BUREAUCRACY AND DEMOCRACY: BUREAUCRATIC PROFILES OF OECD NATIONS AND CITIZENS’ ATTITUDES TOWARD GOVERNMENT By Seo Youn Choi A DISSERTATION Submitted to Michigan State University In partial fulfillment of the requirements For the degree of Political Science—Doctor of Philosophy 2014 ABSTRACT BUREAUCRACY AND DEMOCRACY: BUREAUCRATIC PROFILES OF OECD NATIONS AND CITIZENS’ ATTITUDES TOWARD GOVERNMENT By Seo Youn Choi This dissertation examines how bureaucracies in democratic nations vary across a set of key bureaucratic traits and how these traits relate to citizens’ attitudes toward government. The bureaucracy is an important government institution for exercising public authority, but it has received less scholarly attention relative to other political institutions. There is no consensus on what attributes of public agencies are important in democratic societies. Moreover, previous studies emphasizing the role of bureaucracies often rely on cross-national indicators without rigorous conceptualization. How bureaucracies are similar and different cross-nationally also has not been systematically examined. This dissertation aims to fill this lacuna. To better understand and compare the administrative apparatus of the state, I propose examining how a set of specific traits are presented jointly in a nation’s bureaucracy. For the empirical analysis, I explore the bureaucracies of the member countries of the Organization for Economic Co-operation and Development (OCED). I employ a nonmetric internal analysis of preference data which is called multidimensional preference scaling (MDPREF). The underlying structure of bureaucratic traits and how these traits are presented in national bureaucracies are investigated. The results indicate that a set of key traits—independence from politics, representativeness, impartiality, competency, and career-based system—are emphasized variously in different nations. The findings suggest that bureaucratic characteristics are not unidimensional and that relationships between traits, as well as the relative importance attached to them, vary across nations. Based on estimations from the MDPREF model, I develop a new measure of Bureaucratic Profiles. A bureaucratic profile represents the relative levels of the key bureaucratic traits presented in a national bureaucracy. Substantively, in general, higher values of this measure indicate that the relative level of impartiality in a country’s bureaucracy increases, and the relative level of competency decreases, compared to other traits. Using this new measure, I examine if there is a systematic relationship between bureaucratic profiles and levels of political support. Individual-level survey data from the 2004 International Social Survey Programmes is merged with a country-level dataset—including Bureaucratic Profiles—to investigate this association in a multilevel analysis. The findings suggest that the emphasis on impartiality compared to other bureaucratic traits is important for maintaining public support for government bureaucracy and regime performance. The linkages between political support and the relative levels of other traits are also discussed. This dissertation makes important contributions to previous studies on comparative public bureaucracies by providing systematic evidence on the similarities and differences of bureaucracies across nations. Further, it makes a fruitful addition to a recent debate on how best to measure characteristics of bureaucracies cross-nationally. This project broadens our understanding of the role of bureaucracies in democratic societies by showing what specific set of bureaucratic traits jointly influence citizens’ attitudes toward government in these societies. Copyright by SEO YOUN CHOI 2014 To Mom and Dad: Thanks for all your love, support, and encouragement. v ACKNOWLEDGEMENTS I would like to express my sincere thanks to faculty members and colleagues at Michigan State University. This dissertation would not have been possible without guidance and support from them. I would like to convey my utmost gratitude to Saundra K. Schneider, the chair of my doctoral dissertation committee, for her ceaseless support and encouragement. Her invaluable guidance helped me to overcome many obstacles, grow as a researcher and political scientist, and ultimately, complete my dissertation. I would like to thank William G. Jacoby for his teaching on scaling methods (along with other statistical methods) and public opinion, and for his valuable comments on my work and considerable support over the years. I would like to thank Eric C.C. Chang for his comments and suggestions on my research, and for his warm encouragement over the years. I would also like to thank Sandy Marquart-Pyatt for her time and assistance with my work. I would also like to thank my great friends and fellow graduate students who I met at Michigan State University. Most of them have now graduated and are all over the country, but they are always with me in spirit. My special thanks to Danielle Carter (at St. Mary’s College of Maryland), Nicholas N. Kerr (at University of Alabama), Helen Lee, and Robert N. Lupton, who provided lots of encouragement and great feedback on drafts of my dissertation. I also owe many many thanks to several fellow graduate students, including Shih-Hao Huang, Dominique Lewis, Niccole M. Pamphilis (at University of Edinburgh), Hsin-Hsin Pan, Johann Park (at Mississippi University), and Wen-Chin Wu (at Asiabarometer). Without these people, my life at Michigan State would have been tougher. vi I am also grateful to my teachers at Ewha Womans University in Korea, Tong Hee Park, Sook Yeon Won, and Chung-Lae Cho, who continuously encouraged me to study abroad and pursue doctoral studies in the United States. Finally, I would like to express my deepest thanks to my family. I have been blessed with endless love, warmness, understanding, resources, and encouragement from my family during my studies. I was able to complete my dissertation because of warm and consistent support from my parents, my husband, my brother, my grandmother, my aunts and uncles, and my parents-inlaws. I am very grateful for their support and prayers. Especially, to my parents, SangDae Choi and MeaJa Oh, thank you so much for your support, encouragement, and countless love and sacrifices during my long journey to the Ph.D degree. Without you, I would not have been able to even begin my study abroad. I am so happy that I can finally finish my dissertation and dedicate it to both of you. And to my lovely husband, Joo S. Lee, thank you so much for your love, support, and patience. Without your consistent push, understanding, and supportive presence, I would not have been able to end this journey. Finally, thank you, God, for guiding and helping me through this academic journey. East Lansing, February 2014 Seo Youn Choi vii TABLE OF CONTENTS LIST OF TABLES ......................................................................................................... xi LIST OF FIGURES ...................................................................................................... xii CHAPTER 1 INTRODUCTION .......................................................................................................... 1 1.1 Background ......................................................................................................... 1 1.2 Why Study Bureaucracy? .................................................................................... 3 1.3 How to Study Bureaucracy? ................................................................................ 6 1.4 The Scope of the Study ........................................................................................ 8 CHAPTER 2 TRAITS OF BUREAUCRACIES IN DEMOCRATIC SOCIETIES.............................. 11 2.1 Bureaucratic Traits Required in Democratic Societies........................................ 11 2.1.1 Overview .................................................................................................. 11 2.1.2 Political Independence of Bureaucracy ...................................................... 14 2.1.3 Representativeness of Bureaucracy ........................................................... 18 2.1.4 Impartiality of Bureaucracy....................................................................... 20 2.1.5 Competency of Bureaucracy ..................................................................... 24 2.2 Relationships between Bureaucratic Traits ......................................................... 25 2.3 Measuring Bureaucratic Traits ........................................................................... 28 2.3.1 Cross-national Indicators of Bureaucracies ................................................ 29 2.3.2 Critiques of Cross-national Indicators of Bureaucracies ............................ 30 2.3.1 A Need for an Alternative Way of Measuring Bureaucracies..................... 34 2.4 Conclusion ........................................................................................................ 36 CHAPTER 3 BUREAUCRATIC PROFILES OF OECD NATIONS: A MDPREF ANALYSIS......... 37 3.1 Bureaucratic Profiles ......................................................................................... 37 3.2 Method: A Multidimensional Preference Scaling (MDPREF) Analysis .............. 40 3.3 Data and Measurements ..................................................................................... 44 3.3.1 Data Sources and Measurements of Key Traits.......................................... 44 3.3.2 Dataset ...................................................................................................... 50 3.4 Analysis ............................................................................................................ 52 3.5 Findings: Configuration of the Traits ................................................................. 54 3.6 Findings: Nation Vectors ................................................................................... 58 3.6.1 Interpreting Nation Vectors ....................................................................... 60 3.6.2 Interpreting the Orientations of the Nation Vectors ................................... 61 3.6.2.1 Nation Vectors between “1:00” and “4:00” .................................... 63 3.6.2.2 Nation Vectors between“5:00” and “8:00”..................................... 64 3.6.2.3 Nation Vectors between “9:00” and “10:00” and between “10:00” to “12:00” ........................................................................ 67 viii 3.6.3 Comparing Specific Bureaucratic Profiles ................................................. 69 3.7 Conclusion ........................................................................................................ 73 CHAPTER 4 RELATIONSHIP BETWEEN BUREAUCRATIC PROFILES AND POLITICAL SUPPORT..................................................................................................................... 76 4.1 Political Support for Regime Performance and Regime Institutions ................... 76 4.2 Existing Explanations for Political Support ...................................................... 79 4.2.1 Economic Factors ..................................................................................... 79 4.2.2 Social Factors ........................................................................................... 81 4.2.3 Political Factors ........................................................................................ 83 4.3 Bureaucracies and Political Support ................................................................... 85 4.4 Bureaucratic Profiles as a Measure of Bureaucratic Traits.................................. 88 4.4.1 New Measure of Bureaucratic Profiles ...................................................... 88 4.4.2 Bureaucratic Profiles and Political Support ............................................... 95 4.5 Method: Multilevel Model Analysis ................................................................ 100 4.6 Data and Measurements ................................................................................... 103 4.6.1 Dependent Variable: Public Attitudes toward Bureaucracies and Satisfaction with Democracy................................................................... 103 4.6.2 Key Independent Variable: Bureaucratic Profiles .................................... 106 4.6.3 Control Variables .................................................................................... 106 4.6.3.1 Individual-level Control Variables ............................................... 106 4.6.3.2 Country-level Control Variables .................................................. 108 4.7 Empirical Results and Discussion: Citizens’ Attitudes toward Bureaucracies Models ............................................................................................................ 111 4.7.1 Random Intercept Models: Attitudes toward Bureaucracies ..................... 111 4.7.2 The Relationship between Bureaucratic Profiles and Attitudes toward Bureaucracies ......................................................................................... 117 4.7.3 Random Slope Models: Attitudes toward Bureaucracies.......................... 123 4.8 Empirical Results and Discussion: Citizens’ Satisfaction with Democracy Models .......................................................................................................... 130 4.8.1 Random Intercept Models: Satisfaction with Democracy ........................ 130 4.8.2 The Relationship between Bureaucratic Profiles and Satisfaction with Democracy.............................................................................................. 136 4.8.3 Random Slope Models: Satisfaction with Democracy ............................. 141 4.9 Conclusion ...................................................................................................... 146 CHAPTER 5 CONCLUSION ........................................................................................................... 148 5.1 General Findings and Contributions ................................................................. 148 5.2 Extensions and Implications for Future Studies ................................................ 156 5.3 Conclusion ...................................................................................................... 160 APPENDICES ............................................................................................................ 162 Appendix A. Description of Variables Used in Chapter 3 ............................................ 163 Appendix B. Supplementary Analysis for Chapter 3: Cluster Analysis ........................ 170 ix Appendix C. Supplementary Analysis for Chapter 4 .................................................... 171 Appendix D. Residual Diagnostics for Chapter 4......................................................... 176 REFERENCES ........................................................................................................... 180 x LIST OF TABLES Table 3.1. Hypothetical Example of the Dataset ............................................................ 51 Table 4.1. Bureaucratic Profiles .................................................................................... 92 Table 4.2. Citizens’ Positive Attitudes toward Bureaucracies and Satisfaction with Democracy (mean) ..................................................................................... 105 Table 4.3. Random-Intercept Multilevel Models of Citizens’ Attitudes toward Bureaucracies, with Individual-level Controls ............................................ 112 Table 4.4. Random-Intercept Multilevel Models of Citizens’ Attitudes toward Bureaucracies, with Country-level Controls ............................................... 116 Table 4.5. Random-Slope Multilevel Models of Citizens’ Attitudes toward Bureaucracies ............................................................................................ 124 Table 4.6. Random-Intercept Multilevel Models of Citizens’ Satisfaction with Democracy, with Individual-level Controls ................................................ 131 Table 4.7. Random-Intercept Multilevel Models of Citizens’ Satisfaction with Democracy, with Country-level Controls ................................................... 134 Table 4.8. Random-Slope Multilevel Models of Citizens’ Satisfaction with Democracy ............................................................................................... 142 Table C.1. Descriptive Statistics .................................................................................. 171 Table C.2. Correlation Matrix ..................................................................................... 171 Table C.3. Multilevel Model of Citizens’ Attitudes toward Bureaucracies: Withinversus Between-country Effects ................................................................. 172 Table C.4. Multilevel Models of Citizens’ Attitudes toward Bureaucracies ................. 173 Table C.5. Multilevel Model of Citizens’ Satisfaction with Democracy: Withinversus Between-country Effects ................................................................. 174 Table C.6. Multilevel Models of Citizens’ Satisfaction with Democracy ..................... 175 xi LIST OF FIGURES Figure 3.1. Hypothetical example of the vector representation ....................................... 43 Figure 3.2. The configuration of variable points ............................................................ 55 Figure 3.3. Full model of nations’ vector terminal points and variable points, with five bureaucratic trait points and the mean direction vector ................................. 59 Figure 3.4. Full model, with the vector terminal points and the five trait points. ............ 62 Figure 3.5. Full model, with the vectors for the United States and Germany .................. 70 Figure 4.1. Full MDPREF model, with the mean direction vector as a reference vector for a new measure, Bureaucratic Profiles..................................................... 89 Figure 4.2. Predictions and approximate 95% confidence intervals versus ranking of citizens’ attitudes toward bureaucracies based on Model 4 in Table 4.4...... 118 Figure 4.3. Full MDPREF model, with the vectors for New Zealand and Poland ......... 120 Figure 4.4. Cross-level interaction effect between generalized trust and social trust on citizens’ attitudes toward bureaucracies ..................................................... 126 Figure 4.5. Predictions and approximate 95% confidence intervals versus ranking of citizens’ attitudes toward bureaucracies based on Model 5 in Table 4.5...... 129 Figure 4.6. Predictions and approximate 95% confidence intervals versus ranking of citizens’ satisfaction with democracy based on Model 10 in Table 4.7 ....... 137 Figure 4.7. Full MDPREF model, with the vectors for Denmark and Slovakia ............ 138 Figure 4.8. Cross-level interaction effect between generalized trust and social trust on citizens’ satisfaction with democracy ......................................................... 144 Figure 4.9. Predictions and approximate 95% confidence intervals versus ranking of citizens’ satisfaction with democracy based on Model 11 in Table 4.8 ....... 145 Figure B.1. Tree diagram for the twenty-one variable points ....................................... 170 Figure D.1. Histograms of standardized level-1residuals and standardized level-2 residuals, for the random-intercept model of citizens’ attitudes toward bureaucracies (Model 4)............................................................................. 176 xii Figure D.2. Histogram of standardized level-1residuals, for the random-slope model of citizens’ attitudes toward bureaucracies (Model 5) ................................. 176 Figure D.3. Histograms of random slopes, for the random-slope model of citizens’ attitudes toward bureaucracies (Model 5) ................................................... 177 Figure D.4. Histograms of standardized level-1residuals and standardized level-2 residuals, for the random-intercept model of citizens’ satisfaction with democracy (Model 10) ............................................................................... 178 Figure D.5. Histogram of standardized level-1residuals, for the random-slope model of citizens’ satisfaction with democracy (Model 11) .................................. 178 Figure D.6. Histograms of random slopes, for the random-slope model of citizens’ satisfaction with democracy (Model 11) ..................................................... 179 xiii CHAPTER 1 INTRODUCTION 1.1 Background What do people want bureaucracies to be in democratic societies? Indeed, bureaucracybashing has been prevalent since the late 1970s; media and politicians have described bureaucracies as wasteful, unproductive, inefficient, unresponsive and antidemocratic form of administration (Suleiman 2003; Wilson 1989). As the role of bureaucracy increases in modern society, bureaucracies are often criticized as a problem to be solved, rather than an instrument that can provide essential goods and services to citizens. Moreover, governmental bureaucracy is often blamed for policy failures and framed as obsolete—or, more colorfully, as an “organizational dinosaur” (Olsen 2005, 1)—that must be minimized and reformed (Pierson 1996; Peters 1993). Does the prevalent disdain for bureaucracies indicate that we need to abandon bureaucracies and diminish their role in the society? I argue that the answer is no, and that the challenges posed instead call for our further attention to broadening and deepening our understanding or what “quality” bureaucracies look like in contemporary societies. A closer look at the attacks on bureaucracies reveals that such criticisms are often led by misinformation and a lack of concrete evidence; and that politicians frequently target bureaucracies as scapegoats for unpopular policy makings (Hill 1992; Pierson 1996). Moreover, there is no consensus established in terms of the required characteristics and traits of governmental bureaucracies in democratic societies. Instead, it is often the case that different aspects of bureaucracies are variously identified as an object for attacks (Olsen 2010). For 1 example, as the role of bureaucracies and their discretion in the policy process increases, politicians often criticize an ineffectiveness and unresponsiveness of bureaucracies. And, along with this, the quest for politicization has increased as a way to secure control over bureaucracies (Peters and Pierre 2004). On the other hand, the very politicization is criticized when it does not yield the expected improvements in bureaucratic performance (Lewis 2008). And, instead, the demand for impartiality and expertise has increases. 1 Indiscreet bureaucracy-bashing is not constructive and does not lead to a “better” bureaucracy. If the reason for criticizing bureaucracies is to improve them, then critics should pause to consider what the criteria are for “good” bureaucracies and whether these criteria are being met. Given the significant role of bureaucracies in the policy process and in governance (Kettl 2006; Pierre and Peters 2000), we need to “take bureaucracy seriously” (Hill 1992) and “rediscover bureaucracy” (Olsen 2005). Otherwise, we may end up with the “dismantled state,” a vicious cycle of distrust in the government and a malfunctioning state apparatus (Suleiman 2003). Despite its importance, bureaucracy has received relatively little scholarly attention relative to other political institutions (Hill 1992; Holmberg, Rothstein, and Nasiritousi 2009; Rothstein and Teorell 2008a). For example, comparative studies examining the role of bureaucracies emphasize their importance in democratic consolidation and economic development (e.g., Bratton and Chang 2006; Diamond 1999; Evans and Rauch 1999; Knack 1 For example, politicization has been advocated in many industrialized countries as a way to increase the political responsiveness of bureaucracies (Peters and Pierre 2004). However, in the United States, the increasing number of political appointees was criticized as one of the reasons that the Federal Emergency Management Agency failed to effectively respond to Hurricane Katrina in 2005 (Lewis 2008); the quest for impartiality increases any time impartial treatment of taxpayers by the Internal Revenue Service is suspected (Janofsky 2004; Stevenson 1997; Weisman 2013). For another example, in Britain, they have long been debated allowing ministers to choose their permanent secretaries—the senior officials; advocates emphasize that bureaucrats are unresponsive and ineffectiveness in the current system, while critics are concerned that it will risk impartiality (Economist 2012; BBC News 2012, 2013). 2 2003; Kohli 2004). However, most such studies do not pay sufficient attention to the specific traits these bureaucracies have and how similarly or differently the properties of bureaucracies vary across nations. Scholars instead seem to assume that bureaucratic characteristics are unidimensional and easily measured by simple cross-national indicators of “bureaucratic quality.” Moreover, the discussion of bureaucracies is often conflated with an abstract conceptualization of the state; thus, the specifics of governmental bureaucracies are little studied in the comparative politics literature. Therefore, this dissertation aims at providing more specific understandings of the public bureaucracy in democratic societies. 1.2 Why Study Bureaucracy? Why is it worth studying the bureaucracy? In other words, is the bureaucracy an important institution in the first place? Given negative images of governmental bureaucracies, one may argue that bureaucracies are incompatible in democratic societies and should be minimized to the greatest extent possible. However, governmental bureaucracies play an inevitable role in preparing, developing and implementing public policies in democratic societies (Heclo 1974; Meier and Bothe 2007; Schneider and Jacoby 1996). Bureaucracies are involved in our daily lives in various ways. For example, food and drug safety and water quality are regulated by governmental bureaucracies. Various types of public goods and social services are delivered by public agencies. And, governmental power and authority are exerted through bureaucracies and public policies signed into law are executed by bureaucracies. In this sense, bureaucracies are essential institutions related to the “output” of the political system, which is different from “input” institutions that channel various public demands to the government, such 3 as party systems and electoral rules (Kettl 2006; Rohrschneider 2005; Rothstein and Teorell 2008a). Public bureaucracies, thus, influence ordinary citizens’ quality of life (Coggburn and Schneider 2003; Whiteley et al. 2010). If governmental bureaucracies are incapable of delivering public goods and services to citizens, or if, for example, clean and safe food and housing are available only to a small group of people in power, then citizens rarely will be satisfied with their lives. In addition, how people view and have confidence in their government is closely related to how bureaucracies work. Particularly, bad experiences with public bureaucrats are likely to engender negative feelings about the government (Hill 1992; Pierson 1996). Given that providing citizens with a high quality of life and preserving the popular belief that government is doing the right thing is important in democratic societies, deepening our understandings of bureaucracies is a critical task. Indeed, as Hill (1991) highlights in his “bureaucratic-centered view of governance,” bureaucracy plays a non-negligible role in the policy process. Moreover, as government’s coordinating power is required in a complex and globalized world, how national governmental bureaucracies perform becomes increasingly important (Kettl 2000; Pierre and Peters 2000). Indeed, the public bureaucracy is one of the critical actors in the policymaking process, which is not only influenced by resources availability and various external constraints, but also impacts the other actors in the decision making process (Hill 1991; Schneider and Jacoby 1996; Schneider, Jacoby, and Coggburn 1997). Bureaucracy is not only a “tool for executing the commands of elected leaders,” but also an “institution with a raison d’être and organizational and normative principles of its own” (Olsen 2005, 3). In other words, bureaucrats participate in policy process by applying their own expertise and knowledge with a degree of autonomy and 4 discretion that extends well beyond merely following what is written and directed by their political masters (Hill 1992; Olsen 2005). Comparative studies adopting state-centric approaches also reveal that bureaucracies are important actors in political and social processes (Skocpol 1985). First, a “usable bureaucracy” (Linz and Stepan 1996, 20) is an important element for a strong state (Migdal 1988). Studies of economic development in less-developed countries emphasize that professional and powerful bureaucracies contribute to economic development by securing property rights for individuals and exerting coercive power when necessary (e.g., Evans and Rauch 1999; Keefer and Knack 1997). Kohli (2004), for example, argues that for rapid industrialization, it is crucial that the state has a competent, “public-spirited” bureaucracy, through which the state pursues and implements cohesive and effective economic policy. The role of bureaucracies, conceptualized as a structure of the modern state, is also emphasized in democratic consolidations (e.g., Bratton and Chang 2006; Diamond 1999; Linz and Stepan 1996; Rose and Shin 2001). That is, a new democratic government needs a bureaucracy at its own disposal that has “the effective capacity to command, to regulate, and to extract tax revenues” (Linz and Stepan 1996, 21). The critical role of bureaucracies is emphasized not only within new democracies, but also within advanced industrialized countries. The state-centered approach to welfare state development in Western European countries and in the United States demonstrates that professional and trustworthy bureaucracies have played an important role in social policy making. Heclo (1974) shows that in Sweden and Britain civil service public administrators consistently exert influences on the development of unemployment insurance and old-age pension programs. The autonomous bureaucracy is also important in Skocpol’s (1995) discussion of the development of the American welfare state: The lack of a centralized and professional 5 bureaucracy was one reason that the United States failed in the 1940s to nationalize benefits defined by the Social Security Act. Studies also show that national bureaucracies contribute to welfare state expansion in industrialized countries (Rothstein, Samanni, and Teorell 2012; Schneider and Ingraham 1984). Since the revival of interest in the state since the 1970s (Skocpol 1985), bureaucracies are considered to be an independent variable in several comparative politics literatures. However, our specific understanding of the cross-national variation in bureaucracies has been advanced little. Case studies have examined the characteristics of bureaucracies in a few countries, but they have resulted in parallel descriptions and do not provide a systematic analysis of the similarities and differences of bureaucracies. Moreover, it is unclear what particular attributes of bureaucracies are relevant for cross-national comparison. It is often assumed, especially in largeN studies, that most advanced industrialized countries have “effective” bureaucracies and that there is no significant variation among them. However, as Olsen (2005) and Holmberg et al. (2009) note, it is possible that we have different patterns of bureaucracies, as there are various forms of political institutions in democratic societies. 2 To explore this possibility and to provide better measures for comparing bureaucracies, more studies on bureaucracies are necessary. 1.3 How to Study Bureaucracy? With what criteria can we compare bureaucracies? As mentioned earlier, previous comparative studies of bureaucracies often assume that there is a “good bureaucracy” that all nations strives to establish, but the meaning of a “good” or a “quality” bureaucracy is not 2 For example, nations have different electoral systems, party systems, and forms of government (i.e., executive-legislative relations). 6 specified. On the other hand, studies of comparative public administration stress the great diversity of bureaucracies across nations (e.g., Painter and Peters 2010; Pierre 1995; Pollitt and Bouckaert 201l; Rose 1985). The approach in these studies is very different from that in a largeN study where the role of bureaucracies is examined using a single indicator. Why are the two approaches so different? Given these differences, how can we compare bureaucracies meaningfully and systematically? Is there a single-best-definition of a “good bureaucracy”? How can variations in public administrative structures and cultures be reconciled with a “good bureaucracy”? To address these inquires, I separate the question of “how national bureaucracies differ from each other” from the question of “which characteristics of bureaucracy are viewed more positively by citizens.” Instead of directly defining a “good bureaucracy,” I suggest first specifying a list of possible key bureaucratic traits—all of which have been discussed previously as important in democratic societies—and using them as the criteria for a comparison. In comparing bureaucracies based upon the degree to which they possess these attributes, the first question can be answered. Thus, Chapter 2 reviews and specifies key traits; Chapter 3 systematically compares bureaucracies cross-nationally by focusing on specific bureaucratic attributes. After identifying the key differences and similarities of bureaucracies across nations, the second question can be examined empirically by testing the link between bureaucratic characteristics and popular attitudes toward government. If a nation’s bureaucracy possesses important attributes for democratic governance that can reconcile the institution’s inherent tension with democracy, then it should be linked to higher levels of public support. Therefore, this analysis will provide some suggestions for the meaning of “good bureaucracies,” which 7 refers to bureaucracies that are good for democratic societies from the citizens’ perspective. Chapter 4 investigates how the similarities and differences in the configuration of the key bureaucratic traits are associated with citizens’ attitudes toward their government. In addition to the lack of consensus on the criteria necessary to compare and evaluate bureaucracies, limited data availability is another difficulty involved in studying bureaucracy. Indeed, the lack of available and reliable data is a primary hindrance to the development of comparative public bureaucracy research (Brans 2003; Dahlstrӧm, Lapuente, and Teorell 2010; Van de Walle 2006). Even the number of employees in the governmental bureaucracy—which would seem to be measured straightforwardly—is debatable, because exact counts that are comparable do not exist (OECD 1997). For this reason, many studies turn to existing crossnational indicators of bureaucracy provided by international organizations and commercial groups, despite their acknowledged limitations. These existing indicators of bureaucracies are briefly discussed in Chapter 2. Rather than relying on these indicators, I collect data that measures various aspects of bureaucracies, especially the key traits specified in Chapter 2. Furthermore, I will focus particularly on the relative importance of these key traits in each nation’s bureaucracy. The data and method that ultimately will be used to compare bureaucracies are discussed in Chapter 3. 1.4 The Scope of the Study This dissertation examines bureaucracies in the thirty-four member nations of the Organisation for Economic Co-operation and Development (OECD) (as of 2012). Focusing on OECD member countries has some merits. First, although the organization obviously does not 8 cover all nations on the globe, countries from different regions are included. With its enlargement since 2007, the OECD now includes not only advanced economies, but also some less-developed countries in underdeveloped regions. Thus, by looking at OECD member nations, bureaucracies are compared not only for Western European countries and the United States, but also Asian, Latin American, and Eastern European countries. Second, more practically, data availability is less problematic for OECD member nations (e.g., Demmke and Moilanen 2010; Dahlström, Lapuente, and Teorell 2010; OECD 2009b). Therefore, limiting the scope of the study to OECD member countries enables me to analyze as many national bureaucracies as possible using existing data. National governmental bureaucracies are the focus of this dissertation. Following Hill (1992), I use the term bureaucracy for the institution despite the fact some scholars choose not to use this term because of its negative connotations. The term bureaucracy emphasizes a “political form of organization (emphasis added)” which exercises government authority and is not an “agent… merely acting on behalf of others” (Hill 1992, 3). Bureaucracy is a political institution that exercises public authority. Governmental bureaucracy is used as a collective term for bureaucrats, which broadly include civil servants and public employees in various branches of public administration and at different levels of government. Moreover, importantly, bureaucracy is different from concepts of the state and governance. It is an important element of the state, but the two terms are not interchangeable. Bureaucracy is an administrative apparatus of the state, which provides capacity for the state to “penetrate society, regulate social relationship, extract resources, and appropriate or use resources in determined ways” (Migdal 1988, 4). Bureaucracy also is different from governance, although it plays an important role in governance. Whereas governance is a broader topic 9 emphasizing the networks between various actors involved in the system, bureaucracy is the “instrument through which all governments exercise their authority” (Suleiman 2003, 7) or the “institutional arrangements used for implementation in a governance system” (Meier and O’Toole 2006, 14). 10 CHAPTER 2 TRAITS OF BUREAUCRACIES IN DEMOCRATIC SOCIETIES What are the criteria for comparing and evaluating bureaucracies? Specifying key bureaucratic attributes is the first step toward better understanding and comparing bureaucracies across nations. Previous literature on comparative bureaucracies has yielded no consensus on which bureaucratic traits are important and required in democratic societies. This chapter reviews these previous studies and highlights four traits—political independence, representativeness, impartiality, and competency—as being the most important. I argue that this set of key traits, rather than any single trait, need to be considered jointly in order for researchers to make systematic and meaningful cross-national comparisons of bureaucracies, as the traits are inter-related. Additionally, this chapter discusses existing cross-national indicators of bureaucracies. The important limitations of prior research identified in this chapter further speak to the need for creating an alternative measure of bureaucracies. 2.1 Bureaucratic Traits Required in Democratic Societies 2.1.1 Overview What are the key bureaucratic traits considered to be important, or desirable, in democratic societies? Normative democratic theories do not provide specific guidelines about the role of a public bureaucracy in democracies (Olsen 2010). But, the relationship between bureaucracies and democracies is often described as a dilemma or a paradox (Etzioni-Halevy 1983; Meier and O’Toole 2006; Pierre and Peters 2000; Thomson 1983). That is, a strong and 11 competent bureaucracy is necessary to allocate resources and deliver public services, but such a bureaucracy should not be too strong as to threaten democratic values and democratic development. Balancing the two traits is difficult, and often seems impossible. As a result, prior studies have not reached a consensus on what traits a bureaucracy in democratic societies should possess. Indeed, “[b]ureaucracy is not an unambiguous complement to democracy” (Suleiman 2003, 33). Moreover, due in part to this dilemma faced by democratic governments, views on the proper role of bureaucracies are diverse: some view it as a necessary element and others assume it as inherently ill-suited to democracy. And, these contrasting views have led researchers in different traditions to identify different bureaucratic attributes as important. On the one hand, bureaucracies are understood as essential institutions for democratic government that provide information and expertise for policy making. Thus, capabilities of developing and delivering policies are required for bureaucracies. From this perspective, professional competence and political insulation of bureaucracies are stressed, as Etzioni-Halevy (1983, 91-92) state that “only a full-fledged, politically independent bureaucracy can safeguard full-fledged democratic procedures” because the democratic state must have “an organization that will not only allocate the resources but will do so by non-partisan criteria.” On the other hand, there are concerns that strong bureaucracies may serve as tools for state repression and/or can exempt themselves from the control of elected politicians (EtzioniHalevy 1983; Kohli 2004). Moreover, as the role of bureaucracies increases, it is worried that an independent and powerful bureaucracy could abuse its discretionary power and that its hierarchical structure and rigid operating procedures make a bureaucracy undemocratic. From this view, bureaucracies are considered as a potential challenge to democracy (Meier and O’Toole 2006; Thomson 1983). And, various ways to reconcile the tension between 12 bureaucracies and democracy have been suggested. Some scholars emphasize ‘internal’ bureaucratic traits, such as professionalism (Friedrich 1940; Miller 2000) and impartiality and fairness (Rothstein and Teorell 2008a), while others argue for ‘external’ mechanisms to control bureaucratic behaviors, including politicization. Therefore, I focus in this project on the bureaucratic traits that are suggested most frequently as important and desirable in previous studies—that is, political independence, representativeness, impartiality, and competency. 3 In the next section, I provide detailed discussion of the importance of each trait separately, following previous studies. Then, because my hypothesis underscores the interrelationships of these traits, I examine some possible relationships between them. 3 Some may argue that there are other traits emphasized as a way to reconcile the tension between democracies and bureaucracies. For example, as bureaucracies are attacked for their unresponsiveness and inefficiency, many reforms in the public sector, including downsizing, privatization and emulating a private sector’s managerial practice, were proposed under the name of the New Public Management (NPM) movement. Under this influence, one may argue that “business-likeness” or “managerialism” is an important trait of bureaucracies. However, I do not include these ideas here because, first, they are based on a different assumption about the role of the state bureaucracy in democratic societies. The proponents of business-like bureaucracy assume that a bureaucracy has the same goals and motivations as private sector organizations and that citizens are akin to customers. It is also assumed that emulating business organizations will make governmental bureaucracies compatible with democracy. However, these assumptions are criticized: citizens are not only customers, but also clients who want professional services from the government bureaucracy (Mintzberg 1996; Suleiman 2003). Also, scholars worry that the emphasis on managerial skills and the purchaser-provider relationship can demoralize public bureaucrats who are committed to the public interest (Campbell 2004; Suleiman 2003). Second, I think that it will be more reasonable to discuss how differently nations have responded to those reform ideas, rather than examining how differently they are presented in nations’ bureaucracies as key traits. 13 2.1.2 Political Independence of Bureaucracy One of the important attributes of bureaucracies is an independence from political influence. Since the late nineteenth century, there has been an effort to eliminate political patronage in hiring and firing public employees. Public bureaucracies in Western European countries, for example, were traditionally designed to “reduce … the risk of too much political influence, corruption, misconduct, the exercise of private interests, and instability of government (Demmke and Moilanen 2010, 25).” In the United States, too, civil service reform was aimed at developing a merit system based on three fundamental principles: “competitive examinations for entrance into the public service, … relative security of tenure for [those] employees [and] the [political] neutrality of the civil service” (Van Riper 1958, 98-100). In democratic societies, it is important to allocate resources and deliver public services without bias and favoritism; for this, bureaucracy’s insulation from political pressure is required (Etzioni-Halevy 1983). When bureaucracies fall under political influence—for example, if their employment is totally dependent upon the whims of elected officials—it is possible that the provision of public goods is biased. Incumbents could easily direct bureaucrats to allocate resources in favor of a particular group in exchanging for past or future support from the group. This possibility opens the door to widespread corruption and bribery. Moreover, if elected politicians can freely recruit and dismiss government bureaucrats, then it is possible that public sector jobs will be offered as a reward for political supporters. Thus, bureaucracies that are not independent from political influence are likely to lead to unequal allocation of, and unfair access to, state resources and public services. 14 In addition, under the spoils system—in which bureaucrats are replaced whenever the incumbent party changes—governmental bureaucracies become ineffective and inefficient. If there is a significant turnover at each election, new personnel need to be professionalized too often, reducing bureaucratic effectiveness and morale. On the other hand, when political independence is guaranteed, it contributes to continuity in public administration and the accumulation of professional experience for individual bureaucrats. Also, as decision-makings are not influenced by political pressure, administration of policies will be free from drastic changes. Moreover, bureaucrats who are independent from political influence can be a signal of politicians’ credible commitment to pre-electoral promises (Levi and Sherman 1997). Knott and Miller (2006, 2008) and Miller (2000, 2005) argue that hiring a “trustee-type” bureaucracy— whose key characteristic is political independence—indicates politicians’ attempt to commit themselves to their pre-electoral promises.4 This outcome occurs because a trustee-type bureaucracy is supposed to enforce policies as previously promised even if political masters’ own interests change after an election. However, if politicians hire an “agent-type” bureaucracy, which is not politically independent, but rather responsive to political masters’ short-term interests, citizens’ perception of elected officials’ credibility may be destroyed. 4 They apply Thomas Schelling’s idea of credible commitment to develop their Trustee theory. Schelling-style delegation can be illustrated by the case of a wealthy mother who hires a trustee in order to protect her child from a kidnapper (Knott and Miller 2008). If the mother delegates her fortune to a trustee, then he will be concerned only with her long-term interest, the shrewd management of her wealth. If she delegates to an agent, however, the agent will respond to her short-term interest, which of course is paying a ransom to the kidnapper in the event her child is taken. Therefore, the kidnapper will not attempt to kidnap her son if, and only if, the mother hires a trustee, as hiring a trustee signals her credible commitment not to pay a ransom in the event of a kidnapping. 15 The emphasis since the late 1970s on politicization and political control of bureaucracies stands in stark contrasts to an earlier focus on the establishment of political independence. 5 This more recent emphasis has led to an increasing number of political appointees in the public sector (Peters and Pierre 2004). Politicization efforts are justified as a way of preventing arbitrary bureaucratic behaviors and ensuring political responsiveness and democratic accountability (Moe 1985; Peters and Pierre 2004). However, as Suleiman (2003) states, politicization was not previously proposed as a way to improve bureaucracies. Instead, the need for political control has been stressed only in response to observations of the increased number of political appointees. In other words, the purported merits of politicization are used to justify the reality. Moreover, the effectiveness of politicization remains in question.6 Contrary to the expectation of politicization advocates, the consequences of politicization are often viewed negatively (Peters and Pierre 2004). Scholars argue that increasing politicization may harm the bureaucrats’ expertise (Lewis 2008) and destroy their morale (Plowden 1984; Suleiman 2003). Instead, the 5 Politicization can be practiced in various ways, but “at the most basic level” it can be defined as “the substitution of political criteria for merit-based criteria in the selection, retention, promotion, rewards, and disciplining of members of the public service” (Peters and Pierre 2004, 2). 