principal has to rely on the loyalty of the agent towards him to some extent. A Principal-agent problem can arise in every Principal-agent relationship but the Principal-agent problem in the Principal-agent relationship between the minister and the national civil servant could be intensified because of the Europeanisation of the civil servant. To intensify the Principal-agent problem, the Europeanisation of the civil servant either has to intensify the hidden action and/or hidden knowledge or the different interests between the minister and the national civil servant. Figure 2 Principal-agent problem 22

24 Source: Author s figure (2) based on Lane 2013: pp The national civil servant s reshaped identity could intensify the different interests between the principal and the agent. The national civil servant is exposed to European integration in the same way as an ordinary citizen (e.g. using the Euro, European parliament elections, etc.). But as Mastenbroek and Princen diagnosed, a small group of civil servants deeply involved in EU matters and the majority of those less involved in EU matters (Mastenbroek/ Princen 2010: pp. 154) are additionally exposed to European integration throughout their work (e.g. Special training for the national civil servants to deal with EU matters, The national civil servants have to spend more time on EU matters, etc.) (Mastenbroek/ Princen 2010: pp ). Once the national civil servant is exposed to European integration his/her identity is reshaped (Sittermann 2006: p. 3) either towards an inclusive national identity (national identity + European identity) or his/her national identity is reinforced (exclusive national identity) (Hooghe / Marks 2005: pp ). Either way, the national civil servants communal identities influence (partially) the national civil servants (political) attitudes (Hooghe/ Marks 2005: p.423). In other words, the reshaped identity influences/changes the interests (e.g. different political views) of the national civil servant. But according to theory, a difference in interests between the minister and the national civil servant is one of the two conditions causing a Principal-agent problem and not a simple change in the interests of the national civil servant (Braun/ Guston 2003: p ) (Jordan 2003: pp ). The minister himself is exposed to European integration and is him-/herself Europeanised. His/her identity is influenced in one of two directions as well and therefore his/her interests are changed as well. The condition of a difference in interests between the minister and the national civil servant should therefore only occur if the minister s (principal s) interest for example is EUsceptical and the national civil servants (agent s) interest is EU-enthusiastic (or the other way around). The EU is just one interest of many on which the minister and the civil servant can disagree. Europeanisation in the sense of the cultural perspective does not change the Principal-agent relationship nor does it intensify the Principal-agent problem between the minister and the Europeanised national civil servant. However, the penetration of European rules, directives and norms into the otherwise differentiated domestic spheres (Sittermann 2006: pp. 4-5) onto institutions and actors (Mastenbroek/ Princen 2010: pp ) could intensify the hidden knowledge and/or 23

25 hidden action. As explained before, Europeanisation according to Mastenbroek and Princen occurs in three different dimensions (structure, culture and staff). Two of the three dimensions (structure and culture) could intensify the hidden knowledge and/or the hidden action. It should be pointed out again that Mastenbroek and Princen detected only a small group of national civil servants deeply affected by the EU and a majority of those less so (Mastenbroek/ Princen 2010: pp. 154). In their structural dimension, they examined the personnel allocation which was distinguished in the EU involvement of the national civil servant, time spent on EU matters by the national civil servant and dispersion of EU-related activities (Mastenbroek/ Princen 2010: pp ). A national civil servant becoming more involved in EU matters and spending more time on them will automatically learn more about the EU and its procedures. But it is not unusual for the agent to possess certain skills or to be very specialized. This is in fact one of the reasons why the agent is so valuable to the principal and why he/she was chosen in the first place. The minister, on the other hand, will not be ignorant of the new developments. If the civil servants under his/her command become or already are deeply involved in EU matters, the minister will not only become more involved because the national civil servant reports to him/her but the minister will inform himself/herself as well as this is the only way of controlling the civil servants to a certain degree and make sure that his/her interests are acted upon. This structural dimension does not increase the hidden action and/or hidden knowledge between the minister and the national civil servant. In the cultural dimension, which Mastenbroek and Princen divided into working practices (Training, Recruitment and Career development) and Outlook (Mastenbroek/ Princen 2010: pp ) ( the clarity of negotiation mandates, the priority given to EU matters, and the coordination between negotiation and implementation (Mastenbroek/ Princen 2010: pp. 162)) the national civil servant is taught to think European. The cultural dimension could intensify the hidden action and hidden knowledge side as the national civil servant is specially trained to deal with EU matters or is especially recruited because of his education or work experience in this matter. But as said above, it is not unusual in the Principal-agent relationship for the agent to possess certain skills or to be very specialized. This does not increase the hidden action and/or hidden knowledge. In the second part of the cultural dimension, the outlook could actually improve the Principal-agent relationship by improving 24

26 the loyalty of the national civil servant towards the minister. The principal cannot always ensure that the agent is acting in his/her interest. The principal has to rely on the loyalty of the agent towards him to some extent but for a minister to make his/her attitudes known through clear negotiation mandates, to make his/her priorities clear towards the national civil servant and to coordinate between negotiation and implementation could improve the loyalty of the national civil servant towards his/her minister. If the principal controls the agent too closely it can have the effect of the agent losing all flexibility and thus being unable to use his/her expertise for the principal s benefit. If the principal left the agent some more freedom to act, the possibility that the agent s and the principal s interests overlap would be much bigger with a smaller agency loss as a consequence (Gailmard 2014: pp ). The minister should give clear negotiation mandates, make his/her priorities clear towards the national civil servant, coordinate between negotiation and implementation and improve loyalty towards him/her but take no measures to control the agent too closely. The third and last dimension Mastenbroek and Princen looked at was the staff which they divided into projection ( We distinguish between the tasks of preparation of the Dutch input into EU-level meetings, participation in Council working groups, participation in European Commission meetings, and bilateral consultations with foreign counterparts (Mastenbroek/ Princen 2010: pp. 158)) and reception ( Transposition, the practical application or enforcement of EU policies, and the consideration of EU policies during national policymaking (Mastenbroek/ Princen 2010: pp. 158)). The Europeanisation in this dimension represents one of the greatest challenges of the Principal-agent relationship because it does not intensify the hidden action and/or hidden knowledge, at least not in a different way than the personnel allocation (structural dimension) and the working practices (cultural dimension), nor does it intensify the different interests between the minister and the national civil servant. But the staff dimension could be able to break the whole Principal-agent relationship. At this point, it is necessary to show that the Principal-agent relationship between the minister and the national civil servant is part of an overarching system. The following examination will be executed on the example of the Netherlands. The Netherlands were chosen because the main data sources (Please see chapter 3) are a survey and interviews conducted with national civil servants from the Netherlands. 25

27 The Netherlands are a constitutional monarchy with a parliamentary democracy system. Parliament is divided into the first and second chamber. The first chamber is elected from the provincial assemblies, the second by the Dutch people. The parties in the second chamber decide on the way in which the government is composed, depending on the balance of power in the second chamber following the national elections. The government consists of the king and the ministers. The unwritten rule states that the Council of Ministers (or a single minister) resigns, if the Second Chamber asks them to (Auswärtiges Amt 2015: n.p.). The above said establishes a chain of delegation and accountability for the Netherlands. It is important to note that this chain of delegation and accountability is an oversimplification of the entire Dutch system which includes the courts and some other actors. It is, for example, generally accepted that the national civil servant has more than one principal but it is also generally accepted that at the end of the executive chain, the individual minister is the most relevant actor, especially for the national civil servant (Van Thiel/ Yesilkagit 2008: pp ) (Gailmard 2014: pp ). Additionally, it shall be noted that scholars argue about the position of some actors in the chain. But for the purpose of examining the relationship between the minister and the Dutch national civil servant, this figure suffices. In this chain of delegation and accountability, there are several Principal-agent relationships (voter towards parliament, parliament towards minister, minister towards Dutch civil servant and minister towards managers of state owned enterprises). The principals delegate tasks to the agents and the agents are accountable to the principals (Van Thiel/ Yesilkagit 2008: pp ). Ministers are ultimately responsible for the decisions, actions and behaviour of bureaucrats residing under their authority (Van Thiel/ Yesilkagit 2008: pp. 140). 26

28 Figure 3 Chain of delegation and accountability in the Netherlands Source: Author s figure (3) based on Ennser-Jedenastik 2013: pp Since the EU was established, a completely new chain of delegation and accountability has emerged. Here, it is even more important to point out that this chain is also an oversimplification. On this topic alone, many scholars have written thousands of pages and much still stays in the dark. To be more precise, this chain of delegation and accountability for the EU is what the chain should look like, if it were constructed similarly to the structures found in most of the EU member states, rather than a status quo. For multiple reasons, the EU to date has problems with accountability and democratic input. One example is that the voters elect the European Parliament but the European Parliament does not completely choose or nominate the European Commission. Still for the purpose of the examination in this thesis, the figure suffices and stresses the following point. 27

29 Figure 4 Chain of delegation and accountability in the European Union Source: Author s figure (4) based on Ennser-Jedenastik 2013: pp The author of this text is looking at the Principal-agent relationship between the minister and the national civil servant in the Netherlands. Therefore, it must be differentiated between the Dutch citizens working for the EU as EU civil servants and Dutch national civil servants which includes the average national civil servant working for the Empire with different amounts of EU related tasks in the Netherlands but also the Dutch national civil servants sent to the different EU institutions by the Dutch government to work on their behalf. The former, the EU civil servants, are those who work in the applied chain of delegation and accountability in the European Union. The latter, the Dutch national civil servants, will be the ones examined here and are those working in the chain of delegation in the Netherlands. According to Schmidt and Van den Berg, the group of national civil servants, who are sent to the EU on behalf of the Netherlands, includes even more people than Mastenbroek and Princen operationalized. They work in European Commission expert groups, advisory committees, European Council working parties, comitology committees, independent networks of officials and regulators, bilateral consultations with their counterparts in other member states, etc. (Schmidt/ Van den Berg (in print): n.p.). Of course, European civil servants and the national civil servants interact when the national civil servants are sent to the EU institutions or the 28

30 European civil servants are sent to the member states. In fact, they interact so regularly that some scholars predict the disappearance of the unified traditional diplomatic service in international relations of the EU towards a new model of constituent components filled with specialized civil servants who deal only in their field of activity and work with the civil servants in their field of activity from the other member states (Geuijen, K./ t Hart, P./ Princen, S./ Yesilkagit, K. 2008: p. 15). EU specialists who can combine technical/sectoral expertise with European experience (Laffan 2006: p. 699). The question as to whether these developments caused by the Europeanisation of the staff (projection and reception) break the Principal-agent relationship between the minister and the Dutch national civil servant remains. To break the Principal-agent relationship between the minister and the Europeanised Dutch national civil servant one or more of the three conditions have to be met: The minister does not nor plans to delegate tasks of his/her area of responsibility to the civil servant or the civil servant receives no benefit from the minister or the civil servant is no longer accountable to the minister. All three can be dismissed. As long as the civil servant works for the minister, the minister will delegate tasks to the civil servant even if they are of minor importance. The minister would not let his/her resources dwindle and fire the civil servant instead if he/she was of no use. Second, if the civil servant is not paid or receives any other kind of benefit he/she will not bother to show up for work. The third condition of accountability is the most challenging condition. The minister and the Dutch civil servant are in a chain of delegation and accountability in the Netherlands. As long as both are locked into this chain of accountability nothing changes in the relationship and accountability can generally be assumed. The Dutch civil servant is responsible for all his/her actions towards the minister. The minister can reassign, cut the benefits or fire the national civil servant, if the civil servant does not do his/her job. When the Dutch national civil servant is sent to the EU (projection), he/she is sent on behalf of the minister. The accountability does not change because both are still in the chain of delegation and accountability in the Netherlands and the Dutch national civil servant reports back to the minister and is dependent on his/her good graces. If the minister is dissatisfied, he/she can call the national civil servant back home. In practice, the distance can make control of the Dutch national civil servant more difficult but it is not unusual in the Principal-agent relationship for the agent to be very specialized. In most cases, the minister will not only stay home in the member state. 29

31 The minister gives the national civil servant orders and will be on site when important things are decided. In fact, the theory shows that support from the minister is essential to achieve policy goals on EU and national level in the Netherlands (Schmidt/ Van den Berg (in print): n.p.). The minister has to find the perfect balance between travelling to the EU and staying back home in his/her ministry to oversee all his/her national civil servants (Laffan 2006: pp ). Additionally, the minister and the Dutch national civil servant would have to have different interests to cause a Principal-agent problem in the first place. If a Dutch national civil servant becomes an EU civil servant, he is no longer in a Principal-agent relationship with the Dutch minister but with the EU and therefore part of the chain of delegation and accountability of the EU (Gailmard 2014: pp ). For reception to break the chain of delegation and accountability the penetration of European rules, directives and norms into the otherwise differentiated domestic spheres (Sittermann 2006: pp. 4-5) onto institutions and actors (Mastenbroek/ Princen 2010: pp ) would have to go so far as to result in the breaking of the actual Principal-agent relationship between the minister and the Dutch national civil servant. One example would be an EU regulation stipulating that the minister is no longer accountable to the Dutch national civil servants. This would of course require the EU to have the competences in this area. Another example would be if the Netherlands decided to reorganize their complete administration without including the Principal-agent relationship between the minister and the Europeanised national civil servant. This is very unlikely and there is no indication in the literature as regard such plans because this Principalagent relationship between the minister and the national civil servant is one of the most important and effective ones not only in the Netherlands because it connects policy making and implementation (Gailmard 2014: pp ). The key structures, the ministries, the cabinet and committees served to channel work on Europe in the national administrations (Laffan 2006: pp. 697). The Netherlands took considerable care to make sure that the high degree of necessary inter-ministerial coordination caused by Europeanisation did not undermine the national politico-administrative connection. European policy making and policy implementation have been included in the existing structures with respect to the equality in the Council of Ministers and ministerial autonomy (Harmsen, Robert 1999: pp ). 30

32 This part examined the influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between the minister and the Dutch national civil servant. A small group of Dutch national civil servants are deeply affected by Europeanisation whereas the majority is only shallowly affected. There was a possibility that the Principal-agent problem might have been intensified by the Europeanisation of the Dutch national civil servant. The influence of Europeanisation on the Dutch national civil servants identity was examined. The Dutch national civil servant s identity reshaped but did not intensify the different interests between the principal and the agent. The EU is just another interest among many on which the minister and the Dutch national civil servant may disagree. Europeanisation in the sense of the cultural perspective does not change the Principal-agent relationship nor does it intensify the Principal-agent problem between the minister and the Europeanised Dutch national civil servant. This was followed by the penetration of European rules, directives and norms into the otherwise differentiated domestic spheres (Sittermann 2006: pp. 4-5) onto examined institutions and actors (Mastenbroek/ Princen 2010: pp ). Europeanisation occurs here in three different dimensions (structure, culture and staff). The structural and cultural dimension did not increase the hidden action and/or hidden knowledge between the minister and the Dutch national civil servant because it is not unusual for the agent to possess certain skills or to be very specialized. The outlook of the cultural dimension could actually even improve the Principal-agent relationship by improving the loyalty of the Dutch national civil servant towards the minister (Please see below 2.3). If the principal controls the agent too closely, it can have the effect that the agent loses all flexibility and is unable to use his/her expertise for the principal s benefit. If the principal gives clear negotiation mandates, makes his/her priorities clear towards the national civil servant and coordinates between negotiation and implementation, loyalty towards the minister could be improved. The third and last dimension the staff (projection and reception) did not break the Principal-agent relationship between the minister and the Dutch national civil servant. Even when sent to the EU, the Dutch national civil servants are a part of the chain of delegation and accountability in the Netherlands and are therefore still accountable to the minister. European rules, directives and norms penetrating the domestic spheres onto institutions or single actors do not influence the Principal-agent relationship between the minister and the Dutch national 31

33 civil servant in a way to interrupt the relationship. The Principal-agent relationship still connects policy making and implementation in the Netherlands. Europeanisation of the Dutch civil servant influences the Principal-agent relationship between him/her and the minister without exceeding the norm. The Principal-agent relationship between the minister and the Dutch national civil servant is an average Principal-agent relationship with an average Principal-agent problem and Europeanisation of the Dutch national civil servant does not disrupt this relationship Theoretical model and causal mechanism The Dutch national civil servant works on behalf of the minister ( extension of self ) and receives some benefit from the minister (e.g. payment, social recognition, etc.). The benefits of the minister depend on the condition in which the national civil servant performs the tasks in accordance with the wishes of the minister. The minister is in the dilemma that he/she does not know whether the civil servant is motivated to act as to his/her own interest, contrary to those of the minister (Braun/ Guston 2003: p ) (Gailmard 2014: pp ). The minister cannot always ensure that the national civil servant is acting in his/her interest because of the hidden action and hidden knowledge. The civil servant is able to act in his/her own interest because it can turn out to be very costly for the minister to control the civil servant s actions. The minister therefore needs to rely, to some extent, on the loyalty of the civil servant. Whenever a Principal-agent problem between the minister and the Dutch national civil servant arises, will the agent act disloyally (Braun/ Guston 2003: p ). The Dutch national civil servant should feel loyal to the minister in order to act on his/her behalf. This is one of the basic conditions ensuring the principals benefits and a healthy Principalagent relationship between the minister and the Dutch national civil servant and the reason why the feeling of loyalty of the Dutch national civil servant towards his/her own minister will be used as the dependent variable. In the progress of this chapter based on the literature review, the factors influencing the feeling of loyalty of the national civil servant towards his/her own minister will be examined with the main emphasis on the findings of the influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship. 32

34 The first independent variable will be the feeling of loyalty of the Dutch national civil servant towards his/her own policy views. The theory showed that next to the hidden action and hidden knowledge a second condition, the different interests between the minister and the Dutch national civil servant, is required to cause a problem within the Principal-agent relationship and therefore disloyal behaviour. When the Dutch national civil servant has different interests than the minister, he/she is prone to act against the minister s interests. The different interests of the principal and the agent can be manifold, for example, they can have different political views (Braun/ Guston 2003: p ) (Gailmard 2014: pp ). Unfortunately, no ministers took part in the survey or the interview groups which would have given this thesis unparalleled opportunities to research the influence of different interests between the minister and the national civil servant on the feeling of loyalty of the civil servant towards his/her own minister. Some scholars use policy decisions the minister has made in the past, political party election manifestos, etc. to investigate the minister s interests but this author decided to refrain from using such methods because the hidden powers influencing these political decisions would require a master thesis of their own. For this author, it is rather a question of method than one of the outcome. Nevertheless, what is important here is that a difference in the interest between the minister and the Dutch national civil servant has to exist for it to cause different interests in the sense of the Principal-agent problem and with it one condition for disloyalty. Unless there is a difference, the loyalties of the Dutch national civil servant towards his/her own policy views and his/her own minister do not form a contradiction. The Dutch national civil servant has of course his/her own policy views and he/she can be more or less loyal towards their own policy views, but unless these policy views differ from those of the minister, they do not cause any problems. In other words, even when the Dutch national civil servant is loyal towards his/her own policy views, this does not contradict the loyalty of the Dutch national civil servant towards the minister when no other interests exist. The feeling of loyalty of the Dutch national civil servant towards his/her own policy views should therefore not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister negatively. But how then does the independent variable influence the dependent variable not at all or positively? An answer can be gleaned from a theory that, to this author s knowledge, has never been 33

