THE EFFECT OF SIMILARITIES BETWEEN STATES ON INTEREST GROUP FORMATION AND POLICY INNOVATION By Carl Snook A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Political Science—Doctor of Philosophy 2013 ABSTRACT THE EFFECT OF SIMILARITIES BETWEEN STATES ON INTEREST GROUP FORMATION AND POLICY INNOVATION By Carl Snook This dissertation examines state-level policy innovation and interest group formation in primary and secondary education policy during the 1990s. While focusing specifically on the area of primary and secondary education, this research explores how underlying state characteristics lead to both the formation of interest groups and the enactment of policy innovations. The dissertation utilizes multidimensional scaling techniques to explore both policymaking similarities between states and the similarities of the diversity state interest group communities. I am able to expand on previous research into policy innovation and show that the state demographics and partisanship have a stronger influence on policy innovation than interest groups associated with primary and secondary education. TABLE OF CONTENTS LIST OF TABLES v LIST OF FIGURES vii CHAPTER 1 Introduction 1 Evaluating Differences Between States Overview of Dissertation Chapters 4 6 CHAPTER 2 An Overview of State Primary and Secondary Education Policy and Policy Innovation General Insights through Education Policy A Brief History of Education Policy Education Finance in the 1990s Public Dissatisfaction with Educational Performance and Equity Education Reform: Policy Innovations in the 1990s An Example of 1990s Education Reform: School Choice Conclusion 10 12 14 24 28 30 33 42 CHAPTER 3 State Policy Innovation in the Context of Prior Research Determinants of Policymaking Defining Policy Innovation Explanations for State Policy Innovations Competitive Federalism and State Preferences Interest Groups and Innovation Expected Influences on Policy Innovation and Interest Groups 45 46 48 52 59 62 69 CHAPTER 4 An Empirical Look at State Primary and Secondary Education Policy Innovation 71 iii Spatial Analysis using MDS Methods Data Spatial Results of the MDS Results from the Cluster Analysis Implications 77 78 82 84 93 99 CHAPTER 5 An Empirical Look at State Primary and Secondary Education Interest Group Communities Determinants of Interest Group Formation and Diversity Educational Interest Groups in the American States Methodology Regression Results for Density Spatial Results of MDS Cluster Analysis Discussion Implications 102 104 109 112 112 116 123 128 130 CHAPTER 6 A Configuration Comparison of State Policy Innovation and State Interest Group Communities Comparing Configurations (Congruence) Interest Group Diversity and Policy Innovations Results for the Configuration Comparison Conclusions 132 133 137 140 156 CHAPTER 7 Mass Preferences as an Explanation for State Policy Innovation Minority Diversity, Ideology, and Region in Education Policy The Nature of State Interest Group Communities Interest Groups and Policy Innovation Primary and Secondary Education Policy Innovation Points for Future Research Conclusion 160 162 169 173 175 179 181 BIBLIOGRAPHY 185 iv LIST OF TABLES Table 1 Primary and Secondary Education Policies, 1990-2000, by State and Category 73 Table 2 Correlations between Substantive Variables and Stimulus Points 88 Table 3 Regression Model to Examine Legislative Professionalism Scores 89 Table 4 Regression examining variables associated with Dimension 1 of Policy Innovation 91 Table 5 Regression examining variables associated with Dimension 2 of Policy Innovation 92 Table 6 Cluster results from Hierarchical Cluster Analysis of MDS Stimulus points 94 Table 7 Descriptive Statistics for Characteristics of each Cluster (Medians) 96 Table 8 Regression Models explaining the Density of Interest Group Communities 115 Table 9 Correlations between Substantive Variables and Stimulus points 120 Table 10 Cluster results from Cluster Analysis of MDS Stimulus points 125 v Table 11 Descriptive Statistics for the Characteristics of each Cluster (Means) 127 Table 12 Hierarchical Cluster Analysis of MDS Stimulus Points for Interest Groups 149 vi LIST OF FIGURES Figure 1 Scree Plot of SSTRESS of Policy Innovation 85 Figure 2 Stimulus Points with Dimensions related to State Policy Innovation 87 Figure 3 Stimulus Points with Dimensions related to Minority Diversity and Ideology 90 Figure 4 Scree Plot of SSTRESS for Primary and Secondary Education Interest Group Diversity 117 Figure 5 Two-Dimensional Configuration of Stimulus Points for Interest Group Diversity 118 Figure 6 Regression Lines Showing Dimensions for Interest Group Density and State Employees per 10,000 State Residents 122 Figure 7 Relative Cluster Location with each State’s Cluster Indicated 126 Figure 8 Two Dimensional Configuration of Stimulus Points for Policy Innovation 141 Figure 9 Two Dimensional Configuration of Stimulus Points for Interest Group Diversity 142 Figure 10 Optimally Rotated Interest Group Diversity Configuration and the Original Interest Group Diversity Solution 144 vii Figure 11 Optimally Rotated Interest Group Diversity Configuration and Policy Innovation 146 Figure 12 Interest Group Cluster #1 and Policy Innovation 150 Figure 13 Interest Group Cluster #2 and Policy Innovation 152 Figure 14 Interest Group Diversity for Remaining States and Policy Innovation 154 viii Chapter 1 Introduction th Americans in the second half of the 20 Century became increasingly frustrated with their country’s public education system. Between the 1970s and 1990s, the percentage of Americans who expressed either “a great deal” or “quite a lot” of confidence in public schools fell from 55% to 41% (Loveless 1997). 1 Many believed the system failed the students it was designed to teach and as well as a society dependent on those students for its renewal. The famed 1983 report, A Nation at Risk, by President Reagan’s National Commission on Excellence in Education, complained that “if an unfriendly foreign power had attempted to impose [the current level of educational performance], we might well have viewed it as an act of war” (National Commission on Excellence in Education 1983, 1) Whether quality was measured in terms of student achievement as compared to other countries, or in terms of equity for students whose performance lagged behind the national average, although it was typical within their own “failing schools,” demands for the improvement of primary and secondary education in the United States were widespread and included people of all political viewpoints (National Commission on Excellence in Education 1983; Moe 1995). 