6 Previous studies have surveyed the practice of political appointments in a single country (e.g., Page and Wright 2007; Peters and Pierre 2004), but there is no empirical study that systematically examines the effect of increasing political control over bureaucracies, or, importantly, its consequences for citizens. One reason for this shortcoming is that these studies are based on a principal-agent theory and its assumptions (Eisenhardt 1989; Weingast 1984) and focus only on the relationship between bureaucracies and politicians. These studies assume that bureaucrats are self-interest maximizers and the legitimacy of their behaviors is “automatically suspect,” whereas their principals—elected politicians—provide an accurate representation of the public’s preferences and their decisions are “necessarily legitimate” (Meier and O’Toole 2006, 12). However, other scholars argue that a bureaucracy’s accountability to politicians does not necessarily guarantee good outcomes in the eyes of the public. Studies also suggest that politicians do not always represent the public perfectly; a moral hazard problem exists in their relationships with citizens (Knott and Miller 2008; Miller 2000, 2005; Powell 2004). Their assumptions about bureaucrats also are empirically challenged: for example, some evidence shows that bureaucrats tend to be public-minded rather than self-interested (e.g., Brewer 2003; Campbell 2004; Suleiman 2003). 16 empirical studies provide some supportive evidence for the merits of politically independent bureaucracies. They find its role in promoting economic growth (Keefer and Knack 1997; Evans and Rauch 1999; Rauch and Evans 2000), in consolidating new, transitional democratic states (Kohli 2004; Weimer 2005) and in curbing corruption (Dahlstrӧm, Lapuente, and Teorell 2012).7 Therefore, I specify in this study political independence, and not politicization, as one of the essential traits of bureaucracies. A politically independent bureaucracy is defined as being insulated from political pressures in the decision-making process, as well as in the recruitment and dismissal of bureaucrats. Independent bureaucrats are able to provide policymakers with objective advice and implement public policies, without favoring a particular political ideology. It is important to note that a bureaucracy free from political pressure does not imply the total separation between politics and administration, the so-called “politics-administration dichotomy” (Hill 1992). That is, the trait does not indicate that bureaucrats do not make value-laden decisions or that they implement policies without being personally involved in policymaking process. It also does not necessarily indicate that bureaucrats’ political activities are completely prohibited.8 7 Note that several empirical studies show that the political independence of bureaucracies is highly correlated with a meritocratic recruitment. And, Rauch and Evans (2000), in their study of less developed countries, find some support for the relationship. Dahlstrӧm, Lapuente, and Teorell (2012) find similar results in their study of fifty-two, developed and developing countries. 8 A similar argument is made elsewhere. For example, Asmern and Reis (1996, 8-9: quoted from Svara (2001)) note that “[N]eutrality does not mean that top-level civil servants cannot or should not be involved in the articulation of public policy. Indeed, senior officials are professionally and morally obliged to provide their political leaders with the best policy alternatives based on sound argument. The expectation that they will render these services from a non-partisan position is the crux of the matter.” 17 2.1.3 Representativeness of Bureaucracy Descriptive representation in the bureaucracy is argued to reconcile the tension between bureaucracies and democracy (Meier and O’Toole 2006; Selden, Brudney, and Kellough 1998). When an “upper-class-dominated bureaucracy” was replaced with “performance-oriented recruitment” in Britain and France in the eighteenth and nineteenth centuries, for example, the expectation was developed that greater representation of the working class in the administrative apparatus of the state makes civil servants more responsible (Subramaniam 1967, 1011). In the United States, the idea of recruiting ordinary citizens was used in the nineteenth century as a method for ensuring that the a merit-based civil service system would be truly democratic and fair (Schultz and Maranto 1998; Van Riper 1958). A representative bureaucracy is argued to reconcile tensions with democracy and promote democratic values for several reasons. First, it is claimed to reduce arbitrariness in administrative organizations and to prevent them from violating the norms of democratic governance (Keiser et al. 2002; Meier and Nigro 1976; Meier and O’Toole 2006; Mosher 1982; Selden 1997). That is, descriptive representation can prevent bureaucratic abuse of discretion because a civil servant who shares the same demographic characteristics of the population— regarding gender, race, religion, and class—will share citizens’ values, which will constrain bureaucrats’ behavior. Second, a representative bureaucracy is believed to increase social equality. Recruiting bureaucrats that represent the social and economic structure of society has inherent value: it fosters equal access to bureaucratic positions for all citizens. It also has a symbolic meaning by contributing to the elimination of societal prejudices and promoting 18 economic advancement of minority groups (Peters 2001; Rosenbloom 1974), as the interests of diverse groups will be represented in the policy formulation and implementation process. Prior studies of representative bureaucracies examine the effect of ‘passive’ representation—that is, whether ‘passive’ representation is translated into ‘active’ representation—in the context of the United States.9 They explore racial or gender compositions of bureaucratic organizations and the effect of representativeness on outcomes, especially in education (Keiser et al. 2002; Meier and O’Toole 2006; Meier, Wrinkle, and Polinard 1999; Pitts 2005), housing loan decisions (Selden, Brudney, and Kellough 1998; Sowa and Selden 2003) and child support enforcement (Wilkins and Keiser 2006). The findings generally indicate that minority and female representation in bureaucracies produces policy outcomes representing minority and female interests, respectively. It is assumed that greater benefits distributed to minority groups contribute to economic equity, implying that a representative bureaucracy is compatible with democratic values (Keiser et al. 2002; Meier, Wrinkle, and Polinard 1999). The theory is criticized on grounds that the practice of descriptive representation may cause “partiality” (Lim 2006; Subramaniam 1967), or that the benefits for minority groups are achieved at the expense of non-minority groups. If these criticisms are true, then a representative bureaucracy may not be compatible with democratic values; however, other evidence exist refuting these critics. Meier et al. (1999) find that nonminority students, as well as minority students, perform better in the presence of a representative bureaucracy. Moreover, Nicholson- 9 Note that earlier studies focus on whether and to what extent a national bureaucracy is (demographically) representative, with a normative assumption about a representative bureaucracy. Comparisons across a few number of countries are discussed in Subramaniam (1967) and Aberbach, Putnam and Rockman (1981, Chapter 3), for example. Also, Meier (1975) shows that U.S. Civil Servants better reflect the social characteristics of the American public, in their education and occupation, compared to other countries, but this descriptive representativeness decreases in higher positions in the U.S. Civil Service. 19 Crotty, Grissom, and Nicholson-Crotty (2011) show that even in the case of the minority group benefiting at the expense of a nonminority group, the favoritism happens only until the program benefits are distributed equally across both groups. These studies provide evidence to support the idea that a representative bureaucracy is compatible with democratic values. It is also worth noting that managing diversity in the public sector workforce and providing equal opportunities are important aspects of bureaucracies in developed nations (Auer, Demmke, and Polet 1996; OECD 2009a). In particular, many countries have made efforts to increase women’s participation in the public sector. The percentage of women in the public sector workforce varies considerably across nations and across the level of positions within a country (OECD 2009a). Descriptive representation based on religion or race and ethnicity are also important, but they are very difficult to compare cross-nationally because of the degree of heterogeneity in coding and reporting these statistics. The percentage of female representation in the public sector workforce can be much more easily compared across countries. Therefore, this dissertation focuses on gender representativeness in the public bureaucracy as one of the key bureaucratic traits in democratic societies. 2.1.4 Impartiality of Bureaucracy Procedural fairness or bureaucratic impartiality also has been argued as an important element in democratic societies. Levi (1998) and Levi and Sherman (1997), for example, maintain that voluntary public compliance is achieved through an impartial and professional bureaucracy that implements public policies fairly and predictably. Studies stress that impartiality is an especially important principle for “output” institutions, including bureaucracies 20 (Rohrschneider 2005; Rothstein and Teorell 2008a). Rohrschneider (2005, 853) argues that impartial and fair implementation by output institutions of a regime is perceived by citizens as “the capacity of the entire regime to account for their interests” and, thus, is relevant to citizens’ perceptions of representation. Similarly, some studies emphasizing the role that “quality of government” plays in societies, especially by scholars at the Quality of Government (QoG) Institute at the University of Gothenburg, argue that impartiality is a basic, essential principle for government institutions that exercise public authority (Holmberg and Rothstein 2012; Rothstein and Teorell 2008a; Teorell 2009).10 Rothstein and Teorell (2008a, 170) define bureaucrats as being impartial when they do “not take into consideration anything about the citizen/case that is not beforehand stipulated in the policy or the law” and are not influenced by their own interests in exercising authority. There are two things to stress in this conceptualization. First, it is different from “an old-style Weberian rigid rule following, personal detachment, or the lack of creativity and flexibility by the people working in the public sector” (178). That is, policies are not designed with detailed rules that bureaucrats follow, but rather allow “rooms for flexibility and professional judgment” where impartiality is the basic bureaucratic norm (Rothstein and Teorell 2008b, 203). Second, impartiality is emphasized here as a procedural norm. Thus, it is relevant to how public authority is exercised rather than the contents of policies.11 10 Although their discussion focuses on governance or the “quality of government,” which is broader than the bureaucracy, the core of the output institution—the institution that exercises public authority—is bureaucracies and civil servants. Therefore, their discussion of “quality of government as impartiality” implies that impartiality is one of the key traits of bureaucracies. 11 For example, Rothstein and Teorell (2008a, 170) state, “the enactment of [social policies that provide supports for the poor families with children] would not break the principle of impartiality, while denying such allowances for families from to a certain ethnic groups or parents with a certain sexual orientation when implementing the policy would.” 21 The importance of procedural fairness and bureaucratic impartiality has been examined cross-nationally. For example, Rohrschneider (2005) uses the procedural quality of bureaucracy to explain citizens’ perceptions of the representation of parliaments and governments, and he finds a positive relationship between the two in advanced democracies. With more sophisticated indicator of impartiality, using the QoG Expert Survey, 12 Teorell (2009) examines several theoretical arguments regarding the impact of impartiality on a variety of governmental outcomes. He finds positive associations between impartiality and economic growth, institutional trust and citizens’ personal happiness. At the individual-level, the impact of perceptions of procedural fairness on citizens’ attitudes toward government is investigated in prior studies. Levi and Sacks (2009), in their study of sub-Saharan African countries, find a positive association between citizens’ perceptions of procedural fairness and their willingness to comply with governments’ tax decisions. In the context of the United States, Hibbing and Theiss-Morse (2001) argue that procedural fairness plays an important role in shaping citizens’ orientation toward government; Tyler (1990) discusses its relationship with law compliance. Moreover, Kumlin and Rothstein (2005) argue that perceptions of fairness have an important influence on building social capital in their analysis of Swedish survey data. One of the critiques of impartiality is that it is impractical, and undesirable, to specify all circumstances of policies in advance (Wilson 2008). However, as noted earlier, impartiality is different from mere rule-following or the formalistic and impersonal execution of assigned tasks; it indicates equal respect and concerns for participants in the policy process (Rothstein and 12 The QoG Expert Survey is web-based survey answered by public administration scholars. The data from the QoG Expert Survey will be used for the empirical analysis in this dissertation and more information about the survey data, including the measure of impartiality, will be discussed in Chapter 3. 22 Teorell 2008a, 2008b). Thus, impartiality does not require very detailed policies and rules for all situations. Another critique is that impartiality may be a necessary but not a sufficient condition for bureaucracies (Longo 2008; Stensӧta 2012). I will address this critique later, as I agree with it. And, indeed, impartiality will be considered as one of the traits of a bureaucracy in democratic societies, rather than the only one. An impartial bureaucracy is different from “political impartiality” (Christoph 1957), which indicates prohibiting bureaucrats’ political activity. It is worth mentioning that political independence and impartiality can be related, but they are distinct traits. Political independence may be required for impartiality because non-partisan decisions are difficult to make if bureaucrats are under political pressures. But, insulation from political influence does not guarantee impartiality because such bureaucrats may behave arbitrarily and favor a certain interests over others. Another distinction is that the relationship between politicians and bureaucrats is the key concern in the discussion of political independence, whereas it is the relationship between bureaucrats and citizens for impartiality. Impartiality can be implanted in bureaucracies through professionalism, a merit-based recruitment and a security of tenure (Rothstein and Teorell 2008a). If bureaucrats are employed based on political connections, instead of skills and merit, then it is difficult to expect impartial bureaucratic behavior. Moreover, impartiality will be assured when it is ingrained into the ethos of public service (Rothstein and Teorell 2008a). 23 2.1.5 Competency of Bureaucracy Competency is another key bureaucratic trait because it enables democratic governments to work effectively in developing and implementing public policies. For bureaucracies to produce and deliver public goods and services, a skillful and competent workforce is required (Etzioni-Halevy 1983; Peters 2001). Also, when discretion is authorized, bureaucrats’ decision making is expected to be based on their expert knowledge and professional standards. Although studies often do not explicitly discuss competency, it is taken for granted as an important characteristic of bureaucracies.13 It would be of no use for democratic government if bureaucracies are incompetent in the first place (Aberbach and Rockman 2000; Heclo 1977). There has been a concern that bureaucracies staffed with experts can be too strong to be under popular control and, thus, are incompatible with democratic societies. That is, ‘professionals’ or ‘experts’ are difficult to control because they may believe they know what is best (Kearney and Sinha 1988). If an elite group of people who attain exceptionally high education and professional training work in the governmental bureaucracy, they may not listen to citizens’ opinions. Thus, there is a possibility that such bureaucrats abuse their authority. However, Kearney and Sinha (1988) maintain that competent bureaucrats do not necessarily 13 Basically each country wants to attract and retain intellectually competent and qualified personnel, although the methods of recruitment can vary across nations. Some nations, such as Britain, advocate generalists and select bureaucrats based on their intellectual abilities, whereas other nations, such as the United States, emphasize specialist expertise and training (Peters 2001). Although different types of expertise and training are sought, the recruitment of bureaucrats in most of nations basically requires post-secondary education in order to select skillful and competent bureaucrats. Bonuses and higher salaries are also provided for better performance to keep competent bureaucrats (Peters 2001). 24 threaten democratic government, especially when they are professional. That is, professionalism can provide its own constraints on bureaucrats (Knott and Miller 2008). The importance of competency and expertise is implicitly suggested in previous studies. For example, critiques of politicization disclose its importance. Politicization arguments overlook the fact that political control is valuable for democratic government only if bureaucrats are competent and professional. In practice, tradeoffs between politicization and competency have been observed, which the original argument of politicization does not consider. That is, politicization increases the rate of executive turnover, reduces the incentives for career professionals to develop expertise and also decreases the attractiveness of a public sector career for qualified applicants (Lewis 2008; Suleiman 2003). Furthermore, when there is a small pool of qualified people, it becomes difficult to find a person who has both political loyalty and competency and expertise (Peters and Pierre 2004). These unexpected consequences of politicization reveal the problems governments may face when bureaucratic competency is not secured. 2.2 Relationships between Bureaucratic Traits The four traits outlined in the previous section are analyzed separately in previous studies. However, in order to better understand bureaucracies, we should consider all four traits together. For example, a politically independent but totally incompetent or partial bureaucracy is undesirable; and, a representative but partial bureaucracy is not the one we expect. It is critical to examine how these traits are represented in a bureaucracy simultaneously, because they are not mutually exclusive. Let us look at some of the ways in which these traits might be related. 25 The possible relationships are complex, and there is no one, predetermined relationship that I would expect to uncover. Some traits are closely related, as illustrated earlier in the discussion of political independence and impartiality. Others are potentially in conflict to each other. For example, let us think about representativeness (here, gender representativeness) and impartiality. On the one hand, the representative bureaucracy theory posits that bureaucrats will share the values with the sub-population with whom they share demographic origins. If this shared value is different than impartiality, then there is a possibility that the two traits will be at odds. On the other hand, the potential tension between them will be lessened if females’ values are similar to impartiality. For instance, many studies examining behavioral differences across genders have argued that women are more likely to be fair and honest, and less likely to be individualistic and opportunistic, than men (Dollar, Fisman, and Gatti 2001). If this theory holds, then a bureaucracy with more female workers can also be more impartial. 14 For another example, let us think about the relationship between representativeness and competency. Given that bureaucracies in many European countries are male-dominated (Auer, Demmke, and Polet 1996), it is expected that a strict application of a formal entrance exam and career-based promotions may be a barrier to establishing a representative bureaucracy. Many nations have introduced various measures—including Affirmative Action for Women and gender quotas—designed to increase the number of females in the public sector workforce (Auer, Demmke, and Polet 1996). Thus, it is possible that these means to achieve greater 14 Similarly, studies show that a representative bureaucracy with respect to race is constrained by other traits such as professional norms and distributional equity (Nicholson-Crotty, Grissom and Nicholson-Crotty 2011; Watkins-Hayes 2011) and, thus, does not necessarily conflict with impartiality. 26 representativeness lead to a loosening of the requirements in personnel management systems and less emphasis on competency. The relationship between representativeness and competency is not always understood as conflicting, though. Eliminating prejudices against minority communities through the establishment of a representative bureaucracy is considered components of (or part of the broader meaning) of bureaucratic competence (Peters 2001). Further, as women’s educational opportunities and achievements increase, it may not always be the case that a bureaucracy with more female workers has less expertise and competency than a male-dominated one What about the relationship between competency and other traits? It seems that bureaucratic competency does not always comport with political independence. In order to attract more skilled and competent personnel, some nations have introduced a position-based recruitment system, which is the direct opposite to a career-based one, and open the positions (especially senior positions) to the private sector workers. In a position-based system, the position can be filled with external recruitment to find the best candidate for that position (Auer, Demmke, and Polet 1996). And, in many incidences, this decreases the level of political independence because it could be used as an opportunity to attract loyal partisans to government bureaucracies (Suleiman 2003). Thus, this may make a bureaucracy vulnerable to political influence and favoritism (Suleiman 2003). In addition, bureaucratic competency does not always guarantee impartiality, although the traits are related theoretically. For example, bureaucrats who are recruited through a competitive exam and have university degrees do not necessarily behave impartially. Given the complexity of these possible relationships among traits, which can be compatible or conflicting, and also can depend upon context, I argue that examining 27 bureaucracies in terms of how they reflect these four key traits simultaneously will provide a more comprehensive understanding of bureaucracies. How do we measure and compare the characteristics of bureaucracies we have discussed thus far cross-nationally? The next section reviews the measure of bureaucracies used in prior research and discusses their limitations. 2.3 Measuring Bureaucratic Traits Empirical work analyzing bureaucracy as an independent variable and in the context of a large-N studies usually rely on cross-national indicators of bureaucracies developed by international organizations or commercial groups.15 Indeed, advancement of international indicators for the “quality” of public administration, as a part of state capacity, has allowed researchers to rank bureaucracies across nations and also enabled cross-national studies to take bureaucracies more seriously (Van de Walle 2006). However, despite the attractiveness of these cross-national indicators, they are criticized due to some of their important limitations. Particularly, it is questionable if they appropriately reflect the key traits of bureaucracies in democratic societies and capture the important cross-national variation of bureaucracies. Let us briefly review these existing indicators of bureaucracies. 16 15 Exceptions include studies analyzing the “Weberianness Scale” developed by James Rauch and Peter Evans (Evans and Rauch 1999; Rauch and Evans 2000), which focus on the bureaucratic structure of economic agencies in the less-developed countries, and the study of good government (where “good” is defined as “good-for-economic development”) by La Porta et al.’s (1999). More recently, researchers in the Quality of Government Institute in University of Gothenburg, Sweden, have studied the dimensions of bureaucracy and focused on impartiality as a key principle for good government (e.g., Holmberg and Rothstein 2012). 16 Note that although this chapter focuses on two indicators, the general critiques are applied to other indicators on the quality and performance of bureaucracies. Other measures include the “Government Efficiency” indicator in the International Institute for Management Development (IMD)’s World Competitiveness Yearbook and the “Bureaucratic Delays” indicator from the 28 2.3.1 Cross-national Indicators of Bureaucracies The Government Effectiveness indicator from the World Bank’s Governance Indicators (WGI) is one of the most popularly used measurement of bureaucracies.17 Public accessibility and a wide coverage—about 200 countries are covered over the period of 1996 to 2011—make this indicator attractive to both researchers and practitioners alike. The Government Effectiveness indicator summarizes and “combine[s] responses on the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government’s commitment to policies” (Kaufmann, Kraay, and Mastruzzi 2005, 130). It is an estimated value based on an aggregation of various variables from individual data sources that are relevant to government effectiveness, and ranges from -2.5 (weak) to 2.5 (strong). About thirty individual indicators are aggregated to estimate the values of the WGI measures; these individual items are collected from a number of data sources, from expert evaluations to public opinion surveys (Kaufmann, Kraay, and Mastruzzi 2005, 2008). The aggregation is conducted using a complex model called an ‘unobserved components model,’ where the estimated value is the weighted average of rescaled scores for various data sources (Kaufmann, Kraay, and Mastruzzi 2010). Business Environmental Risk Intelligence’s (BERI) Operation Risk Index. However, those indicators focus on measuring levels of red tape and corruption, and do not provide information about how, exactly, government bureaucracies are characterized. Van de Walle (2006) critically accesses these existing indicators and concludes that their common problems are that they are not clear about what is actually measured and that they provide partial, and often biased, views of bureaucracies. 17 The Government Effectiveness indicator, among other indicators of WGI, specifically measures the “management” dimension, which is relevant to bureaucracies (Pollitt and Bouckaert 2011; Van de Walle 2006). Thus, I focus on this indicator. But, an index, comprised of three or five indicators from WGI, is used most often in previous studies. For example, such an index is used as a measure of “quality of (output) institutions” (Rohrschneider 2005), “general governance” (Gilley 2006) or “stateness” (Bratton and Chang 2006). 29 The Bureaucracy Quality indicator from the International Country Risk Guide (ICRG), conducted by Political Risk Services (PRS) Group, is another frequently used indicator. The use of this indicator is advocated in the economic growth literature as a measure of a state’s capacity to protect property and contractual rights (e.g., Keefer and Knack 1997; Knack and Keefer 1995). It is also used in studies that try to capture the general administrative capacity of the state (e.g., Bäck and Hadenius 2008; Charron and Lapuente 2010; Dahlstrӧm, Lindvall, and Rothstein 2012; Rothstein, Samanni, and Teorell 2012). The ICRG’s Bureaucracy Quality indicator is one of the political risk components that the PRS Group uses to assess the political stability of countries over time, and is available for the time period 1984 to 2012. This indicator ranges from zero (the lowest quality) to four (the highest quality) and the points are assigned by PRS editors based on a series of pre-set questions (PRS). According to their methodology report, Bureaucracy Quality measures “institutional strength and quality of the civil service, [which is] strength and expertise to govern without drastic changes in policy or interruptions in government services.” Specifically, the “quality” of bureaucracy is rated according to an overall evaluation based on a criterion that combines the autonomy, expertise, political independence, and mechanism for recruitment and training (PRS). These elements are assumed as key traits that determine the “quality” of bureaucracy. 2.3.2 Critiques of Cross-national Indicators of Bureaucracies Despite the popular usage of these indicators, there are some important limitations. First, the lack of clear conceptualization and operationalization of “government effectiveness” or the “quality of bureaucracy” is one of the fundamental problems. This issue is more serious in the 30 WGI’s Government Effectiveness indicator (Dahlstrӧm, Lapuente, and Teorell 2010; Kurtz and Schrank 2007a; Pollitt and Bouckaert 201l; Thomas 2010). The concept of “government effectiveness” is not developed based on or associated with established theoretical discussions of bureaucratic quality or effectiveness. Although Kaufmann et al. (2007a, 555) argue that their concept of governance is closely related to the “work of Douglas North: the norms of limited government that protect private property from predation by the state,” it is less clear how this is relevant to their conceptualization of “government effectiveness” (see Kurtz and Schrnak 2007a, 2007b for this critique). In addition, “government effectiveness” is operationalized as “the competence of the bureaucracy and the quality of public service delivery” (Kaufmann, Kraay, and Mastruzzi 2005, 5), but it is vague what “competence” and “quality” of bureaucracy/ public services indicate. Second, related to the first point, the specific properties of bureaucracies are overlooked and the overall evaluation of them is focused based on the assumption that all the relevant characteristics of bureaucracies always go hand in hand. As a result, what these indicators actually measure is obscured. In the case of the WGI’s, a variety of individual variables are aggregated to provide an overall rating. But, these individual variables used seem to be related to a broader concept of “governance” (Langbein and Knack 2010), rather than specific and relevant characteristics of bureaucracies. Also, it is not explicitly discussed whether the variable tries to measure particular bureaucratic characteristics or the performance of bureaucracies. For example, the authors include the quality of general infrastructure and business elites’ ratings of the quality of bureaucracy, as well as the consistency of policy direction and whether or not bureaucracy hinders business activity. In addition, the individual items cover not only how the bureaucracy is 31 working in a country, but also how the public feels about it.18 They are too inclusive and it is unclear why these metrics are chosen in the first place. In spite that there is no coherent logic tying these various items together, it is assumed that all of these individual variables are correlated. The ICRG’s Bureaucracy Quality indicator provides a clearer and narrower operationalization for the concept than the WGI’s does. But, this indicator also relies on overall assessment, without providing any theory or logic behind their overall assessment of “bureaucratic quality.” It simply premises that a “bureaucracy quality” is attained when all the elements—autonomy, expertise, political independence, and mechanism for recruitment and training—are sufficiently represented. However, what if, for example, one country with a higher level of autonomy does not have a comparatively high level of expertise? Is this nation’s bureaucracy rated higher overall than a nation whose bureaucracy has higher levels of expertise but lower levels of autonomy? Or does the former nation receive a lower score than the comparison country? These possible questions are not clearly addressed in their methodology report. The measure wrongly assume that all four elements of “bureaucratic quality” necessarily covary, and, does not provide information about how these individual attributes are presented in bureaucracies. The third limitation is that there is a little variation within advanced democracies according to these summary ratings based upon cross-national indicators. Because the WGI’s indicator is based on an aggregated estimate, the authors recommend considering the “margins of 18 Particularly, this makes it difficult to examine the relationship between the bureaucratic characteristics and related concepts, using the Government Effectiveness indicator. That is, a bureaucracy that meets the expectation in democratic societies will contribute to a higher level of citizen satisfaction with government services. But, this hypothesis cannot be tested using Government Effectiveness, because both a measure of characteristics of bureaucracies and public satisfaction with public service delivery are included in this single aggregate indicator. 32 error” or 90 percent confidence intervals when making comparisons (Kaufmann, Kraay, and Mastruzzi 2010).19 Interpreting their ratings according to this standard implies that bureaucracies in most advanced industrialized nations are all effective and of high quality. For example, the score for 2011 OECD member countries ranges from .32 to 2.25. But, the top fifteen countries, whose scores range from 1.53 to 2.25, are actually not significantly different from each other.20 In turn, it makes difficult to discern any variations in bureaucracies within advanced democratic countries. A similar problem exists with the ICRG’s indicator. OECD member nations are scored between three or four, with more than half of them receiving four points.21 Bureaucracies for all nations democratized before 1975 were assigned a score of four, except for France, which receives three points. Therefore, this indicator may be useful in comparing bureaucratic quality between developed versus developing, or between consolidated versus transitional democratic countries. However, it has a limited leverage when it compares bureaucracies within the tier of advanced nations who are members of the OECD. 19 Indeed, comparing the exact values of the Government Effectiveness scores is not recommended (Thomas 2010). The various individual data sources used to calculate this aggregate score vary by country and year. Thus the estimated score for each country does not rely on the same individual variables, even in the same year. Also, for each country, the individual indicators used are not always the same across time. Although the authors argue that their estimation techniques take this fact into account, this inconsistency of base data sources can be problematic given that the choice of those individual variables is not based on a clear conceptualization. Thus, the comparability of this country score may be put in question. 20 They are: Australia, Austria, Belgium, Canada, Denmark, Finland, Germany, Iceland, Luxembourg, Netherlands, New Zealand, Norway, Sweden, Switzerland, and United Kingdom. 21 Nineteen countries, out of thirty four, receive 4 points and they are Australia, Austria, Belgium, Canada, Denmark, Finland, Germany, Iceland, Ireland, Israel, Japan, Luxembourg, Netherlands, New Zealand, Norway, Sweden, Switzerland, United Kingdom, and United States. 33 2.3.3 A Need for an Alternative Way of Measuring Bureaucracies The cross-national indicators used in prior research are based on an abstract concept of the one best way to measure “good bureaucracy.” These indicators fail to adequately capture the key bureaucratic characteristics proposed in the literature. In assuming that all relevant attributes are compatible, they do not consider the complex relationships that might exist among the traits. Moreover, there is little variation within OECD member nations when these indicators are used to measure bureaucratic characteristics. These limitations speak to a need for an alternative measurement of bureaucratic traits. In addition, the basic assumption that a single description of “good bureaucracy” exists seems to be problematic, and it suggests that we need to approach measuring and comparing bureaucracies in a different way. With the cross-national indicators reviewed here, bureaucratic characteristics are assumed to be one-dimensional. That is, the “best bureaucracy” exists at one extreme of the continuum, and bureaucracies in different nations all pursue the same traits. However, as Andrews (2010) states, “the good governance version of good or effective government is a hollow one imposing a false one-best-way model (7-8).” Some comparative public administration researches have argued for cross-national differences in organizational and administrative cultures, as well as in the politico-administrative relations. In examination of the differences and similarities between politicians and bureaucrats (especially higher civil servants) in Britain, Germany, Italy, Sweden, the Netherlands, France, and the United States, Aberbach, Putnam and Rockman (1981) find that American higher civil servants are very different from other European counterparts—the former are more politicallyoriented and have more direct contacts with politicians and clienteles than the latter. Whereas Aberbach et al. (1981) focus on the relationship between bureaucrats and political elites and 34 emphasize how the American case is different, Pierre (1995) examines administrative systems and shows a greater variation among them than is assumed in one-best-way models. In comparing bureaucracies in Britain, Germany, Sweden, France, the United States, and Japan, Pierre (1995, 8) begins with a notion that “[m]ost public administrative systems seem to be guided either by the Rechtsstaat model or by the Anglo-Saxon notion of the ‘public interest’.” Based on case studies of each nation, he concludes that public administration in these states may be categorized differently: they have different types of political and administrative career patterns and different organizational structure. Efforts to understand the variation of public bureaucracies are ongoing (Brans 2003); a recent study further identifies various administrative cultures and traditions. Painter and Peters (2010) propose that there are nine groups or families of administrative traditions in the world, including Anglo-American, Napoleonic, Germanic, Scandinavian, Latin American, Postcolonial South Asian and African, East Asian, Soviet and Islamic. 22 Although some hybrids can be developed, they argue, administrative cultures are persistent and have “legacy effects.” As a result, nations emphasize different characteristics of administrative systems and react differently to the same problems. This implies that a nation may choose to implement different bureaucratic traits in pursuit of better administrative performance. If this holds true, then it challenges the assumption that the various characteristics purported to constitute “good bureaucracies” always go together. 22 But they only illustrate the differences in various administrative traditions. A rigorous, systematic examination of public administrations in different nations is still lacking. 35 2.4 Conclusion In this chapter, I have discussed four key bureaucratic traits claimed to be important for the administrative apparatus of the state in democratic societies. Each of these characteristics is advocated individually in prior studies as a way to make bureaucracies effective and compatible with democratic values. In order to better understand and compare bureaucracies across nations, I argue that we need to consider all four traits—political independence, representativeness, impartiality, and competency—simultaneously. In the last part of the chapter, the most common existing indicators of bureaucracies were reviewed and assessed according to whether or not they provide adequate tools to measure and compare bureaucracies. It seems that their utility is limited: they do not provide sufficient information about specific attributes of bureaucracies and do not successfully capture the possible variations in the presence of key bureaucratic traits across nations. Given the theoretical and methodological limitations of these indicators, there is a need for a new measure of bureaucracies. How, then, can we measure bureaucracies according to the degree to which they simultaneously possess the four traits discussed earlier? Examining the relationships among these four traits is not straightforward given that these traits are not mutually exclusive, but rather closely related to each other. In this study, I will adopt an approach that examines whether and how bureaucracies in different nations possess a set of traits with different weights. This is the task to which I turn in the subsequent chapter. 36 CHAPTER 3 BUREAUCRATIC PROFILES OF OECD NATIONS: A MDPREF ANALYSIS This chapter proposes to consider the four bureaucratic traits discussed in Chapter 2 by examining the relative levels of each trait presented in OECD member countries. A “bureaucratic profile” for each country is created based upon results from a nonmetric internal analysis of preference data which is often called as the MDPREF model or the vector model (Carroll 1972). The results from the MDPREF model will be discussed in terms of both the clusters of traits underlying bureaucracies and the variations in the configurations of bureaucratic traits across nations. 3.1 Bureaucratic Profiles How do we compare bureaucracies in terms of the presence of a set of key bureaucratic traits? One possible way to answer this question is to focus on the one trait that is the most important. For example, if a consensus shows that competency is the most fundamental and essential characteristic of a governmental bureaucracy, then we could compare bureaucracies across nations according to the degree to which they possess this specific attribute. However, using only one key trait as a yardstick for comparison and overlooking how other traits are simultaneously represented will allow for only a partial understanding of what bureaucracies look like. Thus, instead, I argue that we need to consider several key bureaucratic traits jointly. As discussed in Chapter 2, the four traits are not exclusive to each other, but rather are related in various ways. The complex relationships among traits also suggest a possibility that 37 these traits do not always increase or decrease together, even though they are all important attributes. That is, possessing a high level of one trait does not guarantee exhibiting a high level of another trait. For example, representativeness in a national bureaucracy does not necessarily lead to political independence, and vice versa. They are appreciated for different reasons: representativeness is emphasized as a way to reduce arbitrariness in administration and to increase responsiveness to the public (e.g., Keiser et al. 2002; Selden, Brudney, and Kellough 1998), whereas political independence is appreciated as a mechanism to deter bureaucrats’ involvement in corruption (e.g., Dahlstrӧm, Lapuente, and Teorell 2012; Rauch and Evans 2000). In order to study bureaucratic traits that are interrelated but not always consonantly, I argue for scholars to look at how similarly and/or differently a set of traits are exhibited in each country. I assume that all bureaucracies possess a common set of key traits, although each national bureaucracy may present various levels of each trait.23 Imagine that we have data containing the information about each nation’s scores on a set of traits, where scores reflect the relative levels of each trait presented in each governmental bureaucracy. Then, within each country, we can order the traits in terms of their scores on this measure, and this rank order represents a nation’s “bureaucratic profile.” With the assumption that bureaucracies in all OECD member nations possess some degrees of the same set of traits, we can then compare this “bureaucratic profile” across nations. Bureaucratic profiles thus allow us to examine how several key attributes are collectively and simultaneously presented in governmental bureaucracies cross-nationally. 23 In other words, there will be a common underlying structure of bureaucratic traits in each country. But, the configuration of traits is not given as a priori in the following analysis. Instead, the results from the analysis in this chapter will provide information about their underlying structure. 38 Three things are noteworthy regarding the concept of bureaucratic profiles. First, each of bureaucratic traits is exhibited in degrees (e.g., from lower levels to higher levels). Hence, a nation’s bureaucratic profile is not about which traits are present and which are absent. Rather, the profile is based on the information about the extent to which each trait is exhibited. Second, the profile is based on information about relative levels of a trait, compared to other traits within a bureaucracy, but not necessarily about the absolute level of each trait. In other words, a bureaucratic profile identifies which trait is exhibited in the least extent among all of the key traits. But, the measure does not necessarily indicate that this bureaucracy, compared to other nations’ bureaucracies, is defective or faulty with respect to that trait. It only suggests that this specific trait is emphasized less than other traits in this nation’s bureaucracy. 24 Third, different bureaucratic profiles simply suggest differences across nations’ bureaucracies, but do not imply that one profile is better than another. Thus, instead of considering bureaucratic profiles as ranking governmental bureaucracies cross-nationally, the profiles should be discussed in terms of the similarities and differences in the degree to which key traits are presented in each country. Bureaucratic profiles can be determined by various things, such as administrative legacy or tradition (Painter and Peters 2010; Pierre 1995), public preference, and configurations of political institutions.25 And, bureaucratic profiles reflect the relative importance of bureaucratic traits emphasized in each national bureaucracy. Thus, although the specific bureaucratic profile 24 For example, let say that one trait is presented the least in nation A’s bureaucracy, whereas this trait is presented the most in another nation B’s bureaucracy. That is, this trait will come in the first place in B’s bureaucratic profile, while it is in the last place in A’s profile. In this case, it does not necessarily indicate that the absolute level of this trait in country A is lower than it in country B. If we have data comparing the level of this trait across national bureaucracies, and compare it between A and B, then it is possible that the absolute level of this trait exhibited in A is greater than that in B. But the important focus in the bureaucratic profile is the relative levels of this trait in each country. 25 This study does not address the discussion regarding the determinants of the bureaucratic profiles, although it will be an important topic for a future study. 39 present in each country is not exactly a “choice,” the term “nation’s choice” will be used interchangeably with “bureaucratic profile” throughout this chapter for the convenience. 3.2 Method: A Multidimensional Preference Scaling (MDPREF) Analysis In order to examine and compare the bureaucratic profiles of OECD member countries, this chapter conducts a nonmetric internal analysis of preference data, which is often called as a “multidimensional preference scaling” or MDPREF (Carroll 1972; Davison 1983; Weller and Romney 1990). MDPREF is a special case of the “unfolding” method or a distance model for preference data, which contain the proximity relation between “subjects” and “stimuli” (Carroll 1972; Davison 1983). Both subjects and stimuli are geometrically represented in the same spatial configuration (Weller and Romney 1990, 45). This space where the stimuli and subjects are represented together is called a ‘joint space’ (Jacoby 1991; Weller and Romney 1990).26 MDPREF is also called the vector model because subjects are represented by vectors or directed line segments (Carroll 1972). It is one of the models proposed by psychologists and psychometricians to understand the structure of individuals’ different perceptions of the stimuli, with an assumption that underlying the stimuli is a set of dimensions common to all subjects 26 The MDPREF model analyzes the original variables of the preference data, and it examines the structures of both “the row and column variables” together and presents both in the “jointspace” (Weller and Romney 1990, 45). This is a difference between MDPREF and factor analysis. Factor analysis uses a cross-product matrix—correlation or covariance matrix—to find a basic structure of either “the row variables or the column variables in relation to the latent vectors” (Weller and Romney 1990, 27-45). I conduct the MDPREF and not factor analysis for two main reasons. First, factor analysis focuses on the linear structure of the data. But, I argue that the traits presented in each national bureaucracy do not always go up and down together— they are not linearly related. Therefore, if factor analysis is used to capture such complex relationships between traits, it will produce a less parsimonious scale. Second, and more importantly, I focus on a “bureaucratic profile” and want to differentiate the configuration of traits across nations. The MDPREF model fits better for this purpose. 