35 used in connection with the Principal-agent theory nor with Europeanization because it comes from the field of economics. According to this theory, two types of loyalty can be distinguished: behavioural and emotional loyalty. Behavioural loyalty is the name given to behaviour that proves loyalty to the brand through sales. It means that the customer behaves in the way the brand wants, i.e. by spending money (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). Applied to the case of this thesis, it would mean that the Dutch national civil servant proves his/her loyalty towards the minister by acting on behalf and in the interest of the minister. Emotional loyalty acknowledges that perceptions and emotions drive behaviours. A customer who is emotionally loyal has empathy and attachment with a brand and company and is more likely to recommend it to potential customers (Chaffey/ Ellis- Chadwick/ Johnston/ Mayer 2009: p. 336). Applied to the case of this thesis, this would mean that the Dutch national civil servant is emotionally loyal, has empathy towards and an attachment with his/her own policy views. The two types of loyalty reinforce each other when they can be achieved together (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). If in the case of this thesis behavioural as well as emotional loyalty exist, then this would mean that the feeling of loyalty of the Dutch national civil servant towards his/her own policy views reinforces the feeling of loyalty of the Dutch national civil servant towards his/her own minister (and vice versa). The feeling of loyalty of the Dutch national civil servant towards his/her own policy views should therefore influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister positively. First hypothesis: The higher the feeling of loyalty of the Dutch national civil servant towards his/her own policy views the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The second independent variable will be the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant. Either, once the Dutch national civil servant is exposed to European integration and additionally exposed to European integration during his/her work, his/her national identity and his/her new build European identity reinforce each other. The Dutch national civil servant then has multiple identities (inclusive national identity) (Hooghe/ Marks 2005: p.422) (Mastenbroek/ Princen 2010: pp ). Or, once the Dutch national civil servant is exposed to European integration and opposes it, the national identity is reinforced. The civil 34

36 servant has an exclusive national identity (Hooghe / Marks 2005: pp ). These group loyalties can be extremely powerful in shaping views towards political objects (Hooghe/ Marks 2005: p.423). Communal identities partially influence the citizens (political) attitudes (Hooghe/ Marks 2005: p.423). In other words, the reshaped identity influences/changes the interests (e.g. different political views) of the Dutch national civil servant. But as discussed above, a difference in interests between the minister and the Dutch national civil servant is one of the two conditions causing a Principal-agent problem and leading to disloyal behaviour as opposed to a simple change in the interests of the Dutch national civil servant (Braun/ Guston 2003: p ) (Jordan 2003: pp ). The minister himself is exposed to European integration and Europeanised himself/herself. His/her identity is influenced in one of the two directions as well and therefore his/her interests are changed as well. The condition of a difference in the interests between the minister and the Dutch national civil servant should therefore only occur if the ministers (principal) interest, for example, expresses EU-scepticism and the national civil servants (agent) interest veers towards EUenthusiasm (or the other way around). The EU is just another interest among many on which the minister and the Dutch national civil servant can disagree. But they have to disagree to fulfil the first condition of a Principal-agent problem. It should not matter whether the Dutch national civil servant is more or less EU-enthusiastic or sceptic. The degree of EUenthusiasm/ scepticism of the Dutch national civil servant should therefore not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. Second hypothesis: The degree of EU-enthusiasm of the Dutch national civil servants will not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The third independent variable will be the Dutch national civil servant s amount of time spent on EU matters. In their structural dimension, Mastenbrook and Princen examined the personnel allocation, distinguished in the EU involvement of the national civil servant, time spent on EU matters by the national civil servant and dispersion of EU-related activities (Mastenbroek/ Princen 2010: pp ). A Dutch national civil servant who becomes more involved with EU matters and spends more time on them will automatically learn more about the EU and its procedures. But it is not unusual for the agent to possess certain skills or to be very specialized. This is, in 35

37 fact, one of the reasons why the agent is so valuable to the principal and why he was chosen in the first place. The minister, on the other hand, will not be ignorant to the new developments. If the Dutch national civil servants under his/her command become or are already deeply involved in EU matters, the minister will not only become more involved because the Dutch national civil servant reports to him/her but because the minister will inform himself/herself as well, as this is the only way to control the civil servants to a certain degree and make sure that there is no Principal-agent problem and his/her interests are acted upon. Whether a Dutch national civil servant spends more or less time on EU matters does not intensify the Principal-agent relationship and does certainly not cause one. The Dutch national civil servant should therefore have the same feeling of loyalty towards the minister, however, his amount of time spent on EU matters might change. Third hypothesis: The amount of time the Dutch national civil servant spends on EU matters, will not influence the Dutch national civil servants feeling of loyalty towards his/her own minister. The fourth independent variable will be the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level. Despite the fact that complete control of the principal over an agent can never be achieved nor is it desirable, the literature examines different mechanisms for the principal to make sure that the agent acts in the principal s interest to achieve as little agency loss as possible (Lane 2013: pp ) (Ennser-Jedenastik 2013: pp ) (Van Thiel/ Yesilkagit 2011: pp ). Two negative examples from the literature are review and oversight, which are carried out too closely. Too strict a review and oversight by the minister can be counterproductive to exerting a certain degree of control. Because, if the minister controls the Dutch national civil servant too closely, it can have the effect of the Dutch national civil servant losing all flexibility and not being able to use his/her expertise to the ministers benefit. Furthermore, much more Principal-agent problems arise and with it disloyalty. If the minister were to give the Dutch national civil servant more freedom to act, the possibility that the Dutch national civil servants and the minister s interests overlap would be much greater with smaller agency losses as a consequence. But too much leeway has its downside as well. Without a clear mandate, the Dutch national civil servants lack the necessary ministerial support to achieve their goals on EU and national level (Schmidt/ Van den Berg (in print): n.p.) and will get frustrated. Whereas 36

38 too close a principal control over the agent should have a negative effect on the loyalty towards the minister, a clear mandate should have a positive effect. The Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level should therefore positively influence the feeling of loyalty the Dutch national civil servant has towards his/her own minister. Fourth hypothesis: The higher the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The fifth independent variable will be the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU. The minister has to rely, to some extent, on the loyalty of the Dutch national civil servant (Braun/ Guston 2003: p ). Loyalty towards the minister is one of the basic conditions to ensure the principal s benefits and a healthy Principal-agent relationship between the minister and the Dutch national civil servant. The question forces itself onto the reader as to whether the feeling of loyalty of the Dutch national civil servant to another person or institution influences his/her feeling of loyalty towards the minister. It seems to be common sense that one person cannot be loyal to two superiors. The theory also used for hypothesis 1 from the field of economics supports these assumptions. The theory said that behavioural loyalty and an emotional one reinforce each other. But in this case, the feeling of loyalty of the Dutch national civil servant towards his/her own minister as well as the loyalty towards the whole political and administrative System of the EU are both behavioural loyalties and neither of which are emotional. Behavioural loyalty was the name given to behaviour that proves loyalty to the brand through sales. It means that the customer behaves in the way the brand wants, i.e. by spending money (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). The Dutch national civil servant proves his/her loyalty towards the minister by acting on the behalf and in the interest of the minister. But as the loyalty towards the whole political and administrative System of the EU is behavioural as well proves the Dutch national civil servant his/her loyalty towards the whole political and administrative System of the EU by acting on behalf and in the interest of the whole political and administrative System of the EU. Both contradict each other because a Dutch national civil servant cannot act on behalf and in the interest of the minister and the whole political and administrative System of the 37

39 EU. The feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU should therefore have a negative impact on the feeling of loyalty of the Dutch national civil servant towards his/her own minister Fifth hypothesis: The higher the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU, the lower the feeling of loyalty of the Dutch national civil servant towards his/her own minister. This part presented the theoretical model and the causal mechanism. The next chapter will present the research design, starting with its first part the key concepts and their operationalization into variables. 3. Research design 3.1. Key concepts and their operationalization into variables The dependent variable is the national civil servant s feeling of loyalty towards his/her own minister and the independent variables are the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU, Degree EU-enthusiasm/ scepticism of the Dutch national civil servant, Dutch national civil servants amount of time spent on EU matters, Dutch national civil servants feeling of a clear mandate for negotiations on EU level from the minister. The control variables are age of Dutch national civil servant, sex of Dutch national civil servant, educational level of Dutch national civil servant and years working as a Dutch national civil servant. Loyalty plays a role in three of the above mentioned variables. Loyalty means the adherence to agreements, compliance with laws or to be devoted to an authority (Schubert 2011: p.165) 1 (e.g. a superior [ ]). Loyalty is the technical term for the product of the generation and procurement of follow readiness (Schmidt 2010: p. 476) 2. 1 L. bezeichnet das Festhalten an getroffenen Vereinbarungen, das Einhalten von Gesetzesvorschriften oder die Treue gegenüber einer Autorität. Translated by author. 2 (z.b. ein Vorgesetzter, [ ]). Fachausdruck für das Produkt der Erzeugung oder Schaffung von Folgebereitschaft Translated by author. 38

40 To measure the first variable feeling of loyalty of the Dutch national civil servant towards his/her own minister, the surveyed persons were asked: When I am involved with EU-related activities described in the previous question, I feel loyal to my minister? They had to locate themselves on a 5 point scale with the lowest number representing disloyalty to the minister and the highest representing loyalty to the minister. The possible answers were: 1. Not at all/ 2. To a limited extent/ 3. Reasonably/ 4. To a high degree/ 5. To a very high degree. 3 The second variable in which loyalty plays a role is the feeling of loyalty of the Dutch national civil servant towards his/her own political views. The surveyed persons were asked: When I am involved with EU-related activities described in the previous question, I feel loyal to my own substantive policy views? They had to locate themselves on a 5 point scale with the lowest number representing disloyalty towards his/her own political views and the highest representing loyalty towards his/her own political views. The possible answers were: 1. Not at all/ 2. To a limited extent/ 3. Reasonably/ 4. To a high degree/ 5. To a very high degree. 4 The third variable including loyalty is the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU which has to be elaborated further. A political and administrative system is a conceptual summary of the leadership-, decision-making- and responsibility functions of the elected political actors (of the government in a narrower sense) and the implementation and application functions of public administration, authorities etc. 5 (BpB 2016c: n.p.). To measure the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU the surveyed persons were asked: When I am involved with EU-related activities described in the previous question, I feel loyal to the whole political and administrative system of the EU? They had to locate themselves on a 5 point scale with the lowest number representing disloyalty towards the whole political and administrative system of the EU and the highest representing loyalty towards the whole political and administrative system of the EU. The 3 Wanneer ik mij bezighoud met EU-gerelateerde activiteiten beschreven in de vorige vraag, dan voel ik mij loyaal aan mijn politiek gezagsdrager (minister)? Answers in column: 1. Helemaal niet/ 2. In beperkte mate/ 3. In redelijke mate/ 4. In hoge mate/ 5. In zeer hoge mate. Translated by author. 4 Wanneer ik mij bezighoud met EU-gerelateerde activiteiten beschreven in de vorige vraag, dan voel ik mij loyaal aan mijn eigen beleidsinhoudelijke opvattingen? 1. Helemaal niet/ 2. In beperkte mate/ 3. In redelijke mate/ 4. In hoge mate/ 5. In zeer hoge mate Translated by author. 5 Begriffliche Zusammenfassung der Führungs-, Entscheidungs- und Verantwortungsfunktionen der gewählten politischen Handlungsträger (der Regierung i. e. S.) und der Umsetzungs- und Durchführungsfunktionen der Verwaltung, Behörden etc. Translated by author. 39

41 Possible answers were: 1. Not at all/ 2. To a limited extent/ 3. Reasonably/ 4. To a high degree/ 5. To a very high degree. 6 To measure EU-enthusiasm or EU-criticism of the Dutch national civil servant a new variable was created degree of EU-enthusiasm/ scepticism of the Dutch national civil servant from questions of the survey. The opposite of EU-enthusiasm is euro [ ] [scepticism]. The term euro sceptical is the collective term that has been used in political discussions for almost two decades for a very wide range of attitudes towards the EU, ranging from selective criticism to categorical rejection. Objects of this euro scepticism can vary greatly: the European Integration project as a whole, the EU in its current structure, concrete decisions as to individual EU policies or specific projects of development (depression and extension). Carriers of such adjustments are primarily political parties. The label "euro sceptical is attributed to parties across the entire political spectrum - not just the parties at the two outermost wings - as long as they show the appropriate settings in their behaviour. As the collective name euro scepticism is blurry and not entirely conclusive, the attempt to redefine the concept and to ask about motives for euro scepticism and thus identify different forms and manifestations of euro scepticism is made. A classification test distinguishes between "hard" and "soft" euro scepticism, the latter stands for temperate-critical opposition and relevant opposition, while the former refers to basic, comprehensive and often militant hostile rejection, which manifests itself in the advocacy to withdraw from the EU. As these contrasting intensity levels of euro scepticism are open to interpretation, the classification is based on crucially subjective assessment and is insufficient as to more precise determination of what euro scepticism means for the specific case and applied to a single political party. Other classifications differ by dimensions or the examined motives for euro scepticism. Special importance must be attributed to ideology, the party s basic beliefs, values and goals; an ideologically stamped euro scepticism would state the incompatibility of their own values with the central features of the EU and its policies. A party may, secondly, be guided by strategic considerations; they would use their euro sceptical position opportunistically to 6 Wanneer ik mij bezighoud met EU-gerelateerde activiteiten beschreven in de vorige vraag, dan voel ik mij loyaal aan het gehele politiek-bestuurlijke stelsel van de EU. 1. Helemaal niet/ 2. In beperkte mate/ 3. In redelijke mate/ 4. In hoge mate/ 5. In zeer hoge mate Translated by author. 40

42 obtain a different profile over other parties in favour of domestic objectives in party competition. A third type of euro scepticism goes back to utilitarian considerations; what economic benefits EU membership has had for the country or what it promises or what disadvantages the membership could bring. A fourth expression of euro scepticism is rooted in a fixation on state sovereignty; the integration project would come up against concerns about an almost indispensable core of this sovereignty and would be questioned, even if only in terms of division of sovereignty. But also the substantial strengthening of community institutions - especially the European Commission and the European Parliament - [ ] would as the representatives this group considers be unacceptable. A fifth reason for euro scepticism is the concern about loss of (national) identity as a result of advancing integration or by decisions on policy areas that are seen as a threat to that identity. Parties which are placed at the extreme left or right of the political spectrum and are euro sceptical go back not only for one single reason but for many reasons. Euro scepticism also characterizes the attitude and policies of individuals located in parties at the centre of the political spectrum, regardless of whether they act as government or opposition parties. Parties and politicians, with the attributed label "euro sceptical" often refuse this characterization with the argument that their attitudes and policies reflect euro realism (BpB 2016a: n.p.). To measure the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant, a new variable was created based on two survey questions and their accompanying answers. If the variable degree of EU-enthusiasm/ scepticism of the Dutch national civil servant had been part of the survey the question asked would have been: Are you an EU enthusiast or an EU critic? The respondents would have had to locate themselves on an 18 point scale with the lowest number representing EU critic and the highest EU enthusiasm. The possible answers were: - 9. EU critic/ -8 to +8/ +9. EU enthusiasm/ 99. Missing. This question did not become part of the survey as the intention behind the question would have entailed the risk of a response bias. The variable was therefore measured in two separate questions. The surveyed persons were first asked: Where do you think the EU is now? They had to locate themselves on a 10 point scale with the lowest number representing fully intergovernmental and the highest representing fully supranational. The possible answers were: Possible answers: 1. Fully 41

43 Intergovernmental/ 2-9./ 10. Fully Supranational/ 99. Do not know or no opinion. 7 This question established a status quo of the respondent s belief as to the state the EU is in at the moment. The second question the surveyed persons were asked was: Where do you think the EU should be in 10 years time? The respondents had to locate themselves on a 10 point scale with the lowest number representing fully intergovernmental and the highest representing fully supranational. The possible answers were: 1. Fully Intergovernmental/ 2-9./ 10. Fully Supranational/ 99. Do not know or no opinion. 8 This second question established what the respondent hopes the EU would be in 10 years time. Combined, they both measure questions regarding the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant. For this purpose, the answer to question two from every respondent was divided by the respondent s answer to question number one. A negative outcome suggests EU-criticism and a positive one EU-enthusiasm. A brief introduction (How the EU functions may theoretically vary between a full intergovernmental EU (where the member states have the power and can veto any policy proposal) and a fully supranational EU (where the EU institutions have the power and the member states cannot veto). Answer the following questions and assign each question a number between 1 and 10. Please take the EU as a whole and not only their own policy sector into consideration. 9 ) was given to the respondents to prepare them for the content of the questions above. The next variable had to be created from the survey as well. The task was to measure the Dutch national civil servant s amount of time spent on EU matters. If the variable Dutch national civil servant s amount of time spent on EU matters had been part of the survey, the question asked would have been: If your work has something to do with the European Union what percentage of your working hours on average do you spend per week on the EU-related activities mentioned in the previous question? The respondents would have had to give a 7 Waar staat de EU volgens u op dit moment? 1. Volledig Intergouvernementeel/ 2-9./ 10. Volledig Supranationaal/ 99. Weet ik niet/geen mening Translated by author. 8 Waar zou de EU volgens u over 10 jaar moeten staan? 1. Volledig Intergouvernementeel/ 2-9./ 10. Volledig Supranationaal/ 99. Weet ik niet/geen mening Translated by author. 9 Hoe de EU functioneert, kan in theorie variëren tussen een volledig intergouvernementele EU (waarin de lidstaten de macht hebben en over elk beleidsvoorstel een veto kunnen uitspreken) en een volledig supranationale EU (waarin de EU-instellingen de macht hebben en lidstaten geen veto-recht hebben). Beantwoord de volgende vragen door voor elke vraag een cijfer tussen de 1 en de 10 op dit spectrum te geven, waarbij u de EU als geheel in ogenschouw neemt, niet alleen uw eigen beleidsterrein. Translated by author. 42