1 Loveless (1997) makes the counter-intuitive argument that schools maintain the confidence of the public because more students are staying in school, fewer are attending private schools, funding for education is increasing, and people across the country believe local schools are better than the public schools generally. Even given that confidence is not the same as satisfaction, the problems with selecting these measures are legion. Funding increases would seem to be a remedy for problems at least as much as a sign of confidence. Staying in school, as opposed to dropping out, would seem a good choice in any reasonably safe school provided that any learning occurs. Loveless does not discuss cost and the ability to pay for private school. 1 The central position of education reform stems from concerns about both quality and equality. With the beginning of the cold war in the early 1950s, and especially after the space race made this competition an overtly scientific and educational contest, Americans had begun to see primary and secondary education as a profound national problem related to international competitiveness and national security. During the same decade, equality and educational opportunity became potent political goals with Brown v. Board of Education, 347 U.S. 483 (1954) only one of many events focusing the country’s attention. Although this broad national concern for education led to some national involvement in education, the responsibility for education policy remained primarily with state governments and local school boards who did not always share federal government priorities (Kantor 1991). 2 Therefore, the two competing concerns in education: equality in school performance and student opportunity versus promoting high standards and accountability, have long been at the forefront of education policy debates. Mintrom (2000) noted the importance American presidents have placed on standards in terms of economic and national security grounds (15), even as others have been concerned about the erosion of communities and of students’ educational opportunities based on reforms such as school choice and voucher programs. The Supreme Court in Brown v. Board of Education (347 U.S. 483) made this explicit in noting, “it is doubtful that any child may be expected to succeed in life if he is denied the opportunity of an 2 Because the level of analysis for this study is the state, local school boards are treated primarily as organized interests who have the opportunity to advocate for policy reform at the state level. While local school districts have independent existences as policymakers, the policy innovations examined are statewide in nature. 2 education…it is a right which must be made available to all on equal terms” (493). As such, government is left with two directives, maximize quality in order to achieve economic security and the national defense, and promote equal access in order to secure civil rights under the Constitution. During the late 1980s and the 1990s, the continued dissatisfaction with education policy interacted with a second powerful policy trend: devolution. Devolution, or the process of shifting the focus of policy from a national government to a sub-national government, was a trend that gained support in many different governmental activities, including education. This was especially true during the Reagan Administration but carried forward into the Clinton era in large measure due to state efforts to reform education policy. As political leaders on the national level suggested a new approach to federal-state relations, states sensed a new empowerment in policymaking generally. 3 Therefore, when continued dissatisfaction with primary and secondary education led to a period of dramatic change in education policy, the national government was not at the forefront of this movement. With the central government turning increasing responsibility over to the states during this period, state governments were left to lead primary and secondary reform in the 1990s. This era of state-led policy reform underscored that state policy communities can vary dramatically and their choices often lead to very different outcomes when faced with similar problems. States varied in both 3 In this context, the term devolution might refer to an increased role in policymaking by states without a decreased role by the federal government, if the proportion of the total policymaking by states increases. I am not using devolution to refer either to a formal, constitutional transfer of power to states. Of particular note, the use here of devolution does not refer to a transfer of authority from states to local governments or school boards. 3 the number of policies they enacted and the types. While some states engaged in numerous policy changes ranging from education financing reforms, stricter teacher certification standards, increases in standards and assessments, and even school choice plans, other states were relatively inactive. The exploration of why and how these states varied in policymaking decisions in primary and secondary education is the central question of this dissertation. Evaluating Differences Between States While it is easy to recognize the pervasiveness of the national government and the national political culture by seeing the similarities in policy and social norms throughout the country, it is the differences between the state policy outcomes that hint at the diversity and uniqueness of the states. Certain characteristics render each state different from others, often quite different from nearby neighbors—given that each state is inarguably part of the same country. If these differences were small, such as the hours of operation for state licensing boards or emphasis on tourism campaigns, then this would matter little except to the residents of each state. Education policy, however, is seen as a central role of government. This dissertation explores the linkage between state governments, political culture, interest group communities, and mass public opinion with state primary and secondary education policy. The role of the structure of state governments, an important characteristic in its own right, will receive less attention. This will allow for