40 (Carroll 1972). This model has been used, for example, to examine the variations in national role conceptions of twenty-nine foreign policy decision makers (Wish 1980) and to analyze the structure of individual value choices in the American public (Jacoby 2011a). The data typically used for this model are preferential choice data which contain two different sets of objects—subjects (which here are nations27) and stimuli (which here are bureaucratic traits)—and the cell entries indicate subjects’ preference judgment for the stimuli (Jacoby 1991; Weller and Romney 1990). 28 But, MDPREF is equally applicable to data in which each cell entry includes the extent to which a subject (i.e., nation) possesses a stimulus (i.e., bureaucratic trait), or the level of a trait exhibited in a nation. 29 And, the Eckart-Young procedure, a strategy called “alternating least-squares, optimal scaling” or ALSOS (Carroll 1972, Jacoby 1999), is used for the model estimation. Within a common space, the trait points and the nation vectors are arranged, as much as possible, in a way that the trait a nation’s bureaucracy exhibits to a greater degree comes closer to the terminal points of this nation’s vector.30 As a result, the analysis produces two separate 27 Precisely, “subjects” here are national bureaucracies that possess the specified bureaucratic traits. Since bureaucracies are compared across nations, I will use “nations” or “countries” to represent national bureaucracies in this chapter. 28 This is a difference from a multidimensional scaling analysis, which usually uses similarities data and examines the relationship among stimuli. Both similarities data and preferential choice data contain information about the relationship between points in a pair (i.e., proximity relation). But similarities data compare pairs of points from the same set, whereas preferential choice data compare pairs of points from different sets (Jacoby 1991). 29 Indeed, a data theory can facilitate utilization of scaling methods—here, the vector model (Jacoby 1991). According to a data theory, as Jacoby (1991, 4) states, a researcher observes information in the real world and these empirical observations are transformed to the data, which contain only “the information that [researchers] choose to analyze.” As such, I derive “preferential choice” data from the empirical observations containing information about the extent to which each country possesses various bureaucratic traits. 30 Specifically, the distance between the terminal point of the nation vector and the trait point’s perpendicular projection onto that nation’s vector reflects the relative levels of that trait presented in this nation. The points and vectors are arranged so that the order of trait points’ 41 matrices: One for the coordinates of trait points and another for nations’ vector orientations, which provide information of nations’ locations in relations to trait points (Weller and Romney 1990). Here, the bureaucratic traits are represented in the “center” of the configuration, and countries are shown with their vector terminal points in the “outside” of the space. The vectors originate from the centroid, or from the origin of the space; the extended line for the vectors can also be drawn to facilitate interpretations of bureaucratic profiles. The dimensionality of this space contains the “important sources of variability underlying a set of objects” (Jacoby 1991, 28) and is a choice made by the researcher. One can determine the number of dimensions empirically, after considering a fit measure, such as STRESS formula two, and also interpretability and parsimoniousness (Davison 1983; Jacoby 1991). The dimensionality is a minimum number of important sources of variability, and two-dimensional or three-dimensional spaces are preferred because one can visualize the estimations from the model (Jacoby 1991). An illustration of simple hypothetical data will help understand how the trait points and nation vectors are geometrically represented when MDPRF is used (Jacoby 2011a). Let us assume that we observe the profiles of scores for three countries (labeled “1”, “2”, and “3”) across three bureaucratic traits (labeled “A”, “B” and “C”). That is, the order of the relative levels of traits presented in a nation’s bureaucracy is: A > B > C in county 1, C > A > B in country 2 and B > C > A in country 3. Part 1 of Figure 3.1 shows how these three profiles are represented in a two-dimensional joint space, with the three trait points and the three nation vectors. The distance from the terminal point of a vector to trait points’ perpendicular projections onto that vector reflects the relative levels of traits exhibited by that country. Specifically, as a projections onto a nation vector monotonically reflects the order of the relative level of each trait presented in that nation. 42 Figure 3.1. Hypothetical example of the vector representation Part 1. Geometric representation of hypothetical data for three nations (labeled “1”, “2” and “3”) across three bureaucratic traits (labeled “A”, “B” and “C”), where the order of the relative levels of traits presented in a nation’s bureaucracy is: A > B > C in county 1, C > A > B in country 2 and B > C > A in country 3. B Country 1 A Country 3 C Country 2 Part 2. Perpendicular projections from the three traits onto the vector of country 1. The distance between the terminal point of the vector and each projected point (“A1”, “B1” and “C1”) reflects the order of the relative level of a trait presented in country 1. The smaller distance indicates the more emphasized trait in country 1. B Country 1 A A1 Country 3 B1 C1 C Country 2 43 trait point’s projection onto a vector is closer to the terminal point of that nation vector, this trait is presented more in that nation than other traits that are further from the terminal point of the nation vector. Part 2 of Figure 3.1 shows the perpendicular projections from the three points of bureaucratic traits, A, B and C, onto the vector of country 1 (labeled “A1” “B2” and “C1,” respectively, and marked with an x). The distance from the vector terminus to each of the projected points, A1, B1 and C1, indicates the ordering of three traits exhibited in country 1: The closer the trait is to the vector terminus, the more it is presented in that nation’s bureaucracy. For this country, the projection from the point of trait A onto the country 1’s vector (i.e., A1) is nearer to the vector terminus than the other two projected points, B1 and C1. And, C1 is the least proximal to the terminal point of country 1’s vector. This represents that the country 1 has the highest score on A, the second highest score on B and the lowest score on C. 3.3 Data and Measurements 3.3.1 Data Sources and Measurements of Key Traits In order to measure bureaucratic traits, several data sets are used.31 They include the Quality of Government (QoG) Expert Survey data, national statistics from the OECD and the International Labour Office (ILO) database, the OECD’s Governance at a Glance 2009 dataset, and survey data from the International Social Survey Programmes (ISSP).32 Among others, the 31 Because there is no one concrete measure for each bureaucratic trait, I instead employ several indicators to measure the traits. Specifically, twenty-one variables are used capture bureaucratic traits for thirty-four total countries 32 All data are publicly available. For access to the data and details on collection methodologies, see http://www.qog.pol.gu.se/ (Teorell, Dahlström, and Dahlberg 2011); http://stats.oecd.org/; 44 QoG Expert Survey data provide ratings on various specific characteristics of public administration across countries, which are not available in other data sources. The data are collected by the QoG institute—using a web-based survey of public administration scholars in each country. The respondents of the survey were sampled from “a list of persons registered with four international networks for public administration scholars … and a small snowballing component” (Dahlstrӧm, Lapuente, and Teorell 2010, 20). On average, twelve experts responded in each country in consideration. The survey was conducted in 2008-9 and in 2010, and I assume no significant changes in bureaucratic characteristics since 2004. The survey questionnaires are designed to ask how experts think about their country’s public sector employees 33 on a scale ranging from 1 to 7. The mean estimates for a country, provided in the QoG Country-Level Survey Data, are used. According to Dahlström et al.’s (2010) cross-source validation and assessment of respondent perception bias, they seem to well represent actual workings in a certain country. Other national statistics are obtained through the OECD and the ILO website. Moreover, the OECD’s Governance at a Glance 2009 dataset provides various indicators of the public sector for OECD member countries, which were developed based on statistical data collected through the survey of senior officials in central government personnel department in each http://laborsta.ilo.org/; http://www.oecd.org/gov/indicatorsofgoodgovernment.htm (OECD 2009b); and http://www.issp.org/data, respectively. 33 To define public sector employees, the survey states the following in the beginning of the questionnaire: “All questions in this questionnaire pertain to the public sector employees of a specific country of your choice…When asking about public sector employees in this survey, we would like you to think about a typical person employed by the public sector in your country, excluding the military. If you think there are large discrepancies between branches of the public sector, between the national/federal and subnational/state level, or between the core bureaucracy and employees working with public service delivery, please try to average them out before stating your response.” (Dahlstrӧm, Lapuente, and Teorell 2010, 42). 45 country in 2005 (OECD 2009b). In addition, aggregated data from the ISSP survey data from 2000 to 200434 are used to collect the information on education level of bureaucrats. From these various data sources, twenty-one variables are selected for this study. The degree to which each nation exhibits a trait is observed across OECD member countries. And, the observations are recoded so that higher values indicate that a country exhibits that trait to a greater extent. The exact survey question wordings and further details on the data sources are described in Appendix A. Let us look at the individual variables. First, three variables from the QoG Expert Survey are selected to measure political independence. One variable is used as an overall rating of political independence. The survey question asks respondents the degree to which public sector employees behave to fulfill the political ideology of the incumbent party. This variable inversely represents the degree to which a bureaucracy is independent from politics. 35 Two other variables measure the recruitment system of senior officials—the extent to which senior officials are recruited based on their career, 34 The ISSP surveys are conducted annually and include a uniform set of background questions. Trained professional interviewers interview at least 1,000 randomly selected respondents in each country with the language selected by the respondents. The ISSP surveys are conducted on “a probability-based, nation-wide sample of adults” (ISSP 2009). 35 There is a possibility of different interpretations for what ‘fulfilling political ideology of an incumbent party’ means—that is, partisanships or policy directives. But, I believe this variable captures the partisanship aspect, and comparisons with responses to another question in the same survey provide some support for my interpretation. The question compared asks the extent to which “public sector employees strive to implement the policies decided upon by the top political leadership,” which seems closer to the meaning of ‘following policy directives’ by bureaucrats. The correlation between the two variables (with the country-level, aggregated data) is -0.216 and not statistically significant (p-value = .227). Also, a t-test for the mean difference between responses of these two questions for each nation shows that the mean responses are significantly different in twenty-three nations (one-tailed test, at the .05 significance level). Moreover, it is reasonable to expect that in a highly politicized bureaucracy following policy directives would not be differentiated from fulfilling partisan expectations. And, the eleven countries where the means are not statistically different include Eastern European and Latin counties, which are known for a lack of political independence. Therefore, it seems reasonable to assume that the survey question I used captures the degree of political independence. 46 and not influenced by political partisanship. These variables are included because appointing senior officials for political affinity reduces political independence. Higher values on these three variables indicate a greater level of political independence.36 To capture female representativeness in bureaucracies, four variables are selected. The QoG Expert Survey provides experts’ overall evaluations on women’s representation in the public sector. In addition, statistics from the OECD Government at a Glance data provide detailed information about female representativeness. They are the percentage of females in three different positions in the public sector—in the central government, in senior positions, and in administrative positions in 2005. Because males have dominated public sectors and females are more slowly promoted than males, the number of female workers in the public sector is especially small in senior positions and relatively large at the administrative positions (Auer, Demmke, and Polet 1996; European Commission 2010; OECD 2009a, 2009b). Therefore, measuring the proportion of female workers at different positions in the public sector will help better capture the degree of female representativeness. To measure bureaucratic impartiality, five variables from the QoG Expert Survey are used. The five items are those created by Rothstein and Teorell (2008a) and Teorell (2009), where impartiality is defined as the following: “When implementing laws and policies, government officials shall not take into consideration anything about the citizen/case that is not beforehand stipulated in the policy or the law” (Rothstein and Teorell 2008a, 170). The first question asks experts to rate their bureaucracy in terms of this theoretical definition of impartiality. The next question asks about bureaucrats’ behavior in distributing money in a hypothetical case involving a cash transfer program to the “needy poor.” Then, three questions 36 The proportion of political appointees in the bureaucracy can be another indicator, but such a measure, which is comparable across nations, is not available. 47 ask about the degree to which public employees are not partial and biased in several incidents— public procurement decision-making, implementation of policies, and granting licenses to business startups. In addition, five variables to capture the idea of a merit-based recruitment system and a career-based system are included to operationalize the mechanisms for political independence and impartiality. 37 These variables are collected from two data sources. The four variables from the QoG Expert survey are: The degree to which the public sector employees have a lifelong tenure, enter through formal exams, are recruited based on merits, and are not hired based on political connections. In addition to these variables, a rating on a career-based versus positionbased system, which is provided by the OECD Government at a Glance 2009 data, is included. A career-based system is one in which positions are usually filled within civil service system and based on the career experience of candidates. The opposite recruitment system is a positionbased system, where external recruitment is exercised because finding candidates who best fit a particular position is the main focus (Auer, Demmke, and Polet 1996). Competency would be measured ideally if detailed information about the education and professional training of public bureaucrats are available. But, such direct indicators are not available. Therefore, I instead rely on the ISSP survey data to calculate the proportion of public bureaucrats with a high level of education. 38 To discern a “high level of education,” I decide to look at the proportion of people in the public sector holding university degrees compared to that in the general public. Thus, the average ratio of public sector employees with a bachelor’s degree 37 As noted in Chapter 2, although impartiality and political independence are not the same, they may share some common mechanisms to keep bureaucrats impartial and independent. 38 This education attainment measurement has some limitations but is the only available measure for OECD countries over time (Kahn 2008). Following Kahn (2008), I assume that the statistics for the respondents who are employed by the government reflect the statistics for public bureaucrats in general. 48 or more from 2000 to 2004, compared to that of the general public, is used as a proxy. Further explanation of this variable is included in Appendix A. In addition, two more variables are used as proxies for bureaucratic competency— government employees’ wages and compensation as a percentage of the wages and compensation in the total economy, respectively. These indicators reflect the proportion of the total wages and compensation in the economy used to remunerate government employees. For more detailed descriptions of these two variables, see Appendix A. The data are gathered from the OECD statistics (i.e., Structural Analysis Database in OECD.stat), and the average values between 2000 and 2004 are used in this analysis. I assume that higher wages and compensation will make highly-skilled workers more likely to work in the public bureaucracy. Thus, higher levels of average wage and compensation are assumed to reflect a higher level of bureaucratic competency. Related to competency, I add a variable measuring the size of the government workforce. If higher wages and compensation attract skilled workers to the public sector, then one possibility is the government workers who are well-remunerated enjoy a prestigious status, which should be associated with a lager government labor force. On the contrary, if a nation pursues a smaller government and fewer public sector employees, then the choice often will be associated with reducing benefits for government workers. This, in turn, will render public sector work unattractive for high-skilled workers; this could lead to a loss of competent, qualified employees (Suleiman 2003). To measure this variable, the number of government employees as an average percentage of total employment between 2000 and 2004 is used. The data are collected from ILO statistics. 49 Note that missing data are filled using a single imputation method in order to facilitate the analysis. The imputation was performed by specifying a multiple regression for each of twelve variables that have missing values in order to obtain the “best” predicted value for each missing value. Regression equations producing the largest possible adjusted-R2 are used, where each variable with missing data is regressed on a subset of available variables.39 Then, the predicted value from each regression model is imputed for the missing value. When the predicted value is out of the theoretical range of the variable, the value is recoded. That is, a missing value is imputed with the theoretical minimum value when the predicted value is smaller than this; with the theoretical maximum when the predicted value exceeds it.40 3.3.2 Dataset From the empirical observations described above, I standardize the values for each variable across countries. These standardized scores are used to gain the information on the relative level of the specified traits presented in a bureaucracy. The standardized scores provide us with a dataset including two objects, thirty-four nations and twenty-one bureaucratic traits, which will be analyzed using a MDPREF model. 39 In the final regression equations I found for twelve variables with missing values, the adjusted R ranges from .6398 to .9908. The predictor variables are selected from all available variables from the data sources used, and not limited to the twenty-one variables used in the analysis. This was done in order to minimize a potential problem that the results may be biased if the imputed values falsely increase the relationship between variables. 40 For example, a variable of career-based system from the OECD Government at a Glance theoretically may range from 0 to 1. And, the predicted value of this variable for Israel is -0.017. Therefore, the value for Israel on this variable is imputed with 0, a theoretical minimum, instead of -0.017, which is theoretically impossible. 2 50 Table 3.1. Hypothetical Example of the Dataset Part 1. Cell entries are the ratings for three bureaucratic traits across three hypothetical countries. Countries are rated on traits and a higher value on a trait indicates that a country exhibits that trait to a greater extent. Trait A Trait B Trait C 6 60 60 Country 1 2 50 130 Country 2 1 80 90 Country 3 Part 2. The data in Part 1 are standardized. This format of the dataset is used for the analysis. Trait A Trait B Trait C 1.13 -.22 -.95 Country 1 -.38 -.87 1.04 Country 2 -.76 1.09 -.09 Country 3 Part 3. The relative level of traits exhibited in countries, or “preference ordering” information, can be obtained from the dataset in Part 2. Trait A Trait B Trait C 3 2 1 Country 1 2 1 3 Country 2 1 3 2 Country 3 Part 1 in Table 3.1 shows the hypothetical values for three nations. In the original observation, each bureaucratic trait is rated across countries; a larger value on a bureaucratic trait indicates that a country exhibits that trait to a greater extent. Then, the values are standardized, as shown in Part 2 of Table 3.1. It shows that each nation has a profile of scores on a set of traits. From the standardized values illustrated in Part 2, we can obtain information about the order of the level of traits exhibited in each country’s bureaucracy—the data, as shown in Part 3 of Table 3.1. For country 1, for example, trait A receives the largest value, 1.13, and trait C receives the smallest value, -.95. From this, we can interpret that country 1exhibits trait A to the greatest degree and trait C to the lowest degree. Likewise, we can interpret that trait C is exhibited to the greatest extent and trait B to the lowest extent in country 2, while country 3 51 emphasizes trait B the most and trait A the least. The information about the (order of) relative levels of traits will be analyzed as the data. 3.4 Analysis A nonmetric internal analysis of preference data, or a nonmetric MDPREF analysis, is conducted on these data.41 A nonmetric analysis is employed because the level of measurement for the variables was assumed to be ordinal. And, for the mode of analysis, this chapter employs an internal analysis, in which a stimulus space is not given as a priori. That is, the data are used entirely to estimate both dimensions of bureaucratic traits and the location of nations. A twodimensional model is used; the R2 is 0.719, where the R2 is calculated by comparing the input data and the optimally-scaled predicted values from the model. 42 The analysis produces estimates of the coordinates for trait points and estimates of the orientations for vectors representing different nations in a two-dimensional joint-space. The results are interpreted in two ways: Identifying substantively important groupings of trait points and interpreting the orientations of the nation vectors.43 Section 3.5 discusses the configuration 41 An R-function written by William G. Jacoby, together with R-package, optiscale is used in this portion of the analysis (Jacoby 2011b). 42 This shows an acceptable level of model fit. The analysis also indicates that a two-dimensional model is appropriate for the data. 43 Since clustering of the points seems more substantively interesting than the dimensions themselves in the results, I will focus on interpreting the groupings of the points, as well as the nation vectors. Nevertheless, to understand the “joint-space” within which points and vectors are arranged, it is worth exploring the possible meaning of the dimensions. And, it seems that there are two underling dimensions, a ‘structural-versus-behavioral’ dimension (horizontal dimension) and a ‘traditional-versus-reformative’ one (vertical dimension). That is, some bureaucratic traits are related to the system of recruitment and retention, while others are more about bureaucrats’ behaviors and actions in their relations to politicians and the public. And, if nations have different choices among the traits, there will be nations emphasizing one side over the other. This 52 of the trait points that provides information on the relationship among the twenty-one variables measuring bureaucratic attributes. The locations of the points and their groupings can be interpreted as the “similarities” between the variables. This is because variables exhibited similarly (in terms of their relative levels) across countries will be located closer to each other in the joint space, whereas variables presented differently across nations will be positioned farther away from each other. As a consequence, variables used to capture the same trait are expected to be located near each other. Let us consider the variables designed to measure female representativeness as an example. I expect that the relative level of female representativeness in administrative positions will be similar to that in the central government, compared to the level of other traits presented in a country, if both variables truly capture female representativeness. The results also will enable a test of my argument that bureaucratic traits are not always present together with the same degrees. That is, if my expectation holds, several clusters of variable points will be produced. The clusters will indicate that bureaucracies possess a set of distinct bureaucratic traits in which their relative importance varies across countries. If, however, we see only one cluster of variables in the joint space, then my argument is not supported and the result implies that key bureaucratic traits all go together. If this is the case, then previous measures assigning a single overall rating to national bureaucracies would be considered reasonable because a high level of one bureaucratic attribute always leads to that of another trait. dimension is similar to Painter and Peters’ (2010) distinction between the legalistic and management administration traditions. In addition, in the late 20 th century various models of public management reforms have been proposed to transform an “old” to a “new” one, but their influence has been varying across nations (Pollitt and Bouckaert 2011). And, some bureaucratic traits considered in the analysis are closer to a “new” bureaucracy, while others are the core attributes of an “old” one. Reflecting this, another dimension in the space seems to represent a deviation from, or an adherence to, a traditional, “ideal-type rational/legal bureaucracy” (Demmke and Moilanen 2010; Gualmini 2008; Pollitt and Bouckaert 2011). 53 After discussing the configuration of the traits, Section 3.6 discusses the orientation of vectors representing nations’ bureaucracies. Each vector reflects a nation’s bureaucratic profile. I discuss whether OECD member nations share similar profiles (i.e., similar patterns for the relative levels of bureaucratic traits) or have distinct profiles. If all nations have similar bureaucratic profiles, then the nation vectors all will be oriented in a single direction. If there are various profiles, however, the nation vectors will be spread out. 3.5 Findings: Configuration of the Traits Let us interpret the configuration of trait points. The estimates of the coordinates for trait points are represented in a joint-space, as shown in Figure 3.2. The configuration allows us to understand the similarity and dissimilarity among variables measuring bureaucratic traits. The twenty-one variable points appear to form five clusters, which indicate five different bureaucratic traits.44 The variables whose points can be grouped together indicate that these variables measure the same construct. That is, the ranking of these variables in terms of their relative levels presented in a bureaucracy are similar cross-nationally. The four key traits I originally considered as being important to the structure of OECD member countries’ governmental bureaucracies are found by the results. But, different from my expectations, one more cluster is found; the results suggest that there are five distinct attributes underlying national bureaucracies of OECD member nations. The first grouping of variables can be found in the middle, right-hand side of the figure. Seven variables are grouped together and labeled with one of bureaucratic traits, “impartiality.” 44 The five groups of variables are also supported by the results from the cluster analysis. See Figure B.1 in Appendix B. 54 Figure 3.2. The configuration of variable points female_senior Female Representativeness female_central female_administration women_represnt Impartiality Competency fair_licenses distribute_needy compensation fair_implement impartial wage higher_edu fair_procurement recruit_merit recruit_no_polital size_government senior_no_political no_fulfill_ideology senior_career Independence from Politics career_based tenure recruit_exam Career-based System Note: The twenty-one variable points are shown with closed circles and their names. The points are located in a two-dimensional space based on the estimates from MDPREF. The description of the variables and the data sources can be found in the text and Appendix A. The circles around points (in a dotted line) represent my interpretation of five clusters of points—the five traits— and they are labeled accordingly (in bold). Recall in Chapter 2 I discussed that an impartial bureaucracy is one that does not take into account partisan, personal or a particular group’s interests in decision making. The seven variables include the five variables designed to measure the theoretical construct of impartiality 55 (as Rothstein and Teorell (2008a) suggest), plus two variables capturing merit-based recruitment. That is, the results show that recruitment based on merits and skills, and not on political connections, is related to the degree to which bureaucrats are impartial in the policy process. This finding is reasonable given that impartiality is difficult to preserve if bureaucrats’ employment is determined by politicians. Note that I expected not only merit-based recruitment, but also formal exams, tenure, and a career-based system to be associated with an impartial bureaucracy. But, only merit-based recruitment seems to be associated with impartiality. And, as will be discussed below, other variables intended to capture the common mechanism for impartiality and political independence form a distinct cluster. Second, three variables in the bottom, right-hand side of the figure (in the lower-right area) appear to form a group representing political independence. The proximity of these variables is as expected. Also, as I argue that impartiality and political independence are distinct but closely related characteristics, the first cluster (impartiality) and the second cluster (independence from politics) being located relatively near each other confirms my expectation on these traits’ relationship. This second cluster provides some supportive evidence that recruiting and promoting senior officials based on their career reflects less politicization and more effort to guarantee independence from political influence. Also, it implies that as bureaucracies’ independence from partisan politics is emphasized, senior officials are more likely to be recruited within in the ranks of the public sector and not based on their partisanship. The third cluster, in the bottom, left-hand side corner, includes three variables representing a career-based system. The three variables are the degree to which public sector employees are recruited through formal exams, promoted based on career, and work with a secured tenure. It is different from my expectation that variables measuring career-based system 56 form a distinct group, separated from variables capturing political independence or impartiality. In the sense that they are related to human resource management, the variables in the third cluster (career-based system) are related to the variables forming the second cluster (independence from politics). But, the distinction between the two also makes sense: Variables in the second cluster are especially related to public sector employees’ behavior in their relations to politics, whereas the variables in the third cluster concern the structural and organizational institutions involved with recruiting and retaining public sector employees. 45 Fourth, the four variables in the middle, left-hand side of the figure appear to form a cluster labeled ‘competency,’ as they relate to the retention of talented bureaucrats. The four variables include compensation, wage, education levels of government workers, and the size of the government workforce. Higher compensation and wages allow a public bureaucracy to attract and maintain more competent workers. If the proportion of people with higher education is greater in governmental bureaucracies compared to that in the public, then it indicates that more talented people are recruited and working in the public bureaucracy. One may argue that the size of the governmental workforce is not related to the other three variables within this cluster. However, I think that this clustering is sensible given that downsizing in the public sector is often argued to have a demoralizing and de-professionalizing consequence, which may make 45 This is an interesting finding. With the QoG Expert Survey data, Teorell (2009) develops and measures impartiality, and Dahlström et al. (2010) differentiate “closedness (as opposed to openness)” and “professionalism (as opposed to politicization)” of bureaucratic structure. Although they are conceptually distinct, these three have not been clearly differentiated empirically. When I conduct factor analysis, professionalism and impartiality are not distinguished. On the other hand, the vector model employed here provides more supportive evidence for their theoretical distinctions. It shows that impartiality and political independence (which is similar to Dahlström et al.’s (2010) professionalism) are related to each other, but can be differentiated. 57 bureaucratic jobs less attractive. Thus, as a country focuses more on attracting and retaining talented employees, the size of public sector labor force will tend to increase. Finally, four variables in the upper, left-hand side of the figure can be clustered as female representativeness, consistent with my expectation. The cluster of these four variables is meaningful because it captures the proportion of female bureaucrats using various measures. It is interesting that the four variables are less tightly clustered than one might expect. This reflects that female representativeness is not the same for all levels of positions in the public sector in all nations. For example, for many nations, more female workers in lower level positions may not always guarantee a similar proportion of female workers in more senior positions. 3.6 Findings: Nation Vectors How do national bureaucracies exhibit these five traits? Do the OECD member nations share similar profiles or do they have distinct profiles regarding the five bureaucratic traits? These questions will be addressed by interpreting nation vectors and the relative locations of the traits in the joint space. The terminal points for the nation vectors are presented in Figure 3.3, with the variable points in the same space. The vectors are adjusted to a unit length, so that the vector terminal points fall along a circle (Jacoby 2011a). The variable points are shown with closed circles and the terminal points of nation vectors are marked with open circles. In order to foster easier interpretation, I also present the mean points for each of five groups of variable points—trait points—with open squares and corresponding labels. 58 Figure 3.3. Full model of nations’ vector terminal points and variable points, with five bureaucratic trait points and the mean direction vector EE IS CL SK DK US AU NZ HU female_representativeness SI UK PL CZ GR PT SE NO FI CA AT NL CH impartiality competency MX independence_from_politics IT career_based_system LU TR ES JP IL BE DE IE KR FR Note: The nations’ vector terminal points (open circles) and the bureaucratic trait points (solid circles) are shown in the joint space. The points are estimated from MDPREF. The mean points for the five groups of variable points are presented with open squares and labeled. The arrow oriented about 56 degree below the horizontal dimension represents the mean direction vector, whose length is .079. The thirty-four countries are: AU (Australia), AT (Austria), BE (Belgium), CA (Canada), CL (Chile), CZ (Czech Republic), DK (Denmark), EE (Estonia), FI (Finland), FR (France), DE (Germany), GR (Greece), HU (Hungary), IS (Iceland), IE (Ireland), IL (Israel), IT (Italy), JP (Japan), KR (Korea), LU (Luxembourg), MX (Mexico), NL (Netherlands), NZ (New Zealand), NO (Norway), PL (Poland), PT (Portugal), SK (Slovakia), SI (Slovenia), ES (Spain), SE (Sweden), CH (Switzerland), TR (Turkey), UK (United Kingdom) and US (United States). 59 3.6.1 Interpreting Nation Vectors In order to interpret both the similarities and differences among bureaucratic profiles, we focus on the angular separation between the nation vectors. In fact, the correlation between the predicted bureaucratic profiles of any two countries is equal to the cosine of the angle between the two vectors for those nations. Thus, the smaller is the angle between two vectors, the similar the nations’ choice of bureaucratic traits, and vice versa. For instance, the angle between the vector of Australia and that of New Zealand is very small (i.e., 1.24 degrees); this indicates that the two countries have very similar bureaucratic profiles. Two nations are very different and have an opposite order of relative trait importance if the angle between the vectors for them is 180 degrees. For example, the angle between the vector for the United States and that for Italy is almost 180 degrees (precisely, 175.50 degrees). This means that the bureaucratic profiles are almost opposite in these two nations. Another way to look at the distribution of nation vectors is by taking their value average and calculating the mean direction vector. The arrow shown in Figure 3.3 represents the mean vector of all nation vectors. The length of this mean direction vector, which is called the “mean resultant length” (Gill and Hangartner 2010), is relatively short in Figure 3.3. The mean resultant length theoretically ranges from zero to one, where a smaller number indicates greater variation in the full set of vectors. If there is no variation in vector orientations, then the mean resultant length will be one (Gill and Hangartner 2010; Jacoby 2011a). Therefore, the short length of the mean direction vector in Figure 3.3 indicates that bureaucratic profiles are very diverse across OECD member countries.46 46 The mean resultant length in the result is .079. 60 Overall, the results show that the vectors are spread out around the entire circle. There is a small concentration of vector terminal points in the middle right quadrant, around “3:00” position. However, rather than concentrating only on this position, the nation vectors are dispersed into several different groups. For example, nations are located closely at around the “2:00” position, “5:00” position and “9:00" position. These groupings speak to the diversity of bureaucratic profiles. To understand the profiles in detail, the next section interprets nations’ locations in relations to the trait points. To simplify interpretation, I focus on five trait points, which refer to the means of the variable points within each of five clusters (shown with opensquares in Figure 3.3). 3.6.2 Interpreting the Orientations of the Nation Vectors The terminal points of nation vectors for the thirty-four OECD member countries and the mean points for each of the five bureaucratic traits (i.e., the means of the variable points within each of the five clusters) are presented in Figure 3.4, with open circles and closed circles, respectively. Recall that bureaucracies in nations whose vectors are located close together should actually share similarities; countries located farther away from each other exhibit different bureaucratic profiles. These comparisons are what I focus on in the following discussion. 47 47 The exact predicted profiles are presented and discussed in Chapter 4, where a new measure is developed based on the results interpreted here. 61 Figure 3.4. Full model, with the vector terminal points and the five trait points EE IS CL SK DK US AU NZ HU female_representativeness SI UK PL CZ GR PT SE NO FI CA AT NL CH impartiality competency MX independence_from_politics IT career_based_system LU TR ES JP IL BE DE IE KR FR Note: Open circles represent the nations’ vector terminal points. The solid circles indicate the five traits, which are the mean points for each of the five groups of variable points. The individual variable points are not shown. All the points are located based on the results from MDPREF. See Note in Figure 3.3 for the full names of the thirty-four countries. 62 3.6.2.1 Nation Vectors between “1:00” and “4:00” Nations positioned between “1:00” and “4:00” include Nordic countries and AngloAmerican countries. In these nations, impartiality is exhibited to a greater degree than other traits. Specifically, in countries positioned around “2:00,” female representativeness is the second most exhibited characteristic and career-recruitment system is displayed to the least extent. The United States and Australia are examples of countries fitting this profile. These countries have some commonalities: For instance, they are advocates for New Public Management (NPM) movements and also much influenced by the NPM ideas (Halligan 2010; Pollitt and Bouckaert 2011). Thus, business management skills were introduced and personnel powers were decentralized here. This could explain why the point of career-based system is the furthest away from the terminal points of the vectors for these countries. Along with benchmarking business management, politicization was also emphasized as a way to make bureaucracy more responsive (Pollitt and Bouckaert 2011; Suleiman 2003). And, this seems to be reflected in the bureaucratic profile, which shows that the level of political independence presented is less than that of impartiality and female representativeness in these nations. In nations located around “3:00” position impartiality is again the trait most exhibited, but independence from politics is predicted to be second. And, competency is exhibited to the least extent, relative to other traits. Sweden and Norway are examples of countries fitting this profile, which again are substantively similar. It is reported that there were some reform attempts based on the idea of NPM, but dramatic reforms were not observed in Nordic countries 63 compared to Anglo-Australian countries (Pierre 2010; Pollitt and Bouckaert 2011).48 This could explain why, compared to the United States or Australia, political independence is still relatively salient (in fact, the second important trait) in these countries. In Northern European countries, including Sweden, Finland and the Netherlands, the “anti-government theme” (Pollitt and Bouckaert 2011,166) or the “rhetoric of reforming an old (bad) bureaucracy to a new (good) bureaucracy” (Demmke and Moilanen 2010, 1) has not been as strong as in Anglo-Australian countries, and the political control over public management is less emphasized (Pollitt and Bouckaert 2011). Thus, in nations located around the “3:00” position, it is more likely that career civil servants and political careers are separated and that a non-partisan culture exists, despite the increasing number of political appointees (Pollitt and Bouckaert 2011). Plus, most of these nations’ civil service systems are described as position-based system in which recruitment and promotions are not based on a specific career (Auer, Demmke, and Polet 1996). This fact is consistent with the vector model results that the point of career-based system being located farther away from these nations’ vector terminus. 3.6.2.2 Nation Vectors between“5:00” and “8:00” Vector terminal points in the bottom of the figure, between the “5:00” and “8:00” positions, represent very different bureaucratic profiles from those mentioned above. For these bureaucracies, career-based system is displayed at a higher level than other traits. The nations in this position include Japan, Germany, France, Belgium, and Spain. These nations’ bureaucracies 48 Compared to other Nordic countries, Denmark is described as being most influenced by the NPM ideas (Greve 2006; Jensen 1998). And, the vector model results are consistent with this observation, in that Denmark is located near the Anglo-Australian countries at the “2:00” position, rather than the “3:00” position with other Nordic countries. 64 are often characterized by legal and formal administrative culture (e.g., Painter and Peters 2010).49 Nations located in this area, however, have different profiles regarding the other four bureaucratic traits. The nations whose vector terminal points are around the “5:00” position are often described as having a career-based system (Auer, Demmke, and Polet 1996; Demmke and Moilanen 2010), which is reflected in the results. The results show that the bureaucracy’s independence from politics is highly emphasized, as is career-based system; female representativeness is the least displayed bureaucratic trait. Nation vectors in this position include those for Germany and Japan, for example. These nations seem to share some commonalities. For example, bureaucracies in Korea and Japan are influenced by German-type bureaucracies (Nakamura 2003; Paik 2003; Painter 2010) and civil service systems in Ireland and Germany are considered as career-based and closed-systems (Auer, Demmke, and Polet 1996). The German model of public bureaucracy is understood as a Rechtsstaat-type, where public administrative systems are based on public law and a very ‘traditional bureaucracy’ (Pierre 1995; Pollitt and Bouckaert 2011). In addition, the public sector workforce in these nations is dominated by males; the gap between men and women is still large (Auer, Demmke, and Polet 1996; OECD 2009a). It is reported that although these countries have de jure equal opportunity, women still enter bureaucracy at lower rates than men and move up the ranks more slowly (European Commission 2010; UNPAN 2006). 49 Also note that the terminal points of the vectors for Austria, the Netherlands, and Switzerland are around the “3:00” position, but they are at lower position (thus, closer to the vector terminal points that are around the “5:00” position). It seems that career-based system is emphasized more in these nations than in those around the “3:00” position. This reflects that they have Germanic tradition (Painter and Peters 2010), although these countries have some similarities to Nordic countries. 65 Nations whose vector terminal points are at the bottom of the figure, the “6:00” position, are a little different than countries like Germany, although there are many similarities between them. France and Belgium are the examples of a French-type or a Napoleonic-type bureaucracy (Ongaro 2008, 2010; Painter and Peters 2010). Similar to nations located at the “5:00” position, it appears that these countries assign the highest importance to career-based system, while female representativeness is not highly ranked (Auer, Demmke, and Polet 1996; European Commission 2010; OECD 2009). Moreover, these countries, as well as Germany, are described as being resistant to reforms based on NPM ideas (Auer, Demmke, and Polet 1996; Pollitt and Bouckaert 2011). The main characteristics of a Napoleonic-type bureaucracy are legalism and formalism characterized by a competitive exam and centralized state structure (Demmke and Moilanen 2010; Ongaro 2008, 2010; Painter and Peters 2010). Countries positioned around the lower-left quadrant in the figure, between the “7:00 and 8:00” positions, are Southern European counties, including Greece, Italy, Portugal, and Spain. Bureaucracies in these countries are characterized with a high degree of legal formalism, similar to a French-type bureaucracy (Ongaro 2008, 2010). However, different from it, patronage and politicization are prevalent in these Southern European countries (Painter and Peters 2010). Indeed, increasing political patronage and the existence of the spoils system are serious problems facing the public bureaucracies in Italy and Greece (Demmke and Moilanen 2010; Painter and Peters 2010; Pollitt and Bouckaert 2011). Moreover, difficulties combating these problems have increased as several administrative reforms, largely influenced by the NPM, were introduced (Ongaro 2008, 2010). Studies show that such reforms have been more influential in Southern European counties than in Continental European countries. These characteristics explain that the 66 bureaucratic profiles of nations positioned at between “7:00 and 8:00” emphasize both political independence and impartiality less than other traits. 