44 percentage between 0.1% and 100% to answer the question. The variable was measured in two separate questions. The surveyed were asked at first: Has your work something to do with the European Union (EU)? The respondents had to choose between the answers yes and no. 10 If the respondent answered the question with no he/she was excluded. If the respondent answered the question with yes the second question asked was: What percentage of your working hours on average do you spend per week on the EU-related activities mentioned in the previous question? The respondents had to give a percentage between 0.1% and 100% to answer the question. 11 If somebody answered question one with yes his/her answer to question two became the value of the newly established variable. Combined, they both measure questions regarding the Dutch national civil servant s amount of time spent on EU matters. The last variable used Dutch national civil servants feeling of a clear mandate for negotiations on EU level from the minister. Having a mandate means the authorization to carry out certain interests or accomplish specific tasks to another person, organization, etc. 12 (BpB 2016b: n.p.). To measure the variable Dutch national civil servants feeling of a clear mandate for negotiations on EU level from the minister the surveyed persons were asked: When I participate in EU-level meetings, I receive a clear negotiation mandate? They had to locate themselves on a 5 point scale with the lowest number representing their total disagreement of receiving a clear mandate for negotiations on EU level and the highest representing total agreement. The possible answers were: 1. Totally disagree/ 2. Mostly disagree/ 3. Neither agree nor disagree/ 4. Largely agree/ 5. Fully agree. 13 A brief introduction (The following statements are about how your employer supports EU-related work. This can include career counselling and guidance. To what extent do you agree or disagree with the following 10 Heeft u in uw werk te maken met de Europese Unie (EU)? 1. Nee 2. Ja Translated by author. 11 Hoeveel procent van uw werktijd besteedt u gemiddeld per week aan de in de vorige vraag genoemde EUgerelateerde activiteiten? % Scripter: min. 0.1, max. 100 Translated by author. 12 Die (vertraglich festgelegte) Vollmacht zur Wahrnehmung bestimmter Interessen oder zur Erledigung bestimmter Aufgaben einer anderen Person, Organisation etc. Translated by author. 13 De onderstaande stellingen gaan over de wijze waarop uw werkgever het EU-gerelateerde werk ondersteunt. Dit kan bijvoorbeeld loopbaanbegeleiding en aansturing zijn. In welke mate bent u het met de volgende stellingen eens of oneens? Als ik deelneem aan overleggen in EU-verband beschik ik over een duidelijk mandaat voor de onderhandelingen. 1. Volstrekt oneens/ 2. Grotendeels oneens/ 3. Noch mee eens, noch mee oneens/ 4. Grotendeels mee eens/ 5. Volledig mee eens Translated by author. 43

45 statements? 14 ) was given to the respondents to prepare them for the content of the above question. The control variables (Age of Dutch national civil servant, Sex of Dutch national civil servant, Educational level of Dutch national civil servant and Years in Government) were operationalized by asking the following questions and giving the corresponding answers. Age of Dutch national civil servant: What is your age? Possible answers: years old Sex of Dutch national civil servant: What sex are you? Possible answer: 1. Male/ 2. Female Educational level of Dutch national civil servant: What is the highest degree or level of education you have completed? Possible answer: 0. No schooling completed/ 1. Nursery school to 8th grade/ 2. Some high school, no diploma/ 3. High school graduate, diploma or the equivalent (for example: GED)/ 4. Some college credit, no degree/ 5. Trade, technical, vocational training/ 6. Associate degree/ 7. Bachelor s degree/ 8. Master s degree/ 9. Professional degree/ 10. Doctorate degree Years working as a Dutch national civil servant: How many years have you worked as a national civil servant? Possible answer: years 14 De onderstaande stellingen gaan over de wijze waarop uw werkgever het EU-gerelateerde werk ondersteunt. Dit kan bijvoorbeeld loopbaanbegeleiding en aansturing zijn. In welke mate bent u het met de volgende stellingen eens of oneens? Translated by author. 44

46 Table 1 List of variables Variable Type Dependent variable Independent variables Dimension Variable Survey item(s) National Loyalty Loyalty to own policy views EUenthusiasm/ scepticism Feeling of loyalty of the Dutch national civil servant towards his/her own minister Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil servants feeling of the present degree of supranationalism/ intergovernmentalism in the EU Dutch national civil servants wish of the degree of supranationalism/ intergovernmentalism in the EU in 10 years. A03Q4 A03Q9 EUEnt A04Q1 A04Q2 When I am involved with EU-related activities described in the previous question, I feel loyal to my political minister? When I am involved with EU-related activities described in the previous question, I feel loyal to my own substantive policy views? The variable was not in the survey and is a result of the subtraction of Variables A04Q2 and A04Q1 (Please see below). Where do you think is the EU now? Where do you think should the EU be in 10 years? 45

47 Control Variables Individuals amount of time spent on EU matters Clear negotiation mandate EU Loyalty Age servant civil Dutch national civil servants amount of time spent on EU matters DTSEU The variable was not in the survey and includes only civil servants who answered the question regarding variable A01 with yes (2.) and who answered the question regarding variable A02QPC (Please see below). EU related work A01 Has your work something to do with the European Union (EU)? % of EU related work A02QPC What percentage of your working hours on average do you spend per week on Dutch national civil servants feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU Age Dutch national civil servant A07Q3 A03Q1 Age the EU-related activities mentioned in the previous question? To what extent do you agree or disagree with the following statements? When I participate in the EUlevel meetings, I receive a clear negotiation mandate. When I am involved with EU-related activities described in the previous question, I feel loyal to the whole political and administrative system of the EU? What is your age? 46

48 Sex civil servant Educational level Years working as a Dutch national civil servant. Sex Dutch national civil servant Educational level of Dutch national civil servant Years working as a Dutch national civil servant Sex Edu What sex are you? What is the highest degree or level of school you have completed? YiG How many years have you worked as a national civil servant? Source: Author s table (1) based on Mastenbroek/ Princen/ Van den Berg (in print): n.p. The interviews were operationalized in the same way as the survey, to achieve maximum comparability. The interviewee were asked six questions for this master thesis. Each question was supposed to measure one of the variables mentioned above: 1. How would you describe your relationship with your minister and has this relationship changed since the EU plays an important role in many areas of politics? -> Feeling of loyalty of the Dutch national civil servant towards his/her own minister 2. Are there differences in political views between you and your minister, if so in what way do your political views differ from your minister regarding the EU? -> Feeling of loyalty of the Dutch national civil servant towards his/her own policy views 3. What are your views on the European Union with regards to intergovernmentalism and supranationalism, do you prefer more or less EU and why? -> Degree EU-enthusiasm/ scepticism of the Dutch national civil servant 4. Do you receive a clear mandate for your work from your superior or are you free to adopt your own policy position as you see fit? -> Dutch national civil servants feeling of a clear mandate for negotiations on EU level from the minister 47

49 5. Do you sometimes experience a conflict in loyalty to intergovernmental or supranational institutions? -> Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU 6. When you work, where do you feel that your loyalty lies? -> Question to connect interview questions 1, 2 and 5 about loyalty of the Dutch national civil servant and to bring them in an order. Unfortunately had this author to concentrate in the interviews on the most important variables due to time issues, which made him prioritize. Therefore were the variable Dutch national civil servants amount of time spent on EU matters and all control variables not part of the interviews. But as interviewee are free to answer the question the way they like, it sometimes happened that an interviewee talked about one of the other variables he or she was actually not been asked about. In order not to lose such valuable information, whenever an interviewee used a key word (e.g. clear mandate) or a corresponding expression of any of the used variables in the answers to any of the six questions, was the appropriate part of the answer dedicated to the corresponding variable. Furthermore was a category system developed using the same operationalization as for the survey. This means: To measure for example the variable feeling of loyalty of the Dutch national civil servant towards his/her own minister, the surveyed person was asked: When I am involved with EU-related activities described in the previous question, I feel loyal to my minister? The possible answers were: 1. Not at all/ 2. To a limited extent/ 3. Reasonably/ 4. To a high degree/ 5. To a very high degree. In the interviews they were asked: How would you describe your relationship with your minister and has this relationship changed since the EU plays an important role in many areas of politics? And when you work, where do you feel that your loyalty lies? The answer the Dutch national civil servant gave in the interview was then, when the interview answer matched the measured variable or any other of the variables, classified to one of the categories of the survey question (e.g. 1. Not at all for the survey question: When I am involved with EU-related activities described in the previous question, I feel loyal to my minister?). The benefit of this procedure was that it was possible to compare the answer which a Dutch national civil servant gave in the survey and in the interview. But 48

50 the downside of this categorization system would have been that a lot of additional interview information would have been lost as interviews are about context and subjective meaning and not about dry numbers (e.g. 1. Not at all). Therefore was the categorization system intensively evaluated. This means, if the answer of the interviewee was categorized as for example 1. Not at all to the survey question When I am involved with EU-related activities described in the previous question, I feel loyal to my minister? and with it to the corresponding variable, the entire part of the interviewee s answer was allocated to the category and not just a value (e.g. 1. Not at all). This way it was possible on the one hand to compare interview and survey values and on the other hand to use the additional information of the interviews about the variables. The next part will discuss the method of analysis and explain why the method was chosen Research approach and design Generally, it is possible to differentiate between experiments and ex-post facto designs. In experiments, the conditions of the examination are self-induced which ensures that they are under the control and the exertion of the scientists. The ex-post facto design can be about a questioning, an observation or a content analysis (Gehring/ Weins 2009: pp , 70-91). As the last two methods were not suitable for this master thesis, because both were not able (or only under high costs and effort) to measure all the variables of this thesis, were questionings used. To be more precise two different kind of questionings were used: quantitative survey and qualitative interview. While qualitative research methods (e.g. interview) are more about understanding motivations and basic reasons, are the benefits of quantitative research methods (e.g. survey) to be able to generalize the findings from the statistically valid random sample onto the population (if the sample is big enough) (Gehring/ Weins 2009: pp ) (Schumann 2006: pp ). The analysis of the survey is a secondary analysis and was conducted for a similar purpose (Gehring/ Weins 2009: pp , 70-91). The large number of cases collected in the survey suggest a Large-N design. Common Large-N designs are time series, cross-sectional designs, panel designs and multilevel designs. The survey with this particular group of variables has only been conducted once in 2015 which disqualifies the time series design, panel design and 49

51 multilevel design. Cross sectional designs are characterized by comparing a large number of units (Dutch national civil servants) at one point in time (Toshkov 2016: pp ). Usually, several variables are collected at once from all the units (Baur/ Behnke/ Behnke 2010: pp ). The causal effect is identified by comparing the distribution of the outcome across the sub-sets of units defined by a particular value of the main explanatory variable (Toshkov 2016: p. 231). The usual aim is to ensure the capability of making assumptions about the change in the dependent variable when the independent one of a unit changes (Toshkov 2016: pp ). Even when such a connection is found, there should not be any doubt about the fact that causality in Ex-post facto designs can never be proven empirically (Gehring/ Weins 2009: pp ). Although the control of the third variables and the time sequence of the variables suggest a causal interpretation of a statistical connection (Gehring/ Weins 2009: pp. 27) 15. The second research approach will be interviews. In qualitative research, there are a number of interview techniques with as many different names. The advantage of interviews is that the research subjects speak for themselves (Mayring 2002: p.66) and reconstruct active events, report about experiences and knowledge (Bohnsack 2003: p.95). The qualitative methods share the principle of openness, the principle of communication and the consideration of the researcher s subjectivity. The principle of openness is defined as the degree of the freedom of the respondent. Unlike the closed form, the interviewee can answer freely and is not bound to pre-defined answer formats (Mayring 2002: p 66). Important basic forms of qualitative interviews are the narrative interview and the guided interview. The narrative interview is a nonstandard type. The interviewer has no fixed list of questions, which motivates the interviewee to explain freely. The interviewer gives a short introduction into the topic and allows the respondent to explain freely from there on. The free style of explaining during the interview uncovers subjective meaning that would not come to light through systematic questioning. Inquiries to resolve open issues are only allowed after the actual interview (Mayring 2002: p.72). A guided interview operates without response options from a classical quantitative questionnaire, but is based on an interview guide. The guide consists of questions that largely 15 eine kausale Interpretation eines statistischen Zusammenhangs Translated by author. 50

52 ensure the respondent s freedom in their utterances. This way, the guide keeps the narrative character of the interview but also ensures that all the key issues will be addressed in the interview. In addition, the interviewer can latch onto the answers and expand important aspects with questions. For this reason the technique is referred to as semi-standardized (Bohnsack 2004: p.114). This master thesis will use the guided interview because it has two advantages over the narrative interview: All important aspects will be addressed and comparability with the other interviews is reached. The fact that all aspects will be addressed was especially important because the interviews were conducted in cooperation by a group of 5 students of which each had his/her individual master thesis topic. In other words, although each interview covers the main topic The Europeanisation of Dutch officials, each interview consists of five different subtopics (Please see below). A narrative interview with its unstandardized structure could not cover all five subtopics. The kind of interview used in this master thesis is a derived form of a guided interview (problem centred interview) (Mayring 2002: p.67). The problem centred interview is centred on a particular (social) problem ("The Europeanisation of Dutch Officials"), which has previously been objectively illuminated by the researchers. The problem-centred interviews are particularly suitable in the more theory-driven research, which deals with specific issues (Mayring 2002: p.71) Measurement and data source Two major data sources were used to make this master thesis: survey and interviews. The survey was conducted between 12 November 2015 and 2 December 2015 on the subject The Europeanisation of Dutch officials. The questionnaire used in the survey was developed by the Ministry of the Interior & Kingdom Relations and Leiden University and monitored by the Society for Consumer Research (GfK). The questionnaire comprises a maximum of 14 questions. The time to answer the questionnaire is approximately 6 minutes. The questionnaire was sent to 9109 Dutch officials in Europe. The Dutch officials worked for the Empire, within the Judiciary, municipality, for the provinces, on the water boards, communal regulations and Research Institutions. Depending on the answers given the minimum length 51

53 of the questionnaire was 7 questions and the maximum length 14 questions. Additional background characteristics were collected the age of the respondent, his or her sex, the education of the respondent and how many years he or she had been working within government at the time of the survey Dutch officials completed the questionnaire; that is a response rate of 38.7%. The interviews were conducted between 17 May 2016 and 27 May 2016 on the subject The Europeanisation of Dutch officials. The questionnaire used for the interviews was developed by Arsjalus Amirchanian, Giulia Francaviglia, Bjorn Oud, Sevgi Yilmaz and the author. The questionnaire comprises 20 questions. Each of the students developed 3-6 questions in view of their own master thesis. They were combined into a complete questionnaire by dividing some of the original questions to ensure that the interviewee would not be overwhelmed and bringing them into a logical order. The time per interview amounted to approximately minutes. The invitations for the interviews were developed by the 5 students and sent to 13 selected Dutch civil servants, who had already taken part in the survey and left their address in the survey to be contacted for further questioning. In the run up to the interviews, the interviewee had to choose whether he/she would prefer to conduct the interview face to face at a location of their choosing, via mobile phone or via skype. The interviews were always conducted by at least one of the above mentioned 5 students with (if possible) one Dutch student among them. 5 interviewees chose face to face, 1 mobile phone and 2 skype interviews. Each interviewee had to fill out an interview consent form before the interview could take place. The interviews were electronically recorded and afterwards transcribed according to the stipulated transcription guidelines. 8 Dutch officials gave an interview which equates to a response rate of 61.54%. Dr. van den Berg of the Leiden University was the initiator of the master thesis project on the subject The Europeanisation of Dutch officials and the master thesis supervisor of the 5 students. In the next part, the relevant population of the survey and the interviews are going to be defined and the case selection is explained. 52

54 3.4. Defining the relevant populations, unit of analysis and case selection The population can be defined as all the cases or observations covered by the hypothesis; all the units of analysis to which a hypothesis applies (Johnson/ Reynolds 2005: p. 268). The population in this master thesis is similar for the survey and for the interviews. The findings are applied to all Dutch national civil servants who work in a principal-agent relationship with a minister (working for the empire). The unit of analysis, on the other hand, has to be distinguished for both methods. It consists of all the units from which the sample is actually selected (Baur/ Behnke/ Behnke 2010: pp ). The unit of analysis for the survey is every Dutch national civil servant working for the Empire. The reason for no civil servant from within the Judiciary, municipality, provinces, water boards, communal regulations and Research Institutions being part of the unit of analysis is that they do not (necessarily) work in a Principal-agent relationship with a minister. For the interviews, the unit of analysis was smaller because in order to control the findings of the survey only Dutch national civil servants working for the empire and participating in the survey plus providing their address in order to be contacted for further interviews were part of the unit of analysis. The case selection has to be differentiated as well. The survey selected all elements of the population and became a comprehensive survey. The advantage of this survey was that it was conducted in conjunction with the Ministry of the Interior & Kingdom Relations. Therefore, the contact information of the Dutch national civil servants working for the Empire was available and all of them were invited to participate in the survey (Baur/ Behnke/ Behnke 2010: pp ). The case selection for the interviews has to be elaborated somewhat further. The interviewees were selected from the group of Dutch national civil servants working for the empire and participating in the survey plus providing their address in order to be contacted for further interviews. But qualitative sampling is about doing a criteria based and considered (this means not by pure chance) case selection. Therefore was the selection of the interviewees done by extreme groups (special cases) and modal categories (common cases) of all the operationalized variables for the interviews. The aim is not to achieve statistical representativeness, but to achieve a maximum variation of cases (Kelle/ Kluge 1999: pp ) (Baur/ Behnke/ Behnke 2010: pp ). To give an example, the variable feeling of loyalty of the Dutch national civil servant towards his/her own policy views 53

55 has a range of 1-5. It was therefore tried (if possible) to select on the one hand interviewee with the value of 1 and 5 (extreme groups) and on the other hand interviewee with modal categories (e.g. the most frequent value of the variable feeling of loyalty of the Dutch national civil servant towards his/her own policy views was 3 therefore were interviewee with the value 3 selected). Drawing criteria and sample size of the interviews were determined after the survey and the data was not analysed until after the survey Common threats to inference Unfortunately but understandably, the survey used in this thesis was not specially tailored to examine the topic at hand but covered rather a large variety of different topics surrounding the Europeanisation of Dutch officials. The survey enabled many different topics to be researched and is a gain to the entire field as reliable current statistical data on national civil servants is rare. But this variety prevented this author, on the other hand, from being able to use all the questions asked in the survey. Furthermore, the connections between the master thesis and some of the survey questions might seem a bit constructed. The reason is that the survey was conducted before the master thesis had been started and the author was not able to influence the questions. Nevertheless, this author would not have used this survey had he not been one hundred per cent sure that the survey was up to the task at hand. Another threat to inference could follow from the fact that the survey is very short with only 14 questions. To completely cut out any response bias (incorrect response by the respondent, arbitrary tick, approval trend, social desired answers or the question wording is misleading or confusing) most surveys include questions that ensure that people with response bias are identified and can be excluded afterwards. In the short survey, there was almost no room for such questions. The reason for this is that this survey attached great importance to the time needed to finish the survey. If a respondent needs less time to finish the survey, he/she is more likely to participate in the first place (higher response rate) and to finish the survey completely and not stop half way through (Schumann 2006: pp ) (Toshkov 2016: pp ). At first glance, the response rate of 38.7% seemed to be very low, considering that the Ministry of the Interior & Kingdom Relations to the Society for Consumer Research (GfK) helped to develop the survey and that it was done with their blessing. Two explanations are possible. Either the respondents only perceived that Some company was doing a survey. Or 54