examination of which statewide characteristics have the most impact on primary and 4 secondary education policy and, thereby, will provide some insight into the policy implications of increased state autonomy in a salient issue for national politics. The starting point for this analysis is the assumption that states are likely to have different policy goals and methods, because they represent different people and resources. In 2000, California was 68.6 times as large as Wyoming (U.S. Census 2012) during the period of this research. The African American proportion of the population in Mississippi was 120 times as large as in Montana (U.S. Census 2012). Of course, population size and diversity are just two of numerous ways states might be different. States varied in terms of liberalism in the mass public, minority diversity, and numerous other measures. Dissimilarity in policymaking should be expected if state governments are responding to the desires of their populations. It should not be strange to note that Louisiana (200) adopted over three times as many primary and secondary education policy innovations as Nebraska (66) during a period of reform, because those states differ in many ways other than policymaking (Hurst, Tan, Meek, and Sellers 2003). This is to say nothing of the variation in the types of policies from educational standards, to school funding reforms, and teacher training requirements. Whether measured on the micro or macro level, membership in different social groups, different ideological self-identification, and different experiences are expected to lead to variation in preferred policy alternatives. Because states are aggregations of the individuals they serve, differences in measured characteristics should be expected to lead to different outcomes. At the same time, however, states are also political institutions. If government is viewed as a collection of rules and enforcement mechanisms for a society, then states should be expected to empower political forces 5 within them, based on their access to this political power. Thus, the structure of the state policy community within the primary and secondary education policy areas should be expected to have an influence on the variation between states. Therefore, there is an expectation that state policymaking is a competition between organized and diffuse policy interests (as measured by the influence of the interest group community versus the influence of the size of various social groups and ideological identification). The lack of influence of state interest groups is therefore somewhat surprising. Overview of Dissertation Chapters The dissertation is comprised of seven chapters including this introduction, each serving a different purpose in developing the discussion of state variation in primary and secondary education policymaking. Chapter 2 is designed to show that primary and secondary education policy is an appropriate venue for measuring state policymaking differences. The chapter evaluates primary and secondary education both historically and during the period of state-led reform in the 1990s. The historical background of the chapter shows the evolving nature of federalism as it relates to education policy. The chapter shows that as of the 1990s, however, state and local governments still provided the great majority of educational funding in the United States. The chapter also introduces the competition between two approaches to policy innovation in public education. The approaches are: to advocate the goal of providing equal educational opportunity and funding to children, particularly minority children and those in disadvantage school districts, and advocating the goal of attaining educational 6 excellence through demanding standards and accountability for school districts and teachers. The chapter shows the role of ideology in these approaches. Chapter 3 defines and examines policy innovation, looking to classic literature on the policymaking system. The chapter shows a linkage between policy innovation, as defined by Walker (1969) and others, with policymaking in general (Kingdon 2003), as well as research into policymaking on the state-level (Berry and Berry 1990; Erikson, Wright, and McIver 1993; Gray and Lowery 1996; Hero 1998). The chapter shows that the pro-reform environment of the 1990s was a type of policy diffusion where there was a pressure for reform, coupled with an array of policy alternatives to choose. As expected by Kingdon (2003), the direction of this change, given that both a problem and a set of alternatives existed, was driven by political feasibility. This suggests that the determinants of policy innovation should be the same state characteristics found in research into state policymaking generally. Chapter 4 utilizes a multi-dimensional scaling (MDS) model to explore the similarities between states in the primary and secondary education policies they adopted during the 1990s. In examining the spatial configuration for policy innovation, it becomes clear that state differences are associated with both the number and types of policies adopted. Despite representing only a few of the policies adopted by states governments, school funding levels, mechanisms, and sources represent a great deal of the differences between states. Racial and ethnic diversity and ideology are two dimensions that shed considerable light on those differences. Interestingly, diversity leads not only to differences in policy choices but also to more total innovation. In those states with intermediate levels of diversity, the contest between ideologies as to which 7 innovations to adopt provides considerable insight (although this does not seem to be independent of other characteristics such as legislative professionalism and state wealth). This supports the notion that the interests of large social groupings and the ideological identification of state residents both have an important role in shaping state educational policy in the United States. Chapter 5 conducts a second MDS analysis to determine the differences between states in their primary and secondary education interest group communities. If interest groups are to provide an explanation for policymaking, it is important to first determine how state interest group communities differ. In order to explore interest groups, the chapter utilizes elements of the population ecology model for interest group representation as proposed by Gray and Lowery (1996). The population ecology model proposes that communities (in this case, communities of interest groups—but this approach is borrowed from ecology where it is used for biological systems) vary both in terms of density and diversity. Gray and Lowery (1996) found that interest group communities are seldom diverse unless they are extremely large. In the area of education policy, only a few states have diverse interest group communities, while others have only a handful of interest groups. The number of local school boards in the state dominates even the largest communities. 