3.6.2.3 Nation Vectors between “9:00” and “10:00” and between “10:00” to “12:00” As we move further in a clockwise direction, the nations whose vectors are oriented between the “9:00” and “10:00” position include mostly Central and Eastern European countries.50 Bureaucracies in these nations, such as Hungary and Slovenia, share some similarities. The creation of a professional civil service, and transforming the old, politicized state administration, are among the important goals (Verheijin and Rabrenovic 2007). Although the public management systems in these nations were influenced by Continental European countries, the career-based system is not as strong as in those nations. This is due to the unstable legal basis for civil service, weak job security and no centralized competitive exams in Central and Eastern European countries (Demmke and Moilanen 2010; Meyer-Sahling 2010; Verheijin and Rabrenovic 2007). And, these countries confront a difficult task that they attempt to establish transparent and open recruitment systems, and at the same time create an impartial system (Meyer-Sahling 2010; Verheijin and Rabrenovic 2007).51 50 The terminal points of vectors representing Mexico and Chile are also around this position. The results seem to reflect that Latin American systems are influenced by Spain’s and Portugal’s system, especially regarding surface level legalism and formalism, but also prevalent particularism and clientelism in practice (Painter and Peters 2010). 51 Note that these countries usually face double tasks, to develop democracy and to establish a professional bureaucracy. Different from Western European countries, the state bureaucracies have not been consolidated before democratic politics was introduced. And, the lack of an established legal basis for bureaucracies tend to make these countries vulnerable to patronage politics, especially when the party system is not institutionalized (O’Dwyer 2004). 67 The bureaucratic profiles of these nations reflect these observations. The trait point that is the farthest from the vector terminus of these countries is impartiality. Also, as severe politicization is one of their problems, it is reasonable that political independence is less exhibited than other traits. In addition, their bureaucratic profiles show that competency is presented to a greater extent than other bureaucratic traits. This suggests a bureaucrat in these countries may be considered a prestigious occupation, as more highly educated people work for the government and a greater proportion of workers are in the public sector, which pays well. Finally, four nations are spread out in the upper-left quadrant to the top of the figure, between the “11:00” to “12:00” positions. One of them, Estonia is described as having a very open and highly politicized bureaucracy (Randma 2001). This is reflected in the results that career-based system and independence from politics are farther away from the terminal point of the vector representing Estonia than other traits. One interesting common characteristic of these four countries is that they have a relatively large percentage of women in the senior positions of the central government (European Commission, 2010). In their bureaucratic profiles, female representativeness is the trait present to the greatest extent. But, more importantly, the bureaucratic profiles for these nations show that other traits are exhibited to a lesser degree than is female representativeness. Moreover, the vectors for these nations show that career-based system is the least demonstrated trait. These bureaucratic profiles suggest a possibility that female representativeness conflicts with career-based system. In other words, they may have been able to achieve greater female representativeness because they put less emphasis on a career-based system. A position-based system, which is the opposite of career-based one, could contribute to employing more females who may face some barriers in the entrance to, and promotions in, 68 bureaucracies. Although less obvious, in nations whose vectors located between “1:00” and “2:00,” female representativeness also is presented at a higher levels than other traits, except for impartiality; and career-based system is the least emphasized trait. This result also suggests possible conflicts between female representativeness and career-based system. 3.6.3 Comparing Specific Bureaucratic Profiles Beyond general interpretations of the locations of nation vectors in relation to trait points, let us take a look at specific nation vectors. This example illustrates how a particular nation’s bureaucratic profile can be interpreted and compared with that of another nation. Figure 3.5 shows the vectors for two countries, the United States and Germany, along with the terminal points of the nation vectors (with open circles), and the points for five bureaucratic traits (with closed circles). The vector pointing toward the upper-right represents the United States and the one directed toward the bottom-right is for Germany. If you draw perpendicular projections from the five points onto the vector for the United States, for example, impartiality is shown to be the trait most emphasized, followed by female representativeness, independence from politics, and competency, with career-based system as the least emphasized bureaucratic attribute. The bureaucratic profile looks different for Germany. Career-based system receives the greatest level of importance in Germany, followed by independence from politics, impartiality, and competency; female representativeness is the least emphasized bureaucratic characteristic. The nation vectors seem to reflect the American and German bureaucracies relatively well. For example, in the United States, a merit-based civil service was created by the Pendleton Act of 1883, but a large number of political appointees continue to be employed compared to 69 Figure 3.5. Full model, with the vectors for the United States and Germany Note: All the points are located based on the results from MDPREF. Open circles represent the terminal points of nation vectors and the closed circles indicate the five traits. The vector for the United States is shown with an arrow pointing toward the upper-right in the joint space. The vector orienting toward the bottom-right is for Germany. The vectors are originated from the centroid as shown with solid, directed line segments. The extended lines are drawn for vectors as shown with dashed lines to foster interpretations of bureaucratic profiles. The angle between the two vectors is 90.49 degrees. It is estimated that the bureaucratic traits are presented in the United States in the order of impartiality, female representativeness, independence from politics, competency, and career-based system; in Germany in the order of career-based system, independence from politics, impartiality, competency, and female representativeness. See Note in Figure 3.3 for the full names of the thirty-four countries. 70 other industrialized countries to this day (Pierre 1995; Van Slyke and Riccucci 2003). This could explain why the relative level of political independence presented in the U.S. bureaucracy is in the third place among five traits. Another characteristic of the U.S. bureaucracy is that bureaucrats are recruited not only for the entry level positions, but also for top positions (the Senior Executive Services). Relatedly, career distinctiveness is relatively weak in the United Sates because workers frequently move between the public and private sector (Peters 2001). Hence, it is expected that the relative level of career-based system is the lowest among the five traits in the American bureaucracy. In addition, female representativeness is the second closest point to the vector terminus; it reflects that the U.S. is one of the countries with a relatively large percentage of women in the civil service (Peters 2001). Regarding the German bureaucracy, its vector seems to reflect that there exists a strong norm that bureaucrats should be neutral and independent from political influence; that there is a lower degree of movements between politics and public administrative positions; and that bureaucratic expertise is considered as an important attribute, as opposed to party loyalty (Pierre 1995). Particularly, the historical development of the civil service explains why career-based system and political independence are the highest valued traits in German bureaucracy. In Germany, a professional civil service was “institutionalized” in 1794 by the Prussian civil service code specifying their role as “public (and no longer personal) servants” and additional features, such as guaranteeing lifelong tenure and requiring university training, were added in the 1820s (Derlien 2003, 103). This Prussian civil service code and the “traditional principles” have been the basis for the civil service law in the German republics (Derlien 2003). The constitutional status of civil service provides a “strong safeguard against basis reforms” (Derlien 71 2003, 104, italics are original), and thus the structure and privileged status of the German bureaucracy is hardly changed. Perhaps a more interesting point is implied when we compare the two nation vectors in the joint space, which provides information that the bureaucratic profiles for the United States and Germany are different from each other. The angle between the two vectors is close to 90 degrees (precisely, 90.49 degrees), meaning a close-to-zero correlation between the two profiles. Given that the German and American models of public administration—the Rechtsstaat model and the Anglo-Saxon notion of the ‘public interest’ model, respectively, in previous studies (e.g., Pierre 1995; Pollitt and Bouckaert 2011)—are often described as very different, no correlation between the two bureaucratic profiles is entirely expected. Moreover, as we compare the two bureaucratic profiles, the relationships between some of the bureaucratic traits provide interesting insights. Recall that in Chapter 2, some complex relationships between bureaucratic attributes were discussed. The expectations on cross-national differences in the different potential relationships among these traits were realized empirically in the vector model. For example, three traits in Germany’s bureaucratic profile—impartiality, political independence, and career-based system—are located very near to each other. This may suggest that in Germany, a career-based system is perceived as a trait contributing to, rather than hindering, impartiality and political independence. And, this result is reasonable given that in Germany career-based system is advocated for promoting professionalism, which functions as an internal constraint on bureaucrats’ behaviors (Derlien 2003; Peters 2001). Similarly, professionalism contributes to impartiality. The relationship between the three traits looks quite different in the United States. The relative location of trait points to the vector for the United States shows that impartiality and 72 career-based system are positioned on opposite sides of the joint space, with political independence in the middle. This suggests that, first, career-based system is not linked to impartial bureaucracies in the U.S., different from Germany. Instead, as a way of improving impartiality, the U.S. seems to focus on a position-based system, which has been advocated as a way to make bureaucracies more “democratic” and to improve public responsiveness (Van Slyke and Riccucci 2003). Second, the different location of political independence in the U.S. also implies that it is not perceived as being linked to impartiality, as in Germany. In sum, the results show variations in perspectives on the consequences of a career-based system and in the relationship between impartiality and political independence, in the two countries. 52 3.7 Conclusion In this chapter, the key bureaucratic traits were empirically examined and compared across OECD member countries. Following discussions in the previous chapter, I proposed to examine “bureaucratic profiles” which provided an opportunity to systematically examine the 52 This difference may be due to the different historical development of bureaucracy in each country. For example, Shefter (1977, 423-433) interprets the development of bureaucracy in Germany as the survival of a “constituency for bureaucratic autonomy,” which emerged in the pre-democratic era, throughout the regime changes that Germany experienced. That is, patronage practices and any political interference have been prevented and bureaucratic autonomy has been protected because a professional bureaucracy—where civil service exams are instituted and a lifelong tenure is guaranteed—was established and institutionalized in the process of statebuilding, before the creation of a mass electorate and the development of political parties. On the other hand, the sequence of development was reversed in the United States, as Shefter (1977, 447) notes: “the creation of a mass electorate in the Jacksonian period preceded by half a century the emergence of American’s constituency for bureaucratic autonomy during the Progressive era.” A bureaucracy that institutes formal exams and whose recruitment and dismissal are protected from outside interference was not established before the political parties and candidates competed with each other. And, thus, for the bureaucracy, being democratic and not threatening democratic values is more important (Van Slyke and Riccucci 2003) than being competent and autonomous. 73 presence of various important traits simultaneously and cross-nationally. The analysis using the MDPREF model showed that bureaucracies in OECD nations share a common underlying structure—characterized by a greater or lesser degrees of political independence, representativeness, impartiality, competency, and career-based system—but, more importantly, have very diverse bureaucratic profiles. That is, national bureaucracies present varying degrees of these five traits; each nation seems to emphasize different bureaucratic traits. This finding is important because it suggests that previous measures assigning an overall rating to public bureaucracies, combining all of the important traits, potentially provide a misleading portrait of the degree to which each trait is presented. Indeed, five clusters of traits, instead of four I hypothesized, were found in the two-dimensional joint space. It suggests a complex relationship between impartiality, political independence, merit-based system, and career-based system. That is, in some nations a career-based system is considered very differently from other three traits, whereas all of them are treated as compatible traits in other nations. The complexities of these relationships, and their variation cross-nationally, are precisely what previous, and more normative, assessments of governmental bureaucracies miss. The wide variety in bureaucratic profiles is the most interesting finding in this chapter. There is not a single pattern of relative levels of bureaucratic traits exhibited by all of the OECD countries. This provides empirical support for the various types of “administrative traditions” argued by comparative public administration researches (Brans 2003; Painter and Peters 2010; Pierre 1995). Examining various bureaucratic profiles systematically shows similarities and differences among national bureaucracies. Moreover, the findings speak to the possibilities, as Andrews (2010) and Holmberg et al. (2009) posit, that countries have various institutional configurations of bureaucracy. Do these various bureaucracies lead to different consequences in 74 a society? Particularly, in this dissertation project, I ask whether or not these different bureaucratic profiles are linked to different public attitudes toward government. This will be examined empirically in Chapter 4. 75 CHAPTER 4 RELATIONSHIP BETWEEN BUREAUCRATIC PROFILES AND POLITICAL SUPPORT This chapter examines the linkage between various bureaucratic profiles and citizens’ political support. In order to explore what citizens expect from governmental bureaucracies, I focus on political support as the dependent variable. If the bureaucratic profiles reflect what traits citizens want bureaucracies to possess in democratic societies, and if the relative importance of such traits matches citizens’ expectations, then we should find higher levels of political support in such nations. Thus, using a new measure of bureaucratic traits, Bureaucratic Profiles, which is constructed based on the results from the previous chapter, I will investigate if there is a systematic relationship between a particular configuration of bureaucratic traits and levels of political support. Studying the effect of bureaucratic profiles on political support will provide an additional explanation for cross-national differences in levels of political support that has not been fully accounted for in previous studies. 4.1 Political Support for Regime Performance and Regime Institutions Citizens’ political support is an essential element of democratic consolidation (Andrain and Smith 2006; Chu et al. 2008; Easton 1975; Hetherington 1998), and advanced democratic countries strive to enhance levels of political support (Dalton 2004; Nye, Zelikow, and King 1997; Norris 1999c, 2011). Political support is a multidimensional phenomenon that can be differentiated according to “objects” of political support (Dalton 2004; Easton 1975; Norris 76 1999c; Suleiman 2003).53 In this chapter, I focus on two specific dimensions of political support, political support for regime performance (i.e., satisfaction with operation of democracy) and regime institutions (i.e., attitudes toward bureaucracies, especially). Political support for regime performance is defined as “evaluations of the way the regime works, and particularly satisfaction with the way the democratic process functions in practice” (Norris 1999c, 17). This is different from political support for regime principles which refers to support for democratic values or commitment to democracy. Political support for regime performance is often measured with the survey question that asks how much respondents are satisfied with ‘the way democracy is functioning.’ Studies have shown that levels of citizens’ (dis)satisfaction with regime performance are not the same as their levels of (dis)approval of a democratic regime over other regime types (Dalton 1999; Klingemann 1999).54 Whereas political support for regime performance tries to capture a “middle level of support,” political support for regime institutions focuses more on a “realistic view of democracy” (Norris 1999c, 11). Different from a general evaluation of how democracy is functioning in practice, political support for regime institutions focuses on specific institutions of the state, such as parliaments, the legal system, the state bureaucracy, and political parties. Political support for regime institutions is defined as citizens’ evaluations of “the formal structure, not the specific 53 For example, Norris (1999c), in Critical Citizens, differentiates the five objects of political support: political community, regime principles, regime performance, regime institutions, and political actors. This classification is expanded from the distinctions made by David Easton. Studies also provide empirical evidence that citizens can make a distinction between these objects, and that this distinction is theoretically and practically important (Dalton 1999; Klingemann 1999; McAllister 1999; Suleiman 2003). 54 As some studies note, it may be possible that this question captures other aspects of people’s satisfaction with democracy—democracy as a process, regime principles or performance of incumbents. However, this survey question is generally used in previous studies to gauge individuals’ support for system performance (Dalton 2004; Anderson and Tverdova 2003). For example, studies show that this question measures public attitudes that are different from their trust in politicians or support for regime principles (Dalton 2004; Klingemann 1999). 77 incumbents or office-holders” (Norris 1999c, 19). Some studies focus on public confidence in a particular institution such as parliament and the civil service (e.g., Anderson and Tverdova 2003; McAllister 1999; Van der Meer 2010), while others examine a combined index of citizens’ evaluations of several political institutions (e.g., Chang and Chu 2006; Norris 1999b; Mishler and Rose 2001). In this chapter, I particularly examine citizens’ attitudes toward government bureaucracies.55 Prior studies have observed cross-national variations in levels of the two dimensions of political support examined here (Dalton 1999; McAllister 1999). Klingemann (1999), for example, finds that cross-national differences are greater in these two objects of political support—regime performance and institutions, whereas levels of political support for community (i.e., national pride) and for regime principles (i.e., support for democratic values) are consistently high in most of the countries examined. On average, satisfaction with democratic performance is higher in Western European countries than Eastern European or Latin American countries. Specifically, Denmark and Norway demonstrate relatively higher levels of satisfaction, while Greece and Italy show lower levels. Although studies have examined various factors to 55 Focusing on an individual institution, rather than creating an index of several institutions, is favorable because it is possible that people develop different attitudes toward different institutions. For example, Rothstein and Stolle (2008) argue that popular attitudes are different for the institutions on the “representational” side (i.e., the input institutions, such as parliaments) than for the “implementational” side institutions (i.e., the output institutions, such as bureaucracies). Suleiman (2003) also shows that the level of confidence in the civil service is different (i.e., more favorable) from other political institutions. These studies suggest that the expectations people have of bureaucracies may be different from those they have of other political institutions. 78 explain cross-national differences in levels of political support,56 there remains room for further investigation (Anderson and Tverdova 2003; McAllister 1999; Norris 1999b). 4.2 Existing Explanations for Political Support What are the explanations for cross-national differences in levels of political support? Some studies look at micro-level explanations for political support, such as socioeconomic variables, national attachment, and perceived economic situations (e.g., Andrain and Smith 2006; Berg 2007; Van de Walle 2007; Vigoda-Gadot, Shoham, and Vashdi 2010). These studies examine whether or not the key micro-level factors have consistent effects across nations. On the other hand, other studies examine the effect of macro-level variables, such as economic performance, political institutions, and cultural factors (e.g., McAllister 1999; Miller and Listhaug 1999; Mishler and Rose 2001; Newton 2001; Norris 1999b). Recent studies, moreover, consider both individual- and country-level variables to explain variation in political support (e.g., Anderson and Tverdova 2003; Van der Meer 2010). Let us examine in more detail the determinants of political support which are investigated in previous studies. 4.2.1 Economic Factors First, economic factors are discussed as one of the key explanations for citizens’ political support. This argument stems from the micro-level account that an individual is more likely to 56 In this chapter I use this term, political support, in a broader sense than it has conventionally been used to refer to both popular satisfaction with democratic regime performance and public attitudes toward bureaucracies. 79 have positive attitudes toward the government when she is more economically well-off or when she is satisfied with the national economy (McAllister 1999). The positive relationship between economic satisfaction (whether sociotropic or egocentric) and political support is empirically supported in various studies (e.g., Andrain and Smith 2006; Chang and Chu 2006; Dalton 2004; McAllister 1999; Mishler and Rose 2001). At the country-level, the relationship between economic factors and political support is more complicated. The rationale for the macro-level relationship between a nation’s economic performance and levels of political support is weak, and the empirical evidence is mixed. For example, McAllister (1999) uses the 1990-1 World Value Survey (WVS) data on 24 countries to examine the relationship between economic performance and public confidence in parliaments and in the civil service. While he finds supportive evidence that an individual’s satisfaction with the nation’s economic performance positively influences her confidence in both political institutions, a macroeconomic condition (measured with a nation’s GDP per capita) has a negative relationship with levels of institutional confidence. He suggests that a more affluent country shows less confidence in political institutions because people in old and affluent democratic countries have higher expectations of democratic institutions. However, this negative association between economic performance and political support has not always been confirmed in other studies. For example, Norris (1999b) uses the 1990-3 WVS data on 25 nations and finds that GNP per capita is positively related to institutional confidence. Moreover, the results are mixed in studies that use different economic indicators to capture the health of the economy, such as economic growth and the inflation rate. Although it is argued that the public will feel that government is doing its job properly when the nation’s economic situation is healthy (e.g., a higher growth rate or a lower inflation rate), this argument 80 is not consistently supported in empirical studies.57 Given the limitations of an economicallydriven explanation of political support, other studies have proposed alternate factors that possibly have an impact on levels of political support (Norris 1999b). 4.2.2 Social Factors Another frequently mentioned explanation for political support is social factors. Beyond an individual’s socioeconomic status and demographics such as age, gender and education levels (Andrain and Smith 2006; Van de Walle 2007), previous studies examine the relationship between social trust and political support. The argument is that with generalized trust, people are more likely to participate in organizations and politics and to exert a collective influence on the political process. This, in turn, results in a more responsive and accountable government (Fukuyama 1995; Putnam 1993). Therefore, social trust is expected to lead to a higher level of political support. This relationship has been examined at both the individual- and the countrylevel, yielding some supportive evidence in both instances. In previous studies, survey data about interpersonal trust has largely been used to capture the level of social trust. Social trust or generalized interpersonal trust is considered a core concept of social capital (Freitag and Buhlmann 2009; Newton 1999, 2001). At the individual- 57 For example, Anderson and Tverdova (2003) find a significant positive relationship between public attitudes toward government and the economic growth rate, using the 1996 International Social Survey Program (ISSP) data on 16 nations. But, Van der Meer (2010) finds no significant effect of GDP per capita and economic growth on trust in parliament, using three waves of European Social Survey data on 26 countries. Rohrschneider (2005) presents mixed results regarding this relationship. He finds a negative relationship when he examines people’s representational judgment on national parliament and government, using 1999 Eurobarometer data on 13 nations; and a positive relationship with perceived political credibility of politicians, using 1996 ISSP data on 13 countries. 81 level, studies have found that as a person trusts people in general, she is more likely to have higher levels of political confidence and satisfaction with democracy (e.g., Andrain and Smith 2006; Dalton 2004; Zmerli and Newton 2008). This is because a trusting individual is inclined to trust government officials and view them in a more positive way than those who do not have interpersonal trust. Especially when interpersonal trust is understood as one component of a “civic culture” (Almond and Verba 1963; Inglehart 1988), it can be implied that a trusting person will also trust political institutions (Mishler and Rose 2001). Another explanation is that a trusting person shows higher levels of political support because she is more likely to participate politically—such as by contacting government agencies—which, in turn, increases her levels of political support (Andrain and Smith 2006; Dalton 2004). Although some studies find a weak or non-significant relationship between interpersonal trust and political support at the individual level (e.g., Mishler and Rose 2001; Newton 2001), Zmerli and Newton (2008) argue that the relationship holds when better and sensitive measurements are employed, especially in European countries and the United States. This relationship also applies to the aggregated level. Newton (2001, 207), for example, argues that social trust is a “societal [and] not an individual property.” Using the percentage of trusting people in each country as a measure of social trust, he finds a positive association between social trust and political confidence in his cross-national study based on the 1991-1995 WVS survey data on 42 countries. Although focusing on a single country, the United States, Keele (2007) also argues that the decline of social capital has caused a declining level of political trust in the long-run. Related to this, the relationship between social trust and government performance is also examined. Knack (2002) finds that social trust is associated with better government performance in the American states. Based on data of local government in Germany 82 and the U.S., Tavits (2006) shows that social trust increases policy activism, that is, more people actively participate in demanding public goods and services. Although some scholars question the causal direction between social trust and political support (e.g., Brehm and Rahn 1997), I argue that what generates social trust is not political support itself, but trustworthy state institutions (Levi 1998; Rothstein and Stolle 2008). In other words, it is important to conceptually distinguish between trust-generating state institutions and individuals’ attitudes toward these institutions (i.e., political support), to better understand the relationship between social trust and political support. A government institution can provide credible assurances for interpersonal trust, that is, it (informally) institutionalizes trust by reducing the risks of trusting others (Levi 1998; Levi and Stoker 2000; Rothstein 2011).58 Studies have argued that certain state institutions—such as impartial and professional bureaucracies (Kumlin and Rothstein 2005; Zucker 1986), incorruptible and nonpartisan government institutions (Freitag and Buhlmann 2009), or non-suppressive state (Fukuyama 1995)—can help to generate higher levels of social trust. In turn, higher social trust will lead to higher levels of political support. 4.2.3 Political Factors The third explanation of political support concerns political institutions. Because political institutions function at the national level, their relationship with political support is discussed 58 For example, in a study of the production of trust from1840 to 1920, Zucker (1986, 89-94) argues that the development of bureaucracy and the professionalization of it in the 1800s and the early 1900s was a source of the production of process-based trust in the United States. Moreover, she argues that government officials function as “intermediaries” between individuals and firms who distrusted each other. That is, government was asked to create regulation and legislation which would formalize the patterns of interactions so as to produce (institutionalized) trust. 83 only at the country-level. Several studies argue that the effect of political institutions on political support is important because individuals will develop attitudes about government through their experience with political institutions (Anderson and Guillory 1997; Norris 1999b). One of the most important political factors is the level of democratization. That is, when freedom and political liberty are well protected in a nation, citizens will more positively evaluate regime performance and regime institutions (Norris 1999b; Mishler and Rose 2001). But, empirical findings are mixed: whereas Norris (1999b) shows that levels of democracy, measured with Freedom House scores, have a statistically significant relationship with institutional trust, Anderson and Tverdova (2003) finds no significant effects on system support. In addition, the direct effect of various political institutions—particularly constitutional arrangements—on political support is examined in prior studies. Lijphart (1999) argues that consensual democracies with power-sharing political institutions perform “kinder and gentler” than majoritarian democracies do. He also shows that consensual democracies, particularly on the executive-parties dimension, enjoy higher levels of public satisfaction with democracy. But, Anderson and Guillory (1997) argue that consensual democratic institutions do not necessarily produce higher levels of system support at the aggregated-level than majoritarian institutions. Rather, they are expected to narrow the winner-loser gap in satisfaction with democracy. Because focusing on the type of democracy does not provide information on which political institutions would enhance political support, Norris (1999b) further examines the effect of individual political institutions. Using the 1990-3 WVS data on 25 nations, she finds that nations with higher levels of democratization (according to the Freedom House score), parliamentary systems, unitary systems, two-party systems, and majoritarian electoral systems are more likely to have higher levels of confidence in institutions. The findings provide limited 84 support for the positive effects of consensual democracies. Perhaps surprisingly, power-sharing institutions, such as federal systems, multi-party systems, and proportional representation (PR) electoral systems do not necessarily have higher levels of political support than their counterparts.59 In other studies, the effects of political institutions are mixed. For example, Van der Meer (2010) finds that level of trust in parliament is higher in PR electoral systems. But, Rohrschneider (2005) finds very little support for this relationship between political institutions (i.e., majoritarian versus proportional systems) and citizens’ evaluations of the representativeness of their government, where he controls for the quality of arbitrating institutions, bureaucracies and judiciaries. 4.3 Bureaucracies and Political Support As an explanation of political support, few previous studies have taken into account the role of bureaucracy. However, it seems plausible that a bureaucracy influences how people view their government because it plays a critical role in the policy process. Moreover, citizens come into contact with bureaucrats more frequently than they meet with political representatives over the course of their lives. Given the variety of bureaucratic profiles found in the previous chapter, I further examine in this chapter whether countries with different bureaucratic characteristics have different levels of political support. Before discussing this relationship, let us briefly 59 These power-sharing institutions seem to have stronger effects on democratic consolidation than on popular support for political institutions. Norris (2008) examines the effect of these power-sharing institutions on levels of democracy and finds that countries with PR electoral systems, parliamentary monarchies, and federalism are more likely to have consolidated democracy. 85 discuss limitations in prior studies concerning bureaucracies and its effect on popular attitudes toward the government. Although some studies suggest that bureaucracies influence how people view their government (e.g., Rohrschneider 2005), no empirical studies have provided a comprehensive analysis of the impact of specific bureaucratic characteristics. Studies examining this link between bureaucracy and popular attitudes toward government often fail to specify bureaucratic characteristics that are related to different levels of political support. For example, Gilley (2006) finds a positive relationship between “general governance” and state legitimacy. Even when scholars articulate specific properties of bureaucracies, they end up showing a relationship between a broader concept of “governance” and public attitudes, due to the limited availability of data on bureaucratic characteristics (Brans 2003). Rohrschneider (2005) argues that the impartiality and procedural fairness of bureaucracies and judiciaries will inform citizens that the regime has a capacity to meet their interests. But, as a measure of bureaucracy, he uses an index of three World Bank’s Governance Indicators—rule of law, control of corruption, and government effectiveness—which captures a broad and abstract concept, rather than specific bureaucratic traits such as impartiality and fairness. The same indicator is used in Gilley’s (2006) study. Thus, it is difficult to tell from these researches what characteristics of bureaucracies are contributing to public evaluations of government. Another line of studies focuses on the eroding effect of corruption (in the public sector) on political support. Anderson and Tverdova (2003) argue that country-level corruption in the public sector negatively influences citizens’ attitudes toward government. At the individual-level, studies show that a person who perceives higher levels of corruption among public officials is less likely to have confidence in institutions (Chang and Chu 2006; Mishler and Rose 2001). 86 Given these studies, one could argue that levels of corruption characterize a bureaucracy and link it to political support. However, as stated in Chapter 2, I would argue that the actual level of corruption is not a fundamental bureaucratic trait, but a consequence of bureaucratic profiles. For example, bureaucracies emphasizing political independence and merit-based recruitment will be less involved in corruption (Dahlstrӧm, Lapuente, and Teorell 2012; Rauch and Evans 2000).60 Therefore, studies on the effect of corruption on political support suggest a link between bureaucracies and political support, but do not provide sufficient explanation about what specific bureaucratic characteristics are related to political support. In addition, some studies focus on individuals’ evaluations of government performance to predict levels of political support. For example, Kim (2010) finds that citizens’ evaluations of government performance are associated with trust in government in Korea and in Japan. However, this study does not speak to specific bureaucratic traits that could affect government performance. Vigoda-Gadot et al. (2010) examine this link more specifically in six European nations—Ireland, Israel, Lithuania, Norway, Slovakia, and Spain. They explore how citizens’ perceptions of various aspects of managerial excellence in public sector are associated with public sector image, satisfaction with public service, and confidence in public administration. Although they provide some insights about the relationship between specific bureaucratic characteristics and public attitudes toward bureaucracy, they only focus on micro-level 60 Moreover, I argue that perceived levels of corruption is not always the same as actual levels of corruption. And, perception of corrupted bureaucracies is the attitudes people have about bureaucracies. Thus, as will be discussed in Section 4.6.1, I consider an individual’s perception about the prevalence of corruption in public service as one way to measure (unobserved) overall attitudes toward bureaucracies. On the other hand, one may argue that the perceived levels of corruption should be controlled in predicting levels of public satisfaction with democracy. To account for this, I tested the model controlling for country-level perceived corruption (measured with Corruption Perception Index from Transparency International, following Anderson and Tverdova (2003)) and found a robust relationship between Bureaucratic Profiles and satisfaction with democracy. 87 relationships. Thus, country-level factors, including the different properties of national bureaucracies, are not accounted for. In sum, although prior studies suggest a possible link between a bureaucracy and political support, they do not speak to specific bureaucratic traits and cross-national bureaucratic differences that could be associated with different levels of political support. Therefore, in order to empirically examine what specific configuration of bureaucratic traits contributes to higher levels of political support, I propose a new measure of bureaucratic traits, Bureaucratic Profiles and examine its relationship with political support. Next section describes how I construct this new measure and, then, discusses how different bureaucratic profiles might be related to different levels of political support. 4.4 Bureaucratic Profiles as a Measure of Bureaucratic Traits 4.4.1 New Measure of Bureaucratic Profiles A new measure of bureaucratic traits, which is called “Bureaucratic Profiles,” is developed based on the estimation of nation vectors obtained in the previous chapter. Recall that a nation’s location in relation to trait points reflects the rank order of relative levels of bureaucratic traits presented in each nation. To summarize the data, I look at the mean direction vector, or a mean resultant length (see Section 3.6). This shows how a set of bureaucratic traits are presented in a bureaucracy in OECD member nations, on average. See Figure 4.1 for the location of the mean direction vector. The trait points can be projected onto the (extended) mean direction vector. It shows that the positions of perpendicular projections for three traits—political independence, career-based 88 Figure 4.1. Full MDPREF model, with the mean direction vector as a reference vector for a new measure, Bureaucratic Profiles Note: Open circles and solid circles represent nations’ vector terminal points and the five traits, respectively, which are estimated from MDPREF in Chapter 3. The arrow originated from the centroid represents the mean direction vector and its extend line is drawn with a dotted line. This mean direction vector is used as a reference vector in creating a new measure, Bureaucratic Profiles. The numeric values are assigned for each country according to the angular separation between each nation vector and the reference vector. See Chapter 3 for the details of trait points. See Note in Figure 3.3 for the full names of the thirty-four countries. 89 system, and impartiality—are closer to the terminal point of the mean direction vector. This indicates that these three traits are presented more than the other two traits in OECD member nations, on average. Competency and female representativeness are relatively farther away from the terminal point of the mean direction vector. This means that these two bureaucratic traits are relatively less presented, on average. Specifically, when the traits are ordered in terms of their closeness to the terminal point of the vector (i.e., relative levels presented in the nation represented by the vector), we see the following order: political independence, career-based system, impartiality, competency, and female representativeness. By setting this mean direction vector as a reference vector, I assign the numeric values to create Bureaucratic Profiles, which reflects the positive and negative angular separations of each nation vector from the reference vector in radians. Thus, a (hypothetical) nation represented by the mean direction vector will be coded as zero (0). As a vector orientation moves in a counterclockwise direction, the value for Bureaucratic Profiles increases from 0 to pi (π). For example, a nation whose vector is one radian from the reference vector in a counterclockwise direction will be coded as a positive one (1). And, the nation vector which is located 180 degrees (i.e., π radians) from the mean direction vector in a counterclockwise direction is coded as pi (π). In this nation the rank order of bureaucratic traits, in terms of their relative levels, is the opposite of that for the reference vector because the angle between two vectors inversely reflects the correlation between the two bureaucratic profiles. On the other hand, as a vector orientation moves in a clockwise direction, the value of Bureaucratic Profiles decreases from 0 to negative pi (-π). For example, a nation whose vector is one radian from the reference vector in a clockwise direction will be coded as minus one (-1). And, the nation vector coded as a negative 90 pi (-π) is located 180 degrees from the mean direction vector in a clockwise direction, and has the opposite bureaucratic profile from the one for the reference vector. Table 4.1 describes the values of Bureaucratic Profiles for each nation (see the second column). It ranges from -2.87 (in Slovakia) to 3.09 (in Chile). To understand the substantive meaning of this value, the third column lists the specific profiles for each nation, that is, the traits in the order of relative levels presented in each bureaucracy. The trait exhibited to the greatest degree, compared to other traits, will be in the first place and the one presented to the least extent will be in the last place in the profile. For example, Korea has a value of 0.00, which is the closest value to the mean direction vector. And, the specific bureaucratic profile for Korea indicates that traits are exhibited in the following order: independence from politics (IND), career-based system (CAR), impartiality (IMP), competency (COM), and female representativeness (REP). This is what is described above as the mean direction vector’s profile. I stated earlier that the values of Bureaucratic Profiles reflect each nation vector’s deviation from the mean direction vector, in either counterclockwise or clockwise direction. What does this specifically entail? Both Table 4.1 and Figure 4.1 help to facilitate a substantive understanding of the bureaucratic profile. Using information in Table 4.1, let us explore how the relative levels of bureaucratic traits change as the value of Bureaucratic Profiles changes. Note that the descriptions of nation vectors on the 24-hour clock scale in the following refer to their orientation in the joint space, as shown in Figure 4.1. When closely examining the specific profile of each nation, the first noticeable thing is that, in general, the relative level of IMP presented in a bureaucracy is the smallest when the value of Bureaucratic Profiles is small, and it increases as the value gets larger. More specifically, IMP is the least presented trait when Bureaucratic Profiles is between -1.31 and 91 Table 4.1. Bureaucratic Profiles Nation Bureaucratic The order of relative importance of traits Profiles (radians) Slovakia (SK) -2.867 REP COM CAR IND IMP Hungary (HU) -2.586 REP COM CAR IND IMP Slovenia (SI) -2.447 COM REP CAR IND IMP Poland (PL) -2.335 COM REP CAR IND IMP Czech Republic (CZ) -2.328 COM REP CAR IND IMP Greece (GR) -2.316 COM REP CAR IND IMP Portugal (PT) -2.214 COM CAR REP IND IMP Mexico (MX) -2.095 COM CAR REP IND IMP Italy (IT) -1.773 CAR COM REP IND IMP Luxembourg (LU) -1.513 CAR COM IND REP IMP Turkey (TR) -1.351 CAR COM IND REP IMP Spain (ES) -1.309 CAR COM IND REP IMP Israel (IL) -1.128 CAR COM IND IMP REP Belgium (BE) -0.933 CAR IND COM IMP REP France (FR) -0.804 CAR IND COM IMP REP Germany (DE) -0.132 CAR IND IMP COM REP Ireland (IE) -0.068 CAR IND IMP COM REP Korea, Rep. (KR) -0.000 IND CAR IMP COM REP Japan (JP) 0.070 IND CAR IMP COM REP Switzerland (CH) 0.783 IMP IND CAR REP COM Netherlands (NL) 0.844 IMP IND CAR REP COM Austria (AT) 0.892 IMP IND CAR REP COM Canada (CA) 0.917 IMP IND REP CAR COM Finland (FI) 0.986 IMP IND REP CAR COM Norway (NO) 1.011 IMP IND REP CAR COM Sweden (SE) 1.063 IMP IND REP CAR COM United Kingdom (UK) 1.169 IMP IND REP CAR COM New Zealand (NZ) 1.403 IMP IND REP COM CAR Australia (AU) 1.424 IMP REP IND COM CAR United States (US) 1.447 IMP REP IND COM CAR Denmark (DK) 1.631 IMP REP IND COM CAR Iceland (IS) 2.412 REP IMP COM IND CAR Estonia (EE) 2.524 REP IMP COM IND CAR Chile (CL) 3.089 REP COM IMP IND CAR Note: The third column presents the rank order of the relative levels of five traits presented in each nation’s bureaucracy. For example, the trait which is emphasized the most importantly comes in the first place. “IMP” represents impartiality, “IND” independence from politics, “CAR” career-based system, “COM” competency, and “REP” female representativeness. 