56 many civil servants decided not to respond at all or only answer some questions in order to avoid extra work or some disadvantages if they answered honestly (Schumann 2006: pp ). The next threat would be if the findings in this master thesis were applied to national civil servants from other EU member states working for the empire. There could be country specific differences that have to be taken into account. The findings in this thesis should not be applied to civil servants in other EU member states without additional adaption (Schumann 2006: pp ). Furthermore, the addresses of their employees were delivered from the Ministry of the Interior & Kingdom Relations to the Society for Consumer Research (GfK), which conducted the survey. Depending on the time between the delivering of the addresses and conducting of the survey this could have led to under coverage (if in the meantime new civil servants have been employed) or over coverage (if in the meantime civil servants stopped working for the as civil servants) (Schumann 2006: pp ). A number of threats come from the quantitative nature of the research. The problem with quantitative research like the survey is that it is dominated by the researcher. The researcher dictates the procedure and the respondents are only allowed to react according to the researcher s plan. Furthermore, the situation in quantitative research is artificial. A free portrayal of views as in "ordinary dialogue" is not possible. Additionally, the focus on the examination is rigid. All social phenomena outside the question/answer grid are "hidden". The research is "blind" to phenomena outside the specifications and hypotheses (Diekmann 2007: pp ). But the qualitative research like the conducted interviews have disadvantages as well which can influence inference. The selection of the sample is done through a deliberate selection of cases according to controlled criteria. In other words, no random sample is used which makes the generalization of the findings difficult (Diekmann 2007: pp ). Furthermore did unfortunately only Dutch national civil servants agree to participate in the interviews, who had a high value (4-5) of the variable the Dutch national civil servant s feeling of loyalty towards his/her own minister (complete value range was between 1-5) and additionally took 55

57 no Dutch national civil servants part in the interviews with a high (+9 to +7) and none with a low (-9 to -7) degree of EU-enthusiasm (complete value range between (-9 to +9). Also, "measurement errors", interviewer effects, self-fulfilling prophecy effects, halo effects, social desirability can become a problem in qualitative interviews (Diekmann 2007: pp ). The last threat to inference which can be caused by using the conducted interviews is connected to the data anlysis. When qualitative interviews are conducted, large volumes of data accure with each interviewee. To utilize them and to not loose information in the process can be quite a challenge. It is important to garuantee evaluation objectivity (Diekmann 2007: pp ). The key to assure quality and inference in this master thesis is the mix of methods. By using a survey (quantitative) and interviews (qualitative) the weaknesses of both methods are minimized and the findings can be double checked. 4. Data collection and findings This chapter will present the data collection and findings. An increasing approach will be used, meaning that this chapter starts with basic findings and ends with the complete tested model. All the relevant variables, how often they have been observed in the survey and their minimum and maximum values in the survey are presented in table 2. The first thing that meets the eye is the huge difference in the number of observations. The range measured is between 370 and 1394 observations. Whereas the control variables and the degree EUenthusiasm/ scepticism of the Dutch national civil servant were answered by almost all 1394 Dutch national civil servants working for the empire and part of this examination, the three variables regarding feelings of loyalty towards own minister or own policy views or the whole political and administrative System of the EU were only replied to 370 times. One reason could be that Dutch national civil servants are more comfortable answering how long they have been working for the government rather than where their loyalties lie because they fear reprisals. Nevertheless, the number of observations for each variable suffices for reaching reliable results. The second remarkable observation is that the extremes (min. and max. 56

58 values) of each variable (except age) were chosen by at least one Dutch national civil servant. The sample of the survey seems to represent a wide range of opinions. Table 2 Descriptive statistics of all relevant variables Variable Observations Minimum Maximum Feeling of loyalty of the Dutch national civil servant towards his/her own minister Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil Servant s amount of time % 100% spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU Age Dutch national civil servant Sex Dutch national civil servant Educational level of Dutch national civil servant Years working as a Dutch national civil servant Source: Author s table (2) based on Stata output of the survey examination. The next part analyses the measure of central tendency. 57

59 4.1. Measure of central tendency The arithmetic mean is one of the statistical ratios, which characterizes the central tendency of a distribution (Gehring/ Weins 2009: p ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The arithmetic mean is the average value of all the statistical units (e.g. persons). The ability to calculate the arithmetic mean requires the data to be at least interval scaled (Gehring/ Weins 2009: p.126). 16 The arithmetic mean is calculated by adding all characteristic values of all statistical units and then dividing them by the number of statistical units. The arithmetic mean has two qualities that will be used for calculations in the following parts (e.g. measure of variation) and that are the reason why the arithmetic mean was chosen from the many measures of central tendency (Gehring/ Weins 2009: p ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The first is that the sum of the deviation of all measured values from the arithmetic mean is 0. [ ] [The second is that] the sum of all squared deviations of all measured values from the arithmetic mean [ ] is minimal. [ ] [Minimal means], in the calculation of the squared deviations of the measured values taken from any other value, the result would in any event be larger than in the event using the arithmetic mean (Gehring/ Weins 2009: p.128). 17 The arithmetic mean of each variable can be found in table 3. It is noticeable that the feeling of loyalty of the Dutch national civil servant towards his/her own minister with 3, 61 on a scale of 1-5 has the highest value of the three loyalty variables. The degree of EU-enthusiasm/ scepticism of the Dutch national civil servant is almost 0 on a scale of +9 to -9. About the average of Dutch national civil servant working for the empire can be said to be male, 51 years old, holding an associate degree, having worked as an national civil servant for 3, 5 years and spending 21 % of his work time on EU matters (Gehring/ Weins 2009: p ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The next part will examine the measures of variation. 16 Das arithmetische Mittel ist der Wert, den alle Merkmalsträger, also z.b. Personen, im Durchschnitt aufweisen. Die Berechnung des arithmetischen Mittels setzt mindestens intervallskalierte Daten voraus. Translated by author 17 Die Summe der Abweichungen aller Messwerte vom Mittelwert ist 0. [ ] [Die zweite ist], die Summe der quadrierten Abweichungen aller Messwerte vom Mittelwert [ ] ist minimal. [ ] [Minimal bedeutet], dass bei der Berechnung der quadrierten Abweichungen der Messwerte von irgendeinem anderen Wert das Ergebnis auf jeden Fall größer wäre als bei Verwendung des arithmetischen Mittels. Translated by author 58

60 Table 3 Measurement of central tendency and measure of variation Variable Mean Standard deviation Feeling of loyalty of the Dutch national civil servant towards his/her own minister Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil Servant s amount of time spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU Age Dutch national civil servant Sex Dutch national civil servant Educational level of Dutch national civil servant Years working as a Dutch national civil servant Source: Author s table (3) based on Stata output of the survey examination Measure of variation The standard deviation is one of the statistical ratios characterizing the divergence of the measurement values. The arithmetic mean describes a distribution more aptly when the measurement values are closer together. Therefore, the standard deviation should be collected and taken into account. The standard deviation is calculated in two steps. First, the 59

61 sum of the squared deviation of all measurement values from the arithmetic mean is divided by the number of statistical units. And second, the square root is taken. The advantage of the standard deviation is that it has its original unit of measurement, which simplifies examining the size of the variation (Gehring/ Weins 2009: p. 120, ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The standard deviation of each variable can be found in table 3. What sticks out in a negative sense are the standard deviations of the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant with 3.04 values of its scale, the Age Dutch national civil servant with 8.03 years, the Educational level of Dutch national civil servant with 2.49 of its scale and especially the Dutch national civil Servant s amount of time spent on EU matters with 27%. The larger the standard deviation the worse the arithmetic mean, which means that the values including the variable are not very homogenous. The age of the Dutch national civil servant, for example, with an arithmetic mean of 51 years has a standard deviation of 8.03 years. The measured values of the variable deviate on average by + 8 years or - 8 years (Gehring/ Weins 2009: p. 120, ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). Having learned much about each variable itself, the next part the statistical correlation will start to tackle the relationships between the variables Statistical correlations The strength but not the direction of a connection between two variables is indicated by correlation measures. Which appropriate correlation measure out of several is chosen depends on the scale level. Two different correlations will be used in this thesis, the productmoment correlation (Pearsons r) and eta-square. Pearsons r measures the strength of the linear correlation between two variables, which have at least interval scale. It captures only the linear correlation between two variables and has a range of +1 to -1. If Pearsons r assumes the values of +1 or -1 all measured values are arranged in a straight line. If Pearsons r assumes the value 0 there is no linear correlation (Gehring/ Weins 2009: p ). To measure the correlation between an independent nominal variable and a dependent (at the least) interval scaled variable, eta-square is used. Eta square is a proportional reduction 60

62 in error measure (PRE). The value of the PRE measure eta square represents the extent by which the values of the dependent variable can be forecasted by taking the independent variable into account. Eta square has a value range of +1 to represents a perfect correlation and 0 no correlation. If eta square is multiplied by 100, the result can be interpreted as a percentage. By adding the knowledge about the independent variable, the dependent variable can be predicted at percentage value (Gehring/ Weins 2009: p ). The correlation value between the feeling of loyalty of the Dutch national civil servant towards his/her own minister to each other variable (independently) can be found in table 4. The correlation values have to be split into three groups according to their significance. Group one are those with a significance higher than 0.05, they are not statistically significant from zero (Degree EU-enthusiasm/ scepticism of the Dutch national civil servant; Dutch national civil Servant s amount of time spent on EU matters; Age Dutch national civil servant; Years working as a Dutch national civil servant) and the correlations can therefore not be taken into account. The second group comprises those variables with a significance smaller than 0.05 but higher than 0.01; they are statistically significant from zero on a 5% level but not a 1% level (Sex of the Dutch national civil servant). The last group is made up of variables with a significance smaller than 0.01; they are statistically significant from zero on a 5% level and even on a 1% level (Feeling of loyalty of the Dutch national civil servant towards his/her own policy views; Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister; Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU; Educational level of Dutch national civil servant). A very small correlation between Sex of the Dutch national civil servant and the feeling of loyalty of the Dutch national civil servant towards his/her own minister of is statistically significant at a 0.05 level but not on a 0.01 level. Furthermore, the correlation value of the PRE measure eta² can be interpreted as a percentage value. By adding the information of the Sex of the Dutch national civil servant the values of the feeling of loyalty of the Dutch national civil servant towards his/her own minister can be predicted with 0.7 %. The correlations between the feeling of loyalty of the Dutch national civil servant towards his/her own minister and the feeling of loyalty of the Dutch national civil servant towards his/her own policy views of 0.27 (the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister of 0.31; the feeling of loyalty of the Dutch 61

63 national civil servant towards the whole political and administrative System of the EU of 0.42; Educational level of Dutch national civil servant of 0.15) is statistically significant at a 0.05 level and even at a 0.01 level. None of the correlations has a value of over 0.5. The highest value has the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU with 0.42 and the lowest the educational level of the Dutch national civil servant with Pearsons r has a range of -1 to and +1 represent all the values arranged in a straight line whereas 0 represents no linear correlation. The correlations between the feelings of loyalty of the Dutch national civil servant towards his/her own minister and each of the significant variables are rather small, which means that there is no straight line of the values. The question as to which variable influences which value remains. In other words, does the feeling of loyalty of the Dutch national civil servant towards his/her own minister influence each of the significant variables (independently) or is it the other way around? This question will be answered in the next part, linear regression with dummy variables (Gehring/ Weins 2009: p ). 62

64 Table 4 Statistical correlation Variable (independent or dependent) Correlation dependent and independent (Pearsons R or Eta²) Feeling of loyalty of the Dutch national civil servant X towards his/her own minister (dependent Variable) Significance X Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil Servant s amount of time spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU Age Dutch national civil servant Sex Dutch national civil servant (Eta²) Educational level of Dutch national civil servant Years working as a Dutch national civil servant

65 Source: Author s table (4) based on Stata output of the survey examination Linear regression with dummy variables A bivariate linear regression makes it possible to explain the variation in a dependent variable as a consequence of variation in an independent variable. In contrast to the calculation of a correlation, as a measure of association for metric data, the bivariate linear regression makes it possible to judge the effect of one variable on another by establishing a direction of the relationship between the two variables. Furthermore, the bivariate linear regression assumes a cause-effect relationship. Therefore, changes in one variable (dependent variable) can be attributed to changes in another variable (independent variable). Here, the linearity assumption of the regression analysis refers to the linearity of the parameters - in other words, if the independent variable increases by one unit, the dependent variable changes (linearly) by the regression coefficient. The bivariate linear regression analysis is also a forecasting tool. It estimates an explanatory model in view of the available data and allows for forecasts of expected values of the dependent variable for different observations of the independent variable (Gehring/ Weins 2009: pp ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The findings of the bivariate linear regressions can be found in table 5 below. In each case, according to the hypothesis, the influence of one different independent variable (or control variable) was examined as to the dependent variable the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The independent variables are the feeling of loyalty of the Dutch national civil servant towards his/her own policy views; the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant; the Dutch national civil servant s amount of time spent on EU matters; the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister and the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU. The control variables are the age of the Dutch national civil servant; the sex of the Dutch national civil servant; the educational level of the Dutch national civil servant and the years working as a Dutch national civil servant. The sex of the Dutch national civil servant is a dummy variable, namely a useful extension of the regression model for the bivariate linear as well as the multiple one. Dummy variables are variables which have only two characteristic values (usually 0 and 1). This makes it possible to define a group membership 64

66 for all observations as "true" (1) and "false" (0). Therefore, dummy variables can only increase once by one value. The one-time increase of the dummy variables describes nothing more now than the "group effect" - the change in the average rating of the explanatory variable when it is changed from the reference group (0) to the peer group (1). Dummy variables can be used as independent variables, but not as the dependent variable (Kühnel/ Krebs 2012: pp ). The variable sex of the Dutch national civil servant has the two characteristics male and female, which will allow this author to differentiate between both groups. 65

67 Table 5 Linear regression with dummy variables Feeling of loyalty of the Dutch national civil servant towards his/her own minister R² Coefficient Std. Error P > t 95 % Conf. Int. (Min & Max) Feeling of loyalty of the Dutch national civil servant towards his/her own policy views to Degree EU-enthusiasm/ scepticism of the Dutch national civil servant to Dutch national civil servant s amount of time spent on EU to matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from to the minister Feeling of loyalty of the Dutch national civil servant towards to the whole political and administrative System of the EU Age Dutch national civil servant to Sex Dutch national civil servant to

68 Educational level of Dutch national civil servant Years working as a Dutch national civil servant to to Source: Author s table (5) based on Stata output of the survey examination. The t-test of the regression coefficient of the independent variables has to be split into three groups for their p-values (p > t ). Group one comprises those with an alpha higher than 0.05, they are not statistically significant from zero (the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant; the Dutch national civil servant s amount of time spent on EU matters; the age of the Dutch national civil servant; the sex of the Dutch national civil servant and the years working as a Dutch national civil servant). Since for p > t > 0.05 the null hypothesis cannot be discarded. If the regression coefficient of an independent variable is not statistically significant, like in these cases, then the regression coefficient is not valid for the entire population. The second group comprises those variables with an alpha lower than 0.05 but alpha higher than 0.01; they are statistically significant from zero at a 5% level but not a 1% level (None of the variables used in this case.). The last group is made up of those variables with an alpha smaller than 0.01; they are statistically significant from zero at a 5% level and even a 1% level (the feeling of loyalty of the Dutch national civil servant towards his/her own policy views; the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister; the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU and the Educational level of Dutch national civil servant) (Gehring/ Weins 2009: pp ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). Therefore, the deviations of the regression coefficients of the feeling of loyalty of the Dutch national civil servant towards his/her own policy views with around 0. 27, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister with around 0. 27, the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU with around and the Educational level of 67

69 Dutch national civil servant with around from zero are statistically significant at alpha = 0.05, and even at alpha = 0.01 level. In all four cases, the regression coefficient is positive. Exemplary interpretation of the prediction for the feeling of loyalty of the Dutch national civil servant towards his/her own policy views: If for a Dutch national civil servant the feeling of loyalty towards his/her own policy views increases by one unit on the scale of the variable feeling of loyalty of the Dutch national civil servant towards his/her own policy views (becomes more positive), the civil servant s feeling of loyalty towards his/her own minister increases (improves) by around times of a category on the scale of the feeling of loyalty of the Dutch national civil servant towards his/her own minister (Gehring/ Weins 2009: pp ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The exemplary interpretation can be applied to the other three significant variables with their corresponding regression coefficients. The 95% confidence interval specifies the interval within which the true (but unknown) regression coefficient of the population with a probability of 0.95 (95%) lies. The confidence interval is therefore a measure of the accuracy of the estimate: The narrower the interval the more accurately the (unknown) true value of the regression coefficient can be specified. The findings in this case seem to be relatively accurate: With a high level of security, the confidence interval is moving in the range between 0.17 to 0.37 (for the feeling of loyalty of the Dutch national civil servant towards his/her own policy views), 0.19 to 0.36 (for the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister), 0.33 to 0.51 (for the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU) and 0.11 to 0.56 (for the educational level of Dutch national civil servant) for the unknown regression coefficients of the population of Dutch national civil servants. In reality, of course, the true regression coefficient lies either in the range between the confidence interval (that is, with a probability of 1 or in other words with a 100%) or not (with a probability of 0 or in other words 0%). Specifying a 95% probability is a statistical auxiliary structure to determine the accuracy and not a substantive conclusion (Gehring/ Weins 2009: pp ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). R² has a PRE-interpretation of the regression model and therefore ranks in the range of zero to one. By multiplying R² by 100, the percentage of how much of the variation of the 68

70 explanatory variable can be explained by the regression model is received. The feeling of loyalty of the Dutch national civil servant towards his/her own policy views with an R² of about 0.07, the regression model is capable of explaining 7% of the variation in the dependent variable correctly. For the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister with an R² of about 0.09, the regression model is capable of explaining 9% of the variation in the dependent variable correctly. For the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU with an R² of about 0.18, the regression model is capable of explaining 18% of the variation in the dependent variable correctly. For the educational level of Dutch national civil servant with an R² of about 0.02, the regression model is capable of explaining 2% of the variation in the dependent variable correctly. Since only one variable is used for the prediction, these are high explanatory powers (Gehring/ Weins 2009: pp ) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). This part identified four significant variables each of which have a meaningful influence on the dependent variable. The regression coefficients are small, which means that if the independent variables change the influence on the dependent variable is moderate yet especially the first three independent variables each have a high explanatory power regarding the variation of the dependent variable. In the next part, the bivariate linear regression model will be advanced to the multiple linear regression model Multiple linear regression with dummy variables and beta In reality, a bivariate linear regression is a (mostly inaccurate) simplification. Barely one variable can be explained on its own, only by one other variable. For this reason, several variables are included on the explanatory side (multiple model). The dependent variable is thus explained by the multiple interaction of several independent variables happening simultaneously (Gehring/ Weins 2009: pp ) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). The findings of the multiple linear regressions can be found in table 6 below. The findings of the multiple linear regressions including the control variables (including dummy variable 69

71 sex of the Dutch national civil servant) can be found in table 7 below (Kühnel/ Krebs 2012: pp , ). 70

72 Table 6 Multiple linear regression Feeling of loyalty of the Dutch national civil servant towards his/her own minister Coefficient Std. Error P > t 95 % Conf. Int. (Min & Max) Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil servant s amount of time spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU to to to to to Adj. R² Prob > F Source: Author s table (6) based on Stata output of the survey examination. 71