4 The chapter finds that population and the number of full-time state employees determines interest group density, while interest 4 Local school boards are treated as interest groups, because of two aspects of their roles. First, they are the employers of education professionals, who are responsible for providing a service. They engage in manage/labor relations much as other service industries under state and federal employment regulations. Second, local school boards lobby state governments for everything from educational funding to teacher training and required standards—essentially the reform issues considered by this research. 8 group density (along with the number of state employees per capita) is strongly associated with interest group diversity. Chapter 6 draws the previous two chapters together to explore the influence of interest group diversity on policy innovation. This chapter utilizes a configuration comparison to determine whether there is congruence between the two spatial models developed in Chapter 4 and Chapter 5. If the models are essentially the same, it suggests that something is influencing the models to behave similarly. However, the weakness of the correlation between the points in the two models suggests that the structure of policy innovation and the structure of the interest group community are very dissimilar. As expected from the results of Chapter 4, this suggests that the structure of the interest group community does not influence policy innovation. Chapter 7 summarizes the findings of the previous chapters, and explores their implications while tying the results back to previous literature on state policymaking. This will include an examination of the broader implications of the research and how the research interacts with the policy literature. This will involve a brief reflection on policy innovation as it relates to those who would distinguish it from the broader stream of state policy research. As with most policymaking in a democracy, policy innovation reflects both the biases and fears of the represented public, rather than just its aspirations. 9 Chapter 2 An Overview of State Primary and Secondary Education Policy and Policy Innovation When broadly defined, the education of children in the United States belongs to the small class of valence issues in American policymaking. The widespread debate over education policy stems less from disagreement over whether young people should be educated than over the goals, purpose, and methods of that education. Education policymaking claims to seek everything from political integration to economic opportunity. More importantly, education and education reform is the purported goal of advocates from many divergent political perspectives from teachers unions, such as the National Education Association (NEA), to the charter school movement (Mintrom 2000). Additionally, the agreement on the importance of education policy extends to periodic bouts of dissatisfaction with the quality of educational opportunities and outcomes. However, any perception of agreement falls apart along political and special interest lines when it comes to specific ideas about how to reform the educational system. At different times, educational reform proposals have come from many different groups, and a political divide appears in many of these proposals. Henig (1994) noted the “topsy-turvy” nature of education reform with conservatives calling for the most significant changes (3). While the opposing sides do not warrant each other’s true intentions, both liberal groups (ranging from civil rights groups to teachers’ unions) and conservative groups (ranging from home school advocates to policy think tanks and presidential commissions) have proposed policy alternatives to solve the problem of 10 children receiving what they agree is an inadequate education. Of course, the different groups do not agree that both sides care about education. Milton Friedman (1962), an early voucher proponent, asserted that public schools were largely run by and for the benefit of teachers unions, and were more interested in “schooling” than “education” (86). By this, Friedman meant that education took a backseat to promoting the interest of school employees. On the other side, Henig (1994) argues that the market metaphor and terms such as “crisis” and “choice” is manipulated in the education reform debate in order to radically change the educational system even at the risk of undermining schools’ roles in facilitating democracy. Henig (1994) can be seen as a response against Chubb and Moe (1990), who advanced a market-driven approach to education funding that they claimed would improve the quality of education. While the focus of policy proposals has changed over time, liberal groups have generally advocated desegregation, as well as more funding and more equal funding. These might be termed egalitarian or redistributive concerns. Conservative groups have focused less on the role of increased funding. Instead, their emphasis has been on using standards and accountability for schools and especially teachers to improve public education and to offer educational choice. Because these proposals have not focused on increased funding for public schools, opponents have seen this as neglectful. While both approaches would term educational choice as providing families with the chance to abandon failing schools, the connotation to each of the option to escape public schools is very different. 