92 -2.87 (i.e., when nation vectors are located between around the “7:00” and “10:00” positions). As the value of Bureaucratic Profiles increases, the relative level of IMP increases; and IMP is presented to a greater degree than other traits when Bureaucratic Profiles is between .78 and 1.63 (i.e., located between around the “2:00” and “4:00” positions). Note that the relative level of IMP decreases in countries with the highest values (i.e., that are positioned around “12:00” and “11:00”), which is due to the “circular” nature of the data (Gill and Hangartner 2010).61 The relative level of COM generally decreases as the value of Bureaucratic Profiles increases, specifically when it changes from -2.45 to 1.17 (i.e., as a nation vector moves from the “9:00” to the “2:00” position in a counterclockwise direction). Thus, COM is emphasized more than other traits in nations positioned around “9:00” and “10:00.” And, it is the least exhibited trait in countries located around the “3:00” position. But, also note that the relative level of COM increases as the value of Bureaucratic Profiles increases from 1.17 to 3.09 (i.e., as a nation vector moves from the “3:00” to the “12:00” position in a counterclockwise direction). Next, let us focus on the order of REP and IND in bureaucratic profiles. REP is exhibited with a greater degree than other traits in Slovakia and Hungary, two countries with the lowest 61 That is, IMP is presented to a smaller degree in countries with the highest and lowest values of Bureaucratic Profiles. To understand the “circular” nature of the data, let us look at Slovakia and Chile, which have the lowest and the highest values of Bureaucratic Profiles, respectively. Although they have very different scores on this measure, they are located close to one another in the joint space (see Figure 4.1). This is because, Chile has a bureaucratic profile which is very different from one for the mean direction vector—the angle between the two is 177.01 degrees in the counterclockwise direction—and, thus, Chile has a higher value on this measure. On the other hand, Slovakia has the lowest value on this measure because it too has a very different profile than one for the mean direction vector, but in a clockwise direction (i.e., the angle between the two is 164.27 degrees in a clockwise direction). Thus, when we compare the profiles of these two countries, it shows that REP and COM are placed in the first and the second places in both nations (which reflects that their profiles are almost opposite to the mean vector’s one) but that the order of other three traits are different (which is because they are deviated in the different directions). That is, these two countries’ profiles have some similarities even though they have extreme values on this measure; which shows the circular nature of the data. 93 values of Bureaucratic Profiles; and its relative level decreases as the value of Bureaucratic Profiles increases to zero (i.e., as the vector moves from the “10:00” to the “5:00” position in a counterclockwise direction). Thus, REP is the least exhibited trait in nations located around the “5:00” and “6:00” positions. Then, its relative level increases as the value of Bureaucratic Profiles increases from zero to 3.09. This reflects how nation vectors are different than the mean direction vector. That is, the relative level of REP is the smallest for nations close to the reference vector; and its relative level increases as the nation vector deviates from the reference vector in either a counterclockwise or clockwise direction. As opposed to this pattern, IND is exhibited to a greater degree than other traits in nations with Bureaucratic Profiles of zero (i.e., when a nation vector locates close to the reference vector). The relative level of IND decreases as a nation vector deviates in either a counterclockwise or clockwise direction. Thus, IND is presented to a lesser degree than other traits when Bureaucratic Profiles is between -1.77 and -2.87 (i.e., located between around the “8:00” and “10:00” positions), or when it is between 2.41 and 3.09 (i.e., positioned around “11:00” and “12:00”). Finally, let us explore the changes in the relative level of CAR. In countries with the lower values of Bureaucratic Profiles CAR is displayed more than IND and IMP, but less than REP and COM. The relative level of CAR increases as Bureaucratic Profiles increases to -.07 (i.e., as a nation vector moves from the “10:00” to the “6:00” position in a counterclockwise direction). Thus, CAR is exhibited to the greatest extent in nations positioned around between “6:00” and “7:00.” Then, its relative level decreases as Bureaucratic Profiles increases from -.07 to 3.09. In nations located around the “2:00” position, CAR is exhibited to the least extent, relative to other traits. 94 4.4.2 Bureaucratic Profiles and Political Support What is the relationship between Bureaucratic Profiles and levels of political support? I have proposed to look at this relationship in order to determine which bureaucratic profile reflects what people view as important and desirable in democratic societies. If we find that a particular bureaucratic profile is related to higher levels of political support, it will imply that this particular bureaucratic profile is the one that is desired in democratic societies from the perspective of citizens. This assumes that people want bureaucracies to possess a set of traits in a certain style (i.e., with a particular order of the relative levels of traits) in democratic societies and that this desire is universal. An alternative possibility is that what people want from bureaucracies in democratic societies varies across nations. If each country has designed and developed their output institutions as the best-fit for their nation’s democratic government and to reflect their own citizens’ preference, it is possible that nations with various bureaucratic profiles all enjoy higher levels of political support. If this is the case, no systematic relationship between Bureaucratic Profiles and levels of political support will be found. However, I hypothesize that universally there is a particular bureaucratic profile that people find desirable in democratic government. Which bureaucratic profile will contribute to higher levels of political support? Although all the traits considered to construct a measure of Bureaucratic Profiles are discussed as key attributes for bureaucracies in democratic societies, can we think of differences in political support depending on the configuration of the relative levels of these traits presented in a bureaucracy? Before hypothesizing a pattern between Bureaucratic Profiles and political support, let us think through the possible relationship between the level of each trait represented in a 95 bureaucracy and political support based on discussions in Chapter 2. It will help to draw an expectation about the systematic pattern between the two variables. First, given previous studies, especially Rohrschneider (2005) and Rothstein and Teorell (2008a), impartiality is expected to be the most important trait that impacts levels of political support. If bureaucrats behave impartially—that is, if individuals are treated fairly by bureaucrats and public goods are provided in an unbiased way—people’s experience with such a bureaucracy will enhance how they evaluate their government (Hibbing and Theiss-Morse 2001; Levi 1998; Levi and Sherman 1997; Rohrschneider 2005; Rothstein and Teorell 2008a). However, given that the question of interest here is not about which single trait is most important in its relation to political support, but about which bureaucratic profile is, I posit that levels of political support will be higher for a nation where impartiality is exhibited to a greater degree than other traits in a bureaucracy. Then, how will the relative levels of other traits matter? Independence from politics is also an important trait that contributes to higher levels of political support. But the relationship between its relative level and political support seems more complicated. On the one hand, I expect a positive influence of political independence on political support. This is because political independence is, itself, necessary for a bureaucracy to be impartial. This is also implied by the result of a MDPREF analysis: As depicted in Table 4.1, there is no case where the relative level of political independence is the lowest and that of impartiality is the highest. In other words, when impartiality is presented with the greatest degree, the level of political independence exhibited in that bureaucracy is also relatively high. And, when it is low, independence from politics is also presented in a lower level, compared to other traits. Moreover, as I have argued in Chapter 2, the politicization of bureaucracies, which can undermine political independence, will not necessarily improve bureaucrats’ responsiveness to 96 the public. This is because bureaucrats who follow elected leaders’ partisan directives may not respond to the general public’s demands, but instead work for political masters’ short-term interests. Therefore, in general, higher levels of political independence, as oppose to politicization, will be associated with higher levels of political support. On the other hand, if a politically independent bureaucracy becomes an omnipotent and uncontrollable one, it will not contribute to perceptions that bureaucracies and democratic government are performing well. Considering implications of high levels of political independence, I posit that only when a certain level of political independence is presented in a bureaucracy, and when it is accompanied by impartiality, will people have higher levels of political support. That is, nations where independence from politics is presented in a greater degree than other traits will not necessarily enjoy higher levels of political support, if they are lacking in impartiality. For example, let us consider Korea and Japan. As shown in Table 4.1, the relative level of impartiality in these countries is next to career-based system (i.e., independence from politics is the most exhibited trait and impartiality is the third important one). Hence, I expect that levels of political support in these countries will not be as high as in nations where impartiality is displayed to a greater degree than independence from politics. Next, let us think about female representativeness. A representative bureaucracy will signal the government’s effort for inclusiveness of and responsiveness to the public. Also, women may feel that their voices are equally heard as men’s when female bureaucrats are present. In these ways, female representativeness can contribute to higher levels of political support. However, there are contrasting views on the effect of representative bureaucracies as discussed in Chapter 2. Some critics suggest that it can cause “partiality” (Lim 2006; Subramaniam 1967), whereas others argue that it does not behave in a way that harms 97 democratic values (Meier, Wrinkle, and Polinard 1999; Nicholson-Crotty, Grissom, and Nicholson-Crotty 2011). Particularly with female representativeness, it is argued that more female bureaucrats will be more impartial (Dollar, Fisman, and Gatti 2001). Therefore, I expect that female representativeness will contribute to levels of political support only if it comes with impartiality. That is, when a nation’s bureaucracy possesses representativeness to the greatest extent, but presents lower levels of impartiality, this nation will not enjoy high levels of political support. For example, among five nations where female representativeness is in the first place in the Bureaucratic Profiles (see Table 4.1), impartiality is the least emphasized trait in Slovakia and Hungary. But, Iceland, Estonia, and Chile seem to emphasize impartiality, together with female representativeness. Thus, I expect that the latter three nations will have higher levels of political support than the two nations where impartiality is not well preserved. In these two nations (and also in other nations where impartiality is much less represented than representativeness), it is possible that a representative bureaucracy causes partiality and, in turn, will lead to lower levels of political support. What about the linkage between the relative level of competency and political support? I argued in Chapter 2 that competency is a key bureaucratic trait that is usually taken for granted and not explicitly discussed in prior studies. Does this imply that the relative level of competency presented in a bureaucracy is not an important factor for people evaluating governments? Or will people have higher levels of political support when competency is greatly emphasized? I argue that higher relative levels of competency will lead to more favorable popular views of democratic governance, but only if impartiality is also emphasized in a bureaucracy. 98 The Bureaucratic Profiles presented in Table 4.1 reveals an interesting feature, which makes directly testing such a hypothesis impossible. The bureaucratic profiles in OECD member nations show that, in general, as the relative level of impartiality increases, levels of competency decrease, and vice versa. Specifically, nations located between “7:00” and “10:00” seem to prioritize competency over impartiality; while nations around “1:00” to “4:00” appear to choose impartiality over competency (see Figure 4.1.). Thus, there is no case where the two traits are exhibited to a greater degree than other three traits. Instead, there seems to exist tradeoffs between emphasizing impartiality and competency. Hence, I posit that individuals will be less satisfied with bureaucracies and democratic performance if competency is prioritized over impartiality in their nation’s bureaucracy. Finally, what is expected for the linkage between a nation’s emphasis on career-based system and political support? First of all, the level of career-based system in a bureaucracy is not an important factor in itself; its relationship with other traits will be more important in determining its impact on political support. In Chapter 2, I argued that career-based system will go together with impartiality and political independence. But, the analysis of the MDPREF model shows that, as listed in Table 4.1, there are mixed relationships between them. Among the nations with the greatest emphasis on career-based system (i.e., between the “5:00” and “8:00” positions), impartiality and political independence are preserved only in some nations. For them, I expect that levels of political support will be higher when career-based system is emphasized. On the other hand, I posit that if lower levels of career-based system are accompanied by lower levels of political independence, then levels of political support will be also low. This is because position-based system (which is the opposite of career-based system) will not have an intended effect—increasing effectiveness and efficiency of administration, by recruiting the best people 99 for the positions—if it is used as a way of politicization. In this case, a lack of emphasis on career-based system will decrease the political independence that a national bureaucracy possesses. In turn, levels of political support will decrease. Considering all these expectations together, what do I hypothesize as a general pattern between a measure of Bureaucratic Profiles and levels of political support? A positive and significant relationship between them is expected. This is because as the value of Bureaucratic Profiles increases, the relative level of impartiality increases; and as the value decreases, the relative level of competency increases and that of impartiality decreases. Of course, individuals within a country may feel differently about how bureaucracies are performing, and they may have different attitudes toward government. However, I think that overall levels of political support will be linked to different bureaucratic profiles. 4.5 Method: Multilevel Model Analysis In order to examine the relationship between bureaucratic profiles and political support, a multilevel analysis is used, where level-2 is the country and level-1 is the individual citizen. A multilevel analysis is appropriate because the dependent variable, individuals’ political support, has a clustered nature: Survey respondents are nested within countries. Therefore, it is highly possible that any individual’s opinion is correlated with others within the same country. 62 62 The intra-class correlation (ICC) shows the correlation between the two randomly drawn level1 units (individuals) within the same, randomly drawn, level-2 unit (country). The ICC is .158 for the public attitudes toward bureaucracies model and .161 for the satisfaction with democracy model. This indicates the possibility that the assumption of independence among units of analysis is violated. The ICC value of .158 and .161 are not very small. Snijders and Bosker (2011, 18) note that the ICC values range from .10 to .25 in a number of studies of educational performance in American schools, where a multilevel modeling is used. 100 In addition, the key independent variable is measured at the country-level. Thus, a conventional OLS regression model at the individual-level could lead to erroneous conclusions. If the effects of country-level variables are examined using an OLS regression model with pooled data, all these variables will seem to be significant even when they are not (Snijders and Bosker 2011; Steenbergen and Jones 2002; Weldon 2006). Another option would be a simple country-level analysis based upon aggregated data. But, this would miss any information about individual differences within a country—the effects of the individual-level characteristics are overlooked in such a model (Luke 2004). Indeed, levels of political support vary across individuals within a nation and we need to account this variability. 63 Therefore, we need to control the factors to explain individual variations within a country, as well as country-level variables. 64 In the following sections, I investigate two models, one for examining the relationship between bureaucratic profiles and public attitudes toward bureaucracy; and another looking at the association between bureaucratic profiles and citizens’ satisfaction with democratic performance. The two models are treated separately, because these two dependent variables are 63 Public attitudes toward bureaucracies range from -1.96 to 2.09 (with the mean of 0 and the standard deviation of .83) in pooled data. The standard deviation in each country ranges from .61 (in Poland) to .94 (in Chile). Citizens’ satisfaction with operation of democracy in their country ranges from 0 to 10 (with the mean of 6.01 and the standard deviation of 2.25) in pooled data. The standard deviation in each nation ranges from 1.70 (in Switzerland) to 2.45 (in Mexico). 64 A multilevel modeling is increasingly used in other similar studies examining the effect of country level variables on public attitudes. Anderson and Singer (2008) look at the effect of income inequality on public attitudes toward political institutions, using 2002-3 European Social Survey (ESS) data; Anderson and Tverdova (2003) examine the effect of corruption on public attitudes toward government, using 1996 International Social Survey Programme (ISSP); Van der Meer (2010) conducts a three-level random-coefficient analysis to study citizens’ trust in parliament using 2002-2006 ESS data. 101 intended to capture distinct concepts as discussed in Section 4.1.65 For the multilevel models I use a REML estimation because the number of country-level units is relatively small (RabeHesketh and Skrondal 2008; Snijders and Bosker 2011).66 For a better model specification, I employ two analytical strategies (see Snijders and Bosker 2011). First, for each model, I begin with individual-level variables only and, then, I include aggregated mean of individual-level variables to distinguish between within- versus between-country effects of these variables. If only an individual level variable is included, then its individual- and country-level effects are confounded, which causes “cluster-level confounding or cluster-level omitted-variable bias” problem (Rabe-Hesketh and Skrondal 2008). Thus, it should be tested if one variable’s between-county effect is different from its within-country effect (Bartels 2008; Snijders and Bosker 2011). Therefore, I test the model with aggregated mean of all individual-level variables, and only the statistically significant variables are included in the final model for the consideration of parsimoniousness. 67 Second, I explore possibilities of a random slope, that is, a varying effect of an individual-level variable across countries, following Snijders and Bosker’s (2011) suggestion. It would be the best to choose between a fixed and a random coefficient based on theory, but statistical considerations are also important. Snijders and Bosker (2011, 87) state that one may easily decide not to consider random slopes when no theory suggests any clues; but “this implies 65 Examination of the relationship between these two concepts is beyond the scope of this dissertation and left for future studies. 66 The number of country-level units is twenty-eight. Note that this number of units is not too small to apply a multilevel modeling. According to Stegmueller’s (2013) results from a Monte Carlo experiment, the relative bias of estimates is close to zero when the number of country is greater than 25 (and the ICC is .10). 67 The non-significant, aggregated mean of individual-level variables are dropped because including a number of unnecessary country-level factors can make the model estimation unstable (Hoffman 2010; Snijders and Bosker 2011). 102 a risk of invalid statistical tests, because if some variable [has] a random slope, then omitting this feature from the model could affect the estimated standard errors of the other variables.” Therefore, the possibilities of varying individual-level effects across nations will be explored. 4.6 Data and Measurements 4.6.1 Dependent Variable: Public Attitudes toward Bureaucracies and Satisfaction with Democracy To measure citizens’ attitudes toward bureaucracies and satisfaction with democracy, the survey questions from the 2004 International Social Survey Programme (ISSP) are used. Due to the data availability, only twenty-eight countries among OECD member nations are examined. 68 The first dependent variable gauges public attitudes toward bureaucracies. Based on the three questions that ask how respondents think about public services, I create an index of attitudes toward bureaucracies. The survey questions used to create this index are:69 Q54. Thinking of the public service in (COUNTRY), how committed is it to serve the people? [very committed, somewhat, not very, not at all committed] Q55. When the public service makes serious mistakes in (COUNTRY) how likely is it that they will be corrected? [very likely, somewhat, not very, not at all likely] Q56. How widespread do you think corruption is in the public service in (COUNTRY)? [hardly anyone is involved, a small number, a moderate number, a lot, almost everyone is involved] 68 Six countries that are not included in the ISSP survey data are Estonia, Greece, Iceland, Italy, Luxembourg, and Turkey. 69 Note that in the survey, it tries to clarify the meaning of public service. Translation notes in the codebook say the following: 54-56. Public service should be translated with the appropriate term for government officials. Do not use the term “bureaucrat.” 103 The three survey items are recoded so that higher values indicate more positive attitudes, and then standardized. Cronbach's alpha for the three variables is .656 in pooled data. 70 The index ranges from -1.96 to 2.09. Aggregated at the country-level, Poland has the least favorable attitudes toward public bureaucracies (-.66), while New Zealand has the most favorable (.49). The average value of aggregated mean of this index is .007 (std. = .33), where Hungary is the nation that comes closest to this mean value, at .006 (see Table 4.2). The second dependent variable captures political support for regime performance, which is a level of respondents’ satisfaction with operation of democracy in their country. The survey question asks: “Q57. On the whole, on a scale of 0 to 10 where 0 is very poorly and 10 is very well, how well does democracy work in (COUNTRY) today?” Aggregated mean value at the 70 In each of individual countries, the smallest Cronbach's alpha is .286 in Israel and, then, it ranges from .537 in Mexico to .711 in Slovakia. In Israel, the correlations between three variables are relatively weak; especially, correlation between the first variable (i.e., perceived commitment of bureaucracy) and the third one (i.e., perceived corruption in bureaucracy) is not statistically significant. With this exceptional case of Israel (illustrated below) I check whether the results change when Israel is excluded; the relationship between Bureaucratic Profiles and political support holds for the models examined in this chapter. Also, after excluding Israel, I calculate the factor scores for each country separately and conduct analyses; there are also no substantial changes. In this chapter, I report the results when dependent variable is the factor scores obtained from the pooled data (including Israel). Let me briefly explain Israeli case. A weak relationship among the three variables can be due to diverged attitudes between Israeli Jews and non-Jews (Vigoda-Gadot and Yuval 2004). In the 2004 ISSP data, Israeli Jews and Israeli Arabs are not distinguished as in other years of the ISSP survey, but I am able to compare the differences between these two groups based on respondents’ religious affiliation. For Israelis who are not Jewish, I find that, the correlation between the two variables is the same as in other nations—positive and significant—and the Cronbach’s alpha of the three items is .564. However, for Israeli Jews the correlation between the two variables is negative and significant (i.e., Pearson’s correlation coefficient is -.068, with the p-value of .036). This may be because Israeli Jews perceive bureaucrats’ involvement in corruption differently than Israeli Arabs; and think corrupted bureaucrats are actually committed to serve people. Indeed, in Israel it is noted that Arab citizens generally do not receive equal treatment from the public administrators (Yaroni and Rosenbloom 2003). Moreover, Israeli bureaucracy is highly politicized and partisan politics prevails in personnel management; some types of bureaucratic corruption, such as “petty bribery” and “proteksia” (the use of political connections for favorable treatment by bureaucrats), are condoned (Yaroni and Rosenbloom 2003; Werner 1983). 104 country-level ranges from 4.17 (in Slovakia) to 7.77 (in Denmark). The average value of aggregated mean of this dependent variable is 5.98 (std. = .89), where Spain has the value closest to this mean value, at 6.13 (See Table 4.2). Table 4.2. Citizens’ Positive Attitudes toward Bureaucracies and Satisfaction with Democracy (mean) Positive Attitudes toward Bureaucracies Satisfaction with Democracy New Zealand .49 Denmark 7.77 Switzerland .38 Austria 7.02 Austria .38 New Zealand 6.91 Denmark .34 Switzerland 6.88 Norway .30 Australia 6.88 Ireland .28 Canada 6.78 United Kingdom .27 Finland 6.72 Canada .26 Norway 6.67 Israel .25 Ireland 6.65 Australia .23 Netherlands 6.64 Finland .22 United States 6.58 United States .17 Sweden 6.38 Netherlands .15 United Kingdom 6.13 Sweden .11 Spain 6.13 Spain .06 Israel 5.70 Germany .05 Korea, South 5.68 Belgium .05 Portugal 5.68 Hungary .01 Germany 5.62 France -.04 Chile 5.62 Korea, South -.28 France 5.57 Slovenia -.31 Japan 5.43 Chile -.32 Belgium 5.40 Portugal -.33 Czech Republic 5.30 Mexico -.34 Slovenia 4.97 Slovakia -.46 Hungary 4.95 Japan -.49 Mexico 4.93 Czech Republic -.55 Poland 4.22 Poland -.66 Slovakia 4.17 Note: It describes cross-national differences in levels of political support in the twenty-eight OECD member nations. It shows each country’s mean value of the two dependent variables. The data are obtained from the 2004 ISSP and the survey questions are described in the text. The first column presents each nation’s mean score on the index of attitudes toward bureaucracies. The second column reports a country’s mean values on the question about satisfaction with democracy, on a 11-point scale, from 0 (very poorly satisfied) to 10 (very well satisfied). 105 4.6.2 Key Independent Variable: Bureaucratic Profiles A new measure of Bureaucratic Profiles, created as described in Section 4.4.1, is used to examine which configuration of bureaucratic traits are relevant to higher levels of political support and whether there is a systematic pattern between bureaucratic profiles and levels of political support. Bureaucratic Profiles is a continuous variable which theoretically ranges from (-π) to (π) radians. The larger the absolute value of Bureaucratic Profiles, the more different the bureaucratic profile (i.e., the order of the relative levels of traits) than the one for the reference vector. In the data, this variable ranges from -2.87 radians for Slovakia to 3.09 radians for Chile (see Table 4.1. for the distribution of this variable). In the results below, the coefficient of this variable will indicate the changes in the dependent variable for each one radian change in Bureaucratic Profiles in a counterclockwise direction. But, it will be more important to identify a systematic pattern between the rank order of bureaucratic traits and levels of political support. Therefore, in interpreting the results, I will discuss in more detail which bureaucratic profile is related to higher or lower levels of political support, by looking at the predicted values of dependent variables and their associated Bureaucratic Profiles. 4.6.3 Control Variables 4.6.3.1 Individual-level Control Variables Other factors that have been argued to predict political support are also controlled at both the individual- and the country-level. At the individual level, I control for demographic variables, 106 including age (modeled as a non-linear effect71), education, 72 sex (female), and working status (unemployed). In addition, generalized trust73 and the level of political interest 74 are controlled.75 I hypothesize that the old, males, the employed, and those who attain a higher level of education will be more likely to have higher levels of political support because they tend to be in a socially and economically better condition than vulnerable groups (e.g., the young, females, the unemployed, and the less-educated), as found in previous studies (e.g., Brehm and Rahn 1997; Chang and Chu 2006; Dalton 2004; Hibbing and Theiss-Morse 2001). I also expect that a person who is inclined to trust people in general will be more likely to have favorable attitudes toward bureaucracies and democratic government, following Andrain and Smith (2006). And, an individual with more interests in politics is expected to have higher 71 Age variable is modeled as a potentially non-linear effect, following Van der Meer’s (2010) suggestion. Along with age, an additional variable, which is squared and divided by 100 (to adjust extreme values), is included. Age is recoded so that the value of zero is meaningful: zero indicates a 47-year-old person. 72 Education variable ranges from -3 (“No formal qualification”) to 2 (“University degree completed”). The value of zero indicates “Higher secondary complete.” 73 The responses are dichotomized. The survey question asks “Generally speaking, would you say that people can be trusted or that you can’t be too careful in dealing with people?” Responses are recoded: “People can almost always be trusted” and “People can usually be trusted” are coded as 1, and “You usually can’t be too careful in dealing with people” and “You almost always can’t be too careful in dealing with people” are coded as 0. 74 Political interest variable is measured using the survey question, “How interested would you say you personally are in politics?” This variable ranges from -1.5 (not at all interested) to 1.5 (very interested). 75 It should be noted that some variables are not controlled in this analysis due to the lack of data. For example, as mentioned earlier, Anderson and Guillory (1997) argue that individuals’ political majority/minority status affects levels of satisfaction with democracy and this relationship is mediated by political institutions. But, this political allegiance account is not included in the analyses here because of a large number of missing values in the responses to a party-identification question in the 2004 ISSP data. For example, 40% of respondents do not answer this question in Austria and 53% in Poland. Another factor that cannot be controlled is the perception of personal economic situation. The variable of ‘Top - Bottom self-placement,’ which is available in the 2004 ISSP data, can be used; but I do not include this because it is not asked in the United Kingdom. Although no data are available to directly measure a personal financial situation, the employment status (unemployed) can be considered as a proxy for this variable (McAllister 1999). 107 levels of political support, as prior studies have found (e.g., Anderson and Guillory 1997; Anderson and Tverdova 2003). A dummy variable for public sector workers is also included because an individual who works for the government is more likely to have positive attitudes toward government than one who does not. Additionally, for the model predicting citizens’ satisfaction with democracy, I account for an individual’s perceived level of fairness of election.76 This is because peoples’ view of democracy is closely related to the free and fair elections (Chu et al. 2008). I posit that a respondent is more likely to be satisfied with democracy as she perceives elections as fair. 4.6.3.2 Country-level Control Variables The country mean of the individual-level variables are included in order to discern their between- and within-country effects (see Bartels 2008; Enders and Tofighi 2007). Particularly, following Newton (2001), I expect that the aggregated level of generalized trust in each country (i.e., the percentage of trusting persons in a country) will have a positive association with public attitudes toward bureaucracies and democratic performance. As mentioned earlier, there are several mechanisms that would make this link possible. People in general more actively demand what they want and participate in politics in a high-trust society; governments will perform better due to reduced transaction costs in such a society (Fukuyama 1995; Newton 2001; Putnam 1993; Tavits 2006). And, it needs to be tested empirically whether or not this aggregated level relationship is different from the individual level association. 76 The perceived fairness of election is measured using the following survey question: “Thinking of the last national election in (COUNTRY), how fair was it regarding the opportunities of the candidates and parties to campaign?” The response ranges from -2 (very unfair) to 2(very fair). 108 To account for alternative explanations for the cross-national differences in political support, I consider three other country-level variables.77 First, I control for economic growth rate using the annual percentage growth rate of GDP/per capita from the World Bank. Previous studies show that economic performance could contribute to higher levels of support for governments (Anderson and Tverdova 2003; McAllister 1999; Norris 1999c). Economic growth should be positively related to political support because people link better national economic performance with well-functioning bureaucracies and their democratic government. The level of democracy is another explanation for cross-level variations in levels of political support. For example, McAllister (1999) and Mishler and Rose (2001) suggest that political support in post-Communist countries can be different from other Western European nations because of their relatively new experience with democratic institutions. Also, as individuals’ freedom and political rights are protected in a nation, people will have higher levels of political support. Thus, the level of democracy can be an important factor to explain crossnational differences in levels of political support. To measure the level of democracy, I include the Freedom House index (i.e., the average score of ‘political rights’ and ‘civil liberties’).78 Third, I control for the type of electoral system. The consensual democracy argument implies that the government in countries with PR electoral systems will have higher levels of political support because citizens’ preferences are better represented and public participation is generally encouraged (Lijphart 1999; Norris 1999b; Norris 2008). However, following Rohrschneider (2005), I do not expect that the type of electoral systems will explain the 77 They are measured at the year the ISSP survey was actually conducted in each country. Country-level control variables (except for a dummy variable of PR electoral systems) are grandmean centered so that the value of zero is meaningful. See Table C.2, in Appendix C, for the correlation matrix for these variables. 78 This variable is recoded from the original data so that it theoretically ranges from -3 (least free) to 3 (most free). In the dataset, it ranges from 1.5 to 3. 109 variations of political support across nations—especially in terms of attitudes toward bureaucracies and satisfaction with democracies. This is because majoritarian and PR electoral systems target different properties of democracies—either accountability or representation—and, thus, the PR system will not particularly contribute to higher levels of political support (Golder and Stramski 2010). I measure PR electoral systems using the Database of Political Institutions (DPI) (Beck et al. 2001; Teorell et al. 2011). Note that the consensual democracy argument also suggests that a parliamentary system and a federal system perform better than a presidential system and a unitary system, respectively (Lijphart 1999; Norris 2008). Thus, one may argue that nations with such power-sharing institutions will have higher levels of political support. However, I expect that all these political institutions will not have direct effects on political support, as in the case of electoral systems,79 and thus they are not included in the current analyses. 80 79 Although it is plausible that such power-sharing institutions contribute to the better quality of democracy (Lijphart 1999; Norris 2008), their impact on public attitudes is ambiguous. In fact, studies have not focused on the consequences of these institutions on public attitudes. For example, the merits of each form of government have been discussed (e.g., Cheibub 2006; Gerring, Thacker, and Moreno 2005; Linz 1990), but it is not clear whether parliamentary systems will lead to higher political support than presidential systems. Norris (1999) finds a very small difference between parliamentary and presidential systems in the level of political support. Moreover, Weaver and Rockman (1993) argue that the institutional influences on government effectiveness are more relevant to the second tier institutions, including electoral rules. They also show a greater variation of government effectiveness within both presidential and parliamentary systems. Regarding federal versus unitary systems, the consensual democracy argument predicts that federal systems represent public preference better than unitary systems, because the power is shared between different levels of government and minority communities are provided with autonomy (Lijphart 1999; Norris 2008). However, again, its consequence on public attitudes toward government is unclear. In addition, I think that the forms of government (i.e., presidential or parliamentary) or the systems of the state (i.e., federal or unitary) will influence the variations in the characteristics of bureaucracies; thus, they may have an indirect effect on political support. I tested the model controlling for the forms of government and the systems of the state, and the results supported my expectations. The differences in the level of political support between presidential and parliamentary systems (which is measured with a system variable from the DPI) and between (constitutional) federal and unitary systems (which is measured using a fedtype 110 4.7 Empirical Results and Discussion: Citizens’ Attitudes toward Bureaucracies Models Let us begin with the model that explores public attitudes toward bureaucracies.81 First, because my main interest is to examine the relationship between Bureaucratic Profiles and political support, I start with a simple random intercept model. The results of the random intercept models are discussed in Section 4.7.1. Meanwhile, the relationship between Bureaucratic Profiles and political support is further illustrated in Section 4.7.2. Section 4.7.3 presents random slope models, because it is important to explore the possible random coefficients for a statistical consideration. 4.7.1 Random Intercept Models: Attitudes toward Bureaucracies The results for random-intercept multilevel models for citizens’ attitudes toward bureaucracies are presented in Table 4.3 and 4.4. I begin with the model where only individuallevel variables are included to explain public attitudes (Model 1) and compare it with the model where my key independent variable, Bureaucratic Profiles is added (Model 2) in Table 4.3. The models with other country-level controls are shown in Table 4.4. Let us begin with the model where only individual-level variables are included. variable from Norris (2008)) are not statistically significant, controlling for the individual-level variables. The results are the same when Bureaucratic Profiles is controlled, while the relationship between Bureaucratic Profiles and political support is statistically significant. 80 Descriptive statistics for all variables used in the analysis can be found in Table C.1 in Appendix C. 81 All multilevel models in this chapter are estimated using xtmixed command in STATA SE10.1. The residual diagnostics are discussed in Appendix D. 111 Table 4.3. Random-Intercept Multilevel Models of Citizens’ Attitudes toward Bureaucracies, with Individual-level Controls Only Level 1 Controls (Model 1) Bureaucratic Profiles with Level 1 Controls (Model 2) Country Level .108 (.028)* Bureaucratic Profiles GDP/capita growth rate Democracy (Freedom House) Electoral system: PR Country-level generalized trust Individual Level .264 (.010)* .264 (.010)* Generalized Trust * -.001 (.000) -.001 (.000)* Age .009 (.002)* .009 (.002)* Age2/100 * -.021 (.010) -.021 (.010)* Female .013 (.004)* .013 (.004)* Education * -.070 (.024) -.070 (.024)* Unemployed .189 (.012)* .189 (.012)* Public Sector Worker * .066 (.006) .066 (.006)* Political Interest * -.173 (.055) -.155 (.045)* Constant Variance Components .082 .053 Country-level intercept Individual-level .555 .555 Residuals Proportion Reduction in Error:** 08.11% 12.22% at Level 1 25.13% 51.19% at Level 2 57348.854 57341.774 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 25409. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. ** Proportion Reduction in Error is calculated at both levels, as suggested by Snijders and Bosker (2011). See ft. 82. As shown in Model 1, individual-level variables generally display the expected associations with attitudes toward bureaucracies. Specifically, males and those who attain higher education are more likely to have positive attitudes than their counterparts, on average, 112 everything being equal. This suggests that people who are in a socioeconomically better situation have more positive views on bureaucracies, perhaps because their experience with government agencies has been relatively good. In addition, ‘age’ variable has a non-linear effect: attitudes toward bureaucracies are the most negative among the mid-fifties. It may imply that individuals begin to view bureaucracies more negatively as they expose to more hardship and accumulate social experiences, especially from their twenties to fifties. But, then, as they are retired and get older, they become to have positive views. This non-linear relationship, alternatively, may reflect the generational differences. That is, those who are in mid-fifties in 2004 have more negative attitudes toward bureaucracies than other generations. Other individual-level variables also have expected relationships. As hypothesized, public sector workers have more positive attitudes than a person who does not work in the public sector. Perhaps, this is because a bureaucrat views himself and his colleagues as being committed to the service, and bureaucracies as doing a good job in general. Furthermore, an individual with greater interest in politics is more likely to have positive attitudes toward bureaucracies. This is perhaps because a person is more likely to have accurate information about how government bureaucracies are performing as he gets interested in politics. An individual who is not interested in politics may have a negative image of bureaucracies without concrete evidence. Finally, an unemployed respondent shows less positive attitudes toward bureaucracies. This seems to reflect two things. On the one hand, the unemployed are in an economically vulnerable situation than the employed and, thus, they view governmental bureaucracies more negatively. On the other hand, the unemployed may blame governments’ economic performance for their employment status and, thus, have negative attitudes about bureaucracies. 113 When Bureaucratic Profiles is included, as shown in Model 2 of Table 4.3, the proportion reduction in error (PRE) 82 at the level-2 increases from 25.13% to 51.19%. The positive coefficient of this variable means that as the value of Bureaucratic Profiles increases by one radian, public attitudes toward bureaucracies increase by .108 points, on average, when individual-level variables are held constant. The effect of Bureaucratic Profiles when all controls are included will be discussed in more detail later in Section 4.7.2. As the next step of model-building, I test whether within- and between-country effects are different for all individual-level variables. It turns out that only interpersonal trust has an independent between-country effect (see Table C.3 in Appendix C). Model 3 in Table 4.4 shows the results, when the aggregated-mean of interpersonal trust is included. Other aggregated mean variables are dropped for the parsimoniousness of the model. The t-ratio for the coefficients of this “Country-level generalized trust” variable (t = .012/.003 = 4) indicates that a between-country effect of interpersonal trust is indeed different than its within-country effect. That is, a trusting person is more likely to have positive attitudes toward bureaucracies (by .263 points higher) than a person who does not have generalized interpersonal trust, all things being equal. At the same time if her country has more trusting 82 There are several ways of calculating the variance explained by the model (Hoffman 2010). In Table 4.3 through Table 4.8 (and in Appendix C), I report the proportional reduction of error (PRE) at both level-1 and level-2, as suggested by Snijders and Bosker (1999). PRE for predicting level-1 and level-2 outcomes are calculated as following, respectively: Also note that for calculating PRE for random slope models, they suggest using the parameter estimates obtained from the model with the same fixed part but omitting random slopes (i.e., the random intercept model). Thus, PREs are calculated in this way for the following random slope models, Models 5, 6, 11, and 12. 114 persons than other countries, it will contribute to more positive attitudes, independent from whether or not she trusts people in general. Therefore, an individual in a country with a larger proportion of trusting people (which can be interpreted as a country with higher levels of social trust) is, on average, more likely to have more positive attitudes about bureaucracies than an individual in a nation with lower levels of social trust, all things being equal. Specifically, the between-country effect of interpersonal trust is .275 (=.263 + .012). This finding provides support for the association between social trust and political support at both the individual- and the country-level. Previous studies have debated about the level this relationship holds (e.g., Newton 2001), and this result suggests that both are possible. This leads to an interesting question of whether the effect of an interpersonal trust on attitudes toward bureaucracies varies depending on the level of social trust in society. And this will be discussed further with the random slope models explored in Section 4.7.3. As shown in Model 3 (in Table 4.4), the coefficient for Bureaucratic Profiles is positive and statistically significant, when the aggregated-mean of interpersonal trust in included, although the size of the effect decreases from .108 to .074. When other country-level variables are controlled, the coefficient for Bureaucratic Profiles changes to .053, but is still positive and statistically significant.83 83 One may suspect the multicollinearity between country-level variables or concern that the inclusion of a number of country-level variables causes model instability due to the smaller number of level-2 units. Therefore, I tested the models where the key independent variable, Bureaucratic Profiles, and each one of three county-level control variables are included. The results are presented in Table C.4 in Appendix C. The coefficient for each of county-level control variables does not reach statistical significance, holding constant for Bureaucratic Profiles and other individual-level variables (and aggregated level of generalized trust). But, the coefficient for Bureaucratic Profiles is always statistically significant. These results provide supportive evidence for the relationship between Bureaucratic Profiles and citizens’ attitudes toward bureaucracies. 115 Table 4.4. Random-Intercept Multilevel Models of Citizens’ Attitudes toward Bureaucracies, with Country-level Controls With the Aggregated-Mean of Level 1 Controls (Model 3) With Level 2 Controls (Model 4) Country Level .074 (.025)* .053 (.030)* Bureaucratic Profiles -.041 (.031) GDP/capita growth rate Democracy .147 (.122) (Freedom House) -.048 (.115) Electoral system: PR Country-level .009 (.003)* .010 (.003)* generalized trust Individual Level .263 (.010)* .263 (.010)* Generalized Trust * -.001 (.000) -.001 (.000)* Age .009 (.002)* .009 (.002)* Age2/100 * -.021 (.010) -.021 (.010)* Female .013 (.004)* .013 (.004)* Education * -.070 (.024) -.070 (.024)* Unemployed .189 (.012)* .189 (.012)* Public Sector Worker * .066 (.006) .066 (.006)* Political Interest * -.161 (.038) -.120 (.099) Constant Variance Components .038 .036 Country-level intercept .555 Individual-level .555 Residuals Proportion Reduction in Error:** 14.50% 14.71% at Level 1 65.61% 66.95% at Level 2 57341.816 57347.866 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 25409. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. The results shown in Model 4 (in Table 4.4) indicate that as the value of Bureaucratic Profiles increases by one radian, public attitudes toward bureaucracies increase by .053 points, all things being equal. That is, let say there are two individuals who are the same but different only in their countries’ Bureaucratic Profiles: One person’s country has the value of 116 Bureaucratic Profiles that is one radian greater than the country of another person. This person will have, on average, .053 points more positive attitudes toward bureaucracies than another. Because, in general, as the value of Bureaucratic Profiles increases, a bureaucracy presents greater degree of impartiality and smaller level of competency, relative to other traits, this positive relationship suggests that citizens are more likely to have favorable attitudes toward bureaucracies when bureaucratic impartiality is emphasized more than other traits, especially competency. I provide a more substantive interpretation of the relationship between Bureaucratic Profiles and attitudes toward bureaucracies in the next section. The results in Model 4 also indicate that, holding constant for Bureaucratic Profiles and other individual-level variables (and aggregated level of generalized trust), economic growth, levels of democracy, and PR electoral systems do not have statistically significant relationships with public attitudes toward bureaucracies.84 Further explanations for these country-level control variables are discussed in the random slope models (in Section 4.7.3). 4.7.2 The Relationship between Bureaucratic Profiles and Attitudes toward Bureaucracies Let us look at in more detail the relationship between my key independent variable, Bureaucratic Profiles, and public attitudes toward bureaucracies. The results reported in Model 4 (in Table 4.4) show that there is a positive and statistically significant relationship between Bureaucratic Profiles and popular attitudes toward bureaucracies, as expected, when both individual- and country-level variables are controlled. In order to interpret this relationship substantively and in more depth, let us explore which bureaucratic profiles are associated with 84 The results are the same when different economic indicators—GDP per capita (logged) and inflation rate—are used. 117 the highest and lowest predicted values of dependent variable, public attitudes toward bureaucracy. Figure 4.2 shows the mean of the predicted values for each country, along with 95% confidence intervals. The predicted values are calculated by considering linear predictors of the fixed part and the contributions based on predicted random effects (which here is a random intercept). 1 Figure 4.2. Predictions and approximate 95% confidence intervals versus ranking of citizens’ attitudes toward bureaucracies based on Model 4 in Table 4.4 NZ .5 CH AT BE DE ES US IL DK MXPT SI CL KR 0 Prediction FRHU FR HU SE NL NO AU FI CA UK IE SK JP CZ -1 -.5 PL 0 10 20 30 Rank Note: The predicted values are calculated based on a random-intercept model for citizens’ attitudes toward bureaucracies, which are shown in Model 4 in Table 4.4. The predictions for public attitudes toward bureaucracies are aggregated at the country-level. These values are ranked and plotted with approximate 95% confidence intervals. Country names are shown on top of confidence intervals. See Note in Figure 3.3 for the full names of the thirty-four countries. 118 Which countries have the highest and the lowest predicted values of attitudes toward bureaucracies? As shown in Figure 4.2, New Zealand has the highest predicted value. The bureaucratic profile of this country indicates that impartiality is emphasized the most and careerbased system is presented the least in a bureaucracy. The specific order of the relative levels of bureaucratic traits exhibited in this nation is: Impartiality, independence from politics, female representativeness, competency, and career-based system (see Table 4.1). That is, in a nation with the most positive attitudes toward bureaucracies, political independence and female representativeness are also emphasized to some extent, but less than impartiality. And, careerbased system and competency are the least presented attributes. On the other hand, Poland has the lowest predicted value of public attitudes toward bureaucracy. And Poland’s bureaucratic profile shows that bureaucratic traits are presented in the following order: Competency, female representativeness, career-based system, political independence, and impartiality (see Table 4.1). First, the comparison of these two nations’ bureaucratic profiles suggests that citizens are more likely to view their bureaucracy positively when impartiality is presented to the greater extent than other traits in a bureaucracy. This is because New Zealand emphasizes impartiality more than other four traits, whereas impartiality is the least exhibited trait in Poland. The comparison also suggests that when competency is emphasized more than other traits and, at the same time, when impartiality is the least presented trait, citizens perceive a bureaucracy more negatively. To understand this visually, let us look at Figure 4.3. New Zealand is around the “2:00” position and Poland is around the “9:00” position. And, the closest point to the vector terminus is impartiality in case of New Zealand, whereas it is competency for Poland. Moreover, Figure 4.3 shows that as a nation vector moves in a counterclockwise direction from Poland to New 119 Figure 4.3. Full MDPREF model, with the vectors for New Zealand and Poland Note: Open circles and closed circles represent the nation vector’s terminal points and the five traits, respectively, that are estimated from the MDPREF analysis as discussed in Chapter 3. The vectors for New Zealand and Poland, nations with the highest and the lowest predicted value of citizens’ attitudes toward bureaucracies, respectively, are presented with arrows (in solid lines) and their extended lines (in dashed lines). The vector for New Zealand is shown with an arrow pointing upper-right in the space. It is estimated that the bureaucratic traits are presented in New Zealand in the order of impartiality, independence from politics, female representativeness, competency, and career-based system. The vector directing upper-left is for Poland, where the bureaucratic traits are exhibited in the order of competency, female representativeness, careerbased system, independence from politics, and impartiality. See Note in Figure 3.3 for the full names of the thirty-four countries. 120 Zealand,85 the relative importance of competency decreases and that of impartiality increases. In turn, it indicates that individuals view bureaucracies more favorably when impartiality is presented to a greater degree than competency in their national bureaucracies. Second, the result suggests that when independence from politics is emphasized, in addition to bureaucratic impartiality, citizens are more likely to have positive attitudes toward bureaucracies. In New Zealand’s bureaucratic profile, independence from politics is the second most important trait. But, it is the fourth one in Poland. On the one hand, this may reflect that political independence generally present together with impartiality in a bureaucracy. And, it provides some support that a sore emphasis on politicization will not have a positive impact on popular attitudes toward bureaucracy. As political independence is exhibited to a lesser degree than other traits, which may occur with increasing politicization, citizens tend to have negative attitudes toward bureaucracies. On the other hand, it does not necessarily mean that people view bureaucracies the most favorably when political independence is emphasized more than other traits in a bureaucracy: nations around the “5:00” position do not enjoy the most positive public attitudes of bureaucracies. Therefore, these findings suggest that independence from politics is an important bureaucratic trait, but it should not undermine impartiality in a bureaucracy. Is the relative level of female representativeness also closely related to different levels of public attitudes toward bureaucracies? It is the third most important trait in New Zealand and it is the second most important one in Poland. The relative level of female representativeness itself appears not to be a critical factor shaping popular view about bureaucracies. But, as closely examining bureaucratic profiles, it shows that the rank order between impartiality and 85 Recall that as the value of Bureaucratic Profiles increases, a nation vector in the joint space moves in a counterclockwise direction. 121 representativeness changes as a nation vector moves from Poland to New Zealand (in a counterclockwise direction). That is, impartiality is exhibited less than representativeness in Poland, and it is the opposite in New Zealand. It suggests that, therefore, bureaucracies are viewed favorably by citizens when impartiality is emphasized over representativeness. Finally, how is the relative level of career-system exhibited in a bureaucracy associated with public attitudes toward bureaucracies? It is the least presented trait in New Zealand, while the third most important trait in Poland. Does this mean that people view their bureaucracy more positively when career-based system is displayed at a lower level than other traits? The relationship does not seem to be straightforward. On the one hand, if both bureaucratic impartiality and political independence are secured, as in New Zealand, then attitudes toward bureaucracies become more positive as career-based system is exhibited less than other traits. On the other hand, this relative emphasis on position-based system (i.e., lack of career-based system) does not always have a positive effect on public attitudes: for nations where bureaucracies are not impartial and under greater political influence, emphasis on position-based system seems to be associated with more negative popular attitudes toward bureaucracies. This may reflect problems faced by some countries that implemented public management reforms, while basic levels of impartiality and political independence are not established, as in Central and Eastern European countries (Raadschelders, Toonen, and Van der Meer 2007). Taking all the interpretations discussed previously into consideration, I conclude that citizens are more likely to view a bureaucracy positively when impartiality (together with political independence) is presented more than other traits and, at the same time, competency is relatively less emphasized than other attributes. This implies that people may want bureaucracies prioritize impartiality over other traits. 122 In addition to this, more importantly, the findings provide information about how other bureaucratic traits should be presented, relative to impartiality, in a national bureaucracy. In other words, to be a bureaucracy that is suitable in democratic societies and, more specifically, one that is publicly supported, some degree of independence from politics should be guaranteed. An exclusive emphasis on political independence is not desirable, but lack of political independence is also problematic. Moreover, prioritizing competency over other bureaucratic traits is also not viewed positively by citizens. When a bureaucracy has high levels of competency, but it is not impartial, and not free from political influence, people may think that such a bureaucracy is not serving the public interest. Instead, this type of bureaucracy only enjoys a predominant status but behaves arbitrarily. Finally, only when impartiality is emphasized more than other traits, a higher level of female representativeness and a lower degree of career-based system seem to be associated with positive popular attitudes toward bureaucracies. 4.7.3 Random Slope Models: Attitudes toward Bureaucracies This section examines the possibilities of varying individual-level coefficients across nations, following Snijders and Bosker’s (2011) suggestion. The previous model (Model 4) is reexamined by allowing all individual-level variables to vary across nations. As done previously, I only retain significant random slopes in the model, based on the deviance difference test. Model 5 in Table 4.5 shows the results for the random-slope model when Bureaucratic Profiles and other country-level control variables are included. 123 Table 4.5. Random-Slope Multilevel Models of Citizens’ Attitudes toward Bureaucracies Random Slopes, with Level 2 Controls (Model 5) Cross-level Interaction: Interpersonal Trust and Social Trust (Model 6) Country Level .046 (.025)* .070 (.024)* Bureaucratic Profiles .016 (.024) GDP/capita growth rate Democracy .322 (.099)* (Freedom House) -.117 (.092) Electoral system: PR Country-level .014 (.003)* .007 (.003)* generalized trust Individual Level .261 (.018)* .259 (.017)* Generalized Trust -.000 (.001) -.000 (.001) Age 2 * .009 (.002) .009 (.002)* Age /100 -.022 (.010)* -.022 (.010)* Female * .014 (.004) .014 (.004)* Education -.069 (.023)* -.069 (.023)* Unemployed * .197 (.018) .197 (.018)* Public Sector Worker .061 (.010)* .062 (.010)* Political Interest Cross-level Interactions Generalized Trust * Country-level .003 (.001)* generalized trust Generalized Trust * Democracy (Freedom House) -.068 (.088) -.162 (.043)* Constant Variance Components .062 .048 Country-level intercept Individual-level .006 .005 Generalized Trust .000 .000 Age .004 .005 Public Sector Worker .002 .002 Political Interest .549 .549 Residuals Proportion Reduction in Error: 14.71% 14.32% at Level 1 66.95% 64.20% at Level 2 57193.39 57197.1 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 25409. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. 124 The results indicate that four variables have random slopes. The effects of interpersonal trust, age, public sector workers, and political interest vary across nations. Whereas the reason why these effects vary is another interesting topic to investigate in future studies, in this chapter I will focus on the effect of interpersonal trust. Indeed, this is one of interesting findings: interpersonal trust has both the within- and the between-country effects, and the within-country effect varies across countries. This indicates that a trusting person is more likely to have positive attitudes toward bureaucracies than a person who does not have interpersonal trust, but the size of this effect (precisely, difference between the two) is not the same for all nations. One possible explanation for the random slope of interpersonal trust is levels of social trust in a country. A trusting person views bureaucracy more positively, but this is more likely to happen when other people in her county also tend to trust others— which means, higher levels of social trust in her country. Recall that prior studies posit that interpersonal trust increases a person’s engagement in social and political activities, and this makes the government more trustworthy (e.g., Fukuyama 1995; Putnam 1993; Tavits 2006). But, in order to make this happen, there should be a relatively large number of people with generalized interpersonal trust. If a trusting person lives in a nation with lower levels of social trust (i.e., if a smaller percentage of people have interpersonal trust in her country), the impact of interpersonal trust on political support may be less recognizable. In order to take this possibility into account, I test the model including a cross-level interaction effect between interpersonal trust and the percentage of trusting persons in a nation. In the model, I only include the main independent variable, Bureaucratic Profiles, and other country-level independent variables are not included. This is because of a possible multicollinearity problem and also because too many level-2 variables can cause model 125 instability, especially when cross-level interactions are considered (Hoffman 2010; Snijders and Bosker 2011). The results are shown in Model 6 in Table 4.5. .2 .4 .6 Figure 4.4. Cross-level interaction effect between generalized trust and social trust on citizens’ attitudes toward bureaucracies 0 Trusting person -.4 -.2 Untrusting person 0 20 40 60 Percentage of trusting persons 80 100 Note: It shows the changes in the difference between trusting and untrusting person in their attitudes toward bureaucracies across different levels of social trust (i.e., percentage of trusting persons in a nation). Two reference lines (i.e. dotted lines) indicate the range of observed values of social trust. The results are based on Model 6 in Table 4.5. When the multiplicative term is included in the model, the cross-level interaction effect is statistically significant. This implies that the individual-level effect of interpersonal trust varies across nations and the effect gets greater if an individual is living in a country with higher levels of social trust. Specifically, a trusting person in a country where 47% of people have generalized trust has more positive attitudes about bureaucracies than an untrusting person, by .259 higher 126 points, all other things held constant. But, when there are 10% more trusting people in her country, everything being equal, she is more likely to view bureaucracies positively by .289 higher points (= .259 + .003*10), compared to an untrusting person. This suggests that the individual-level effect of interpersonal trust becomes greater in a high-trust society.86 This interaction effect is depicted in Figure 4.4. For the effects of country-level independent variables, let us look at Model 5 in Table 4.4 again. In the random-slope model (Model 5), the relationship between Bureaucratic Profiles and public attitudes toward bureaucracies holds as in the previous models, although the size of the effect changes to .046. What about country-level control variables? Similar to the random intercept model with the country-level controls (Model 4), economic growth does not have a statistically significant relationship with attitudes toward bureaucracy. The sign of the coefficient for economic growth is positive in the random slope model (Model 5), but it is not statistically different than zero. Interestingly, the relationship is found to be negative in the random intercept model (Model 4) and also in other models with a smaller number of country-level variables.87 When the economic condition is the only country-level independent variable, it has a negative and statistically significant relationship with public attitudes. It could be a result of higher expectation of people in nations with better economic conditions, as McAllister (1999) argues. However, when Bureaucratic Profiles is included in the model, this relationship becomes non-significant. This suggests the properties of bureaucracy may matter more for citizens than a 86 This interesting result and implication need to be explored further in future studies. Especially, because ‘the percentage of trusting people’ is a very limited measurement for social trust, future studies to reexamine this relationship using better indicators of social trust will be fruitful. For example, additional data on the reciprocity norms and network organizations could improve the measurement of social trust (Paraskevopoulos 2010). 87 This association is negative and not significant when two other country-level control variables are excluded from Model 4 (see Appendix C, Table C.4). Moreover, it is the same when other economic indicators, GDP per capita (logged) and inflation rate, are used. 127 nation’s economic situation because bureaucracies influence their daily lives and well-being through the policy making and service delivery (Coggburn and Schneider 2003; Whiteley et al. 2010). PR electoral systems have a negative effect—that is, people have more negative attitudes when their nation adopts PR electoral systems—but, it is not statistically different than zero in the random slop model (Model 5), the random intercept model (Model 4), and the model where other country-level controls are not included (see Table C.4 in Appendix C). The absence of a significant effect for PR systems lends support to my expectation. The negative relationship could be observed because PR electoral systems are likely to produce a multiparty parliament with no single chain of accountability (Norris 2004). But, it is not statistically significant effect.88 Different from the random intercept model (Model 4), the level of democracy now has a statistically significant relationship with public attitudes toward bureaucracies. More democratic nations enjoy higher levels of support for bureaucratic institutions, on average, if economic conditions, electoral systems, levels of social trust, and a bureaucratic profile are held constant. Specifically, compared to a citizen in a country with a Freedom House score of 2, a person in a country with the score of 3, on a scale ranging from -3 (least free) to 3 (most free), is more likely 88 One may argue that electoral systems influence public attitudes indirectly through levels of corruption in the government. Chang and Golden (2006) argue that corruption increases with district magnitudes under open-list PR systems. Given this argument, it is possible that public attitudes toward bureaucracies decrease when the country adopts open-list PR systems and has a large number of district magnitudes. To account for this, I tested the model with three variables—open-list PR systems, the average number of representatives elected by each electoral district to the lower house, and their multiplicative term—instead of one variable, PR systems (data are collected from DPI 2004 (Beck et al. 2001)). The multiplicative term is statistically significant and negative in the random intercept model. It seems that people have more negative attitudes about bureaucracies when district magnitudes increase, under open-list PR systems, but not in closed-list PR systems. But, it is not significant in the random slope model (and also in the model for satisfaction with democracy) and, thus, it will need a further investigation. Note that the association between Bureaucratic Profiles and public attitudes holds in these model specifications. 128 to have more positive attitudes toward bureaucracies by .322 points, all things being equal. When freedom and civil liberty are less protected in a nation, individuals appear to view bureaucracies as being more arbitrary and less committed to serve people. 1 Figure 4.5. Predictions and approximate 95% confidence intervals versus ranking of citizens’ attitudes toward bureaucracies based on Model 5 in Table 4.5 FI .5 SE FR Prediction HU NO DK AUCA NL US BE DE ES AT NZ CH UK IE IL 0 CL CZ SK MX PT SI KR JP -1 -.5 PL 0 10 20 30 Rank Note: Based on a random-slope model for citizens’ attitudes toward bureaucracies, which are shown in Model 5 in Table 4.5, the predicted values are calculated. The predictions for public attitudes toward bureaucracies are aggregated at the country-level. These values are ranked and plotted with approximate 95% confidence intervals. Country names are shown on top of confidence intervals. See Note in Figure 3.3 for the full names of the thirty-four countries. The comparison of the random-slope model with the random-intercept one provides further support for the positive relationship between Bureaucratic Profiles and public attitudes toward bureaucracies: it is robust when random slopes are considered. Figure 4.5 presents the 129 mean of the predicted values for each country, along with 95% confidence intervals, based on Model 5 in Table 4.5. The ranking of the country’s mean predicted values is the same as that obtained from the random-intercept model. 4.8 Empirical Results and Discussion: Citizens’ Satisfaction with Democracy Models Let us turn to the next model that examines political support for the regime performance, satisfaction with operation of democracy. The models are analyzed and discussed in the same order as in the models for attitudes toward bureaucracies. Section 4.8.1 discusses random intercept models; the findings of the relationship between Bureaucratic Profiles and satisfaction with democracy are further explained in Section 4.8.2. The results from random slope models are presented and discussed in Section 4.8.3. The cross-level interaction between interpersonal trust and social trust is also considered. The findings are generally similar to the models for attitudes toward bureaucracies; the similar results are discussed briefly. 4.8.1 Random Intercept Models: Satisfaction with Democracy Table 4.6 and Table 4.7 present four random-intercept models. Model 7 in Table 4.6 presents the model when only the individual-level controls are included; it is compared to Model 8, where Bureaucratic Profiles is added. In addition, I test the within- versus between-country effect of individual level variables and include aggregated-mean of interpersonal trust in Model 9 in Table 4.7. The model when all country-level controls are included is shown in Model 10 in Table 4.7. Let us begin with interpreting the effects of individual-level variables. 130 Table 4.6. Random-Intercept Multilevel Models of Citizens’ Satisfaction with Democracy, with Individual-level Controls Only Level 1 Controls (Model 7) Bureaucratic Profiles with Level 1 Controls (Model 8) Country Level .359 (.054)* Bureaucratic Profiles GDP/capita growth rate Democracy (Freedom House) Electoral system: PR Country-level generalized trust Individual Level .525 (.027)* .527 (.027)* Generalized Trust -.000 (.001) -.000 (.001) Age .010 (.005)* .010 (.005)* Age2/100 * -.116 (.025) -.116 (.025)* Female .096 (.010)* .095 (.010)* Education * -.209 (.063) -.208 (.063)* Unemployed .019 (.032) .019 (.032) Public Sector Worker .155 (.016)* .155 (.016)* Political Interest Perceived Election .503 (.013)* .502 (.013)* Fairness * 5.504 (.140) 5.561 (.090)* Constant Variance Components .533 .204 Country-level intercept Individual-level 3.783 3.783 Residuals Proportion Reduction in Error: 13.71% 20.28% at Level 1 33.82% 74.63% at Level 2 102369.03 102346.46 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 24513. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. As shown in Model 7 in Table 4.6, individual-level variables mostly have expected relationships with citizens’ satisfaction with democracy, except that there is no statistically significant difference in levels of satisfaction between public sector workers and those who are 131 not. It shows that individuals’ opinion about the operation of democracy in their nation is not influenced by the work place, especially whether they work for government or not. Perhaps, this is because the ways that people participate in politics and their demands are channeled to the government are not critically different between governmental and non-governmental workers. In regards to other variables, males and those who attain higher-level education are more likely to be satisfied with democratic performance than their counterparts. Age has a non-linear effect: levels of satisfaction with democracy are the lowest among the late-forties. As in the models for attitudes toward bureaucracies, it seems that people become less satisfied as they get older from their twenties to late-forties. But the older generations (e.g., the sixties in the year of the survey) generally have higher levels of satisfaction than the younger generations. In addition, the unemployment status has a negative association, as expected. The unemployed may blame the government for their personal financial situation or feel that the government does not care about their interest. In turn, this will make them less satisfied with democratic performance in their country. Levels of political interest have a positive relationship. A person who is interested in politics will be better informed about how political process works and also have more chance to input her demands. In turn, she will be more satisfied with democratic performance in her country. Moreover, the more a person perceives the last elections as fair, the more likely she is satisfied with democratic performance of her country. This linkage is reasonable given that elections are important components of democracy. As my key independent variable, Bureaucratic Profiles, is added (Model 8), it explains a relatively large proportion of variance at level-2: PRE at level-2 increases from 33.82% in Model 7 to 74.63% in Model 8. Bureaucratic Profiles has a positive association with citizens’ 132 satisfaction with democracy; this is statistically significant. The interpretation of this effect is further discussed in Section 4.8.2. When the between-country effects of individual-level variables are tested, only interpersonal trust shows a country-level effect that is different from the within-country effect (See Table C.5 in Appendix C). Thus Model 9 in Table 4.7 shows the results when only aggregated-mean of interpersonal trust is included. The effect of interpersonal trust can be interpreted as the following. A trusting person has higher levels of satisfaction with democracy (by .522 points). Independent of this effect, a person living in a country with a greater number of trusting people (i.e., greater social trust) has higher levels of satisfaction with democracy compared to an individual from a low-trust country, all things being equal. The size of this between-country effect is .556 (= .522 + .034). These findings, again, support the notion that the relationship between social trust and political support exist at both the individual- and the country-level. Bureaucratic Profiles still has a positive and statistically significant association with citizens’ satisfaction with democracy. As the aggregated-mean of interpersonal trust is included in the model, only the size of the effect decreases from .359 in Model 8 to .274 in Model 9. Further, the results in Model 10 in Table 4.7 show that this positive effect of Bureaucratic Profiles holds when other country-level variables are included. 89 89 Similar to the models explaining public attitudes toward bureaucracies, I tested the models where my key independent variable, Bureaucratic Profiles, and one of three county-level control variables are included. The results are presented in Table C.6 in Appendix C. When one of country-level controls is added to Model 9, the relationship between Bureaucratic Profiles and satisfaction with democracy is always statistically significant. Holding constant for Bureaucratic Profiles and other individual-level variables (and aggregated-mean of generalized trust), only economic growth has a statistically significant effect. The coefficients for Freedom House score and that for PR electoral systems are not statistically different from zero. These results provide 133 Table 4.7. Random-Intercept Multilevel Models of Citizens’ Satisfaction with Democracy, with Country-level Controls With the Aggregated-Mean of Level 1 Controls (Model 9) With Level 2 Controls (Model 10) Country Level .274 (.042)* .251 (.042)* Bureaucratic Profiles -.149 (.043)* GDP/capita growth rate Democracy .239 (.171) (Freedom House) .133 (.161) Electoral system: PR Country-level .023 (.005)* .019 (.005)* generalized trust Individual Level .529 (.027)* .523 (.027)* Generalized Trust -.000 (.001) -.000 (.001) Age .010 (.005)* .010 (.005)* Age2/100 * -.116 (.025) -.115 (.025)* Female .095 (.010)* .095 (.010)* Education * -.207 (.063) -.207 (.063)* Unemployed .017 (.032) .017 (.032) Public Sector Worker .154 (.016)* .153 (.016)* Political Interest Perceived Election .502 (.013)* .501 (.013)* Fairness * 5.547 (.067) 5.452 (.140)* Constant Variance Components .103 .067 Country-level intercept Individual-level 3.783 3.783 Residuals Proportion Reduction in Error: 22.31% 23.02% at Level 1 87.23% 91.65% at Level 2 102337.03 102331.97 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 24513. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. The results in Model 10 indicate that as the value of Bureaucratic Profiles increases by one radian, citizens’ satisfaction with operation of democracy increases by .251 points, on support for robustness of the relationship between Bureaucratic Profiles and citizens’ satisfaction with democracy. 134 average, controlling for all individual- and country-level variables. Let say there are two persons who are the same but different only in their countries’ bureaucratic profiles: One person’s country has the value of Bureaucratic Profiles that is one radian greater than the country of another person. This person will have, on average, .251 points higher levels of satisfaction than another. And, this is statistically different than zero. I provide a more substantive interpretation of this relationship in the next section. Next, let us look at the effects of country-level control variables in Model 10 (in Table 4.7). Similar to the model explaining citizens’ attitudes toward bureaucracies, the level of democracy and PR electoral systems do not have statistically significant relationships with levels of satisfaction with democracy (see Section 4.8.3 for more discussion of these variables). Contrary to my expectation, the results in Model 10 show that economic growth has a negative relationship with levels of satisfaction with democracy. But, note that when different economic indicators—GDP per capita (logged) or inflation rate—are used, no statistically significant effect is observed. This is in line with previous studies finding mixed support for the effect of economic performance (Rohrschneider 2005). Further studies will be needed to discern this unexpected relationship. One possible interpretation of this finding is that individuals do not link improved national economic situation with better government performance. An individual’s personal financial situation may influence her level of political support, as it is implied by the negative effect of unemployment. But, an individual’s economic situation or satisfaction with national economy can be different from how national economy actually does. Moreover, some other factors may exist that undermine levels of political support in nations with higher economic growth rates. For example, in the data, most of countries with higher economic growth rate in the 135 survey year are Eastern European and Latin American countries. 90 Thus, it suggests that Eastern European and Latin American countries enjoy higher economic growth rates than other counties, but this better economic performance does not translate into higher levels of political support. If such a high economic growth rate has been achieved by sacrificing democratic values and procedures, for instance, this could lead to lower levels of satisfaction with democracy in these countries. Further studies should investigate this puzzling relationship between economic performance and political support. 4.8.2 The Relationship between Bureaucratic Profiles and Satisfaction with Democracy Let us interpret in more detail the relationship between Bureaucratic Profiles and satisfaction with democracy. Examining bureaucratic profiles in the two countries with the highest and lowest predicted values of dependent variable, satisfaction with democracy, will help to interpret the positive relationship between the two variables. Figure 4.6 shows the mean of the predicted values for each country, along with 95% confidence intervals. Figure 4.6 shows that Denmark has the highest predicted value of satisfaction with democracy and Slovakia has the lowest predicted value. In Denmark, impartiality is presented to the greatest extent compared to other bureaucratic traits, and female representativeness is the second most presented trait. And, career-based system is the least emphasized one. That being said, in nations with the highest predicted level of satisfaction with democracy, the bureaucratic 90 For example, countries whose economic growth rate is greater than one standard deviation from the mean (i.e., 4.7% of economic growth rate) include Czech Republic, Hungary, Israel, Mexico, Chile, and Slovakia. 136 traits are exhibited in the following order: Impartiality, female representativeness, independence from politics, competency, and career-based system (see Table 4.1). On the other hand, Slovakia seems to emphasize female representativeness and competency compared to other traits. And, political independence and impartiality are presented to lesser degrees than other traits. The specific order of the relative levels of traits exhibited in Slovakia is: Female representativeness, competency, career-based system, independence from politics, and impartiality (see Table 4.1). 10 Figure 4.6. Predictions and approximate 95% confidence intervals versus ranking of citizens’ satisfaction with democracy based on Model 10 in Table 4.7 ES UK SE FR JP PT KR AU AT CL IL DE 6 SI MXCZ HU BE US IE FI NL NO CA NZ CH PL 4 SK 2 Prediction 8 DK 0 10 20 30 Rank Note: The predicted values are calculated based on a random-intercept model for citizens’ satisfaction with democracy, which are shown in Model 10 in Table 4.7. The predictions for citizens’ satisfaction with democracy are aggregated at the country-level. These values are ranked and plotted with approximate 95% confidence intervals. Country names are shown on top of confidence intervals. See Note in Figure 3.3 for the full names of the thirty-four countries. 137 Figure 4.7. Full MDPREF model, with the vectors for Denmark and Slovakia Note: Open circles and solid circles represent the nation vector’s terminal points and the five traits, respectively, that are estimated from the MDPREF analysis as discussed in Chapter 3. The vectors for Denmark and for Slovakia, nations with the highest and the lowest predicted value of citizens’ satisfaction with democracy, respectively, are presented with arrows (in solid lines) and their extended lines (in dashed lines). The vector pointing upper-right in the space represents Denmark, where bureaucratic traits are presented in the order of impartiality, female representativeness, independence from politics, competency, and career-based system. The vector directing upper-left is for Slovakia, where the traits are exhibited in the order of female representativeness, competency, career-based system, independence from politics, and impartiality. See Note in Figure 3.3 for the full names of the thirty-four countries. 138 Let us look at Figure 4.7 to visually compare the Bureaucratic Profiles for Slovakia (around the “10:00” position) and Denmark (around “2:00”). 91 First, the comparison between the two profiles suggests that citizens are more likely to perceive democracy functioning satisfactorily when impartiality is the most important feature of a country’s bureaucracy. If the priority is on securing impartiality, then a nation’s bureaucracy, perhaps, provides public goods and services in a fair and unbiased way. Such a bureaucracy would contribute to citizens’ equal access to public goods and services (Rothstein and Teorell 2008a). In turn, these citizens are more likely to be satisfied with the performance of democracy in their country. These findings also suggest that individuals are less likely to be satisfied with the function of democratic regimes, when competency is emphasized more than other traits. In such a nation, bureaucrats perhaps have prestigious status. In other words, if a country put exclusive emphasis on competency, people attaining higher-level education are likely to work in the government and get paid relatively well. The size of government workforce will also be relatively large. Thus, if these privileges of bureaucrats are not accompanied by their impartial behaviors in the policy process, people may feel that their “competent” bureaucrats do not serve citizens, but instead abuse their power without popular control. In turn, this will make people less satisfied with their democratic system. In addition, political independence is also not much presented in Slovakia, together with impartiality. That being said, it seems that bureaucracies from countries with lower levels of democratic satisfaction emphasize competency without securing impartiality and political independence. The greater level of competency compared to other traits may be related to a lower degree of political independence in these countries. If this is 91 Recall that as the value of Bureaucratic Profiles increases, a nation vector in the joint space moves in a counterclockwise direction. Figure 4.7 helps to visualize the changes in the relative levels of traits presented in a national bureaucracy as a nation vector moves from Slovakia (around the “10:00” position) to Denmark (around “2:00”) in a counterclockwise direction. 139 the case, then too much emphasis on competency is perhaps related to patronage, and this decreases citizens’ satisfaction with democracy. Career-based system is the third most displayed bureaucratic trait in Slovakia and the least exhibited trait in Denmark. Does this indicate the negative impact of career-based system on citizens’ satisfaction with democracy? Although the rank order of career-based system in a bureaucratic profile decreases when comparing Bureaucratic Profiles for Slovakia and Denmark, recall that the rank order of career-based system does not continuously decrease as the value of Bureaucratic Profiles increases. It seems that, instead, which traits are emphasized together with a career-based system in bureaucracies is more important factor for citizens’ democratic satisfaction. For example, let us look at nations whose vectors located between the “7:00” and the “11:00” position in Figure 4.7. In these nations, competency is greatly emphasized and impartiality is not presented as much as other traits in a bureaucracy. For them, having relatively greater levels of career-based system seems to help enhance democratic satisfaction. On the other hand, let us compare bureaucracies where nation vectors are positioned between “1:00” and “5:00.” The priority is given to impartiality over competency in these nations—that is, levels of impartiality increase and levels of competency decreases. Among these nations, relatively lower levels of career-based systems contribute to greater satisfaction with democratic government. These findings suggest that bureaucracies emphasizing position-based systems (i.e., less degree of career-based system) are likely to contribute to democratic governance only when impartiality is presented to a greater extent. Finally, how is the relative level of female representativeness presented in a bureaucracy associated with public satisfaction with democracy? Female representativeness is the second 140 most exhibited bureaucratic trait in Denmark, whereas it is presented to a greatest degree compared to other traits in Slovakia. Why do these two countries have very different levels of public satisfaction with democracy, while both emphasize female representativeness more than other bureaucratic traits? The rank order between female representativeness and impartiality is in stark contrasts between two nations’ profile. In Denmark, impartiality is presented to a greater degree than female representativeness, whereas it is the opposite in Slovakia. Thus, the results suggest that citizens are satisfied with democracy when female representativeness in a bureaucracy is emphasized, but only when impartiality is prioritized over female representativeness. If impartiality is not presented more than other traits in a bureaucracy, a greater emphasis on female representativeness perhaps decreases public satisfaction with democracy. 4.8.3 Random Slope Models: Satisfaction with Democracy Next, the possibilities of random slopes are explored. Only significant random slopes are retained in the model based on the deviance difference test. Model 11 reported in Table 4.8 shows the results for the random-slope model, when Bureaucratic Profiles and all country-level control variables are included. The results show that three variables have the random slopes. That is, the effects of interpersonal trust, age, and perceived fairness of elections on citizens’ satisfaction with operation of democracy vary across nations. Different from the results in the model explaining public attitudes toward bureaucracies, the effect of political interest does not vary across nations and public sector workers have no significant effect. 141 Table 4.8. Random-Slope Multilevel Models of Citizens’ Satisfaction with Democracy Random Slopes, with Level 2 Controls (Model 11) Cross-level Interaction: Interpersonal Trust and Social Trust (Model 12) Country Level .267 (.036)* .275 (.040)* Bureaucratic Profiles * -.158 (.037) GDP/capita growth rate Democracy .341 (.149)* (Freedom House) .168 (.136) Electoral system: PR Country-level .017 (.004)* .020 (.005)* generalized trust Individual Level .508 (.037)* .505 (.037)* Generalized Trust -.000 (.002) -.000 (.002) Age 2 * .008 (.005) .008 (.005)* Age /100 -.115 (.025)* -.116 (.025)* Female * .097 (.010) .097 (.010)* Education -.220 (.063)* -.220 (.063)* Unemployed .025 (.032) .026 (.032) Public Sector Worker * .148 (.016) .150 (.016)* Political Interest Perceived Election .529 (.032)* .528 (.032)* Fairness Cross-level Interactions Generalized Trust * Country-level .004 (.003) generalized trust * 5.408 (.129) 5.527 (.077)* Constant Variance Components .113 .142 Country-level intercept Individual-level .018 .016 Generalized Trust .000 .000 Age Perceived Election .024 .024 Fairness 3.737 3.737 Residuals Proportion Reduction in Error: 23.02% 22.29% at Level 1 91.65% 87.03% at Level 2 102131.62 102149.26 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 24513. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. 142 I explore one possible explanation for varying effect of interpersonal trust in this chapter.92 It is possible that levels of social trust in a country influence the effect of an individual’s interpersonal trust on satisfaction with democracy. In a low-trust society, a trusting person’s participation in political and communal organizations may not be effectively translated into a collective influence on the political process. Thus, the cross-level interaction effect between interpersonal trust and social trust (i.e., the percentage of trusting people) is examined. The results are reported in Model 12 in Table 4.8. It shows that a trusting person has on average .505-point greater satisfaction with democracy than an untrusting person, all things being equal, when their country has the 47% of trusting people. And, if that country has 10% more trusting people, the difference between trusting and untrusting person in the level of satisfaction with democracy becomes greater, which is .549-point (= .505 + .004*10). But, the current data do not provide sufficient evidence that this increased difference is statistically different than zero.93 This interaction effect is visualized in Figure 4.8. Do the effects of country-level variables hold the same in the random slope models? Let us look at Model 11 again. The effect of Bureaucratic Profiles and economic growth are the same as in the previous random-intercept model (Model 10), although the size of their effects is 92 Similarly, future study should explore the reasons of varying effects of age and perceived election-fairness. 93 In order to investigate the interactive relationship between individual’s interpersonal trust and social trust in more detail, the improved measurement of on social trust will be needed (see ft 86). Future studies should also consider alternative explanations for varying effects of generalized trust. Perhaps, the effect of interpersonal trust depends on the existence of trust-generating institutions. For example, Jamal and Nooruddin (2010) argue that the “democratic utility of trust”—that is, whether generalized trust is linked to support for democratic principles—depends on a country’s level of democracy. The mechanism is that people will be more likely to trust each other when political institutions can secure their interest. Therefore, they argue that interpersonal trust will be associated with support for democratic principles, but only when the country enjoys higher levels of democracy. 143 slightly changed. And, PR electoral systems have no statistically significant effect on democratic satisfaction, as expected.94 7 8 Figure 4.8. Cross-level interaction effect between generalized trust and social trust on citizens’ satisfaction with democracy 6 Trusting person 4 5 Untrusting person 0 20 40 60 Percentage of trusting persons 80 100 Note: It shows the changes in the difference between trusting and untrusting person in their satisfaction with democracy across different levels of social trust (i.e., percentage of trusting persons in a nation). Two reference lines (i.e. dotted lines) indicate the range of observed values of social trust. The results are based on Model 12 in Table 4.8. 94 Note that this relationship is positive and statistically significant when Bureaucratic Profiles is not included in the model. The positive relationship may reflect that PR electoral systems are designed to focus on representativeness of voters’ preferences (Norris 2004). In turn, citizens could be more satisfied with functioning of democracy when their interests are represented in politics. Alternatively, this may be because the winner-loser gap is smaller under PR systems (Anderson and Tverdova 2003; Norris 2004). However, this relationship is not statistically significant when Bureaucratic Profiles is included. 144 Interestingly, the relationship between the level of democracy and satisfaction with democracy becomes statistically significant when random slopes are considered in the model (Model 11). For example, a person in a country with the score of 3 (most free) is more likely to be satisfied with democracy by .341 points when compared to a citizen in a country with a Freedom House score of 2, all things being equal. It seems plausible because more democratic nations will be able to preserve democratic values and, in turn, it will influence how citizens view the performance of democratic government. 10 Figure 4.9. Predictions and approximate 95% confidence intervals versus ranking of citizens’ satisfaction with democracy based on Model 11 in Table 4.8 DK AU SE 8 UK BE FI NL NO CA NZ AT CH DE ES CL IL PT KR SI MX HU PL 4 6 SK JP IE 2 Prediction CZ FR US 0 10 20 30 Rank Note: Based on a random-slope model for citizens’ satisfaction with democracy, which are shown in Model 11 in Table 4.8, the predicted values are calculated. The predictions for citizens’ satisfaction with democracy are aggregated at the country-level. These values are ranked and plotted with approximate 95% confidence intervals. Country names are shown on top of confidence intervals. See Note in Figure 3.3 for the full names of the thirty-four countries. 