73 Table 7 Multiple linear regression with control variables Feeling of loyalty of the Dutch national civil servant towards his/her own minister Coefficient Std. Error P > t 95 % Conf. Int. (Min & Max) Feeling of loyalty of the Dutch national civil servant towards his/her own policy views to Degree EU-enthusiasm/ scepticism of the Dutch national civil servant to Dutch national civil servant s amount of time spent on EU to matters Dutch national civil servant s feeling of a clear mandate for to negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards to the whole political and administrative System of the EU Age Dutch national civil servant to Sex Dutch national civil servant to Educational level of Dutch national civil servant to

74 Years working as a Dutch national civil servant to Adj. R² Prob > F Source: Author s table (7) based on Stata output of the survey examination. The first value to look at is R² or in the case of the multiple linear regression the corrected R². The corrected R 2 (adj R-squared) adjusts the calculation of the determination coefficient by the number of regressors. Although the discrepancy between R² and the corrected R² decreases with increasing sample size, the corrected version should still be used as a PRElevel in the multiple case. The interpretation is identical to the "normal" R²: With an "Adj R- squared" of around the multiple regression model can explain, 23.8 % of the variation of the dependent variable. With the control variables included, 23.7 % can be explained. The Prob > F value is the p-value of the F-test of the regression model. This tests whether the regression model as a whole (the independent variables in their combination) has a statistically significant explanatory power for the dependent variable. It occurs with a value of in the case in both models, which makes it statistically significant. Technically, it is tested whether the coefficient of the determination coefficient R² is statistically significantly different from zero, and thus the present explanation of the variation in the dependent variable by the regression model does not just occur by chance (Gehring/ Weins 2009: pp ) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). The p-values of the variables the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister and the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU are below the alpha = 0.01 level in both models and are therefore, at this level, statistically significant - their regression coefficients have (as all other remaining regressors are checked for) a statistically significant deviation from zero at a significance level of All statistically significant explanatory variables have a positive regression coefficient and are thus exerting a positive 73

75 influence on the dependent variable (Gehring/ Weins 2009: pp ) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). The general interpretation for all significant regressors is: If the respective independent variable increases by one unit (on their scale), then the dependent variable increases (on its scale) by the regression coefficient, assuming all remaining independent variables are kept constant (marginal partial influence). Furthermore, the degree of EU-enthusiasm/ scepticism of the Dutch national civil servant, the Dutch national civil servant s amount of time spent on EU matters, the age of the Dutch national civil servant, the sex of the Dutch national civil servant, the educational level of Dutch national civil servant and the years working as a Dutch national civil servant seem to offer no (additional) explanatory power alongside the other regressors. The regression coefficient has no statistically significant deviation from zero, since the p-value is well above 0.05 (Gehring/ Weins 2009: pp ) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp , ). The regression coefficient of the feeling of loyalty of the Dutch national civil servant towards his/her own policy views is (first model)/ (second model), the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister is (first model)/ (second model) and the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU is (first model)/ (second model) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ) (Gehring/ Weins 2009: pp ). Exemplary interpretation of the prediction for the feeling of loyalty of the Dutch national civil servant towards his/her own policy views: If for a civil servant, the feeling of loyalty towards his/her own policy views increases by one unit on the scale of the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the civil servant s feeling of loyalty towards his/her own minister increases by approximately (first model)/ (second model) times of a category on the scale of the feeling of loyalty of the Dutch national civil servant towards his/her own minister, provided that no other variable changes on the independent side. The exemplary interpretation can be applied to the other two significant variables with their corresponding regression coefficients (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ) (Gehring/ Weins 2009: pp ). 74

76 So far, this part has shown that each of three of the independent variables examined here influence the dependent variable and are able to explain 23.8 % of the variation of the dependent variable. Two of the independent variables and all of the control variables do not improve the model. The question as to which of the three significant variables influences the dependent variable the most remains. The problem is that the regression coefficients are not directly comparable because they are dependent on the original scale of the underlying basic variable. In order to fix that, the regression coefficients will be standardized through the Stata command beta. Beta standardizes the regression coefficients and thus allows them to be directly compared because they are now independent of the original scale of the underlying basic variable. The findings are portrayed in table 8. According to the findings, the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU with a standardized regression coefficient of 0.33 is the most relevant regressor, followed by the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister with a standardized regression coefficient of 0.19 and last the feeling of loyalty of the Dutch national civil servant towards his/her own policy views with a standardized regression coefficient of The three variables contribute to the explanatory power for the prediction of the feeling of loyalty of the Dutch national civil servant towards his/her own minister in that order (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). 75

77 Table 8 Standardised multiple linear regression with control variables Feeling of loyalty of the Dutch national civil servant towards his/her own minister Coefficient Std. Error P > t Beta Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil servant s amount of time spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU Age Dutch national civil servant Sex Dutch national civil servant Educational level of Dutch national civil servant Years working as a Dutch national civil servant Adj. R² Prob > F 76

78 Source: Author s table (8) based on Stata output of the survey examination. The direct comparison between the bivariate linear and multiple linear regression model is allowed here, since the identical dependent variable is explained. A better explanatory power is shown by the larger and more complex multiple regression model. Whereas the best significant bivariate linear regression was only able to explain 18 %, the multiple regression model explains almost 24 % of the variation of the dependent variable (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). This part explained the dependent variable by the multiple interaction of three significant independent variables simultaneously and allowed their regression coefficients to be directly compared through standardization. The next part checks this multiple interaction model for interaction effects Multiple linear regression with interaction effects An interaction effect in a regression model is none other than the product of two variables. This allows the effect of a simultaneous change of two independent variables on the dependent variable to be examined, in other words, whether the influence of two independent variables systematically condition each other (or at least one independent variable the other) in their influence on the dependent variable. Where an interaction effect exists between the two independent variables on the dependent variable three different effects influencing the dependent variable can be ascertained: the two effects of each independent variable and additionally the mutual interaction effect. The interaction effect can be calculated between two dummy variables, between one dummy variable and one metric variable or between two metric variables (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). This author tested each of the three significant independent variables of the multiple linear regression model above, the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister and the feeling of loyalty of the Dutch national civil 77

79 servant towards the whole political and administrative System of the EU for interaction effects on the dependent variable the feeling of loyalty of the Dutch national civil servant towards his/her own minister with each other. None of the assumed interaction effects were significant (the feeling of loyalty of the Dutch national civil servant towards his/her own policy views and the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister (0.912 P> t )/ the feeling of loyalty of the Dutch national civil servant towards his/her own policy views and the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU (0, 166 P> t )/ the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister and the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU (0, 802 P> t )). Exemplary interpretation of the non-existing interaction effect between the feeling of loyalty of the Dutch national civil servant towards his/her own policy views and the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU: Both independent variables have an effect on the dependent variable but no interaction effect. In other words, there are no effects of the feeling of loyalty of the Dutch national civil servant towards his/her own policy views on the dependent variable, which can be attributed to the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU or the other way round. The developed model in part 4.5 does not have to include interaction effects between the independent variables on the dependent variables to improve the model. The next part will go one step further and check the model for non-linear influences of the independent variables and even more importantly test whether the model is missing any variables (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ) Non-linear influences of the independent variables + test for missing variables Non-linear influences of the independent variables (e.g. square effects) can become a problem for the multiple linear model developed above. The problem is that the multiple linear model relies on the linearity of its variables for its calculations. Therefore, if non-linear influences of the independent variables are part of the model, it could compromise the 78

80 findings. One way to check for non-linear effects of the independent variables is a graphical test of nonlinearity the Augmented Component Plus Residual Plot (ACPR plot) (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp , ). The ACPR plot below was used for each of the three significant independent variables of the model. Figure 5 ACPR plot for the feeling of loyalty of the Dutch national civil servant towards his/her own policy views Civil servants feeling of loyalty towards own policy views Source: Stata figure (5) based on Stata output of the survey examination. 79

81 Figure 6 ACPR plot for the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative system of the EU Feeling of loyalty of the civil servant towards the whole political and administ Source: Stata figure (6) based on Stata output of the survey examination. 80

82 Figure 7 ACPR plot for the feeling of a clear mandate for negotiations on EU level from the minister Clear mandate for negotiations on EU level for the civil servant from the minist Source: Stata figure (7) based on Stata output of the survey examination. The stronger the "curved" line ("Median spline") deviates from the (linear) estimate the more prominent the problem of non-linearity, either over the total number of observations or from a specific expression of each independent variable. In all three cases, the linear estimate and the curved line are very close but not a perfect match which, in most cases, cannot be expected as it is only an estimate. So far, no non-linear influences of the independent variables could be discovered (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp , ). Another way to test for non-linear influences on the independent variables and, even more importantly, for generally relevant variables which are still missing in the model is the "Regression Equation Specification Error Test (RESET). This test is a very general way to estimate missing sizes (e.g. third variables, quadratic terms, etc.) in a regression model. The RESET test tests the null hypothesis (The model is complete) against the alternative hypothesis (The model is missing "something"). The advantage is the general testing for 81

83 missed sizes. The downside is that the test indicates whether something or not is missing but not what it is exactly that is missing. The test is nevertheless a good indication. The outcome of the test is not significant on a 0.05 level (0.9551). No real problem seems to exist - the null hypothesis (The model is complete) can be maintained. Our model does not seem to have missed a (statistically) relevant variable which effects the linearity assumption of the linear regression model fundamentally. However, this is of course just statistical evidence and no content-theoretical one (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp , ). The tests shows that the model is neither missing variables nor adulterated by non-linear influences of the independent variables. Up to this point, this author is able to say that the model itself seems accurate and complete for the purposes of the following analysis but to be valid each regression model has to fulfil six conditions which will be examined in the next part Regression diagnostics Six assumptions must each be met before a regression model is "valid". The first condition is that the dependent variable has to be interval scaled. The dependent variable needs a metric scale, so that the regression analysis can be used. In our case, this condition is fulfilled as the dependent variable in all cases the feeling of loyalty of the Dutch national civil servant towards his/her own minister on a scale of 1-5 is an interval scaled variable. The second condition is fulfilled, if all independent variables have a variance. In other words, each independent variable must have more than one characteristic and thus be able to develop a variance. The reason is that no arithmetic mean can be formed for an independent variable with only one characteristic. But an arithmetic mean is essential for the regression analysis. In this case, all independent variables have several characteristics. The independent variable with the least characteristics is sex of the Dutch national civil servant with just the two characteristics male and female. The third condition is that the correlations between two or more independent variables are not too high. Otherwise, the effects of the two variables are no longer clearly independent because they have much "common information and distort the R². It is not important that multicollinearity does not at all exist, which is very unlikely, but it is not allowed to exceed a certain degree. Only if multicollinearity exceeds the certain degree, 82

84 counter measures are necessary. The VIF (Variance Inflation Factor) calculates the multicollinearity. If the VIF value is below 5, multicollinearity has an acceptable degree. In our case, all VIF values can be found in table 9. All VIF values are way below 5 and multicollinearity therefore not exceeds a certain degree. The condition is fulfilled (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). Table 9 Regression diagnostic VIF test Variable (independent or dependent) VIF Feeling of loyalty of the Dutch national civil servant X towards his/her own minister (dependent Variable) Feeling of loyalty of the Dutch national civil servant 1.11 towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national 1.10 civil servant Dutch national civil Servant s amount of time spent on EU 1.07 matters Dutch national civil servant s feeling of a clear mandate 1.12 for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant 1.32 towards the whole political and administrative System of the EU Age Dutch national civil servant 1.70 Sex Dutch national civil servant 1.07 Educational level of Dutch national civil servant 1.10 Years working as a Dutch national civil servant 1.73 Mean 1.26 Source: Author s table (9) based on Stata output of the survey examination. 83

85 The fourth condition is the lack of covariance or correlation between the independent variable and the error term. If there were a correlation or covariance, this could constitute a third variable effect. Since the third variable was not taken into account, influences from the error term affect the independent variable. In our case, all correlations and covariance between the independent variables and the error term adopted a value of around 0. Therefore, no third variable which was not taken into account from the error term influences the independent variable. No correlation or covariance between the independent variables and the error term can be found and the condition is fulfilled (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). The fifth condition demands homoscedasticity. Homoscedasticity refers to the assumption of constancy of the variance of the error terms. This means that the strength of the fluctuation of the error (its variance / standard deviation) has to be constant for all predicted values of the dependent variable and should be independent of the specific values of the dependent variable. To test for homoscedasticity, the "Breusch-Pagan" test (hettest) is used. The null hypothesis "No, heteroscedastic structure of the residuals - constant variance" is tested against the alternative hypothesis "Yes heteroscedastic structure exists no constant variance". If the test is statistically significant, (the p-value (Prob> chi2) is lower than 0.05 or even 0.01) heteroscedasticity exists and homoscedasticity should be discarded. In our case, the hettest significant (0.0000) and homoscedasticity must be rejected. But in reality (especially at cross-sectional data), heteroscedastic error terms are more the rule. Therefore, it needs to be checked further whether the effects of heteroscedasticity are serious enough to invalidate the regression model. To examine the seriousness of the heteroscedasticity, the normal regression coefficients and standard errors of the independent variables are compared to regression coefficients and standard errors, which have been calculated as heteroscedasticly robust. The normal and the robust values can be found in table 10. In our case, no significant change between the normal and the robust variant is found. Heteroskedastic effects do not appear too strongly and can therefore be disregarded. The condition is fulfilled (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). 84

86 Table 10 Regression diagnostics heteroscedasticity test Feeling of loyalty of the Dutch Coefficient Std. Error Coefficient Std. Error national civil servant towards (original) (original) (test) (test) his/her own minister Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil Servant s amount of time spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU Age Dutch national civil servant Sex Dutch national civil servant Educational level of Dutch national civil servant Years working as a Dutch national civil servant Source: Author s table (10) based on Stata output of the survey examination. 85

87 The last condition is the assumption that the error terms are normally distributed. This is a problem as regards samples smaller than 100. If the number of cases is sufficiently high, according to the central limit theorem, an approximate normal distribution can be assumed, meaning, although the normal distribution is not empirically present, it is nevertheless given in the population and the condition is fulfilled. Thus even if the error terms are not normally distributed, the problem becomes obsolete when the sample is high enough. The sample in the survey are the Dutch national civil servants working for the empire. The sample size consists of 1394 and is therefore much larger than necessary to fulfil the assumption (Kühnel/ Krebs 2012: pp ) (Kohler/ Kreuter 2012: pp ). All six conditions have been fulfilled, the regression model is therefore valid. The next part will discuss the data collection and findings of the interviews Data collection and findings interviews This part will present the data collection and findings of the interviews. This part will be kept to a minimum for several reasons. First, the data described below is already part of the survey and therefore part of the findings above. All interviewees were chosen because they conducted the survey and agreed to further interviews, therefore, their categorized answers are already in the survey and contribute no additional information. Second, in order to achieve convincing findings with the methods used above, the group of participants would have had to be much larger. Third, the interviewees were especially chosen to achieve a maximum variation of cases; the selection of the interviewees was done by extreme groups (special cases) and by modal categories (common cases). Therefore, the interview group is in fact representing the whole spectrum of the survey, but it is not representative of the survey group. Therefore, all measurement methods used above are unsuitable. But why then use interviews or a categorization system in the first place? First, what is meant by the categorization system? As explained in the research design section, each interview was divided into different parts and each part was classified by the author to the accompanying value of the accompanying variable. The categorization system is irreplaceable as to achieving the maximum comparability between the findings in the survey and the interviews. Furthermore, the categorization system is used as a method to bring order into the interviews 86

88 in order to make it easier to apply the corresponding part of the interview to the variable used in this thesis. The categorization of the different parts of the interviews is merely a starting point. It is not useful to present, at this point, what each interviewee said in detail because it would go beyond this thesis remit and distract the reader from the findings of the survey. Of course, the full interviews are part of the annex and included to achieve maximum transparency. The goal of the interviews is not to create a categorization system but to collect the extensive answers and the reasoning behind it. The interviews enabled the author of this thesis to verify and enhance the analysis in the next chapter. The table 11 presents the data collection and data findings of the conducted interviews. They match most extensively the values the same Dutch national civil servants gave in the survey. Small differences between the answers in the survey and the interviews could be expected because of changes in the interviewee s opinion between the time of the survey and the interviews or more likely because of categorization errors by this author. It can be challenging to interpret, for example, the feeling of loyalty of an interviewee towards his/her own minister based on what the interviewee said into the according value range of the variable. All cells marked with an X represent an interviewee who did not talk about the specific variable and could therefore, as based on the interviews, not be classified. All cells marked with two values in it, one within brackets ( ( ) ) and one without, represent an interviewee who gave in the interview a different value for the specific variable than in the survey. The original survey value was put in brackets. 87

89 Table 11 Data collection and findings interviews Variable Pers Pers Pers Pers Pers Pers Pers Pers on 1 on 2 on 3 on 4 on 5 on 6 on 7 on 8 Feeling of loyalty of the Dutch national civil servant towards his/her own minister Feeling of loyalty of the Dutch national civil servant towards his/her own policy views Degree EU-enthusiasm/ scepticism of the Dutch national civil servant Dutch national civil Servant s amount of time spent on EU matters Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister Feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU (0) 2 1 X 5 (3) 80 X X X 20 X (80) (15) 5 X X (4) 4 Source: Author s table (11) based on Stata output of the survey examination. The next part will present a brief summary of the parts discussed in this chapter. 88

90 4.10. Brief summary of the chapter This chapter presented the data collection and findings. An increasing approach was used, which means that this chapter started with basic findings and ended with the complete tested model. In the first two parts, all relevant variables were presented and their central tendency of distribution and the measures of variation examined. In the next step, the relationships between the variables were examined, first, more generally within the part about correlations followed by the bivariate linear regression between each independent variable and the dependent variable. The fifth part developed a complete multiple model with several independent variables and one dependent variable rather than just one independent and one dependent variable. Afterwards, the model was tested negatively for interaction effects within the independent variables, non-linear influences of the independent variables and for entire missing variables within the model. In a last step, it was made sure that the regression model is valid and did not violate certain conditions. In the last part, the data collection and findings of the interviews were discussed and presented. The next chapter will use the findings in this chapter in order to present the analysis. 5. Analysis This last chapter will present the analysis, which discusses the findings in relation to the theoretical framework by drawing on the hypotheses. The aim of all 5 hypotheses is to determine what factors explain a higher or lower degree of loyalty of the Dutch national civil servant towards the minister. It is in the minister s interest to ensure the highest possible degree of loyalty of the Dutch national civil servant towards him/her, as the minister is not able to control the civil servant completely and therefore has to rely, to some extent, on the Dutch national civil servant s loyalty as to him/her acting in the minister s interest Hypothesis 1 According to the Principal-agent theory whenever a Principal-agent problem between the minister and the Dutch national civil servant arises, the agent acts disloyally and in his/her own interest. A Principal-agent problem has two conditions, differing interests between the 89