5 5 In the paragraph above, the terms liberal and conservative are not necessarily meant to connote psychological constraints on political behavior. Instead, they reference the 11 In the 1990s, the states became the primary policy venue for the debate over education reform. This period was the high-water mark of state autonomy in a number of key policy areas in addition to education as President Clinton continued the devolution advocated by Presidents Nixon and Reagan. While the United States government did not ignore education policy between the advent of Nixon’s New Federalism and the passage of the No Child Left Behind Act in 2001—in fact, the Elementary and Secondary Education Act was reauthorized every five years and President Clinton enacted his Goals 2000 plan in 1994—most of the key issues of the 1990s were debated on the state level. This led to a diverse patchwork of policy innovations that provided considerable insight into both education policy and potentially the political landscape of each state. Although the specific purpose of this dissertation is to examine education policy on the state-level, it is important to reflect on a history where the United States and the various states have redefined their relationship in this policy area many times. General Insights through Education Policy Even if education policy proved to be a unique issue in American politics, it would be worthy of considerable research. A majority of Americans are concerned about the quality of education in the United States, especially in urban school districts (Loveless 1997). Even if a large number of people are satisfied with their own districts, many researchers (Chubb and Moe 1990; Moe 1995; Mintrom 2000) have noted objective measures where students have fallen short of expectations. This is a salient policy and identification of the groups within their broader advocacy coalition as envisioned by Sabatier and Jenkins-Smith (1993). 12 a policy where state governments have a great deal of latitude. Therefore, education as a unique policy is of considerable importance. However, primary and secondary education policy provides an ability to look closely at several key influences over state policymaking. Primary and secondary education policy provides an ideal focus for any study of government policymaking on the state level for three reasons. First, education policy is salient for the American public. Most Americans expect children to have an adequate public education designed to prepare them to be socialized citizens and productive workers both for the benefit of those children and to serve the purposes of the society in general. Second, states have a great deal of autonomy in this policy area. Americans understand and accept that most government policymaking in children’s education has traditionally and continues to come from state and local governments (Wong 2004). 6 In fact, public school expenditures represented nearly 20% of the all state and local expenditures as of the mid-1990s (Moser and Rubenstein 2002). Third, education policy is fluid, with change on the state level occurring during a recent period of general policy decentralization. States were extremely active in policymaking in the 1990s, and while they were influenced by national trends, this did not minimize their autonomy. Even by the early 1980s, researchers were remarking on the unprecedented activity of state governments in education policy (Timar and Kirp 1988). This pattern continued throughout the 1990s with many policy innovations (Hurst, Tan, Meek, and Sellers 2003; Wong 2004). 6 While local school boards exercise considerable influence on primary and secondary policymaking, they are fundamentally creatures of state governments with the nature of their independence both regulated by state governments and a characteristic of states for study. Local school boards, in fact, represent much of the interest group community, as will be shown in Chapter 5. Chapter 5 evaluates the influence of the interest group community they dominate. 13 Although the discussion of public education was conducted in numerous policy venues during the 1990s, action on state policy reforms preceded the federal reforms seen in No Child Left Behind. In fact, the reforms of 2001 were in large part advocated as a nationalization of innovative reforms already in existence in many states. The 1990s showed some states engaging in numerous reforms, while others showed less enthusiasm for policy innovation. Even those states that engaged in considerable education reform differed sharply over the types of policies they adopted. Thus, education policy is an area where public beliefs about state autonomy and perceived importance are expected to cause states to act in the ways most consistent with their natural inclinations during a period when reform is taking place. Although recent discussions in education policy have focused on reforms directed in large part by mandates from the national government following the No Child Left Behind Act, the 1990s provide a useful snapshot of what state governments did on this very important subject when they were given a window of opportunity (cf. Kingdon 2003). This provides insight into the diversity of their policymaking preferences, and possibly the characteristics that influence this difference. A Brief History of Education Policy Grozdins (1966) noted, “In virtually no field does the complete body of law…have its source in [only] one of the so-called levels of government” (80), it was no longer reasonable to discuss state and federal powers as distinct. Despite this warning, it is fair to assert that education policy, more than most, has been thought to be a state and local responsibility. Wong (2004) suggests a pendulum between dual and cooperative 14 approaches to federalism in modern American policymaking history. While there is support for this view in history, it is an exaggeration to suggest that the federal th government’s role in education is new or even an innovation of the 20 Century. Dual and cooperative federalism might also be better viewed as heuristics for understanding the difference between a legal and policymaking view of federalism than explanations of the ongoing tension between nationalization and decentralization. Instead, a useful analysis might begin with the understanding that the national government has a smaller influence on education policy than most major policies of interest to Americans. This limited role is strongly influenced by American political history. With the national government’s limited constitutional role in supervising state th activities and limited budget through most of the 19 Century, there was very little direct influence on state education policy. While the Constitution gives Congress significant power in