145 The comparison between the random-slope and the random-intercept model provides further evidence for the positive relationship between Bureaucratic Profiles and public attitudes toward bureaucracies. Figure 4.9 presents the mean of the predicted values for each country, along with 95% confidence intervals, based on Model 11 in Table 4.8. The ranking of the country’s mean predicted values is the same as those obtained from the random-intercept model, except the ranking of Mexico and Hungary. 4.9 Conclusion This chapter has explored the linkage between bureaucracies and citizens’ political support, using a new measure of Bureaucratic Profiles created from the estimations obtained in the previous chapter. Specifically, I have examined which bureaucratic profile, or what combination of bureaucratic traits, contributes to public satisfaction with democracy and attitudes toward bureaucracies, using a multilevel model. The results show that there are particular profiles that are associated with higher/ lower levels of political support, which implies that not all nations’ bureaucratic profiles perfectly represent what citizens want their bureaucracies to possess in democratic societies. I find that bureaucratic profiles generally have a similar relationship with citizens’ attitudes toward bureaucracies, and with citizens’ democratic satisfaction. In countries with higher levels of political support, impartiality are importantly emphasized but relatively little emphases are placed on competency and career-based system in their bureaucracies. As relative levels of impartiality decrease, and, at the same time, as relative levels of competency increase, levels of political support decrease. Thus, the empirical findings in this chapter suggest that a 146 bureaucracy that implements policies fairly and participates in policy making without favoritism or bias is perhaps what citizens want. In addition to this, more interestingly, the results suggest that different relationships between traits, in terms of their relative levels presented in a national bureaucracy, are associated with different levels of political support. The findings speak to the importance of balancing tradeoffs of bureaucratic traits. For example, political independence is important, but a sore emphasis on this trait does not seem to be desirable from the perspective of citizens: people view bureaucracies more positively only when impartiality is presented more than political independence. Less emphasis on career-based system in a bureaucracy leads to higher levels of political support only when impartiality is preserved: for countries with lower degrees of impartiality, it seems to have a negative effect on political support. For another example, female representativeness is another important trait influencing public satisfaction with democracy; but levels of public satisfaction with democracy decrease if female representativeness is emphasized more importantly than impartiality. 147 CHAPTER 5 CONCLUSION The aim of this dissertation has been twofold. First, it proposes an alternative way to compare and understand bureaucracies in OECD member nations by developing a new measure of Bureaucratic Profiles. I claim that bureaucratic characteristics are multi-dimensional and that not all nations have the same attributes of bureaucracy. The empirical results show that bureaucratic profiles vary across OECD member nations. Second, this study examines citizens’ attitudes about bureaucracies and democratic regime performance. The results demonstrate how various bureaucratic profiles are linked to levels of political support, and also suggest the importance of relatively high levels of impartiality in bureaucracies. This dissertation adds insights to current debates on cross-national indicators of ‘governance’ and puts forth a better measure of bureaucratic traits. In this concluding chapter, I summarize the general findings and contributions that this dissertation makes. I also suggest directions for future study. 5.1 General Findings and Contributions In this dissertation, I argue to examine a set of key bureaucratic traits simultaneously, rather than focusing on a single aspect of bureaucracies as previous studies have suggested. In prior studies, different characteristics of public agencies have emphasized to reconcile the potential tensions between bureaucracy and democracy, but the complex relationships between these traits have been overlooked. Without both establishing the criteria for judging “good” bureaucracies and understating possible tradeoffs between different key traits, conventional 148 indicators of bureaucracies used in large-N studies have failed to compare national bureaucracies in a meaningful way. Unlike previous studies, I do not assume that there is a single bureaucratic attribute (or dimension) that is normatively “good” for all nations. Instead, I choose four key bureaucratic traits, all of which have been (separately) proposed as an important bureaucratic characteristic in previous studies. I then examine how these key attributes were presented simultaneously in nations’ bureaucracies, using multidimensional preference scaling (MDPREF). That is, I focus on the “bureaucratic profile” which shows the relative levels of each trait presented in a national bureaucracy. Moreover, based on the estimations from the MDPREF analysis, Bureaucratic Profiles, a new measure of bureaucratic traits, is developed, and its linkage to popular attitudes toward bureaucracies and democratic performance is examined. By investigating bureaucratic profiles across OECD member nations, I am able (1) to examine and identify key bureaucratic traits and (2) to explore similarities and/or differences in the relative levels of these key traits presented in national bureaucracies. I find five clusters of trait variables—which I named impartiality, political independence, career-based system, competency, and female representativeness—as an underlying structure of bureaucratic traits in OECD member countries. Furthermore, and more importantly, the results show that different nations have different configurations of a set of traits represented in a bureaucracy. With these findings, this dissertation contributes to our broader and deeper understanding of bureaucracies from a comparative perspective. The results are based on a new way of looking at bureaucracies comparatively. Focusing on bureaucratic profiles, instead of a single bureaucratic trait, allows us consider that bureaucratic attributes are interrelated and also sometimes contradict to each other. That is, nations may make decisions to balance a tradeoff between different traits (Peters 2001) and the relative importance, or weight, of each trait can 149 vary across nations. 95 The various bureaucratic profiles found in this dissertation demonstrate that each nation chooses to emphasize different traits. Some nations seem to focus more on impartiality and less on career-based system, while in other nations, the order of the relative levels of traits is the opposite. For example, Anglo-Saxon countries seem to choose impartiality over career-based system, while Southern European countries put more emphasis on careerbased system than impartiality. It also appears that bureaucracies in Anglo-Australian countries share similarities with those in Nordic countries, in that they emphasize impartiality over competency. This is the opposite in Central and Eastern European countries. Moreover, this dissertation makes an important contribution to studies of comparative public administration: It provides quantitative data on variation in bureaucracies. Case studies illustrating bureaucracies across a small number of nations have suggested that there are various “traditions of public administration” (e.g., Painter and Peters 2010; Pierre 1995; Tummala 2003). 95 Related to this, it is worth mentioning some studies discussing that organizations choose between and balance the tradeoffs of different bureaucratic characteristics. Herbert A. Simon, in “The Proverbs of Administration” (1946), suggests that some “principles” that are widely accepted for improving administrative efficiency—specialization, unity of command, and span of control—are contradictory to each other, but are all necessary. He notes that, “[n]o single one of these items [can be a sufficient] guiding principle” in designing administrative organization. More importantly, he notes that “[m]utually incompatible advantages must be balanced against each other” to maximize over-all efficiency, “just as an architect weighs the advantages of additional closet space against the advantages of a larger living room” (62). Thus, he proposes exploring alternative administrative arrangements that assign different weights to these principles. Some empirical studies have shown such cases, although the organizations examined were not governmental bureaucracies. For example, Hall (1963) compares ten organizations in the United States in terms of the degree to which a series of dimensions of ‘bureaucracies’—hierarchy of authority, division of labor, system of rules, impersonality, and technical competence—were presented in organizations; he finds variations in the configuration of “dimensional magnitude” across ten organizations. Reimann (1973), in studying nineteen business organizations in the United States, finds various structural arrangements along three bureaucratic structural dimensions—decentralization, specialization, and formalization. Similar to these studies, but with a slightly different focus, I examine various configurations of a set of key traits presented in a bureaucracy, which may reflect nations’ choices to balance the advantages and disadvantages of several key traits. 150 But, with descriptive accounts, it is difficult to compare similarities and differences of bureaucratic characteristics across nations in a systematic way. The comparative bureaucracy research does not provide the quantitative data necessary to compare bureaucracies across a large number of nations (Brans 2003). And, the existing cross-national indicators of “bureaucratic quality” or “effective bureaucracy” fail to capture various configurations of the machinery of government that can produce the same, high-quality outcomes (Andrews 2008, 2010; Holmberg, Rothstein, and Nasiritousi 2009). This dissertation, especially Chapter 3, fills this gap by providing empirical evidence and quantitative data of wide variations in the configuration of bureaucratic traits. By examining how a set of traits are represented within each bureaucracy with various relative importance, this dissertation is able to systematically show the variety of bureaucratic characteristics across nations and develops a single scale to represent such information. As the second part of the analysis, Bureaucratic Profiles, a new measure of bureaucratic traits, is used to examine which bureaucratic profile is related to higher levels of political support, specifically, citizens’ attitudes about bureaucracies and democratic performance. The results from the multilevel models show that when the relative level of impartiality exhibited in a bureaucracy is greater than those of other traits, citizens are more likely to have higher levels of political support. The findings suggest that impartial behaviors of public agent in policymaking processes are what people want for their bureaucracies in democratic political systems. The results also show that citizens do not favorably view a government bureaucracy and regime performance when competency is the main concern in national bureaucracies, but bureaucratic impartiality and independence from politics are understated. It is important to note that this does not necessarily indicate that people do not want competent public agencies. Recall 151 that merit-based recruitment, which also contributes to having skillful and competent workers in public agencies, is clustered together with other indicators of impartiality in all nations. Also, given that competency in this dissertation is about the retention of skilled agents with higher wages and compensation, the findings imply that people do not want bureaucracies that enjoy privileged status in society but not care much about providing impartial services to citizens. Moreover, this finding demonstrates the dilemma that democratic government confronts: bureaucracies should be strong enough to effectively carry on public policies, but they should not be too strong as to threaten democratic values and democratic development. That is, people may want their national bureaucracies to be competent in delivering public goods and services, but this should not be something that ignores other key traits. Some interesting insights about other key traits are also demonstrated in the results. Female representativeness seems to play an important role in shaping public attitudes toward bureaucracies and regime performance only if it does not harm impartiality. In addition, emphasizing career-based system enhances political support if a nation’s bureaucracy possesses relatively lower levels of impartiality; while a relatively lower emphasis on career-based system helps if impartiality is relatively well secured. This implies that changing a bureaucracy from a career-based system to a position-based one will not always contribute to public satisfaction with government. Without preserving impartiality, this could cause a problem. Thus establishing career-based systems may work better to enhance political support, when the degree to which impartiality is exhibited in a bureaucracy is smaller than other traits. Another interesting finding is that a relatively greater emphasis on independence from politics does not seem to be a problem unless impartiality is present along with this trait. 152 These findings provide support for previous research criticizing politicization and emphasizing others means of democratic control of bureaucracies. Indeed, one dominant view on bureaucracies in political science focuses on the relationship between bureaucracies and their political masters—whether they are responsive to the elected politicians and/or how much discretion is given to bureaucracies. This line of studies has largely been influenced by the rational choice approach and the principal-agent theory since Barry M. Mitnick’s (1980: cited in Waterman and Meier 1998) and Barry Weingast’s classic (1984) study. Within this literature, political responsiveness is considered as a desired characteristic of bureaucracies in democratic societies. However, the assumption that bureaucracies are “agents” of politicians and an institution to be tightly controlled leads to a very limited understanding of diverse bureaucratic characteristics that may exist. Moreover, scholars have criticized this perspective on bureaucracies, and have suggested different characteristics, such as a representativeness (e.g., Meier and O’Toole 2006) and impartiality and professionalism (e.g., Rothstein and Teorell 2008b; Suleiman 2003). And, the findings in this dissertation suggest that impartiality and representativeness can be better tools for democratic control of bureaucracies. The emphasis on politicization could harm political insulation of bureaucracies, which, in turn, destructs citizens’ evaluations on bureaucracies and regime performance. Generally these findings corroborate Suleiman’s (2003) argument that politicization and business-emulating reforms have resulted in deprofessionalization of bureaucracies, and this will lead to “dismantle” the state with lowering trust in government. From the analyses in this dissertation, we can see that different nations make different choices about which traits to emphasize and that a set of bureaucratic traits have different relationship with public support for bureaucracies and regime performance. With these 153 discussions, this dissertation makes an important addition to a recent lively debate on measuring “the quality of government”—sometimes termed as “the quality of state” or “governance.” There have been increasing efforts to understand and measure ‘output institutions’ in a more precise and substantive way (Fukuyama 2013; Rothstein and Teorell 2008b). Although some scholars focus on a broader concept such as governance or state capacity, the bureaucracy is at the core of these discussions. One very recent scholarly debate has been initiated by Francis Fukuyama’s (2013) commentary, “What is Governance,” in Governance (See http://governancejournal.net/ category/commentary/ for scholars’ exchange of opinions on this commentary). Although the title implies that it is about governance, he focuses on how to conceptualize and measure the state apparatus, where he suggests autonomy and capacity as key components of the quality of government bureaucracies. This is not a new debate; there have been some scholarly attempts to develop a better indicator of bureaucracy. For example, the QoG Institute produces the Expert Survey data examining structural dimensions of bureaucracies and focusing on impartiality as a norm for “quality of government.” Two dimensions of bureaucratic structure are suggested, that is, professionalism and closedness (Dahlstrӧm, Lapuente, and Teorell 2010). Also, the OECD Governance at a Glance provides statistical data on various aspects of government performance in OECD member nations, including government revenues, government expenditures, public employment, human resource management practices, budget practices and procedures, regulatory management, and integrity (OECD 2009, 2011). For the 27 EU member nations, Demmke and Moilanen’s (2010), which is updated from the study of Auer et al. (1996), provide detailed data on specific features of civil service systems, based on the “Traditional – Post-bureaucracy 154 continuum model,” which consists of legal status, career-structure, recruitment, salary system, and tenure system. This dissertation makes a contribution to this debate and the efforts to measure bureaucracies, by utilizing, but also going beyond, the data provided by the QoG Institute and the OECD Governance at a Glance. Consistent with other studies (e.g., Dahlstrӧm, Lapuente, and Teorell 2010; Fukuyama 2013), I show that bureaucratic characteristics are better understood as multi-dimensional concepts—that is, it is not only about the continuum of “bureaucraticness,” but about capturing several important attributes. Moreover, this dissertation is different from the studies of “good governance” and their measures, in that Bureaucratic Profiles focus specifically on governmental bureaucracies and it does not assume that one profile is “better” than another in priori. Instead of normatively defining “goodness” of the state administrative apparatus, on which these debates focus, I examine bureaucratic characteristics that are associated with citizens’ positive attitudes toward public agencies. Finally, this dissertation contributes to the study of political support, by explaining crossnational differences in levels of public support for bureaucracies and regime performance. It broadens our understandings of the role of bureaucracies in democratic societies, especially from the perspective of citizens. In the literature, cross-national variation in levels of political support has not been fully explained, although economic, political, and social factors have been suggested as key determinants. I argue that, particularly in Chapter 4, bureaucratic characteristics in a nation can influence people’s attitudes toward government institutions and performance. This is because a bureaucracy is an important institution that participates in the policy making process and delivers public goods and services to the public (Rohrschneider 2005). Overall, the findings suggest that the relative level of impartiality compared to other key bureaucratic traits 155 represented in a national bureaucracy matters with regards to citizen’s attitudes about governments. With Bureaucratic Profiles, a new measure of bureaucratic traits, I am able to show that variation in bureaucratic characteristics explains levels of political support: various configurations of a set of key bureaucratic traits have different effects on how people view their government bureaucracies and their democratic system. In addition to the relationship between bureaucracies and political support, I also find an interesting association between interpersonal trust and political support. The findings suggest that there are both individual- and country-level effects of generalized trust and also the possibility of an interaction effect between them. That is, in a high-social trust society, individuals are more likely to have higher levels of political support and the difference between a trusting and an untrusting person is also amplified in this setting. Although the measurement of social trust needs to be improved for further examination of this relationship, the findings provide interesting insights for future studies of political support. 5.2 Extensions and Implications for Future Studies This dissertation provides a new measure, Bureaucratic Profiles, and shows a variety of bureaucracies within OECD nations. Future studies will need to examine the determinants, or the correlates, of variations in bureaucratic profiles. For example, why do some nations choose to emphasize impartiality over competency, while others do the opposite? Why is political independence emphasized along with impartiality in some bureaucracies but not in others? When are female representativeness and career-based system presented in a way that they do not decrease levels of impartiality in a bureaucracy? Such questions can be explored. Various 156 administrative traditions (Painter and Peters 2010), special training within bureaucracies, and demand from the public can be considered as potential explanations. Moreover, the configuration of political institutions and veto points (Knott and Miller 2006; Tsebelis 2002), the competition of political parties (Geddes 1991), and the sequence of state and party building relative to the development of a bureaucracy (O’Dwyer 2004; Shefter 1977) may also be explored as determinants of bureaucratic profiles. This project examines popular attitudes toward government as a possible consequence of various bureaucratic profiles. Studies can be extended to examine different outcomes. For example, one may look at which profiles are related to lower levels of corruption. Some studies argue that emphasizing merit-based recruitment systems (Dahlstrӧm, Lapuente, and Teorell 2012; Rauch and Evans 2000) and impartiality (Rothstein and Teorell 2008b, Teorell 2009) are important in curbing corruption. But, other studies have showed that increasing civil-service pay can also be helpful (Van Rijckeghem and Weder 2001). If this is true, do countries that focus more on increasing competency of bureaucracies by paying higher wages and compensation have lower levels of corruption? Using Bureaucratic Profiles, one may probe which pattern of the relative levels of a set of traits—especially between impartiality and competency—contributes to lower levels of corruption. Studying the relationship between Bureaucratic Profiles and the development of different policies will be another interesting extension. One of the interesting findings was the correspondence between variation in welfare-state regimes and variation in bureaucratic profiles. Specifically, I find that Social-democratic welfare states share a similar bureaucratic profile, as do Liberal and Conservative welfare states. In the comparative welfare state literature, as mentioned in Chapter 1, the “state-centric” approach (Huber and Stephens 2001, 15) focuses on 157 the capacity of the state apparatus to explain the development of welfare states. It suggests that the existence of autonomous bureaucracies have contributed to introductions and expansions of social policies that benefit the broader public. More recently, Dahlstrӧm, Lindvall, and Rothstein (2012) argue that bureaucratic capacity, which is defined as the “competence and reliability” of national bureaucracies, is related to the development of policies that entail a relatively large amount of discretion in their implementation. They show that bureaucratic capacity has a positive relationship with spending on active labor market policies (which require discretion), but not with spending on parental leave benefits (which need less discretion in implementation) in advanced democracies. The mechanism of this relationship is that people will approve the expansion of spending on policies involving bureaucratic discretion only if their bureaucracies have the capacity to implement programs in a reliable manner.96 These studies pose an interesting question, which is whether different bureaucratic profiles are connected to nations’ differential focus on various policies. Given these studies, I would argue that a bureaucracy that puts a higher priority on preserving political independence and career-based recruitment systems will play an important role in such a policymaking process. Based on studies of government spending priorities, 97 futures studies will investigate whether a greater emphasis on the political independence of bureaucracies is associated with government’s 96 Moreover, studies have shown that policy adoptions and expansions are influenced by bureaucratic institutions (Schneider and Ingraham 1984; Schneider and Jacoby 1996; Schneider, Jacoby, and Coggburn 1997). A different level of emphasis on female representativeness may also influence policy implementation (Wilkins and Keiser 2006) and government spending priorities (Dolan 2002). 97 For example, “government spending priorities” between policies targeting subsets of the population and those providing broader collective goods can effectively capture government’s differential emphasis on policies (Jacoby and Schneider 2001, 2009). The spending priority scores developed for OECD member nations by Pamphilis (2012) will provide an opportunity to explore the relationship between various bureaucratic profiles and different spending patterns. 158 priority on policies for the general public compared to policies for targeted groups within the OECD member nations. In addition, future studies examining bureaucratic profiles in different sets of nations, especially including developing countries will broaden our understanding of government institutions. It would be an interesting first step to look at whether the underlying structure of bureaucratic traits in developing countries is the same as what I have found with OECD member nations. For example, Daslstrӧm et al. (2011) report that one of the dimensions of bureaucracy, “closedness,” is not applicable to developing countries. If this is because bureaucratic dimensions in these countries are indeed different from those in Western Europe, then it suggests that the configuration of bureaucratic traits is different in the rest of the globe. Another venue for extending this study of bureaucratic profiles is to measure and compare them across different levels of government (e.g., bureaucracy in the central government versus one in the local government) and across various departments (e.g., bureaucracies mainly in charge of regulations versus ones deliver services). For example, previous studies suggest that the configuration of relative levels of bureaucratic traits can be different in law-enforcing agencies from that in service-delivery agencies. Gingerich (2012), in studying Bolivia, Brazil, and Chile, argues for disaggregating bureaucracies to measure bureaucratic capacity because the elements of capacity have a greater cross-agency variation within a country. Thus, it would be a fruitful addition to compare bureaucratic profiles of different agencies within a nation, and also compare them with an overall profile that this dissertation has presented. 159 5.3 Conclusion Bureaucracy is an important and complicated institution that influences citizens’ life. Bureaucracy bashings without understanding key bureaucratic traits and their complex relationships, as many political elites and media have done, are not fruitful ways to improve bureaucracies to do what it is supposed to do in democratic societies. Instead, in order to enhance bureaucratic capabilities in preparing and delivering public policies—with necessary discretion but not threatening democratic values—it is important to acknowledge that there is no single, best model for bureaucracy. There are different aspects of bureaucracies that are important for democratic governing. Moreover, increases or decreases in one specific bureaucratic trait do not always lead to the same change in another key trait. As the findings in this dissertation suggest, different nations seem to make different choices to balance between various bureaucratic traits. Different configurations of bureaucratic traits may lead to different policy developments and different capacity in public service deliveries. I believe that this dissertation takes an important step to understand the consequences of the variety of bureaucratic profiles by focusing on the perspective of citizens. The analysis has shown that there is a particular pattern of bureaucratic profile that is preferred by citizens. It points to the importance of impartiality, which is in line with the findings of several other studies. As an output institution, impartiality and fairness may be the most important trait for the bureaucracy in democratic societies. However, more importantly, what this dissertation has further revealed is that it is not only impartiality, but also its relative importance and its relationships with other key bureaucratic traits that influence people’s views about democratic government. I hope future researches can provide further 160 understandings of the causes and consequences of various bureaucratic profiles. This will broaden our understanding of bureaucracies and its important role in democratic societies. 161 APPENDICES 162 Appendix A. Description of Variables Used in Chapter 3 This Appendix describes the variables used in Chapter 3. It would help to understand the observations used, although the data used for the multidimensional preference scaling analysis is the combination of these observations, as discussed in Section 3.3.2. Here, I discuss the values of the original observations—before being imputed for missing values and recoded. These variables are all standardized before being used for the analysis. Independence from politics In order to measure political independence, I include three variables from the QoG Expert survey. All three variables’ response ranges from 1 (Not at all) to 7 (To a very large extent). The responses are aggregated for each nation and I use the aggregated version of the survey data provided by the QoG institute (i.e., the QoG Country-Level Survey Data). Note that this applies to all variables collected from the QoG Expert survey. One variable (no_fulfill_ideology_govt) is obtained from the question: Q8. To what extent would you say the following applies today to the country you have chosen to submit your answers for? e. Public sector employees strive to fulfill the ideology of the party/parties in government? The mean score for each country ranges from 2.88 to 5, with an average value of 3.92 and a standard deviation of .59 (where Korea and Poland are around this mean value). That is, bureaucrats are the least likely to behave in a partisan manner (which may indicate the highest level of independence from politics) in Japan (2.88), while a bureaucracy is the most likely to be influenced by partisanship (possibly implying the least degree of independence from politics) in Estonia (5). I assume that fulfilling the political ideology of incumbent parties indicate a lower degree of political independence. Thus, this variable is recoded; a nation where political independence is presented to a greater degree is scored high on this variable. Two other variables (senior_career and senior_no_political) are about the recruitment of senior officials. They are measured using the following questions, respectively: Q2. Thinking about the country you have chosen, how often would you say the following occurs today: e. Senior public officials are recruited from within the ranks of the public sector? d. The top political leadership hires and fires senior public officials? For the first variable (senior_career), the mean score for each country ranges from 4.2 to 6.67 and the average value is 5.25 (with a standard deviation of .68), where Austria, Switzerland, and Turkey are close to this mean value. It shows that senior officials are usually recruited based on career in France (6.67), whereas it is the least likely in Poland (4.2). I assume that as senior officials are recruited based on career and from within the ranks, a bureaucracy is more likely to 163 be independent from political pressures. Thus, a higher score on this variable indicates a greater degree of political independence exhibited in that nation. Next, the country’s mean score of the second variable (senior_no_political) has an average value of 4.34, with a standard deviation of 1.25, where Israel and Korea have values close to the mean. It ranges from 1.56 to 6.14, which indicates that recruitment of senior officials is mostly dependent on the top political leadership in Mexico (6.14), whereas it is the least likely to happen in Ireland (1.56). I assume that as recruitment of senior officials is dependent on the top political leadership, a bureaucracy is less likely to be independent from political influence. Thus, this variable is recoded so that higher values correspond to a greater degree of political independence. Female Representativeness To capture female representativeness, four variables are used. A nation achieving greater levels of female representativeness receives a higher score on these variables. Three variables are from the OECD Government at a Glance data and one variable is from the QoG Expert survey. One variable from the QoG Expert survey (women_represnt) is from the following question: Q8. To what extent would you say the following applies today to the country you have chosen to submit your answers for? i. Women are proportionally represented among public sector employees? The response ranges from 1 (Not at all) to 7 (To a very large extent). The mean score for each country ranges from 2.07 to 5.89. That is, female representativeness among public sector workers is the highest in Estonia (5.89), whereas it is the lowest in Mexico (2.07). The average value is 4.47 (and a standard deviation is .87), where Switzerland and Poland are around this mean value. Three other variables (female_central, female_senior and female_administration) are from the OECD Government at a Glance data. The data provide the proportion of female workers in “the core civil service in central government” in three different positions (OECD 2009b). The first one is the percentage of females among central government employees. It ranges from 12% (in Turkey) to 69% (in Poland) and the average value is 46.18, with a standard deviation of 14.47 (where Norway is close to the mean value). The figures are different for different positions. The percentage of females in senior positions in central government ranges from 2% (in Japan) to 38% (in Greece). Female representativeness in the senior positions is generally lower, compared to that in the central government. The average value is 23.68, with a standard deviation of 11.07, which is close to the percentage observed for Norway and Finland. Again, when it comes to the percentage of women in administrative positions, it ranges from 23% (in Germany) to 83% (in Portugal and in Norway) and its average value is 54.47%, with a standard deviation of 18.76 (where the percentage in Belgium is near this value). 164 Impartiality For measuring impartiality, five variables from the QoG Expert Survey are used. As stated in Section 3.3.1, the five items are developed by Rothstein and Teorell (2008a) and Teorell (2009). The response to the questions, except for the second one, ranges from 1 (Hardly ever) to 7 (Almost always); the response to the second question ranges from 0 to 100 (percentage). The first variable (impartial) is the response to the following question: Q4. By a common definition, impartiality implies that when implementing policies, public sector employees should not take anything about the citizen/case into consideration that is not stipulated in the policy. Generally speaking, how often would you say that public sector employees today, in your chosen country, act impartially when deciding how to implement a policy in an individual case? This question is designed to ask respondents to rate their bureaucracy in terms of the theoretical definition of impartiality (Rothstein and Teorell 2008a). The mean score for each country ranges from 3.5 to 6.4, and its average value is 5.13, with a standard deviation of .80. Bureaucracies in France, Germany, Luxembourg and Sweden receive a score close to the mean value. According to this question, Australia (6.4) has the most impartial bureaucracy, whereas level of impartiality is the lowest in Slovakia (3.5). The second variable (distribute_needy_to_needy) measures a percentage reaching “the needy poor” in the following question: Q6. Hypothetically, let’s say that a typical public employee was given the task to distribute an amount equivalent to 1000 USD per capita to the needy poor in your country. According to your judgment, please state the percentage that would reach: (Six response categories for which the respondents could fill in a number from 0 to 100 percent.) Rothstein and Teorell (2008a) design this question assuming that the percentage of money actually distributed to the targeted population, the needy poor, will reflect impartial behaviors of bureaucrats. The other categories include “People with kinship ties to the public employee,” “Middlemen/consultants,” “The public employee’s own pocket,” “The superiors of the public employee” and “Others.” The percentage reaching “the needy poor” ranges from 30% (in Luxembourg) to 92.44% (in Norway). The average value is 68.82, with a standard deviation of 16.34, where the percentages in Estonia and Germany are close to this average value. The next three variables (not_favor_procurement, not_favor_implementation and not_favor_licenses) are measured with the following questions, respectively: Q2.Thinking about the country you have chosen, how often would you say the following occurs today? g. Firms that provide the most favorable kickbacks to senior officials are awarded public procurement contracts in favor of firms making the lowest bid? 165 h. When deciding how to implement policies in individual cases, public sector employees treat some groups in society unfairly? i. When granting licenses to start up private firms, public sector employees favor applicants with which they have strong personal contacts? The mean score of the first variable (not_favor_procurement) ranges from 1 to 5, with an average value of 3.00 and a standard deviation of 1.25 (where Chile, France, and Korea are rated closer to the mean value). It shows that the public procurement contracts are conducted the most fairly and impartially in New Zealand (1), while it is done the least fairly in countries like Greece, Luxembourg, and Mexico (5). The mean score for the second variable (not_favor_implementation) ranges from 1.6 to 5 and has an average value of 3.10 (and a standard deviation is .87), where Poland is rated closer to the mean value). This tells us that the extent to which people are treated fairly and impartially in implementation of policies is the highest in Switzerland (1.6), whereas it is the lowest in Luxembourg (5). The country’s mean score on the final variable (not_favor_licenses) ranges from 1 to 5.54, and its average value is 2.97, with a standard deviation of 1.18 (where Finland and the United States are rated close to the average value). It indicates that the degree of impartiality in issuing licenses for business start-ups is the highest in Iceland (1), while it is the lowest in Mexico (5.54). In order for higher values of each variable to correspond to a greater degree of impartiality, these three variables are recoded. Mechanisms for impartiality and political independence In addition to the variables for impartiality and independence from politics, I include five additional variables to capture the idea of a merit-based system and a career-based system that are discussed as the mechanisms for political independence and impartiality. The response to the first four questions ranges from 1 (Hardly ever) to 7 (Almost always); the fifth variable ranges from 0 to 1. The first four variables (recruit_merit, recruit_no_political, recruit_exam and tenure) are from the response to the following questions in the QoG Expert survey data, respectively: Q2. Thinking about the country you have chosen, how often would you say the following occurs today: a. When recruiting public sector employees, the skills and merits of the applicants decide who gets the job? b. When recruiting public sector employees, the political connections of the applicants decide who gets the job? c. Public sector employees are hired via a formal examination system? f. Once one is recruited as a public sector employee, one stays a public sector employee for the rest of one’s career? The first variable’s (recruit_merit) mean score for a country ranges from 3 to 6.5, and its average value is 5.07, with a standard deviation of 1.01 (where Israel is rated at this mean value). It 166 indicates that the recruitment of bureaucrats in New Zealand (6.5) is based on merit system most of time, whereas it is not always the case in Luxembourg or in Mexico (3). The country’s mean score of the second variable (recruit_no_political) ranges from 1.67 to 6, with an average value of 3.57 and a standard deviation of 1.24 (where France and Estonia are rated at this mean value). The recruitment of bureaucrats in Luxembourg (6) most of time relies on the political connections of the applicants, whereas recruitment is the least likely to be influenced by political connection in Japan (1.67). This variable is recoded to make its higher values indicate that recruitment in a nation’s bureaucracy is not likely to be based on political connection. The third variable’s (recruit_exam) mean score has an average value of 4.65, with a standard deviation of 1.35, where the nations rated around this mean value is the United States. It ranges from 1.75 (in New Zealand) to 7 (in Luxembourg), which indicates that a formal examination is usually conducted to recruit bureaucrats in Luxembourg, whereas this is not always the case in New Zealand. The country’s mean score of the fourth variable (tenure) ranges from 2.7 (in Estonia) to 6.48 (in Greece), with its average value is 4.91 and a standard deviation of .90 (where Sweden and Slovenia are the nations rated around this mean value). It shows that bureaucrats in Greece enjoye a guaranteed tenure most of time, whereas they are least likely to do so in Estonia. The fifth variable (career_based_hrm) is obtained from the OECD Government at a Glance 2009 data. This variable measures the degree to which each nation has a career- versus position-based system. It is developed based on policies for becoming a civil servant in general and for recruiting senior civil servants, and systems for appointing entry-level positions, and for allocating posts across departments (OECD 2009b). The original indicator ranges from 0 (careerbased) to 1 (position-based). The Netherlands’s system is the closest to an ideal position-based system (.78), while the recruitment system in France is the closest to an ideal career-based one (.05). The German system is around the average of the countries in consideration. And this variable is recoded: Smaller value represents a system open to external recruitment and larger values indicate a system where most of bureaucrats are recruited at lower levels and move upward the rest of their career. That is, after being recoded, countries with a career-based system are scored high on this variable. Competency For competency, four related variables are used. One variable measures bureaucrats with a higher level of education; the next two variables are about the wages and compensation of bureaucrats; and the fourth variable is the size of the government workforce. The first variable (higher_edu_ratio) measures the ratio between the percentage of government workers with higher education and the percentage of everyone in the country with higher education, as an average between 2000 and 2004. Using the ISSP survey data and background variables, I first create a variable, the percentage of government employees with a higher education—bachelor’s degree or more. Based on the question about type of work place, 167 respondents who answer that they “work for government or public sector” are considered government employees. Using the same criteria of higher education, the percentage of general public with a higher education is also calculated. After that, the ratio between the two is calculated for the five years from 2000 to 2004; then, the average ratio though 2000 to 2004 is used for this analysis. This variable ranges from .93 (in Denmark) to 3.45 (in Chile). This indicates that in Chile the percentage of people attaining a bachelor’s degree or more is much greater in the government than in the general public. On the other hand, in Demark the percentage of people attaining a bachelor’s degree or more is similar in the government and in the public (precisely, a little smaller in the government than in the public). The average value is 2.08 (with a standard deviation of .66) and Czech Republic has the value closest to this. The next two variables (compensation_to_total and wage_to_total) are obtained from the OECD’s STAN database (Structural Analysis Database in OECD.stat), as an average between 2000 and 2004. The variables are measured based on the System of National Accounts (SNA), which is “the international statistical standard for the national accounts, adopted by the United Nations Statistical Commission” (United Nations 2009). I use two variables, “LABR” and “WAGE,” provided by the STAN database, which are the compensation of employees and the wages and salaries of employees paid by producers, respectively. Compensation of employees (LABR) is defined as “the total remuneration, in cash or in kind, payable by an enterprise to an employee in return for work done by the latter during the accounting period.” (United Nations 2009, 131). And wages and salaries (WAGE) are the remuneration payable by employers at regular intervals in cash or in kind, excluding social insurance benefits paid by employers (United Nations 2009, 140). They are measured for the total economy which can be grouped into five mutually exclusive institutional units, including non-financial corporations, financial corporations, government units, non-profit institutions serving households, and households (United Nations 2009). Thus, I calculate the compensation of employees, and wages of employees, in the government units as a percentage of that in the total economy. The government units are defined as “unique kinds of legal entities established by political process that have legislative, judicial or executive authority over the other institutional units within a given area” (United Nations 2009, 62). And, based on the Untied Nation’s classifications registry of ISIC Rev. 3, government units are operationalized as the “Division 75: public administration and defence; compulsory social security” (UNSD). In the data, the compensation variable (compensation_to_total) ranges from 6.38% (in Sweden) to 16.98% (in Greece), with an average of 9.40% (where Slovakia and Estonia are around this value) and a standard deviation of 2.21. It shows that, for example, in Sweden, out of the money spent for compensation of employees in the total economy, 6.38% of it is consumed by government units. Higher values indicate that in this country, compared to other countries, government units constitute greater proportions in the final consumption of income in the total economy, specifically in terms of compensation of employees. The wage of employees variable 168 (wage_to_tatal) ranges from 6.17% (in Sweden) to 15.8% (in Greece). It indicates that, for example, in Greece, out of the money paid for wages and salaries in the total economy, 15.8% of it is spent in government units. Higher values indicate the proportion of the government units to the total economy in terms of wages and salaries paid is greater in this nation than other nations. The average value is 9.13% (with a standard deviation of 2.02) and Slovakia (9.10%) is around this value. The fourth variable (size_govt_to_total_ilo) is obtained from the ILO database on labour statistics, LABORSTAT. This is a ratio of the general government sector employment to the total employment. In this data, the general government sector employment refers to “all of the government units, social security funds and nonprofit, non-market public or private institutions which are controlled and mainly financed by public authority” (Hammouya 1999, 3). It ranges from 3.36% (in Korea) to 9.82% (in Belgium). It indicates that the proportion of government workers in the total employment is the smallest in Korea and the largest in Belgium. The average value is 6.05% (with a standard deviation of 1.50) and Estonia and Denmark (5.93%) are around this value. 