91 minister and the Dutch national civil servant and hidden action and/or hidden knowledge (Braun/ Guston 2003: p ) (Gailmard 2014: pp ). Whenever one of the two condition is missing, no Principal-agent problem will occur and the Dutch national civil servant will act loyally and in the interest of the principal. Different interests should not be mistaken for personal policy views. The Dutch national civil servant has of course his/her own policy views and he/she can be more or less loyal towards them but unless these policy views differ from those of the minister they do not cause any issues and do not fulfil the condition of the Principal-agent problem. In other words, even when the Dutch national civil servant is loyal towards his/her own policy views, this does not contradict the loyalty of the Dutch national civil servant towards the minister when no different interests between the minister and the Dutch national civil servant exist. Therefore, the feeling of loyalty of the Dutch national civil servant towards his/her own policy views has no negative effect on the feeling of loyalty of the Dutch national civil servant towards his/her own minister. Furthermore, the theory from the field of economics implies a positive effect of the dependent on the independent variable. Two types of loyalty can be differentiated: behavioural and emotional loyalty. Behavioural loyalty is the name given to behaviour that proves loyalty to the brand through sales. It means that the customer behaves in the way the brand wants, i.e. by spending money (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). Applied to the case of this hypothesis, this means that the Dutch national civil servant proves his/her loyalty towards the minister by acting on behalf and in the interest of the minister. Emotional loyalty, on the other hand, acknowledges that perceptions and emotions drive behaviour. A customer who is emotionally loyal has empathy and attachment with a brand and company and is more likely to recommend it to potential customers (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). Applied to the case of this hypothesis, this means that the Dutch national civil servant is emotionally loyal, feels an empathy and attachment towards his/her own policy views, which seems only naturally. The two types of loyalty reinforce each other when they can be achieved together (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). If in the case of this hypothesis behavioural as well as emotional loyalty exist, then this would mean that the feeling of loyalty of the Dutch national civil servant towards his/her own policy views reinforces the feeling of loyalty of the Dutch national civil servant towards his/her own minister (and vice versa). The feeling of loyalty of the Dutch national civil servant towards 90

92 his/her own policy views should therefore influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister positively. First hypothesis: The higher the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The findings support the hypothesis. If in a Dutch national civil servant the feeling of loyalty towards his/her own policy views increases by one unit on the scale of the variable feeling of loyalty of the Dutch national civil servant towards his/her own policy views (becomes more positive), the civil servant s feeling of loyalty towards his/her own minister increases (improves) by around 0.27 times of a category on the scale of the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The regression model with an R² of around 0.07 shows that the feeling of loyalty of the Dutch national civil servant towards his/her own policy views can explain 7% of the variation in the dependent variable correctly. The multiple regression model, although it predicts a smaller regression coefficient than the bivariate linear regression, confirms the findings of a positive regression coefficient. Therefore, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The interviewees confirm these results as well. Almost all interviewees say that they have their own policy views but this does not cause them to be disloyal towards the minister (Interviewee ) unless a conflict of interests arises (Interviewee 7+8). In other words, unless their own policy views and those of the minister differ, they will be loyal. Even when there are smaller differences of interest between the minister and them (interviewee 7+8), they will act loyally because the intensity is not high enough to cause different interests in the sense of the Principal-agent problem. So it is possible for the Dutch national civil servants to put their own opinions aside even when there are small differences because they have not reached an intensity (not important enough) to cause different interests in the sense of the Principal-agent problem and to induce them to act against the interests of the minister (interviewee ). The most effective way of defusing different interests before a Principal-agent problem can arise is that the Dutch national civil servants simply explain to their superiors why they recommend a certain action or are of a special opinion, which generally causes their superiors to follow their recommendation 91

93 (Interviewee 2+7+8). Now it is easier to understand how the Dutch national civil servant can be emotionally loyal towards his/her own policy views but still act on behalf and in the interest of the minister. It seems therefore not to be a problem for the Dutch national civil servant to harmonize the feeling of loyalty of the Dutch national civil servant towards his/her own policy views and the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The first hypothesis is true. To reinforce the feeling of loyalty of the Dutch national civil servant towards his/her own minister, the minister has to strengthen the feeling of loyalty of the Dutch national civil servant towards his/her own policy views. As the influence of this independent variable on the dependent variable is rather small, this will never become top priority for the minister. The minister should concentrate on other more easily influenced variables, which have a bigger effect on the dependent variable Hypothesis 2 Communal identities like a European identity can be extremely powerful in shaping views towards political objects (Hooghe/ Marks 2005: p.423). In other words, communal identities can partially influence the Dutch national civil servant s interests, for example, his/her political views (Braun/ Guston 2003: p ) (Hooghe/ Marks 2005: p.423). But what applies in this regard as well is that there must be a difference between the interests of the minister and the Dutch national civil servant to cause different interests in the sense of the Principal-agent problem and with it one condition of disloyal behaviour (Braun/ Guston 2003: p ) (Jordan 2003: pp ). The EU is just one interest among many on which the minister and the civil servant can disagree. But they have to disagree according to Principal-agent theory to fulfil the first condition of a problem. Whether the Dutch national civil servant is an EU enthusiastic or an EU sceptic does not influence loyalty towards the minister on its own accord. The degree of EU-enthusiasm/ scepticism of the Dutch national civil servant should therefore not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. 92

94 Second hypothesis: The degree of EU-enthusiasm/ scepticism of the Dutch national civil servants will not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister be. The findings support the hypothesis. If in a Dutch national civil servant the degree of EUenthusiasm/ scepticism increases (or decreases) by one unit on the scale of the variable degree of EU-enthusiasm/ scepticism of the Dutch national civil servants, the civil servant s feeling of loyalty towards his/her own minister does not increase (decrease) significantly. If the regression coefficient of an independent variable is not statistically significant, as in this case, then the regression coefficient is not valid for the entire population. The multiple regression model confirms the findings of insignificance as already predicted by the bivariate linear regression. Therefore, the degree of EU-enthusiasm/ scepticism of the Dutch national civil servants does not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The interviewees confirm these results as well. The situation is similar to the situation in hypothesis 1. Almost all interviewees say that their own opinion about the EU does not cause them to be disloyal towards the minister (Interviewee ) unless it develops into a conflict of interest (Interviewee 7+8). In other words, unless their own policy views about the EU and those of the minister differ, they will act loyally. Even when there are smaller differences in the interests between the minister and them (interviewee 7+8), they will act loyally because the intensity is not high enough to cause different interests in the sense of the Principal-agent problem. So it is possible for the Dutch national civil servants to put their own opinion regarding the EU aside even when there are smaller differences because they have not reached an intensity (not important enough) to cause different interests in the sense of the Principal-agent problem and to induce them to act against the interests of the minister (interviewee ). The second hypothesis was validated. The degree of EU-enthusiasm/ scepticism of the Dutch national civil servant does not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. To reinforce the feeling of loyalty of the Dutch national civil servant towards his/her own minister, this is one aspect which can be neglected by the minister. 93

95 5.3. Hypothesis 3 A national civil servant who becomes more involved in EU matters and spends more time on them will automatically learn more about the EU and its procedures. But it is not unusual for the agent to possess certain skills, to be very specialized or to have more information than the principal. This is, in fact, one of the reasons why the agent is so valuable to the principal and why he/she was chosen in the first place. The minister, on the other hand, will not be ignorant of new developments. If the Dutch national civil servants under his/her command become or are already deeply involved in EU matters, the minister will not only become more involved because the Dutch national civil servant reports to him/her, but the minister will also inform himself/herself as well as this is the only way to control the civil servants to a certain degree and make sure that there is no Principal-agent problem and his/her interests are acted upon. But even if the minister were ignorant, the hidden knowledge on its own would not suffice to cause a Principal-agent problem because the minister and the Dutch national civil servant would still need to have different interests. Whether a Dutch national civil servant spends more or less time on EU matters does not intensify the hidden knowledge within the Principal-agent relationship and certainly does not cause a Principal-agent problem on its own accord. The Dutch national civil servant should therefore, however his amount of time spent on EU matters might change, have the same feeling of loyalty towards the minister (Mastenbroek/ Princen 2010: pp ). Third hypothesis: The Dutch national civil servant s amount of time spent on EU matters will not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The findings support the hypothesis. If the Dutch national civil servant s amount of time spent on EU matters increases (decreases) by one unit on the scale of the variable Dutch national civil servant s amount of time spent on EU matters, the civil servant s feeling of loyalty towards his/her own minister does not increase (decrease) significantly. Because the regression coefficient of the independent variable is, in this case, also not statistically significant, it is not valid for the entire population. The multiple regression model confirms the findings of insignificance, which the bivariate linear regression had already predicted. The interviewees confirmed the findings as well. The Dutch national civil servants use all their hidden knowledge to cater to the different needs of the minister. The minister, on the other 94

96 hand, uses the different expertise and certain skills of the Dutch national civil servants to always stay on top of things. Of course, each minister has his/her priorities, but nevertheless he/she is informed about everything and has to nod everything through (interviewee 1+4+5). In the end, the minister is the one who is accountable to parliament (interviewee 1+4). The third hypothesis was verified. The Dutch national civil servant s amount of time spent on EU matters does not influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister. This is the second variable which can be neglected by the minister for reinforcing the feeling of loyalty of the Dutch national civil servant towards his/her own minister Hypothesis 4 Despite the fact that the complete control of the principal over an agent can never be achieved nor is it desirable, the literature examines different mechanisms for the principal to make sure that the agent acts in the principal s interest to achieve as little agency loss as possible (Lane 2013: pp ) (Ennser-Jedenastik 2013: pp ) (Van Thiel/ Yesilkagit 2011: pp ). Two negative examples from the literature are review and oversight carried out too closely. A review and an oversight by the minister that are too strict can be counterproductive to exerting a certain degree of control. Since, if the minister controls the Dutch national civil servant too closely it can have the effect that the civil servant loses all flexibility and is not able to use his/her expertise to the minister s benefit. Furthermore, significantly more different interests between the minister and the Dutch national civil servant will arise and with it the chance of a Principal-agent problem. If the minister left the civil servant some more freedom to act, the possibility that the Dutch national civil servants and the minister s interests overlapping would be much greater with a smaller agency loss as the consequence. But too much leeway has its downside as well. Without a clear mandate, the national civil servants lack the necessary ministerial support to achieve their goals on EU and national level (Schmidt/ Van den Berg (in print): n.p.) and will get frustrated. Whereas too close a principal control over the agent should have a negative effect on loyalty towards the minister, a clear mandate has a positive effect. It is not relevant for that matter whether the clear mandate is given for negotiations on the national level (e.g. between ministries) or, as in the case of this thesis, on EU level (e.g. within the Council of ministers); both should have 95

97 the same effect. The Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level should therefore influence the feeling of loyalty of the Dutch national civil servant towards his/her own minister positively. Fourth hypothesis: The higher the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The findings support the hypothesis. If in a Dutch national civil servant the feeling of a clear mandate from the minister for negotiations on EU level increases by one unit on the scale of the variable, the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level (becomes more positive), the civil servant s feeling of loyalty towards his/her own minister increases (improves) by around 0.27 times of a category on the scale of the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The regression model with an R² of about 0.09 can explain 9% of the variation in the dependent variable correctly by using the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level. The multiple regression model thus predicts a smaller regression coefficient than confirmed by the linear regression regarding its findings of a positive regression coefficient. Therefore, the higher the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The interviewees confirm these results. All interviewees pointed out that they had been given a clear mandate from the minister for all negotiations on EU level but that they still have some room to manoeuvre. As long as they do not cross some red lines or screw up they are free to act as they see fit within these boundaries (Interviewee ). The forth hypothesis has come true. To reinforce the feeling of loyalty of the Dutch national civil servant towards his/her own minister, the minister has to strengthen the Dutch national civil servant s feeling of a clear mandate from the minister for negotiations on EU level. In this area, the minister can make a big difference. The key here is to understand what his/her agents need and to have a good Principal-agent relationship based on the right amount of flexibility but also clear boundaries. 96

98 5.5. Hypothesis 5 Loyalty towards the minister is one of the easiest ways of ensuring the minister s benefits in his relationship with the Dutch national civil servants. But what if it is no longer only the minister and the Dutch national civil servant? The question as to whether the feeling of loyalty of the Dutch national civil servant to another person or institution influences his/her feeling of loyalty towards the minister forces itself onto the reader; in this case, the other institution in question is the EU. In other words, on the one hand, there is the minister part of the executive of the Netherlands and, on the other hand, the whole political and administrative System of the EU. It seems to be common sense that the Dutch national civil servant cannot be loyal to two masters and that when there is more loyalty to the political and administrative System of the EU, there must be less loyalty towards the minister. The theory from the field of economics used for hypothesis 1 supports these assumptions. The theory said that behavioural and emotional loyalties reinforce each other. Behavioural loyalty was the name given to behaviour that proves loyalty to the brand through sales. It means that the customer behaves in the way the brand wants, i.e. by spending money (Chaffey/ Ellis- Chadwick/ Johnston/ Mayer 2009: p. 336). Emotional loyalty, on the other hand, acknowledges that perceptions and emotions drive behaviours. A customer who is emotionally loyal has empathy and attachment with a brand and company and is more likely to recommend it to potential customers (Chaffey/ Ellis-Chadwick/ Johnston/ Mayer 2009: p. 336). But in this case, the feeling of loyalty of the Dutch national civil servant towards his/her own minister as well as the loyalty towards the whole political and administrative System of the EU are both behavioural loyalties only. The Dutch national civil servant proves his/her loyalty towards the minister by acting on behalf and in the interest of the minister. And as the loyalty towards the whole political and administrative System of the EU is a behavioural loyalty as well, the Dutch national civil servant proves his/her loyalty towards the whole political and administrative System of the EU by acting on behalf and in the interest of the whole political and administrative System of the EU. Both contradict each other because a Dutch national civil servant cannot act on behalf and in the interest of the minister and the whole political and administrative System of the EU. The feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU should therefore have a negative impact on the feeling of loyalty of the Dutch national civil servant towards his/her own minister. 97

99 Fifth hypothesis: The higher the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU, the lower the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The findings disproved the hypothesis. If in a Dutch national civil servant the feeling of loyalty towards the whole political and administrative system of the EU increases by one unit on the scale of the variable feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU (becomes more positive), the civil servant s feeling of loyalty towards his/her own minister increases (improves) by around 0.42 times of a category on the scale of the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The regression model with an R² of about 0.18 is capable of explaining 18% of the variation in the dependent variable correctly by using the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU. The multiple regression model, although it predicts a smaller regression coefficient than the bivariate linear regression, confirms the findings of a positive regression coefficient. Therefore, the higher the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister. The interviewees confirm the findings and therefore also disprove the hypothesis. Both loyalties reinforce each other. All interviewees say that their loyalty is foremost directed towards the minister (Interviewee ) but, on the other hand, the Netherlands is part of the EU and it is therefore in the Netherlands own interest to achieve unity between all member states. To achieve compromises between all member states means to be able to speak with one voice, which is much more influential internationally compared to each single member state alone (Interviewee 4+7+8). Both loyalties complement each other (Interviewee 7). But how can it be explained that both loyalties reinforce rather than reduce each other? The most plausible explanation is to accept that this author was wrong about his assessment of the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU as a behavioural loyalty rather than an emotional one. An emotional loyalty acknowledges that perceptions and emotions drive behaviours. A customer who is emotionally loyal has empathy and attachment with a brand and company and is more likely to recommend it to potential customers (Chaffey/ Ellis-Chadwick/ 98

100 Johnston/ Mayer 2009: p. 336). Applied to the case of this thesis, this means that the Dutch national civil servant is emotionally loyal, has empathy and attachment towards the whole political and administrative System of the EU. The findings and especially the interviews support this view because the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU reinforces the feeling of loyalty of the Dutch national civil servant towards his/her own minister (and vice versa) so one has to be behavioural loyalty and the other emotional loyalty. In the understanding of the Dutch national civil servant, the behavioural loyalty towards the minister and the emotional loyalty towards the whole political and administrative System of the EU cooperate with each other. In other words, the Dutch national civil servant is, on the one hand, able to act on behalf and in the interest of the minister and, on the other hand, he/she is emotionally loyal, has empathy and attachment towards the whole political and administrative System of the EU. Both things can coexist in the mind of the Dutch national civil servant. The fifth hypothesis had to be adjusted. To reinforce the feeling of loyalty of the Dutch national civil servant towards his/her own minister the minister has to strengthen the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU. Compared to the other two independent variables, this constitutes the biggest influence on the dependent variable. This is one independent variable the minister should concentrate on and try to enforce the Dutch national civil servant s feeling of loyalty towards the whole political and administrative System of the EU Discussion analysis When the findings of the linear regression were presented, one of the control variables, the Educational level of Dutch national civil servant, influenced against all expectations the feeling of loyalty of the Dutch national civil servant towards his/her own minister. Therefore, this author had the opportunity, in hindsight, to create a sixth hypothesis, which would have said: The higher the educational level of Dutch national civil servant, the higher the feeling of loyalty of the Dutch national civil servant towards his/her own minister is. This bivariate linear regression coefficient was positive and the bivariate linear regression model was capable of explaining at least 2% of the variation in the dependent variable correctly. But not only could no indicators be found in theory as to the educational level influencing the feeling of loyalty 99

101 to the principal, but more importantly the influence of the educational level of Dutch national civil servant on the dependent variable in the multiple regression model vanished. To remind the reader, a multiple linear regression model was used because in reality, a bivariate linear regression is a (mostly inaccurate) simplification. One variable cannot really be explained on its own, only by one other variable. For this reason, several variables are included on the explanatory side (multiple model). The dependent variable is thus explained by the multiple interaction of several independent variables simultaneously (Gehring/ Weins 2009: pp ) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). Each established multiple linear regression model is unique, the effects of each independent variable on the dependent variable occur in the particular form only in combination with the other variables. Once more variables are included as explanatory variables or existing variables are omitted, the (partial) effects of all other variables change. The multiple regression model thus considers the impact of each variable by controlling (keeping constant) all other variables of the multiple regression model. For each variable is thus indicated by the regression coefficient, namely, what each independent variable can contribute to explain the dependent variable in other words, whatever is not explained by another variable (partial regression coefficient) (Gehring/ Weins 2009: pp ) (Kohler/ Kreuter 2012: pp ) (Kühnel/ Krebs 2012: pp ). When an independent variable significantly influences a dependent variable in a bivariate linear regression without doing so in a multiple linear regression model, as in the case of the educational level of Dutch national civil servant, this is because the other significant variables in the multiple linear model (the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister, the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU) already explain the part the dependent variable, the educational level of Dutch national civil servant, also wanted to explain. In other words, our multiple regression model with the three significant independent variables (the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister, the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU) explains the dependent variable so well that it does not need the educational level of the Dutch national civil servant. Therefore, as the three significant 100