economic matters through the Commerce Clause, this was not initially seen as extending into education policy. The national government had no constitutional authority to require equality of any kind until after the civil war. However, the national government has sought to have an impact from the earliest days, especially when it comes to advocating educational opportunity. At first, a primary venue for this influence on education was the federal government’s role in the creation of states and territories. The importance of “place” as a conscious creation appears odd, given the perception of permanence of states and their boundaries, fifty years after the creation of Hawaii (cf. Dreier, Mollenkopf, and Swanstrom 2004; Burns 1994; Cox and Katz 2002). Nevertheless, state boundaries preserve past political calculations. Later, these boundaries can alter policy outcomes 15 when some event arises. Although common views on education were not the primary purpose of these boundaries, this fact does not lessen the impact of having certain interests drawn into or out of each state. The national government provided boundaries in order to mediate between or advantage interests. Behind borders, slave owners, silver miners, Mormons, apple growers, ship owners, and countless other groups were given a set of advantages and disadvantages in pursuing their policy interests. Government institutionalizes the meaning of place and those institutions have policy ramifications. Among these interests were many that favored or opposed policies toward the education, mis-education, or non-education. In some cases, the federal government took an active role in promoting education, while forming the territories that became certain states. The Land Ordinance of 1785 and the Northwest Ordinance of 1787, together often considered one of the few accomplishments of Congress under the Articles of Confederation, set aside land to provide for education in every surveyed township in the Northwest Territories (Hegreness 2011). 7 When these territories became states, the advantaged place of education had already been determined. Other states, primarily those bordering the Missouri Compromise line and southward were designed to create a hierarchical plantation society based on slavery. Regional differences in the attitudes about education by state elites were inevitable. Not surprisingly, unfavorable attitudes toward the education of minorities developed in this area of the country. Because state 7 The primary purpose of these acts was to establish a system to create parcels of land in the unincorporated lands north of the Ohio River and east of the Mississippi. This gave Congress the opportunity to create a system it deemed ideal, including such new concepts as straight line township and parcel boundaries and a community provision for education. 16 boundaries are not redrawn regularly, their influence locks ancient advantages into policymaking that temporary boundaries do not. Subsequent education policy in the states created from territories should be seen as an extension of the initial action that created them and therefore a federal influence. Once the states were created, however, education policy—despite early federal decisions to set aside land in what must have seemed like a boundless frontier or to re-enforce the political power of southern planters—prior to the civil war was left to the states (Tyack and James 1986). Unlike the federal constitution, state constitutions dedicated considerable attention to education policy. States established jurisdictions for education and mediated between local interests. While the period between the founding through the mid-Nineteenth Century was not a period of large-scale government regulation in any policy area by any level of government (and many schools were run on behalf of communities by religious or philanthropic organizations), states were active in distributing benefits to local schools. This was also the period when states were establishing large numbers of new schools and school districts. Just as the federal government created the boundaries for the states, the states authorized the creation of the boundaries within them and determined the students who would be educated within those boundaries. Because states enjoyed almost complete autonomy at this point, the actions of the states were largely unfettered (Tyack and James 1986). During and after the civil war, the federal government took a few steps toward developing its own education policy. Congress briefly established a Department (later Office) of Education in 1867 and took an interest in the literacy of freed slaves. Reconstruction also led to a short period of public education in the South (as each 17 southern state established a plan for free education). The ultimate failure of federal education policy during reconstruction shows—among other things—the difficulty the federal government has had in changing state political cultures (Franklin 1961, pg 110; Foner 1990, pg. 156-159). These reconstruction efforts to make education the centerpiece of egalitarianism (Foner 1990) shows a connection between crisis and policy centralization that would appear again in education policy when external threats would again lead the United States to education policy innovation aimed at improved standards. The creation of the Office of Education, along with the Morrill Act of 1862 (which led to the formation of land-grant agricultural colleges across the United States), th marked a high-water mark for federal education policy in the 19 8 century. The national government had made forays into both providing postsecondary education and education for the disadvantaged, but these were very tentative steps. Even so the national government began to retreat from its role in primary and secondary education in the 1870s (Eastman 1998; Tyack and James 1986). After the federal government’s role again receded, the state administration of schools provided such quality of education as was desired by the most powerful groups within each state political community. For the most part, states did a very good job. During the period soon after the civil war, the United States had a higher percentage of children in school, and the United States paid more for that education than the leading 8 While the Morrill Act was crucial both for its creation of much of the backbone of the American university system and as a massive set aside of real estate wealth for education, it is slightly outside the scope of this study of primary and secondary education. 