169 Appendix B. Supplementary Analysis for Chapter 3: Cluster Analysis Figure B.1. Tree diagram for the twenty-one variable points females in senior positions females in central govt females in administrative positions women's representation career-based senior officials not strive to fulfill govt ideology non-political senior officials impartial in distribution impartial in granting licenses overall impartiality impartial in implementation non-political recruitment impartial in procurement merit based recruitment tenure career-based HRM formal entrance exam compensation higher-education attained size of government wage 0 .1 .2 .3 Linakge Distance .4 .5 Note: The dendrogram from the cluster analysis of twenty-one variable points shows the possible five groupings. The average-link clustering is used. 170 Appendix C. Supplementary Analysis for Chapter 4 Table C.1. Descriptive Statistics Variable Attitudes toward Bureaucracy Satisfaction with Democracy Bureaucratic Profiles Generalized trust Age (0: 47-year-old) Female Education (0: Higher secondary complete) Unemployed Public sector worker Political interest GDP/capita growth rate Democracy (Freedom House) Electoral system: PR Mean 0.00 6.01 -.13 .47 .28 .53 -.29 .05 .21 -.04 2.80 2.90 .83 Std. Deviation .83 2.25 1.57 .50 17.20 .50 1.49 .22 .41 .86 1.25 .30 .38 Minimum -1.96 0.00 -2.87 0.00 -32.00 0.00 -3.00 0.00 0.00 -1.50 .97 1.50 0.00 (3) (4) Maximum 2.09 10.00 3.09 1.00 51.00 1.00 2.00 1.00 1.00 1.50 6.56 3.00 1.00 Table C.2. Correlation Matrix Variables (1) (2) (1) Bureaucratic Profiles 1.000 (2) Country-level generalized trust .394** 1.000 * (3) GDP/capita growth rate -.323 -.257 (4) Democracy (Freedom House) .232 .124 (5) Electoral system: PR -.389** .214 ** * Note: significant at .05 level, significant at .10 level 171 1.000 -.171 .120 1.000 -.162 (5) 1.000 Table C.3. Multilevel Model of Citizens’ Attitudes toward Bureaucracies: Within- versus Between-country Effects Bureaucratic Profiles Generalized Trust Age Age2/100 Female Education Unemployed Public Sector Worker Political Interest Within- versus Between-country Effects .074 (.041)* Within Effect .263 (.010)* -.001 (.000)* .009 (.002)* -.021 (.010)* .013 (.004)* -.070 (.024)* .189 (.012)* .066 (.006)* Between Effect .009 ( .004)* .006 ( .016) -.359 -.039 -.000 -.134 .076 (1.333) ( .115) ( .032) ( .553) ( .263) -.174 (.059)* Constant Variance Components .048 Country-level intercept .555 Individual-level Residuals ** Proportion Reduction in Error: 12.93% at Level 1 55.66% at Level 2 57354.826 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 25409. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. 172 Table C.4. Multilevel Models of Citizens’ Attitudes toward Bureaucracies Control: Economic Growth Control: Democracy Control: PR system Country Level .064 (.025)* .067 (.025)* .064 (.030)* Bureaucratic Profiles -.046 (.030) GDP/capita growth rate Democracy .168 (.121) (Freedom House) Electoral system: -.073 (.118) Proportional Country-level .009 (.003)* .009 (.003)* .010 (.003)* generalized trust Individual Level .263 (.010)* .263 (.010)* .263 (.010)* Generalized Trust -.001 (.000)* -.001 (.000)* -.001 (.000)* Age 2 * * .009 (.002) .009 (.002) .009 (.002)* Age /100 -.021 (.010)* -.021 (.010)* -.021 (.010)* Female * * .013 (.004) .013 (.004) .013 (.004)* Education -.070 (.024)* -.070 (.024)* -.070 (.024)* Unemployed * * .189 (.012) .189 (.012) .189 (.012)* Public Sector Worker .066 (.006)* .066 (.006)* .066 (.006)* Political Interest * * -.158 (.037) -.161 (.037) -.103 (.102) Constant Variance Components .036 .036 .039 Country-level intercept Individual-level .555 .555 .555 Residuals Proportion Reduction in Error:** 14.77% 14.70% 14.35% at Level 1 66.88% 67.33% 64.71% at Level 2 57344.734 57342.28 57343.882 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 25409. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. 173 Table C.5. Multilevel Model of Citizens’ Satisfaction with Democracy: Within- versus Between-country Effects Bureaucratic Profiles Generalized Trust Age Age2/100 Female Education Unemployed Public Sector Worker Political Interest Perceived Election Fairness Within- versus Between-country Effects .244 (.066)* Within Effect .522 (.027)* -.000 (.001) .010 (.005)* -.115 (.025)* .095 (.010)* -.207 (.063)* .018 (.032) .155 (.016)* .502 (.013)* Between Effect .023 ( .006)* -.005 ( .026) -.241 -.118 -.058 .144 -.500 .037 (2.153) ( .183) ( .052) ( .881) ( .407) ( .314) 5.574 (.223)* Constant Variance Components .120 Country-level intercept 3.783 Individual-level Residuals Proportion Reduction in Error:** 21.96% at Level 1 85.09% at Level 2 102341.65 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 24513. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. 174 Table C.6. Multilevel Models of Citizens’ Satisfaction with Democracy Control: Economic Growth Control: Democracy Control: PR system Country Level .242 (.036)* .262 (.042)* .283 (.051)* Bureaucratic Profiles -.153 (.043)* GDP/capita growth rate Democracy .285 (.203) (Freedom House) Electoral system: .072 (.199) Proportional Country-level .021 (.004)* .023 (.004)* .022 (.005)* generalized trust Individual Level .523 (.027)* .523 (.027)* .523 (.027)* Generalized Trust -.000 (.001) -.000 (.001) -.000 (.001) Age 2 * * .010 (.005) .010 (.005) .010 (.005)* Age /100 -.115 (.025)* -.116 (.025)* -.116 (.025)* Female * * .095 (.010) .095 (.010) .095 (.010)* Education -.206 (.063)* -.208 (.063)* -.207 (.063)* Unemployed .017 (.032) .017 (.032) .017 (.032) Public Sector Worker * * .152 (.016) .154 (.016) .153 (.016)* Political Interest Perceived Election .501 (.013)* .502 (.013)* .502 (.013)* Fairness 5.559 (.057)* 5.548 (.065)* 5.553 (.061)* Constant Variance Components .068 .099 .106 Country-level intercept Individual-level 3.783 3.783 3.783 Residuals Proportion Reduction in Error:** 22.99% 22.39% 22.23% at Level 1 91.48% 87.73% 86.76% at Level 2 102330.89 102336.42 102338.31 -2* Log-Likelihood Note: The restricted maximum likelihood (REML) estimator is used. Standard errors are in parentheses. Number of observations = 24513. Number of countries = 28. * Coefficient is statistically significant at .05 level, directional hypothesis test. 175 Appendix D. Residual Diagnostics for Chapter 4 Inspections of residuals are conducted to check whether level-1 and level-2 residuals are normally distributed (Rabe-Hesketh and Skrondal 2008; Snijders and Bosker 2011). If the model is true, estimated residuals at both levels have normal sampling distributions. Overall, it seems residuals at both levels do not violate normality assumptions. I first estimate residuals for the citizens’ attitudes toward bureaucracies model, the random-intercept model with country-level controls (Model 4).98 Figure 1 shows the histograms of standardized level-1 and level-2 residuals, respectively. The normality assumption for the residuals at both levels does not seem to be violated severely. 0 0 .1 .2 .2 .4 .3 .6 .4 Figure D.1. Histograms of standardized level-1residuals and standardized level-2 residuals, for the random-intercept model of citizens’ attitudes toward bureaucracies (Model 4) -4 -2 0 Standardized level-1 residuals 2 4 -2 -1 0 Standardized level-2 residuals 1 0 .1 .2 .3 .4 Figure D.2. Histogram of standardized level-1residuals, for the random-slope model of citizens’ attitudes toward bureaucracies (Model 5) -4 -2 0 2 Predicted level-1 standardized residuals 4 98 To estimate residuals at both levels, gllapred command is used after estimating models using gllamm command in STATA SE10.1. 176 In addition, level-1 residuals are estimated from the random-slope model (Model 5) of citizens’ attitudes toward bureaucracies.99 Figure 2 shows its histogram. In the random-slope models, it is assumed that level-2 random coefficients are normally distributed. Figure 3 shows the histograms for random slopes of interpersonal trust, age, public sector workers, and political interest. 0 0 2 50 4 100 6 150 Figure D.3. Histograms of random slopes, for the random-slope model of citizens’ attitudes toward bureaucracies (Model 5) -.1 0 .1 Predicted random slope: Interpersonal Trust .2 -.005 0 .005 Predicted random slope: Age .01 0 0 2 4 5 6 8 10 10 -.2 -.1 -.05 0 .05 Predicted random slope: Public Sector Worker .1 -.1 -.05 0 .05 Predicted random slope: Political Interest .1 Next, residuals at both levels are estimated for the citizens’ satisfaction with democracy model, the random-intercept model with country-level controls (Model 10). Figure 4 shows the histograms of standardized level-1 and level-2 residuals, respectively. Both histograms look normal and the normality assumption for the residuals at both levels does not seem to be violated severely. 99 The level-2 residuals can only be estimated when the model is estimated with gllamm, and not with xtmixed. But, the random slope model with four random slopes (Model 4) was not estimated using gllamm. Therefore, I report the histogram of level-1 residuals for the random slope model here. 177 .6 0 0 .1 .2 .2 .3 .4 .4 .5 Figure D.4. Histograms of standardized level-1residuals and standardized level-2 residuals, for the random-intercept model of citizens’ satisfaction with democracy (Model 10) -4 -2 0 Standardized level-1 residuals 2 4 -2 -1 0 Standardized level-2 residuals 1 2 In addition, level-1 residuals are estimated from the random-slope model (Model 11) of citizens’ attitudes toward bureaucracies. Figure 5 shows its histogram. In the random-slope models, it is assumed that level-2 random coefficients are normally distributed. Figure 6 shows the histograms for random slopes of interpersonal trust, age and perceived fairness of elections. 0 .1 .2 .3 .4 .5 Figure D.5. Histogram of standardized level-1residuals, for the random-slope model of citizens’ satisfaction with democracy (Model 11) -4 -2 0 2 Predicted level-1 standardized residuals 4 Finally, diagnostics for influential cases are conducted. To estimate diagnostics for influential cases, MLT package (Möehring and Schmidt 2013) is used, after estimating models using xtmixed command in STATA 12. Overall, the relationship between Bureaucratic Profiles and political support (in both public attitudes toward bureaucracies and satisfaction with democracy) still holds in the model where potential influential cases are excluded from the data. For a further robustness check, I examine the models only with the twenty original member nations of the OECD; the relationship holds. 178 0 0 1 20 2 40 3 60 4 5 80 Figure D.6. Histograms of random slopes, for the random-slope model of citizens’ satisfaction with democracy (Model 11) -.1 0 Predicted random slope: Interpersonal Trust .1 -.02 0 1 2 3 -.2 -.3 -.2 -.1 0 .1 Predicted random slope: Fairness of Elections .2 179 -.01 0 Predicted random slope: Age .01 .02 REFERENCES 180 REFERENCES Aberbach, Joel D., and Bert A. Rockman. 2000. In the Web of Politics: Three Decades of the U.S. Federal Executive. Washington, DC: Brookings Institution Press. Aberbach, Joel D., Robert D. Putnam, and Bert A. Rockman. 1981. Bureaucrats and Politicians in Western Democracies. Cambridge: Harvard University Press. Almond, Gabriel, and Sidney Verba. 1963. The Civic Culture. Newbury Park, Calif.: Sage Publications. Anderson, Christopher J., and Matthew M. Singer. 2008. “The Sensitive Left and the Impervious Right Multilevel Models and the Politics of Inequality, Ideology, and Legitimacy in Europe.” Comparative Political Studies 41(4–5): 564–599. Anderson, Christopher, and Christine Guillory. 1997. “Political Institutions and Satisfaction with Democracy: A Cross-national Analysis of Consensus and Majoritarian Systems,” American Political Science Review 91(1): 66–81. Anderson, Christopher, and Yuliya Tverdova. 2003. “Corruption, Political Allegiances, and Attitudes toward Government in Contemporary Democracies.” American Journal of Political Science 47(1): 91–109. Andrain, Charles F., and James T. Smith. 2006. Political democracy, trust, and social justice: a comparative overview. Boston: Northeastern University Press. Andrews, Matthew. 2010. “Good Government Means Different Things in Different Countries.” Governance: An International Journal of Policy, Administration, and Institutions 23(1): 7–35. Andrews, Matthew. 2008. “The Good Governance Agenda: Beyond Indicators without Theory.” Oxford Development Studies 36(4): 379–407. Auer, Astrid, Christoph Demmke, and Robert Polet. 1996. Civil services in the Europe of Fifteen: Current Situation and Prospects. Maastricht, the Netherlands: European Institute of Public Administration. Bäck, Hanna, and Axel Hadenius. 2008. “Democracy and State Capacity: Exploring a J-Shaped Relationship.” Governance 21:1–24. Bartels, Brandon L. 2008. “Beyond “Fixed versus Random Effects”: A Framework for Improving Substantive and Statistical Analysis of Panel, Time-Series Cross-Sectional, and Multilevel Data.” Paper prepared for the Faculty Poster Session, Political Methodology Conference, Ann Arbor, MI, July 9–12, 2008. 181 BBC News. 2012. “Politically neutral civil service 'in peril', warns Labour.” December 13, 2012. http://www.bbc.co.uk/news/uk-politics-20716245. BBC News. 2013. “Civil service criticised over 'culture of amateurism.'” Februaray5. 2013. http:// www.bbc.co.uk/news/uk-politics-21331042. Beck, T., Clarke, G., Groff, A., Keefer, P. and Walsh, P. 2001. “New Tools in Comparative Political Economy: The Database of Political Institutions.” World Bank Economic Review 15(1): 165–176. Berg, Linda. 2007. Multi-level Europeans: The Influence of Territorial Attachments on Political Trust and Welfare Attitudes. PhD diss., Goteborg University. Brans, Marleen. 2003. “Comparative Public Administration: from General Theory to General Framework” In Handbook of Public Administration, eds. Guy B. Peters and Jon Pierre. London: SAGE Publications. Bratton, Michael, and Eric C.C. Chang. 2006. “State Building and Democratization in SubSaharan Africa: Forward, Backward, or Together?” Comparative Political Studies 39(9): 1059–1083. Brehm, John, and Wendy Rahn. 1997. “Individual-Level Evidence for the Causes and Consequences of Social Capital.” American Journal of Political Science 41:999–1023. Brewer, Gene A. 2003. "Building Social Capital: Civic Attitudes and Behavior of Public Servants." Journal of Public Administration Research and Theory 13(1): 5–26. Campbell, Colin. 2004. "Innovation, Coordination and Accountability: A Comparative Overview with an Emphasis upon Anglo-American Systems." Presented at the Annual Meeting of the American Political Science Association, Chicago, IL, September 2–5, 2004. Carroll, J. Douglas. 1972. “Individual Differences and Multidimensional Scaling.” In Multidimensional Scaling: theory and applications in the behavioral sciences, eds. Roger N. Shepard, A. Kimball Romney, and Sara Beth Nerlove. New York: Seminar Press. Chang, Eric C. C., and Miriam A. Golden. 2007. “Electoral Systems, District Magnitude and Corruption.” British Journal of Political Science 37(1): 115–137. Chang, Eric C.C., and Yun-han Chu. 2006. “Corruption and Trust: Exceptionalism in Asian Democracies?” The Journal of Politics 68:259–271. Charron, Nicholas, and Victor Lapuente. 2010. “Does Democracy Produce Quality of Government?” European Journal of Political Research 49:443–470. Cheibub, Jose Antonio. 2006. “Presidentialism, Electoral Identifiability, and Budget Balances in Democratic Systems.” American Political Science Review 100(3): 353–68. 182 Christoph, James B. 1957. “Political Rights and Administrative Impartiality in the British Civil Service.” The American Political Science Review 51(1): 67–87. Chu, Yun-han, Michael Bratton, Marta Lagos, Sandeep Shastri, and Mark Tessler. 2008. “Public Opinion and Democratic Legitimacy.” In How People View Democracy, eds. Larry Diamond and Marc F. Plattner. Baltimore: Johns Hopkins University Press, 31–44. Coggburn, Jerrell D., and Saundra K. Schneider. 2003. “The Relationship between State Government Performance and State Quality of Life.” International Journal of Public Administration 26(12): 1337–1354. Dahlstrӧm, Carl, Johannes Lindvall, and Bo Rothstein. 2012. “Corruption, Bureaucratic Failure and Social Policy Priorities.” Political Studies. doi: 10.1111/j.1467-9248.2012.00998.x. Dahlstrӧm, Carl, Victor Lapuente, and Jan Teorell. 2010. “Dimensions of Bureaucracy: A CrossNational Dataset on the Structure and Behavior of Public Administration.” QoG Working Paper Series 2010:13, The Quality of Government Institute, University of Gothenburg. Dahlstrӧm, Carl, Victor Lapuente, and Jan Teorell. 2012. “The Merit of Meritocratization: Politics, Bureaucracy, and the Institutional Deterrents of Corruption.” Political Research Quarterly 65(3): 656–668. Dalton, Russell J. 1999. “Political Support in Advanced Industrial Democracies.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 57–77. Dalton, Russell J.2004. Democratic Challenges, Democratic Choices: The Erosion of Political Support in Advanced Industrial Democracies. New York: Oxford University Press. Davison, Mark L. 1983. Multidimensional Scaling. New York: John Wiley & Sons. Demmke, Christoph, and Timo Moilanen. 2010. Civil Services in the EU of 27. New York: Peter Lang Derlien, Hans-Ulrich. 2003. “German Public Administration: Weberian Despite “Modernization.” In Comparative Bureaucratic Systems, ed. Krishna K. Tummala. Lanham, MD: Lexington Books, 97–122. Diamond, Larry. 1999. Developing Democracy: Toward Consolidation. Baltimore:Johns Hopkins U. Press. Dolan, Julie. 2002. “Representative Bureaucracy in the Federal Executive: Gender and Spending Priorities.” Journal of Public Administration Research and Theory: J-PART 12(3): 353– 375. 183 Dollar, David, Raymond Fisman, and Roberta Gatti. 2001. “Are Women Really the “fairer” Sex? Corruption and Women in Government.” Journal of Economic Behavior and Organization 46(4): 423–429. Easton, David. 1975. “A Re-Assessment of the Concept of Political Support.” British Journal of Political Science 5(4): 435–457. Economist. 2012. “Changing Whitehall: Sir Humphrey, your time is up.” June 9, 2012. Eisenhardt, Kathleen M. 1989. “Agency Theory: An Assessment and Review.” Academy of Management Review 14(1): 57–74. Enders, Craig K., and Davood Tofighi. 2007. “Cantering Predictor Variables in Cross-Sectional Multilevel Models: A New Look at an Old Issue.” Psychological Methods 12(2): 121–138. Etzioni-Halevy, Eva.1983. Bureaucracy and Democracy: a Political Dilemma. Boston: Routledge & K. Paul. European Commission. 2010. More Women in Senior Positions: Key to Economic Stability and Growth. Luxembourg: Publications Office of the European Union. www.ec.europa.eu/ social/BlobServlet?docId=4746&langId=en. Evans, Peter, and James E. Rauch. 1999. “Bureaucracy and Growth: A Cross-National Analysis of the Effects of “Weberian” State Structures on Economic Growth.” American Sociological Review 64(October): 748–765. Freitag, Markus, and Marc Bühlmann. 2009. “Crafting Trust: The Role of Political Institutions in a Comparative Perspective.” Comparative Political Studies 42(12): 1537–1566. Friedrich, Carl. 1940. "Public Policy and the Nature of Administrative Responsibility." Public Policy 1:3–24. Reprinted in Francis Rourke, ed. Bureaucratic Power in National Politics. 3rd ed. Boston: Little, Brown. 1978. Fukuyama, Francis. 1995. Trust: Social Virtues and the Creation of Prosperity. New York: the Free Press. Fukuyama, Francis. 2013. “What Is Governance?” Governance 26:347–368. Geddes, Barbara. 1991. "A Game Theoretic Model of Reform in Latin American Democracies." American Political Science Review 85(2):371–392. Gerring, John, Strom Thacker, and Carola Moreno. 2005. “Centripetal Democratic Governance: A Theory and Global Inquiry.” American Political Science Review 99(4): 567–81. Gill, Jeff, and Dominik Hangartner. 2010. “Circular Data in Political Science and How to Handle it.” Political Analysis 18:316–336. 184 Gilley, B. 2006. “The Determinants of State Legitimacy: Results for 72 Countries.” International Political Science Review 27(1): 47–71. Gingerich, Daniel W. 2013. “Governance Indicators and the Level of Analysis Problem: Empirical Findings from South America.” British Journal of Political Science 43(3): 505–540. Golder, Matt, and Jacek Stramski. 2010. “Ideological Congruence and Electoral Institutions.” American Journal of Political Science 54(1): 90–106. Greve, Carsten . 2006. "Public Management Reform in Denmark." Public Management Review 8(1): 161–169. Gualmini, Elisabetta. 2008. “Restructuring Weberian Bureaucracy: Comparing Managerial Reforms in Europe and the United States.” Public Administration 86 (1): 75–94. Hall, Richard H. 1963. “The Concept of Bureaucracy: An Empirical Assessment.” American Journal of Sociology 69(1): 32–40. Halligan, John. 2010. “The Fate of Administrative Tradition in Anglophone Countries during the Reform Era.” In Tradition and Public Administration, eds. Martin Painter and B. Guy Peters New York: Palgrave Macmillan, 129–142. Hammouya, Messaoud. 1999. “Statistics on Public Sector Employment: Methodology, Structures and Trends.” International Labour Organisation Sectoral Activities Programme working paper series, 144. www.ilo.org/public/english/bureau/stat/download/wp_pse _e.pdf. Heclo, Hugh. 1974. Modern Social Politics in Britain and Sweden; from Relief to Income Maintenance. Yale University Press. Heclo, Hugh. 1977. A Government of Strangers: Executive Politics in Washington. Washington DC: Brookings. Hetherington, Marc J. 1998. “The Political Relevance of Political Trust.” American Political Science Review 92:791–808. Hibbing, John R., and Elizabeth Theiss-Morse. 2001. “Process Preferences and American Politics: What People Want Government to Be.” American Political Science Review 95:145–154. Hill, Larry B. 1991. “Who Governs the American Administrative State? A BureaucraticCentered Image of Governance.” Journal of Public Administration Research and Theory: J-PART 1(3): 261–294. 185 Hill, Larry B. 1992. “Taking Bureaucracy Seriously.” In The State of Public Bureaucracy, ed. Larry B. Hill. Armonk: M.E. Sharpe, 15–58. Hoffman, Lesa. 2010. Multilevel Models using SAS and SPSS. ICPSR Summer Workshop at the University of Michigan. July 19 -23, 2010. http://psych.unl.edu/hoffman/Sheets/ICPSR.htm. Holmberg, Soren, Bo Rothstein, and Naghmeh Nasiritousi. 2009. “Quality of Government: What you get.” Annual Review of Political Science 12:135–161. Holmberg, Sӧren, and Bo Rothstein,.eds. 2012. Good Government: The Relevance of Political Science. Massachusetts: Edward Elgar Publishing. Huber, Evelyne, and John D. Stephens. 2001. Development and Crisis of the Welfare State. Chicago: University of Chicago Press. ILO (International Labour Office). LABORSTA Internet. http://laborsta.ilo.org/. Inglehart, Ronald. 1988. “The Renaissance of Political Culture.” American Political Science Review 82 (4): 1203–1230. ISSP Research Group. 2003. International Social Survey Programme: Social Relations and Support Systems / Social Networks II - ISSP 2001. GESIS Data Archive, Cologne. ZA3680 Data file Version 1.0.0. ISSP Research Group. 2003. International Social Survey Programme: Environment II - ISSP 2000. GESIS Data Archive, Cologne. ZA3440 Data file Version 1.0.0. ISSP Research Group. 2004. International Social Survey Programme: Family and Changing Gender Roles III - ISSP 2002. GESIS Data Archive, Cologne. ZA3880 Data file Version 1.0.0. ISSP Research Group. 2007. International Social Survey Programme: Citizenship - ISSP 2004. GESIS Data Archive, Cologne. ZA3950 Data file Version 1.2.0. ISSP Research Group. 2007. International Social Survey Programme: National Identity II - ISSP 2003. GESIS Data Archive, Cologne. ZA3910 Data file Version 2.0.0. Jacoby, William G. 1991. Data Theory and Dimensional Analysis. Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-078. Newbury Park, CA: Sage. Jacoby, William G. 1999. “Levels of Measurement and Political Research: An Optimistic View.” American Journal of Political Science 43(1): 271–301. 186 Jacoby, William G. 2011a. “Is There a Culture War? Heterogeneous value choices and American public opinion.” Presented at the Annual Meetings of the American Political Science Association. Seattle, WA, September 1– 4, 2011. Jacoby, William G. 2011b. Optiscale. R package, Version 1.0-2. Unpublished. Jacoby, William G., and Saundra K. Schneider. 2001. “Variability in State Policy Priorities: An Empirical Analysis.” Journal of Politics 63:544–568. Jacoby, William G., and Saundra K. Schneider. 2009. “A New Measure of Policy Spending Priorities in the American States.” Political Analysis 17(1): 1–24. Jamal, Amaney, and Irfan Nooruddin. 2010. “The Democratic Utility of Trust: A Cross-National Analysis.” The Journal of Politics 72(1): 45–59. Jancofsky, Michael. 2004. “Citing July Speech, I.R.S. Decides to Review N.A.A.C.P.” New York Times. http://www.nytimes.com/2004/10/29/politics/29probe.html. Jensen, Lotte. 1998. “Interpreting New Public Management: The Case of Denmark.” Australian Journal of Public Administration 57:54–65. Kahn, Lawrence M. 2008. “The Impact of Wage-Setting Institutions on the Incidence of Public Employment in the OECD: 1960–1998.” Industrial Relations 47(3): 329–354. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2005. Governance Matters VI: Governance Indicators for 1996-2004. Washington DC: World Bank, Policy Research Paper. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2007a. “Growth and Governance: A Rejoinder.” The Journal of Politics 69(2): 570–572. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2007b. “Growth and Governance: A Reply.” The Journal of Politics 69(2): 555–562. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2008. Governance Matters VII: Aggregate and Individual Governance Indicators, 1996-2007. Washington DC: World Bank, Policy Research Paper No. 4654. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2010. The Worldwide Governance Indicators: Methodology and Analytical Issues. SSRN Scholarly Paper. Rochester, NY: Social Science Research Network. Kearney, Richard C., and Chandan Sinha. 1988. “Professionalism and Bureaucratic Responsiveness: Conflict or Compatibility?” Public Administration Review 48(1): 571– 579. 187 Keefer, Philip, and Stephen Knack. 1997. “Why Don’t Poor Countries Catch Up? A CrossNational Test of an Institutional Explanation.” Economic Inquiry 35:590–602. Keele, Luke. 2007. “Social Capital and the Dynamics of Trust in Government.” American Journal of Political Science 51: 241–254. Keiser, Lael R., Vicky Wilkins, Kenneth J. Meier, and Catherine Holland. 2002. “Lipstick and Logarithms: Gender, Identity, Institutional Context, and Representative Bureaucracy.” American Political Science Review 96(3): 553–564. Kettl, Donald F. 2000. “The Transformation of Governance: Globalization, Devolution, and the Role of Government.” Public Administration Review 60(6): 488–497. Kettl, Donald F. 2006. “Public Bureaucracies.” In The Oxford Handbook of Political Institutions, eds. Rhodes, R. A. W., Sarah A. Binder, and Bert A. Rockman. New York: Oxford University Press. Kim, Soonhee. 2010. “Public Trust in Government in Japan and South Korea: Does the Rise of Critical Citizens Matter?” Public Administration Review 70(5): 801–810. Klingemann, Hans-Dieter. 1999. “Mapping Political Support in the 1990s: A Global Analysis.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 31–56. Knack, Stephen, and Keefer, Philip, 1995. “Institutions and Economic Performance: CrossCountry Tests Using Alternative Institutional Indicators.” MPRA Paper 23118, University Library of Munich, Germany. Knack, Stephen, ed. 2003. Democracy, Governance, and Growth. Ann Arbor, MI: University of Michigan Press. Knack, Stephen. 2002. “Social Capital and the Quality of Government: Evidence from the States.” American Journal of Political Science 46(4): 772–785. Knott, Jack H., and Gary J. Miller. 2006. “Social Welfare, Corruption and Credibility.” Public Management Review 8(2): 227–252. Knott, Jack H., and Gary J. Miller. 2008. “When Ambition Checks Ambition: Bureaucratic Trustees and Credible Commitment.” American Review of Public Administration 38(4): 387–411. Kohli, Atul. 2004. State-Directed Development: Political Power and Industrialization on the Global Periphery. New York: Cambridge University Press. Kumlin, Staffan, and Bo Rothstein. 2005. “Making and Breaking Social Capital: The Impact of Welfare-State Institutions.” Comparative Political Studies 38(4): 339–365. 188 Kurtz, Marcus J., and Andrew Schrank. 2007b. “Growth and Governance: A Defense.” The Journal of Politics 69(2): 563–569. Kurtz, Marcus, and Andrew Schrank. 2007a. “Growth and Governance: Models, Measures and Mechanisms.” Journal of Politics 69(2): 538–554. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny. 1999. “The Quality of Government.” Journal of Law, Economics, and Organization 15(1): 222–279. Langbein, Laura, and Stephen Knack. 2010. “The Worldwide Governance Indicators: Six, One, or None?” Journal of Development Studies 46(2): 350–370. Levi, Margaret, and Audrey Sacks. 2009. “Legitimating Beliefs: Sources and Indicators.” Regulation & Governance 3(4): 311–333. Levi, Margaret, and Laura Stoker. 2000. “Political Trust and Trustworthiness.” Annual Review of Political Science 3:475–508. Levi, Margaret, and Richard Sherman. 1997. “Rational Compliance with Rationalized Bureaucracy.” In Institutions and Economic Development: Growth and Governance in Less-Developed and Post-Socialist Countries, ed. Christopher Clague. Baltimore: Johns Hopkins University Press, 316–341. Levi, Margaret. 1998. “A State of Trust.” In Trust and Governance, eds. Valerie Braithwaite and Margaret Levi. New York: Russell Sage Foundation, 77–101. Lewis, David E. 2008. The Politics of Presidential Appointments: Political Control and Bureaucratic Performance. Princeton, N.J: Princeton University Press. Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Performance in ThirtySix Countries. New Haven, CT: Yale University Press. Lim, Hong-Hai. 2006. “Representative Bureaucracy: Rethinking Substantive Effects and Active Representation.” Public Administration Review 66(2): 193–204. Linz, Juan J., and Alfred Stepan. 1996. “Toward Consolidated Democracies.” Journal of Democracy 7(2): 14–33. Linz, Juan. 1990. “The Perils of Presidentialism.” Journal of Democracy 1:51–69. Longo, Francisco. 2008. “Quality of Governance: Impartiality Is Not Enough.” Governance 21(2): 191–196. Luke, Douglas A. 2004. Multilevel Modeling. Sage University Paper Series on Quantitative Applications in the Social Sciences, 143. Thousand Oaks, CA: Sage Publication. 189 McAllister, Ian. 1999. “The Economic Performance of Governance.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press. Meier, Kenneth J. 1975. “Representative Bureaucracy: An Empirical Analysis.” American Political Science Review 69:526–542. Meier, Kenneth J., and John Bothe. 2007. Politics and the Bureaucracy: Policymaking in the Fourth Branch of Government. 5th ed. Belmont, CA: Thomson/Wadsworth. Meier, Kenneth J., and Laurence J. O'Toole Jr. 2006. Bureaucracy in a Democratic State: A Governance Perspective. The Johns Hopkins University Press. Meier, Kenneth J., and Laurence J. O'Toole Jr. 2006. Bureaucracy in a Democratic State: A Governance Perspective. The Johns Hopkins University Press. Meier, Kenneth J., and Lloyd G. Nigro. 1976. “Representative Bureaucracy and Policy Preferences.” Public Administration Review 36(July/August): 458–469. Meier, Kenneth J., Robert D. Wrinkle, and J. L. Polinard. 1999. “Representative Bureaucracy and Distributional Equity: Addressing the Hard Question.” The Journal of Politics 61(04): 1025–1039. Meyer-Sahling, Jan-Hinrik. 2010. “In Search of the Shadow of the Past: Legacy Explanations and Administrative Reform in Post-Communist East Central Europ.” In Tradition and Public Administration, eds. Martin Painter and B. Guy Peters New York: Palgrave Macmillan, 203–216. Migdal, Joel. 1988. Strong Societies and Weak States: State-Society Relations and State Capabilities in the Third World. Princeton, N.J: Princeton University Press. Miller, Arthur, and Ola Listhaug. 1999. “Political Performance and Institutional Trust.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 204–216. Miller, Gary J. 2000. “Above Politics: Credible Commitment and Efficiency in the Design of Public Agencies.” Journal of Public Administration Research and Theory 10:289–328. Miller, Gary J. 2005. “The Political Evolution of Principal-Agent Models.” Annual Review of Political Science 8:203–225. Mintzberg, Henry. 1996. “Managing Government: Governing Management.” Harvard Business Review May-June, 1996:75–83. 190 Mishler, William, and Richard Rose. 2001. “What Are the Origins of Political Trust? Testing Institutional and Cultural Theories in Post-communist Societies.” Comparative Political Studies 34(1): 30–62. Moe, Terry. 1985. “The Politicized Presidency.” In New Directions in American Politics, eds. John E. Chubb and Paul E. Peterson. Washington, D.C.: The Brookings Institution, 235– 272. Möehring, Katja, and Alexander Schmidt. 2013. “MLT: Stata Module to Provide Multilevel Tools.” Statistical Software Components S457577, Boston College Department of Economics, revised 29 Jan 2013. Mosher, Frederick C. 1982. Democracy and the Public Service. New York and Oxford: Oxford University Press. Nakamura, Akira. 2003. “Public Administration in Japan: The Need for Reform.” In Comparative Bureaucratic Systems, ed. Krishna K. Tummala. Lanham, MD: Lexington Books, 143–163. Newton, Kenneth. 1999. “Social and Political Trust in Established Democracies.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 169–187. Newton, Kenneth. 2001. “Trust, Social Capital, Civil Society, and Democracy.” International Political Science Review 22(2): 201–214. Nicholson-Crotty, Jill, Jason A. Grissom, and Sean Nicholson-Crotty. 2011. “Bureaucratic Representation, Distributional Equity, and Democratic Values in the Administration of Public Programs.” The Journal of Politics 73(2): 582–596. Norris, Pippa. 1999a. “Conclusion: The Growth of Critical Citizens and its Consequences.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 257–272. Norris, Pippa. 1999b. “Institutional Explanations for Political Support.” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 217–235. Norris, Pippa. 1999c. “Introduction: The Growth of Critical Citizens?” In Critical Citizens: Global Support for Democratic Governance, ed. Pippa Norris. New York: Cambridge University Press, 1–27. Norris, Pippa. 2004. Electoral Engineering: Voting Rules and Political Behavior. Cambridge: Cambridge University Press. 191 Norris, Pippa. 2008. Driving Democracy: Do power-sharing institutions work? New York: Cambridge University Press. Norris, Pippa. 2011. Democratic Deficit: Critical Citizens Revisited. New York: Cambridge University Press. Nye, Joseph S. Jr., Philip D. Zelikow, and David C. King, eds. 1997. Why people don't trust government. Cambridge: Harvard University Press. O'Dwyer, Conor. 2004. “Runaway State Building: How Political Parties Shape States in Postcommunist Eastern Europe.” World Politics 56(4): 520–553. OECD. 1997. Measuring Public Employment in OECD Countries: Sources, Methods, and Results. Head of Publications Service, OECD. www.oecd.org/dataoecd/48/7/ 1910752.pdf. OECD. 2009a. Fostering Diversity in the Public Service. Public Employment and Management Working Party. GOV/PGC/PEM(2009)1/FINAL. http://www.oecd.org/governance/ publicemploymentandmanagement/44860884.pdf. OECD. 2009b. Government at a Glance 2009, OECD Publishing. OECD. 2011. Government at a Glance 2011, OECD Publishing. OECD. OECD.Stat Extracts. http://stats.oecd.org. Olsen, Johan P. 2005. “Maybe It is Time to Rediscover Bureaucracy.” Journal of Public Administration Reach and Theory 16(March): 1–24. Olsen, Johan P. 2010. Governing through Institution Building: Institutional Theory and Recent European Experiments in Democratic Organization. Oxford : Oxford University Press. Ongaro, Edoardo. 2008. “Introduction: the Reform of Public Management in France, Greece, Italy, Portugal and Spain.” International Journal of Public Sector Management 21(2): 101–117. Ongaro, Edoardo. 2010. “The Napoleonic Administrative Tradition and Public Mangement Reform in France, Greece, Italy, Portugal and Spain.” In Tradition and Public Administration, eds. Martin Painter and B. Guy Peters New York: Palgrave Macmillan, 174–190. Page, Edward C., and Vincent Wright, eds. 2007. From the Active to the Enabling State: the Changing Role of Top Officials in European Nations. New York: Palgrave Macmillan. Paik, Wanki. 2003. “The Korean System of Bureaucracy.” In Comparative Bureaucratic Systems, ed. Krishna K. Tummala. Lanham, MD: Lexington Books, 185–205. 192 Painter, Martin, and B. Guy Peters, eds. 2010. Tradition and Public Administration. New York: Palgrave Macmillan. Painter, Martin. 2010. “Legacies Remembered, Lessons Forgotten: The Case of Japan.” In Tradition and Public Administration, eds. Martin Painter and B. Guy Peters New York: Palgrave Macmillan, 84–98. Pamphilis, Niccole. 2012. Government Spending Priorities: A Cross-national Perspective. PhD diss., Michigan State University. Paraskevopoulos, Christos J. 2010. “Social Capital: Summing up the Debate on a Conceptual Tool of Comparative Politics and Public Policy.” Comparative Politics 42(4): 475–494. Peters, B. Guy, and Jon Pierre, eds. 2004. Politicization of the Civil Service in Comparative Perspective: the Quest for Control. New York: Routledge. Peters, B. Guy. 1993. “Searching for a Role: The Civil Service in American Democracy” International Political Science Review 14(4): 373–386. Peters, B. Guy. 2001. The Politics of Bureaucracy. 5th ed. New York: Longman. Pierre, Jon, and B. Guy Peters. 2000. Governance, Politics and the State. New York: St. Martin’s Press. Pierre, Jon, ed. 1995. Bureaucracy in the Modern State: an Introduction to Comparative Public Administration. Brookfield: E. Elgar Pub. Co. Pierre, Jon. 2010. “Administrative Reform in Sweden: The Resilience of Administrative Tradition?” In Tradition and Public Administration, eds. Martin Painter and B. Guy Peters New York: Palgrave Macmillan, 191–202. Pierson, Paul. 1996. “The New Politics of the Welfare State.” World Politics 48(1): 143–179. Pitts, David W. 2005. “Diversity, Representation, and Performace: Evidence about Race and Ethnicity in Public Organizations.” Journal of Public Administration: Research and Theory 15:615–631. Plowden, William.1984. “The Higher Civil Service of Britain.” In The Higher Civil Service in Europe and Canada, ed. Bruce L. R. Smith. Washington, D.C.: The Brookings Institution. Pollitt, Christopher, and Geert Bouckaert. 2011. Public Management Reform: a Comparative Analysis. 3rd ed. New York: Oxford University Press. Powell, G. Bingham. 2004. “The Chain of Responsiveness.” Journal of Democracy 15(4): 91– 105. 193 PRS (Political Risk Services). International Country Risk Guide. The Political Risk Services Group. http://www.prsgroup.com/ PDFS/icrgmethodology.pdf. Putnam, Robert. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton, N.J: Princeton University Press. Rabe-Hesketh, Sophia, and Anders Skrondal. 2008. Multilevel and Longitudinal Modeling Using Stata. 2nd ed. Stata Press. Randma, Tina. 2001. “A Small Civil Service in Transition: The Case of Estonia.” Public Administration and Development 21:41–51. Rauch, James, and Peter Evans. 2000. “Bureaucratic Structure and Bureaucratic Performance in Less Developed Countries” Journal of Public Economics 75:49–71. Reimann, Bernard C. 1973. “On the Dimensions of Bureaucratic Structure: An Empirical Reappraisal.” Administrative Science Quarterly 18(4): 462–476. Rohrschneider, Robert. 2005. “Institutional Quality and Perceptions of Representation in Advanced Industrial Democracies.” Comparative Political Studies 38(7): 850–874. Rose, Richard, and Doh Chull Shin. 2001. “Democratization Backwards: The Problem of ThirdWave Democracies,” British Journal of Political Science 31:331–354. Rose, Richard, ed. 1985. Public Employment in Western Nations. New York: Cambridge University Press. Rosenbloom, D. H. 1974. “Forms of Bureaucratic Representation in the Federal Service.” American Review of Public Administration 8(3): 159–177. Rothstein, Bo, and Dietlind Stolle. 2008. “The State and Social Capital: An Institutional Theory of Generalized Trust.” Comparative Politics 40(4): 441– 459. Rothstein, Bo, and Jan Teorell. 2008a. “Impartiality as a Basic Norm for the Quality of Government: A Reply to Francisco Longo and Graham Wilson.” Governance 21(2): 201– 204. Rothstein, Bo, and Jan Teorell. 2008b. “What Is Quality of Government? A Theory of Impartial Government Institutions.” Governance 21(2): 165–190. Rothstein, Bo, Marcus Samanni, and Jan Teorell. 2012. “Explaining the Welfare State: Power resources vs. the Quality of Government.” European Political Science Review 4(1): 1–28. Rothstein, Bo. 2011. “The Low Trust-Corruption-Inequality Trap.” In The Quality of Government: Corruption, Social Trust, and Inequality in International Perspective. Bo Rothstein. Chicago, IL: University of Chicago Press, 145-163. 194 Schneider, Saundra K., and Patricia Ingraham. 1984. “The Impact of Political Participation on Social Policy Adoption and Expansion: A Cross-National, Longitudinal Analysis.” Comparative Politics 17(1): 107–122. Schneider, Saundra K., and William G. Jacoby. 1996. “Influences on Bureaucratic Policy Initiatives in the American States.” Journal of Public Administration Research and Theory: J-PART 6(4): 495–522. Schneider, Saundra K., William G. Jacoby, and Jerrell D. Coggburn. 1997. “The Structure of Bureaucratic Decisions in the American States.” Public Administration Review 57(3): 240–249. Schultz, David A., and Robert Maranto. 1998. The Politics of Civil Service Reform. New York: Peter Lang. Selden, Sally Coleman, Jeffrey L. Brudney, and J. Edward Kellough. 1998. “Bureaucracy as a representative institution: Toward a reconciliation of bureaucratic government and democratic theory.” American Journal of Political Science 42:717–44. Selden, Sally Coleman. 1997. The Promise of Representative Bureaucracy. Armonk, NY: M.E. Sharpe. Shefter, Martin. 1977. “Party and Patronage: Germany, England, and Italy.” Politics and Society 7(3): 403–51. Simon, Herbert A. 1946. “The Proverbs of Administration.” Public Administration Review 6(1): 53–67. Skocpol, Theda. 1985. “Brining the State Back in: Strategies of Analysis in Current Research.” In Bringing the State Back in, eds. Peter B. Evans, Dietrich Rueschemeyer and Theda Skocpol. New York: Cambridge University Press, 3–37. Skocpol, Theda. 1995. Social Policy in the United States. Princeton, N.J: Princeton University Press. Snijders, Tom A. B., and Roel Bosker. 1999. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. Thousand Oaks, CA: SAGE Publications. Snijders, Tom A. B., and Roel Bosker. 2011. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. 2nd ed. Thousand Oaks, CA: SAGE Publications. Sowa, Jessica E., and Sally Coleman Selden. 2003. “Administrative Discretion and Active Representation: An Expansion of the Theory of Representative Bureaucracy.” Public Administration Review 63(6): 700–710. STAN database (Structural Analysis Database in OECD.stat). www.oecd.org/sti/stan. 195 Steenbergen, Marco R., and Bradford S. Jones. 2002. “Modeling Multilevel Data Structures.” American Journal of Political Science 46(1): 218–237. Stegmueller, Daniel. 2013. “How Many Countries for Multilevel Modeling? A Comparison of Frequentist and Bayesian Approaches.” American Journal of Political Science 57(3): 748– 761. Stensӧta, Helena Olofsdotter. 2012. “Impartiality and the need for a public ethics of care.” In Good Government: The Relevance of Political Science, eds. Sören Holmberg and Bo Rothstein. Northampton, MA: Edward Elgar Pub. Stevenson, Richard. 1997. “White House says it supports a bill to overhaul I.R.S.” New York Times. http://www.nytimes.com/1997/10/22/us/white-house-says-it-supports-a-bill-tooverhaul-irs.html. Subramaniam, V. 1967. “Representative Bureaucracy: A Reassessment.” The American Political Science Review 6(4): 1010–1019. Suleiman, Ezra N. 2003. Dismantling Democratic States. Princeton, N.J: Princeton University Press. Svara, James H. 2001. “The Myth of the Dichotomy: Complementarity of Politics and Administration in the Past and Future of Public Administration.” Public Administration Review 61(2): 176–183. Tavits, Margit. 2006. “Making Democracy Work More? Exploring the Linkage Between Social Capital and Government Performance.” Political Research Quarterly 59:211–225. Teorell, Jan, Carl Dahlström, and Stefan Dahlberg. 2011. The QoG Expert Survey Dataset. University of Gothenburg: The Quality of Government Institute. http://www.qog.pol.gu.se. Teorell, Jan, Marcus Samanni, Sören Holmberg and Bo Rothstein. 2011. The Quality of Government Dataset, version 6Apr11. University of Gothenburg: The Quality of Government Institute, http://www.qog.pol.gu.se. Teorell, Jan. 2009. “The Consequences of Quality of Government as Impartiality: Theory and Evidence.” QoG Working Paper Series 2009: 25. The Quality of Government Institute, University of Gothenburg. Presented at the Annual Meeting of the American Political Science Association, Toronto, ON, Canada. September 3–9, 2009. Thomas, M. A. 2010. “What Do the Worldwide Governance Indicators Measure?” European Journal of Development Research 22(1): 31–54. Thompson, Dennis. 1983. “Bureaucracy and democracy” In Democratic Theory and Practice, ed. Graeme C. Duncan. Cambridge: Cambridge University Press, 235–250. 196 Tsebelis, George. 2002. Veto Players: How Institutions Work. Princeton, N.J: Princeton University Press. Tummala, Krishna K, ed. 2003. Comparative Bureaucratic Systems. Lanham, MD: Lexington Books. Tyler, Tom R. 1990. Why People Obey the Law. New Haven, CT: Yale University Press. United Nations. 2009. System of National Accounts 2008, New York: United Nations. http:// unstats.un.org/unsd/nationalaccount/sna2008.asp. UNPAN (United Nations Public Administration Networks). 2006. Public Administration Country Profiles. United Nations’ Division for Public Administration and Development Management (DPADM) and Department of Economic and Social Affairs (DESA). http:// www.unpan.org/DPADM/ProductsServices/ThematicPortals/PublicAdministrationCountr yProfiles/tabid/677/Default.aspx. UNSD (United Nations Statistics Devision). UN Classifications Registry. http://unstats.un.org/ unsd/cr/registry. Van de Walle, Steven. 2006. “The State of the World’s Bureaucracies.” Journal of Comparative Policy Analysis 8(4): 437–448. Van de Walle, Steven. 2007. “Determinants of Confidence in the Civil Service: An International Comparison,” In Cultural Aspects of Public Management Reforms, eds. Kuno Schedler and Isabella Proeller. Amsterdam; London: Elsevier JAI Press, 171–201. Van der Meer, Tom. 2010. “In What We Trust? A Multi-level Study into Trust in Parliament as an Evaluation of State Characteristics.” International Review of Administrative Sciences 76(3): 517–536. Van Rijckeghem, Caroline, and Beatrice Weder. 2001. “Bureaucratic Corruption and the Rate of Temptation: Do Wages in the Civil Service Affect Corruption, and by How Much?” Journal of Development Economics 65(2): 307–331. Van Riper, Paul P. 1958. History of the United States Civil Service. Illinois; New York: Row, Peterson and Company. Van Slyke, David M., and Norman M. Riccuci.2003. “The Politicized Administrative Environment of Public Administration in the United States.” In Comparative Bureaucratic Systems, ed. Krishna K. Tummala. Lanham, MD: Lexington Books, 81–96. Verheijin, Tony J.G., and Aleksandra Rabrenovic. 2007. “Civil Service Development in Central and Eastern Europe and the CIS: Swimming with the tide.” In The Civil Service in the 21st Century: Comparative Perspectives, eds. Jos C.N.Raadschelders, Theo A.J. Toonen, and Frits M. Van der Meer. Basingstoke; New York: Palgrave Macmillan, 17–33. 197 Vigoda-Gadot, Eran, and Fany Yuval. “Research Note: The State of Bureaucracy: Public Opinion about the Performance of Government Agencies in Israel.” International Journal of Public Opinion Research 16(1): 63–80. Vigoda-Gadot, Eran, Aviv Shoham, and Dana R Vashdi. 2010. “Bridging Bureaucracy and Democracy in Europe: A Comparative Study of Perceived Managerial Excellence, Satisfaction with Public services, and Trust in Governance.” European Union Politics 11(2): 289–308. Waterman, Richard W., and Kenneth J. Meier. 1998. “Principal-Agent Models: A Expansion?” Journal of Public Administration Research and Theory 8(April): 173–202. Watkins-Hayes, Celeste. 2011. “Race, Respect, and Red Tape: Inside the Black Box of Racially Representative Bureaucracies.” Journal of Public Administration Research and Theory 21(suppl 2): i233–i251. Weaver, R. Kent, and Bert A. Rockman, eds. 1993. Do Institutions Matter?: Government Capabilities in the United States and Abroad. Washington, D.C.: The Brookings Institution. Weimer, David L. 2005. “Institutionalizing Neutrally Competent Policy Analysis: Resources for Promoting Objectivity and Balance in Consolidating Democracies.” Policy Studies Journal 33(2): 131–146. Weingast, Barry R. 1984. "The Congressional-Bureaucratic System: A Principal-Agent Perspective (with Applications to the SEC)." Public Choice 44(1):147–191. Weisman, Jonathan. 2013. “Management flaws at I.R.S. cited in Tea Party scrutiny.” New York Times. http://www.nytimes.com/2013/05/15/us/politics/report-on-irs-audits-citesineffective-management.html. Weldon, Steven A. 2006. “The Institutional Context of Tolerance for Ethnic Minorities: A Comparative, Multilevel analysis of Western Europe.” American Journal of Political Science 50(2): 331–349. Weller, Susan C., and A. Kimball Romney. 1990. Metric Scaling: Correspondence Analysis. Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-075. Newbury Park, CA: Sage. Werner, Simcha B. 1983. “The Development of Political Corruption: A Case Study of Israel.” Political Studies 31(4): 620–639. Whiteley, Paul, Harold D. Clarke, David Sanders, and Marianne C. Stewart. 2010. "Government Performance and Life Satisfaction in Contemporary Britain." The Journal of Politics 72(3): 733–746. 198 Wilkins, Vicky M., and Lael R. Keiser. 2006. “Linking Passive and Active Representation by Gender: The Case of Child Support Agencies.” Journal of Public Administration Research and Theory 16(1): 87–102. Wilson James Q. 1989. Bureaucracy: What Government Agencies Do and Why They Do It. New York: Basic Books. Wilson, Graham. 2008. “The Quality of Government.” Governance 21 (2): 197–200. Wish, Naomi Bailin. 1980. “Foreign Policy Makers and Their National Role Conceptions.” International Studies Quarterly 24(4): 532–554. Yaroni, Allon, and David H. Rosenbloom. 2003. “Public Administration in Israel” In Comparative Bureaucratic Systems, ed. Krishna K. Tummala. Lanham, MD: Lexington Books, 355–384. Zmerli, Sonja, and Ken Newton. 2008. “Social Trust and Attitudes Toward Democracy.” Public Opinion Quarterly 72(4): 706–724. Zucker, Lynne G. 1986. “Production of Trust: Institutional Sources of Economic Structure 18401920.” Research in Organizational Behavior 8:53–111. 199