102 independent variables just mentioned and the dependent variable are part of the model, the educational level of the Dutch national civil servant does not need to be considered in this analysis. In other models with other independent variables, this might be completely different and the educational level might have to be considered. 4 of the 5 hypotheses could be confirmed, one had to be corrected. Three independent variables (the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister, the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU) do, when they increase significantly, influence a higher degree of the dependent variable (the feeling of loyalty of the Dutch national civil servant towards his/her own minister), which means that these findings can be expected within the population as well. If the minister should achieve a positive influence on each of the three independent variables, a higher degree of a feeling of loyalty of the Dutch national civil servant towards the minister can be expected. Of course, if the value of one of the independent three variables shrinks, the value of the dependent variables shrinks as well. Two different independent variables (the degree of EU-enthusiasm/ scepticism of the Dutch national civil servants and the national civil servant s amount of time spent on EU matters) had no significant influence on the dependent variable. A positive or negative change of either incurs no significant change in the feeling of loyalty of the Dutch national civil servant towards the minister. It does not matter for the feeling of loyalty of the Dutch national civil servant towards the minister whether a Dutch national civil servant is an EU enthusiast or an EU sceptic and it does not matter either whether he/she spends a lot of time on EU matters or less. In the closing remarks of the analysis, it should be said that the outcome of the variables is very impressive given the fact that the survey used was not especially tailored to the topic of this master thesis and therefore the author was limited to a certain selection of variables. But unfortunately, good data is rare and therefore the survey used was a very good choice measured on the results. The multiple linear regression model was able to explain 23.8% of the dependent variable, which is high given the fact that the model was completely newly developed. 101

103 Based on the theory, this author is confident to say that future examinations on the topic of this thesis could reach an even better explanatory power by including further mechanisms the minister uses to control the Dutch national civil servant, similar to the feeling of a clear mandate for negotiations on EU level from the minister used by this author in his model. The control mechanisms have been discussed extensively in chapter two. Some examples are to establish decision making criteria, regular meetings with the Dutch national civil servant to make him/her feel valued and respect his/her expertise, etc. But scholars have to be aware that many of the different control measures mentioned in chapter two restrict the civil servant s flexibility too much and could therefore cause the exact opposite desired, namely, a shrinking of the feeling of loyalty of the Dutch national civil servant towards the minister (e.g. the responsibility of the agent to make reports about all the collected information and the possible courses of action) (Kassim/ Menon 2003: pp ) (Lane 2013: pp ) (Van Thiel/ Yesilkagit 2008: pp ) (Ennser-Jedenastik 2013: pp ). This thesis should therefore only be seen as a first step to achieve the highest possible feeling of loyalty of the Dutch national civil servant towards the minister and through it minimize agency loss. 6. Conclusion This master thesis combined the Principal-agent relationship between a minister and a Dutch national civil servant and the Europeanisation of the Dutch national civil servant. To the knowledge of this author, an in depth analysis with this particular focus on the Principal-agent relationship between a minister and a Dutch national civil servant with such a broad understanding of Europeanisation has never been done before. The literature gap this master thesis has filled, describes the influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between a minister and a Dutch national civil servant. During the theoretical examination, the centrality of the feeling of loyalty of the national civil servant towards his/her own minister for the Principal-agent relationship was established and it was therefore further analysed what factors might explain a higher or lower degree of loyalty of the agent to the principal (minister) particularly in the context of the 102

104 influence of the Dutch national civil servant s Europeanisation as to the Principal-agent relationship. The question this master thesis has wished to answer is: In what way is there a problem in the principal-agent relationship between the home minister and the Europeanised civil servant and what factors explain higher or lower degrees of loyalty of the agent to the principal? Principal-agent problems form a basic part of Principal-agent relationships. A Principal-agent problem in a Principal-agent relationship between the Dutch national civil servant and the minister arises when both have different interests and hidden action and/or hidden knowledge is discernible. Principal-agent problems can arise in every Principal-agent relationship. When a problem arises, the Dutch national civil servant acts disloyally and in his/her own interest, contrary to that of the minister. The minister will suffer agency loss. The minister will try to control the Dutch national civil servant in order to prevent Principal-agent problems from arising, but this is only possible to a certain degree because control is costly (time and money). The minister is therefore, to some extent, dependent on the loyalty of the Dutch national civil servant (Braun/ Guston 2003: p ) (Gailmard 2014: pp ). So a Principal-agent problem in the Principal-agent relationship between the minister and the Dutch national civil servant can generally arise in their Principal-agent relationship but was the Principal-agent relationship changed by the Europeanisation of the Dutch civil servant, causing an ( additional ) problem in the Principal-agent relationship between the home minister and the Europeanised civil servant? This was rejected as no unusual problems in the principal-agent relationship between the home minister and the Europeanised civil servant occurred. Two changes were examined. First, the general Principal agent relationship could have been broken by the Europeanisation of the Dutch national civil servant. Europeanisation in this dimension means that the Dutch national civil servant, on the one hand, prepares Dutch input for EU level meetings, participates in working groups in the Council of the European Union, participates in meetings of the European Commission or consults his/her counterparts in the other member states and, on the other hand, considers EU policies during national policy-making or supervises the practical application or enforcement of EU policies, and the consideration of EU policies during national policy-making (Mastenbroek/ Princen 2010: pp. 158)). The minister and the 103

105 Dutch civil servant are locked in a chain of delegation and accountability in the Netherlands. As long as both are inside this chain of delegation and accountability, nothing changes in the relationship and accountability can generally be assumed. The Dutch civil servant is responsible for all his/her actions towards the minister. Even when sent to the EU, the Dutch national civil servant stays a part of the chain of delegation and accountability in the Netherlands and is therefore still accountable to the minister. European rules, directives and norms, on the other hand, penetrating the domestic spheres do not influence the institutions or the single actors in a way that might interrupt the Principal-agent relationship between the minister and the Dutch national civil servant. The Principal-agent relationship between the minister and the Dutch national civil servant connects policy making and implementation. It is important that this connection is not interrupted by the Europeanisation of the Dutch national civil servant. But the EU has not the necessary competences (nor the interest) to change this connection. The key structures, the ministries, the cabinet and committees served to channel work on Europe in the national administrations (Laffan 2006: pp. 697). The Netherlands took considerable care to ensure that the high degree of necessary interministerial coordination caused by the Europeanisation did not undermine the national politico-administrative connection. European policy-making and policy implementation have been included in the existing structures with respect to the equality in the Council of Ministers and ministerial autonomy (Harmsen, Robert 1999: pp ). Europeanisation of the Dutch national civil servant did not break the Principal-agent relationship between the minister and the Dutch national civil servant. Second, the Principal-agent problem could have been intensified through Europeanisation and the Principal-agent relationship could thus have changed. To intensify the Principal-agent problem, the Europeanisation of the Dutch national civil servant would either have to intensify the hidden action and/or hidden knowledge or the different interests between the minister and the Dutch national civil servant. The different interests between the minister and the Dutch national civil servant could have been intensified by Europeanisation process because it causes an identity change in the Dutch national civil servant. The Dutch national civil servant is exposed to European integration in the same way as an ordinary citizen (e.g. using the Euro, European parliament elections, etc.). But as Mastenbroek and Princen diagnosed, a small group of civil servants who become 104

106 deeply involved in EU matters and the majority of those who become shallowly involved in EU matters (Mastenbroek/ Princen 2010: pp. 154) are additionally exposed to European integration in the course of their work (e.g. special training for the national civil servants to deal with EU matters, national civil servants having to spend more time on EU matters, etc.) (Mastenbroek/ Princen 2010: pp ). Through this kind of Europeanisation, the identity of the Dutch national civil servant is reshaped. In other words, the Dutch national civil servant s identity is either reinforced towards a more exclusive national identity or an inclusive national identity (Hooghe / Marks 2005: pp ). Either way, this did not intensify the Principal-agent problem. The change in identity (partially) influences the Dutch national civil servants (political) attitudes (Hooghe/ Marks 2005: p.423) but does not cause different interests nor any hidden action or hidden knowledge as regards the minister and the Dutch national civil servant. The Principal-agent problem is not changed by the reshaped identity caused by Europeanisation. But the hidden action and/or hidden knowledge between the minister and the Dutch national civil servant could have been intensified through Europeanisation. If the Dutch national civil servant, on the one hand, becomes more involved in EU matters, spends more time on EU matters and is allocated in more EU related activities (Mastenbroek/ Princen 2010: pp ), he will automatically learn more about the EU and its procedures. This could have increased the hidden knowledge and/ or hidden action. But it is not unusual for the agent to possess certain skills or to be expertly specialized. This is, in fact, one of the reasons why the agent is so valuable to the principal and has been chosen in the first place. If his/her ministry deals with more tasks connected with the EU, the minister needs Dutch national civil servants that are up to the task. The minister, on the other hand, will not be ignorant to the new developments and become more involved himself in order to make sure that his/her interests are acted upon. If the Dutch national civil servant, on the other hand, is specially recruited because of EU connected skills or specially trained for EU related tasks or his/her career development is EU connected (Mastenbroek/ Princen 2010: pp ) the hidden action and/or hidden knowledge could be intensified, as the Dutch national civil servant has additional knowledge at his/her disposal. But again, it is not unusual for the agent to possess certain skills or to be very specialized. The minister obviously needs Dutch national civil servants who can deal with EU matters otherwise he/she would not pick or train Dutch 105

107 national civil servants accordingly. In both situations, the hidden action and/or hidden knowledge did not increase between the minister and the national civil servant and therefore the Principal-agent problem was not intensified. All in all, the Principal-agent problem has not been intensified and the Principal-agent relationship has therefore not changed as to creating additional problems. Europeanisation of the Dutch national civil servant certainly influences the Principal-agent relationship between him/her and the minister but does not exceed the norm. The Principalagent relationship between the minister and the Dutch national civil servant is an average Principal-agent relationship with average Principal-agent problems and the Europeanisation of the Dutch national civil servant does not create further problems to this relationship. Three independent variables (the feeling of loyalty of the Dutch national civil servant towards his/her own policy views, the Dutch national civil servant s feeling of a clear mandate for negotiations on EU level from the minister, the feeling of loyalty of the Dutch national civil servant towards the whole political and administrative System of the EU) explain a higher and a lower degree of the feeling of loyalty of the Dutch national civil servant towards his/her own minister (dependent variable). In other words, when the independent variable increases, so does the dependent variable. If the minister should achieve to positively influence each one of the three independent variables, a higher degree of a feeling of loyalty of the Dutch national civil servant towards the minister can be expected. Of course, if the value of one of the independent three variables decreases, the value of the dependent variables shrinks as well. Two different independent variables (the degree of EU-enthusiasm/ scepticism of the Dutch national civil servants and the national civil servant s amount of time spent on EU matters), on the other hand, have no significant influence on the dependent variable. A positive or negative change of either influences no significant change as to the feeling of loyalty of the Dutch national civil servant towards the minister. It does not matter for the feeling of loyalty of the Dutch national civil servant towards the minister whether a Dutch national civil servant is an EU enthusiast or an EU sceptic nor does it matter whether he/she spends a lot of time on EU matters or less. All these findings can be expected within the population (all Dutch national civil servants who work in a principal-agent relationship with a 106

108 minister (working for the empire)). Therefore, only three of the 5 factors (independent variables) explain higher or lower degrees of the loyalty of the agent to the principal. The influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between the minister and the Dutch national civil servant is therefore not as distinctive as expected. That is not to say that the EU in general has little influence on the administration of the member states. On the contrary, the impact of European integration occurs throughout almost all areas of administration but in each area at different intensities and in different forms (Mastenbroek/ Princen 2010: pp. 154) (Schmidt/ Van den Berg (in print): n.p.) (Meyer-Sahling/ Van Stolk 2015: pp ). The limited effect of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between the minister and the Dutch national civil servant should be seen as an advantage. The Principal-agent relationship between the minister and the national civil servant is one of the most important and central connections in the whole apparatus of the state because it ensures a connection between policy making and implementation. To jeopardise this connection would endanger the effectivity and accountability of the whole state. It is therefore crucially important to ensure the functionality of the Principal-agent relationship between the minister and the national civil servant. The Netherlands took considerable care to make sure that the influences caused by the Europeanisation did not undermine the national politico-administrative connection (Harmsen, Robert 1999: pp ). Although the Principal-agent relationship stays unbroken even when a Dutch national civil servant becomes highly Europeanised, a minister should not underestimate the influence loyalty in his/her Principal-agent relationship with the Dutch national civil servant has and how much the minister actually depends on it. The minister cannot always ensure that the Dutch national civil servant acts in the minister s interest because complete control would be too costly and eat up the benefits the minister can gain from making use of the Dutch national civil servant. The principal has therefore, to some extent, to rely on the loyalty of agents towards him to protect his/her interests. The key for loyalty as to the Dutch national civil servants are good relations. Some interviewees mentioned that they had never seen their own minister. This is foolish because the minister as shown in this examination loses a lot of potential, even if the minister has a good command structure and is capable to direct subordinates. The minister is the leader of his part of the administration. He/she is 107

109 accountable to parliament. Good relations do not mean that the minister should waste his valuable time meeting each of his/her civil servants for hours on end. But even if the policy area the relevant Dutch national civil servant is working on does not have priority, the minister should show his/her presence (e.g. by attending a department meeting with all the civil servants in that department). The ministers seem to underestimate the importance of such seemingly trivial things, as it is important that the Dutch national civil servants knows the minister s intentions and that the minister s interests are realized as best as possible. The strength of this master thesis is not one but many little aspects. The run up to the literature review was done by a scoping study. The scoping study allowed the author to examine the extent, range and nature of research activity (Arksey/ O'Malley 2005: pp ), to extract the essence of the literature body, to summarise and disseminate research findings and to identify research gaps in the existing literature (Arksey/ O'Malley 2005: pp ). Additional strength for this master thesis comes from its mix of methods. By using a survey (quantitative) and interviews (qualitative), the weaknesses of both methods were minimized and the findings could be double checked. This approach guaranties a high degree of validity and reliability. Furthermore, the collected data was evaluated with computational statistics. Thanks to the extensive computer controlled examination, many sources of error and alternative explanations could be eliminated from the outset. Last but not least, the mix between the computer controlled analysis of the survey and the human evaluated interviews allowed the contribution of findings of big data as well as personal knowledge, experiences and active events. The theory used proved to be appropriate for the research of this master thesis despite its shortcomings. As mentioned, two different kind of theories were combined. Critics reproach the Principal-agent relationship for taking only the principal and the agent into consideration and abstracting third parties. To some extent, this is correct. Although the Principal-agent relationship can theoretically be enhanced by several principals and agents, it still gives just a black box model which simplifies assumptions about the principal(s) and the agent(s). Therefore, it needs to be clearly expressed that the Principal-agent relationship cannot provide the complete picture. Furthermore, the Principal-agent theory is, in many aspects, very vague (e.g. What is a principal and what is an agent?), which makes it possible to subsume a lot into the broad concept and very difficult to disprove it. The broad concept is, 108

110 of course, one of its advantages because a lot of phenomena can thus be explained by the Principal-agent theory. But, on the other hand, is it important not to deviate too far from the original theory and misuse it. Nevertheless, the Principal-agent relationship offers a most appropriate starting point to theorize and analyse the relation between a principal and his/her agent (Pollack 2007: pp. 3-21). Europeanisation is the subject of many different disciplines causing a huge number of different definitions. At the beginning of his thesis, the author therefore faced a basic decision: whether to choose a more limited approach based on the political perspective or a wider approach including several perspectives. The author felt the need to take the wider approach in order to be able to analyse the all-around influence of the Europeanisation of the Dutch national civil servant on the Principal-agent relationship between the minister and the Dutch national civil servant. Furthermore, the author has the impression that the different conceptions complement rather than exclude each other (Olsen 2002: ). Unfortunately, this research is also subject to limitations. The findings in this master thesis can only be applied to all Dutch national civil servants working in a principal-agent relationship with a minister (working for the empire). It might be assumed in the light of the findings here that the same results can be found in all EU member states but this thesis provides no proof of this. For further research, a cross sectional design or even a pooled (panel) design on the same topic with several EU member states could be recommended. But to analyse the topic of this master thesis further, in the long run, this author is of the strong opinion that only a closer extensive look into a few Principal agent relationships between ministers and Dutch national civil servants can affect a breakthrough. In other words, this analysis should, instead of looking at the macro level, which concentrated on the Principal agent relationship between the minister and the Dutch national civil servant in more general, have a closer look on the micro level into each single decision between a minister and a Dutch national civil servant. For this purpose, the Principal agent relationship will either have to be developed or connected with yet another theory. 109

111 110

112 3 List of references Adams, Gerald R. / Marshall, Sheila K. (1996): A developmental social psychology of identity: understanding the person-in-context, in: Journal of Adolescence (19), pp Arksey, Hilary/ O'Malley, Lisa (2005): Scoping studies: Towards a Methodological Framework, in: International Journal of Social Research Methodology (8): pp.19 32, ( ). Auswärtiges Amt (2015): Startseite/ Außen- und Europapolitik/ Länderinformationen/ Niederlande/ Innenpolitik, ( ). Bach, Tobias/ Ruffing, Eva/ Yesilkagit, Kutsal (2015): The Differential Empowering Effects of Europeanization on the Autonomy of National Agencies, in: Governance (28), pp , ( ). Baur, Nina/ Behnke, Joachim/ Behnke, Nathalie (2010): Empirische Methoden der Politikwissenschaft, Paderborn: Ferdinand Schöningh Verlag. Beblav y, Miroslav (2009): Europeanization and Bureaucratic Autonomy in the Newmember States: A Case Study of the Agricultural Paying Agency in Slovakia, in: Public Administration (87), pp , ( ). Bohnsack, Ralf/ Marotzki, Winfried/ Meuser, Michael (2003): Hauptbegriffe Qualitativer Sozialforschung, Stuttgart: Barbara Budrich Verlag. Bovens, Mark/ Christensen, Jörgen Grönnegaard/ Jenny, Marcelo/ Müller, Wolfgang C./ Yesilkagit, Kutsal (2010): Legal Europeanization: Comparative Perspectives, Public Administration (88), pp , ( ). 111

113 Braun, Dietmar/ Guston, David H. (2003): Principal agent theory and research policy: an introduction, in: Science and Public Policy (30), pp , ( ). Bundeszentrale für politische Bildung (2016a): Nachschlagen/ Lexika/ Das Europalexikon/ E/ Euroskeptizismus, ( ). Bundeszentrale für politische Bildung (2016b): Nachschlagen/ Lexika/ Das Politiklexikon/ M/ Mandat, ( ). Bundeszentrale für politische Bildung (2016c): Nachschlagen/ Lexika/ Das Politiklexikon/ P/ Politisch-administratives System, ( ). Chaffey, Dave/ Ellis-Chadwick, Fiona/ Johnston, Kevin/ Mayer, Richard (2009): Internet Marketing - Strategy, Implementation and Practice, London: Pearson Education Limited. Dresing, Thorsten / Pehl, Thorsten (2013): Praxisbuch Interview - Transkription & Analyse. Anleitungen und Regelsysteme für qualitativ Forschende, Marburg: Eigenverlag, Transkription.pdf ( ). Diekmann, Andreas (2007): Empirische Sozialforschung, Hamburg: Rowohlt Verlag. Ennser-Jedenastik, Laurenz (2013): Political Control and Managerial Survival in State-Owned Enterprises, in: Governance (27), pp , ( ). Gailmard, Sean (2014): Accountability and Principal-Agent Models, in: Bovens, Mark/ Goodin, Robert E./ Schillemans, Thomas (Editor): The Oxford Handbook of Public Accountability, Oxford: Oxford University Press, pp , 112