18 industrialized countries of the day: England, France, and Germany (Tyack and James 1986). If funding and attendance are measures of performance—as they remain for many to this day—then the United States had among the best schools in the world. Two factors make this particularly notable. First, education spending in the United States was particularly uneven with African Americans receiving very little education spending. This means that the relatively high percentage of children receiving an education came from only a subset of the population (racially and regionally defined). Second, those Western European countries in the analysis were much more centralized than the United States. At least on average, the decentralized model appeared to work. The rest of the century saw a continued decline in federal interest, as education policy was left to the states. Fortunately, most state governments continued to provide financial support for education, even though government funding of the public sector in general decreased dramatically (Tyack and James 1986). State government support for public education was the “largest part of the public sector in a period marked by distrust of government” (Tyack and James 1986, 54). Still, education policy remained unequal with wide disparities between states. th Through the rest of the 19 century the importance of requiring universal attendance for secondary education was controversial, especially in the South and in some agricultural and ethnic communities in other parts of the country (Eastman 1998). Typically, state governments enacted laws suggesting that education was important, but they seldom required communities to provide education (Eastman 1998). While there was some impetus for reform during the progressive era (roughly 1890-1920), this reform did not change the venue or dramatically alter the level of governmental control for education 19 policy. The progressive impetus in education led to professionalization and a sense that those educational professionals should be left to promote educational improvement without governmental intervention (Timar 1997). State education departments consisted primarily of representatives of the upper-class and other elites until the 1960s. Therefore, the desires of those whose children were receiving the education were seldom considered. As such, publicly-funded education was often seen as job training for the poor, a re-enforcement of the civic culture especially amongst immigrant populations, and a state rather than national priority (Timar 1997). 9 Despite the progressive reforms, many state governments continued to require segregation by race into the 1950s (and even beyond on a de facto basis), and enforced existing racial and class distinctions within communities. Where civic leaders deemed education of ethnic and racial minorities less important or even undesirable, school-aged children received inferior educations. In the South, the segregated education of African Americans lagged behind even the generally poor standards of the southern states in general. As Eastman (1998) noted access to education was not independent of the need for the equal protection of the laws. Once again, education policy became a national policy issue when the cold war made it a national security issue. The 1950s and 1960s saw both an increasing role for the central government and the genesis of the two competing views for the purpose of education policy that dominated the remainder of the century. In 1953, President Eisenhower created the Department of Health, Education, and Welfare. Timar (1997) 9 Olneck (1989) argues specific efforts at civic education for immigrants were largely unsuccessful between 1900 and 1925. However, he does note that attending public schools did have a positive acculturation influence on immigrant children. 20 credits a growing sense throughout the 1950s that the United States had won the Second World War primarily through its superior use of science and technology, but that American scientific superiority had eroded in the years since the war. Conversely, the location of education policy within the confines of a cabinet department engaged in promoting public health and social welfare did not firmly establish the priority of federal education policy. The ultimate responsibility for the solidification of education policy as a federal priority can be credited to the launch of Sputnik, a Russian satellite on October 4, 1957. The Soviet Union, America’s cold war enemy had launched the first artificial satellite to orbit the Earth on a rocket originally designed to launch nuclear missiles. Within a single year, the United States had enacted the National Defense Education Act (NDEA). While much of the law provided aid to colleges and universities in an effort to provide advanced science and scientists for defense industries, Title III was designed to provide additional scientific, mathematical, and foreign language education for primary and secondary schools. In keeping with long-term trends in education policy, states were given the lead in implementing Title III. Though federal interest in education in the years since Sputnik has varied, the expectation that government should act to encourage academic improvement and excellence has not. 10 10 The importance of Sputnik in forcing the government to take the lead in promoting educational attainment can be found in the tenor of articles such as Smith, Charles E. (1983) “Sputnik II: Where are You When We Need You?” indicating a sense that some new focusing event was needed to return the national consensus to the promotion of education. If such articles are no longer found, it might be a commentary on history education. Even as late as 2010, President Obama sought to advance education funding by using the metaphor of the “Sputnik moment” as a time when education reform, aimed at excellence was needed (Obama 2011). 