114 Agent_Models%282%29.pdf ( ). Gehring, Uwe W./ Weins, Cornelia (2009): Grundkurs Statistik für Politologen und Soziologen, Wiesbaden: VS Verlag. Geuijen, K./ t Hart, P./ Princen, S./ Yesilkagit, K. (2008): The New Eurocrats: national civil servants in EU policy-making, Amsterdam: Amsterdam University Press. Google books (2016): Search books, ( ). Google Deutschland (2016): Google Search, ( ). Google Scholar (2016): Google Scholar search, ( ). Government of the Netherlands (2016): Home/ Government/ Contents/ Members of Cabinet, ( ). Harmsen, Robert (1999): The Europeanization of National Administrations: A Comparative Study of France and the Netherlands, Governance (12), pp , ( ). Hooghe, Liesbet / Marks, Gary (2005): Calculation, Community and Cues: Public Opinion on European Integration, in: European Union Politics, (6), pp Johnson, Janet B./ Reynolds, H.T. (2005): Political Science Research Methods, Washington: CQPress. Jordan, Andrew (2003): The Europeanization of National Government and Policy: A Departmental Perspective, British Journal of Political Science (33), pp , d=33&issueid=02&aid=144651&bodyid=&membershipnumber=&societyetocsession= ( ). 113

115 Kassim, Hussein/ Menon, Anand (2003): The principal agent approach and the study of the European Union: promise unfulfilled?, in: Journal of European Public Policy (10), pp , pal_agent.pdf ( ). Kelle, Udo/ Kluge, Susanne (1999): Vom Einzelfall zum Typus, Opladen: Leske + Budrich Verlag. Kohler, Ulrich/ Kreuter, Frauke (2012): Datenanalyse mit Stata Allgemeine Konzepte der Datenanalyse und ihre praktische Anwendung, München: Oldenburg Verlag. Kühnel, Steffen-M./ Krebs, Dagmar (2012): Statistik für die Sozialwissenschaften Grundlagen, Methoden, Anwendungen, Hamburg: Rowohlt Taschenbuch Verlag. Laegreid, Per/ Steinthorsson, Runolfur Smari/ Thorhallsson, Baldur (2004): Europeanization of Central Government Administration in the Nordic States*, Journal of Common Market Studies (42), pp , ( ). Laffan, Brigit (2006): Managing Europe from Home in Dublin, Athens and Helsinki: A Comparative Analysis, West European Politics (29), pp , ( ). Lane, Jan-Erik (2013): The Principal-Agent Approach to Politics: Policy Implementation and Public Policy-Making, in: Open Journal of Political Science (3), pp , ( ). Mayring, Philipp (2002): Einführung in die qualitative Sozialforschung, Weinheim: Beltz Verlag. Mastenbroek, Ellen/ Princen, Sebastiaan (2010): Time for EU Matters: The Europeanization of Dutch Central Government, in: Public Administration (88), pp Mastenbroek, Ellen/ Princen, Sebastiaan/ Van den Berg, Casper (in print): Euroscepticism and Euro-enthusiasm among Dutch civil servants: In charge but out of touch?, no place: no publisher. 114

116 Meyer-Sahling, Jan-Hinrik/ Van Stolk, Christian (2015): A case of partial convergence: The Europeanization of central government in central and Eastern Europe, Public Administration (93), pp , ( ). Olsen, Johan P. (2002): The Many Faces of Europeanization*, Journal of Common Market Studies (40), pp , ( ). Pollack, Mark (2007): Principal-Agent Analysis and International Delegation: Red Herrings, Theoretical Clarifications, and Empirical Disputes, in: Bruges Political Research Papers (2), pp. 3-21, ( ). Ruffing, Eva (2015): Inside regulatory bureaucracy: When Europe hits home in pharmaceuticals and chemicals, Public Policy and Administration, pp. 1-21, ( ). Schmidt, E./ Van den Berg, C.F. (in print): The Europeanization of the Dutch civil service (for volume for EU presidency), no place: no publisher. Schmidt, Manfred G. (2010): Wörterbuch zur Politik, Stuttgart: Alfred Kröner Verlag. Schubert, Klaus/ Klein, Martina (2011): Das Politiklexikon, Bonn: Dietz Schumann, Siegfried (2006): Repräsentative Umfrage, München: R. Oldenbourg Verlag. Scott, John (2000): Rational Choice Theory, in: Browning, G./ Halcli, A./ Webster, F. (Editor): Understanding Contemporary Society: Theories of The Present, London: Sage Publications, pp Sittermann, Birgit (2006): Europeanisation - A Step Forward in Understanding Europe?, in: Nachwuchsgruppe Europäische Zivilgesellschaft und Multi-Level Governance der Westfälische Wilhelms-Universität Münster, pp

117 Sittermann_Literature_Review_Europeanisation_FINAL2006.pdf ( ). Toshkov, Dimiter (2016): Research Design in Political Science, London: Palgrave Macmillan. Van Thiel, Sandra/ Yesilkagit, Kutsal (2008): Political Influence and Bureaucratic Autonomy, in: Public Organization Review (8), pp , link.springer.com/content/pdf/ /s pdf ( ). Van Thiel, Sandra/ Yesilkagit, Kutsal (2011): Good Neighbours or Distant Friends?, in: Public Management Review (13), pp , equence=1. ( ). 116

118 4 Annex 4.1 Stata do file * ========================================================== * Master Thesis Syntax = * Leiden University = * Master of Science in Public Administration 2015/2016 = * Ulf Schoenert S Ulfschoenert@mailbox.org = * ========================================================== *1. Introduction *The survey was conducted between 12 November 2015 and 2 December 2015 *on the subject The Europeanisation of Dutch officials. The questionnaire *used in the survey was developed by the Ministry of the Interior & Kingdom *Relations and Leiden University and monitored by the Society for Consumer *Research (GfK). The questionnaire comprises a maximum of 14 questions. The *time to answer the questionnaire is approximately 6 minutes. The questionnaire *was sent to 9109 Dutch officials in Europe. The Dutch officials worked for the *Empire, within the Judiciary, municipality, for the provinces, on the water *boards, communal regulations and Research Institutions. Depending on the answers *given the minimum length of the questionnaire was 7 questions and the maximum *length 14 questions. Additional background characteristics were collected the *age of the respondent, his or her sex, the education of the respondent and how *many years he or she had been working within government at the time of the survey. *3529 Dutch officials completed the questionnaire; that is a response rate of 38.7%. * ========================================================== *2. Basic settings 117

119 clear set more off cd "\\uni-mainz.de\dfs\profiles\settings\ulfschoe\desktop\safety copy remotedesktop originale" pwd dir use "Data set Master original.dta", clear * ========================================================== * 3. Delete unused Variables drop a02_1 a02_2 a02_3 a02_4 a02_5 a02_6 a02_7 a02_8 a03_2 a03_3 a03_5 a03_6 a03_7 a03_8 a03_10 a04_3 a05_1 a05_2 a05_3 a05_4 a05_5 a05_6 a05_7 a05_8 a05_10 a05_11 a06 a07_1 a07_2 a07_4 a08 a09 a10 a11 a12_1 a12_2 a12_3 a12_4 a13x1 a13x2 * ========================================================== * 4. Describe Data describe summarize codebook 118

120 * ========================================================== * 5. Converting string variables with non-numeric values into numeric values encode s01, generate(s01) encode a01, generate(a01) encode a03_4, generate(a03q4) encode a03_9, generate(a03q9) encode a04_1, generate(a04q1) encode a04_2, generate(a04q2) encode a03_1, generate(a03q1) encode a07_3, generate(a07q3) encode jaren_overheid_klas, generate(yig) encode Geslacht, generate(sex) encode Opleiding, generate(edu) tab A01, nolab * ========================================================== * 6. Drop all interviewees (civil servants) who do not work at National level (Rijk) drop if S01~=6 * ========================================================== * 7. Rename Variables 119

121 rename a02x2_1 A02QPC rename Leeftijd Age * ========================================================== * 8. Variablelabels tab A01 label variable A01 "EU related work (Part1)" label variable A02QPC "% of EU related work (Part2)" label variable A03Q4 "Civil servants feeling of loyalty towards own minister" label variable A03Q9 "Civil servants feeling of loyalty towards own policy views" label variable A04Q1 "EU-enthusiasm of the civil servant (Part1)" label variable A04Q2 "EU-enthusiasm of the civil servant (Part2)" label variable A03Q1 "Feeling of loyalty of the civil servant towards the whole political and administrative System of the EU" label variable A07Q3 "Clear mandate for negotiations on EU level for the civil servant from the minister" label variable YiG "Years he or she was working within government at the time of the survey" label variable Sex "Sex of the respondent" label variable Age "Age of the respondent" label variable Edu "Education of the respondent" tab A01 * ========================================================== 120

122 * 9. Data manipulation tab A01 tab A01, nolab recode A01 (1=2)(2=1) tab A03Q4 tab A03Q4, nolab recode A03Q4 (2=1)(3=2) (5=3) (4=4) (6=5) (1=.) tab A03Q9 tab A03Q9, nolab recode A03Q9 (2=1)(3=2) (5=3) (4=4) (6=5) (1=.) tab A04Q1 tab A04Q1, nolab recode A04Q1 (1=2) (2=3) (3=4) (4=5)(5=6) (6=7) (8=9) (9=1)(10=10) (11=.) tab A04Q2 tab A04Q2, nolab recode A04Q2 (1=2) (2=3) (3=4) (4=5)(5=6) (6=7) (8=9) (9=1)(10=10) (11=.) tab A03Q1 tab A03Q1, nolab recode A03Q1 (2=1)(3=2) (5=3) (4=4) (6=5) (1=.) tab A07Q3 tab A07Q3, nolab 121

123 recode A07Q3 (2=4)(3=2) (4=3) (5=5) (6=1) (1=.) tab YiG tab YiG, nolab recode YiG (5=1)(4=2) (1=3) (2=4) (3=5) (6=.) * ========================================================== * 10. Valuelabels tab A01 label dir label define container1 /// 1 "No" /// 2 "Yes", replace label value A01 container1 tab A03Q1 tab A03Q4 tab A03Q9 label dir label define container2 /// 1 "Not at all" /// 2 "To a limited extent" /// 3 "reasonable" /// 4 "In the high degree" /// 5 "In very high degree" /// 122

124 99 "missing", replace label value A03Q1 container2 label value A03Q4 container2 label value A03Q9 container2 tab A04Q1 tab A04Q2 label dir label define container3 /// 1 "Fully intergovernmental" /// 2 "2" /// 3 "3" /// 4 "4" /// 5 "5" /// 6 "6" /// 7 "7" /// 8 "8" /// 9 "9" /// 10 "Fully supranational" /// 99 "missing", replace label value A04Q1 container3 label value A04Q2 container3 tab A07Q3 123

125 label dir label define container4 /// 1 "Totally disagree" /// 2 "Mostly disagree" /// 3 "Neither agree nor disagree" /// 4 "Largely agree" /// 5 "Fully agree" /// 99 "missing", replace label value A07Q3 container4 tab Sex label dir label define container5 /// 1 "Male" /// 2 "Female", replace label value Sex container5 tab YiG label dir label define container6 /// 1 "Less than 5 years" /// 2 "5-9 years" /// 3 "10-19 years" /// 4 "20-29 years" /// 5 "30 years or longer" /// 99 "missing", replace 124

126 label value YiG container6 tab Edu label dir label define container7 /// 1 "Otherwise" /// 2 "Primary Education" /// 3 "HAVO, VWO" /// 4 "Higher professional education (HBO)" /// 5 "Exam candidates, University Bachelor, mo B" /// 6 "Lower vocational education (LBO, VMBO)" /// 7 "MAVO, VMBO not exceeding 3 years" /// 8 "Secondary vocational education (MBO)" /// 9 "Masters Degree, KIM, KMA" /// 10 "Masters Degree: PhD", replace label value Edu container7 * ========================================================== * 11. Create new Variable EUEnt(EU-enthusiasm of the civil servant)=a04q2-1 * + Variablelabels + Valuelabels gen EUEnt = A04Q2-A04Q1 replace EUEnt =. if A04Q1==. & A04Q2==. tab EUEnt 125

127 label variable EUEnt "EU-enthusiasm of the civil servant" tab EUEnt label dir label define container8 /// -9 "EU critic" /// -8 "-8" /// -7 "-7" /// -6 "-6" /// -5 "-5" /// -4 "-4" /// -3 "-3" /// -2 "-2" /// -1 "-1" /// 0 "0" /// 1 "1" /// 2 "2" /// 3 "3" /// 4 "4" /// 5 "5" /// 6 "6" /// 7 "7" /// 8 "8" /// 9 "EU-enthusiasm" /// 99 "missing", replace label value EUEnt container8 126

128 * ========================================================== * 12. Create new Variable CSEU (Civil Servants degree of Europeanisation)= Civil servants who answered the question regarding variable A01 with yes (2.) * and who answered the question regarding variable A02QPC + Variable labels gen DTSEU = A02QPC if A01==2 replace DTSEU =. if A01==. & A02QPC==. tab DTSEU label variable DTSEU "Civil Servants degree of Europeanisation" * ========================================================== * 13. Desciptive statistics on the relevant variables sum A03Q4 DTSEU EUEnt A03Q1 A03Q9 A07Q3 Age Sex Edu YiG * ========================================================== * 14. Correlation eta² + Pearsons r *eta²: reg Sex A03Q4 estat esize 127

129 *Pearsons r: correlate DTSEU A03Q4 pwcorr DTSEU A03Q4, sig correlate EUEnt A03Q4 pwcorr EUEnt A03Q4, sig correlate A03Q1 A03Q4 pwcorr A03Q1 A03Q4, sig correlate A03Q9 A03Q4 pwcorr A03Q9 A03Q4, sig correlate A07Q3 A03Q4 pwcorr A07Q3 A03Q4, sig correlate Age A03Q4 pwcorr Age A03Q4, sig correlate Edu A03Q4 pwcorr Edu A03Q4, sig correlate YiG A03Q4 pwcorr YiG A03Q4, sig * ========================================================== * 15. Linear Regression (+ dummy Variables (Sex & CSEU)) reg DTSEU A03Q4 reg EUEnt A03Q4 reg A03Q1 A03Q4 reg A03Q9 A03Q4 reg A07Q3 A03Q4 reg Age A03Q4 reg Edu A03Q4 128

130 reg YiG A03Q4 reg Sex A03Q4 * ========================================================== * 16. Multible Regression (+ dummy Variable (Sex)) (+ beta) (+ regression diagnostic) reg A03Q4 DTSEU EUEnt A03Q1 A03Q9 A07Q3 reg A03Q4 DTSEU EUEnt A03Q1 A03Q9 A07Q3 Age Sex Edu YiG estat vif // (-> estat vif = Test for questionable Multikollinearität) predict fehler, residual // (-> test for: no covariance / correlation between the regressors and the error term (part1)) corr DTSEU EUEnt A03Q1 A03Q9 A07Q3 Age Sex Edu YiG fehler // (-> test for: no covariance / correlation between the regressors and the error term (part2)) corr DTSEU EUEnt A03Q1 A03Q9 A07Q3 Age Sex Edu YiG fehler, covariance // (-> test for: no covariance / correlation between the regressors and the error term (part3)) estat hettest //(-> Test for homoskedasticity (part1)) reg A03Q4 DTSEU EUEnt A03Q1 A03Q9 A07Q3 Age Sex Edu YiG, robust //(-> Test for homoskedasticity (part2)) reg A03Q4 DTSEU EUEnt A03Q1 A03Q9 A07Q3 Age Sex Edu YiG, beta * ========================================================== 129

131 * 17. Multible linear regression with interaction term reg A03Q4 A03Q1 A03Q9 A07Q3 sum A03Q1 if!missing(a03q4,a03q9, A07Q3) return list gen A03Q1mean=A03Q1-r(mean) sum A03Q9 if!missing(a03q4,a03q1, A07Q3) return list gen A03Q9mean=A03Q9-r(mean) sum A07Q3 if!missing(a03q4,a03q9, A03Q1) return list gen A07Q3mean=A07Q3-r(mean) gen A03Q1meanXA03Q9mean=A03Q1mean*A03Q9mean gen A03Q9meanXA07Q3mean=A03Q9mean*A07Q3mean gen A07Q3meanXA03Q1mean=A07Q3mean*A03Q1mean reg A03Q4 A03Q1mean A03Q9mean A03Q1meanXA03Q9mean reg A03Q4 A03Q9mean A07Q3mean A03Q9meanXA07Q3mean reg A03Q4 A03Q1mean A07Q3mean A07Q3meanXA03Q1mean * ========================================================== * 18. Non linear influences of the independent variables + missing variables reg A03Q4 A03Q1 A03Q9 A07Q3 acprplot A03Q1, mspline jitter(10) reg A03Q4 A03Q1 A03Q9 A07Q3 acprplot A03Q9, mspline jitter(10) 130

132 reg A03Q4 A03Q1 A03Q9 A07Q3 acprplot A07Q3, mspline jitter(10) reg A03Q4 A03Q1 A03Q9 A07Q3 ovtest * ========================================================== * End. Save data save Data set Master original.dta, replace 4.2 Survey explainaire Onderzoeksverantwoording Flitspanel Nederlandse ambtenaren en Europa December 2015 Onderwerp Doelgroep Veldwerkperiode Steekproefkader Bruto steekproef 9109 Netto steekproef 3529 De Europeanisering van Nederlandse ambtenaren Rijk, Rechterlijke macht, Gemeente, Provincies, Waterschappen, Gemeenschappelijke regelingen en Onderzoeksinstellingen 12 november 2 december Flitspanel Responspercentage 38,7% Vragenlijst Datum van opleveren 3 december 2015 Wijze van opleveren De vragenlijst is aangeleverd door opdrachtgever BZK & de Universiteit van Leiden en gecontroleerd door GfK, waarna in onderling overleg een definitieve versie is vastgesteld. Voor aanvang van het veldwerk heeft de opdrachtgever een testlink ontvangen. De lengte van de vragenlijst was t = 6 minuten. De totale vragenlijst bestaat uit 14 vragen. Door routing is de minimale lengte 7 vragen en de maximale lengte 14 vragen. SPSS bestand + Excel bestand 131

133 Toegevoegde achtergrondkenmerken: leeftijd, geslacht, opleiding en jaren werkzaam binnen overheid Evaluatie vragenlijst door panelleden: Zinvol onderwerp Leuk om in te vullen Duidelijke vragen De invulduur was niet te lang % 20% 40% 60% 80% 100% 1 - Helemaal niet van toepassing Helemaal van toepassing 4.3 Survey questions English <Basic survey information> Length of interview: 6 minutes Start fieldwork: End fieldwork: 2-12 Master Questionnaire Template Europeanisering van Nederlandse ambtenaren Explanation of codes to indicate type of question: S = Single question - only one answer possible M = Multiple question - several answers possible Q = numerical question - a number must be filled O = Open question - free input field in which text is typed OL = Open Listing - small text fields to fill in names / brands / products V. MAIN QUESTIONNAIRE S SCREENING S01 [S] Are you working in any of the following sectors? 1. National level 2. Judiciary 132