21 Even as educational excellence among the most gifted was seen as one key to winning the cold war, a national consensus also developed during the 1960s about the need to overcome poverty and discrimination in order to promote national security and economic growth (Kantor 1991). Among the assumptions was the need to equalize education spending in order to equalize economic conditions. Thus, the federal government’s primarily role was to redistribute wealth in the form of educational funding between the rich and poor. From these two veins arose the debate over whether education policy should focus on the government requiring high standards and accountability or whether American interests were best served by promoting educational equality. (Kantor 1991). The vehicle for the federal government’s increased role in promoting educational equality was the Elementary and Secondary Education Act (ESEA). This act, reauthorized several times since its initial passage in 1965, provided for a program of federal grants, tied both to the wealth of the school districts and the students the districts served. ESEA Title I provided funding designed to assist “local education agencies [such as school boards] for the education of children [from] low income families” (Public Law 89-10). Enacted at a time when the quality of most American schools was not yet questioned, the primary goal of the initial legislation was to provide the same quality to all schools. The increased activity of the federal government in education policy was similar to the centralization of governmental decision-making seen in most policy areas during the New Deal, Cold War, and especially in the Great Society eras. Designed to spread the advantages of most segments of American society to everyone, the legislation of this period should be seen as setting a goal of equity. 22 During this period, support for state control of education waned as the national government and most educational interests supported national solutions, while other groups began to distinguish between local and state solutions in seeking policy reform (Wong 2004). In a sense, states were too small to solve the large national problem and too large to represent community concerns. The preference for national solutions occurred at the same time as a general feeling that the educational system was failing students (Mintrom 2000), although some subsequent research would tie decreased overall performance to the efforts to provide a single national solution (Chubb and Moe 1990). The state role in education would not rebound until the height of the New Federalism era in the 1980s and 1990s. The 1980s and 1990s saw a debate emerge over the proper division of power within the federal system. Ronald Reagan ran for President in 1980 promising to return power to the states. Although the extent to which this occurred is still debated, states had increasingly come to be considered appropriate venues for policy change (Schram and Weissert 1999; Hanson 2004). The ongoing desire to reform the educational system within the states and the growing push for decentralization left state governments in a position to engage in education reform during the 1990s. Even with this large degree of federal control, each state provided different educational policy decisions to its citizens. Wong (2004) refers to this diversity as a result of the “permissive attitude…consistent with…dual federalism” on the part of the federal government (358). Of course, national government acquiescence to state policies because they do not run afoul of the national education policy is very consistent with cooperative federalism. Instead Wong (2004) makes the point that federal 23 education policy provides more latitude to states than other policy areas. Future chapters will explore the wide range of policy innovations designed to solve problems in education policy. However, these options range from standards, assessment, and accountability to teacher training, finance reform and even school choice options (Hurst, Tan, Meek, and Sellers 2003). Education Finance in the 1990s Given its history as an advocate of educational reform, a good place to look for increased federal involvement would be in the area of federal education spending. However, the evidence suggests the federal government controls very few aspects of education funding. First, most school financing still came from state and local sources throughout the 1990s (and even to the present). Second, most of this money (at least before 2002) came in the form of categorical grants designed to promote equality and access to education for minorities, those with disabilities, non-native speakers, and the poor and disadvantaged (Wong 1999; 2004). State and local authorities were responsible for the quality of the education their schools provided. While the state contributions varied during the 1990s, it remained between 45% and 49%. Local and private funds (with private funds by far the smaller component) remained more than 40% of the total, though the level was falling at the end of the decade. The relative levels of spending were consistent with the traditional view of states and local governments as the primary education providers. Therefore, the preferences and priorities of state governments remained crucial. Even if federal money was meant to encourage state and local spending, over 90% of education 24 spending resulted from state and local choices. As noted by the United States Department of Education, the entire federal contribution to primary and secondary education during the 1990s, including programs administered by all federal departments and agencies was roughly seven to nine percent of the money spent at all levels of government and the private sector for education (Hurst, Tan, Meek, and Sellers 2003). During the 1990s, the amount spent on students varied dramatically from stateto-state, but the average current expenditures rose from $6402 per pupil to $6925 per pupil in constant 1998-1999 dollars (Snyder and Hoffman 2003; Hurst et al.). 11 This reflected an average increase in spending for all of the levels of government in real terms, though much of the increase occurred during the second half of the decade. Instead, it is clear that in actual money spent, the federal increase was dwarfed by a corresponding increase on other levels of government. The difference is even more telling when the difference in the percentage of total state spending went to the increase in education funding versus the percentage of total federal government spending went to its portion of the increase. However, the national average did not tell the complete story. State spending on education per pupil varied dramatically across the United States. For the 1989-1990 school year, students in the state with the highest per pupil expenditure (Alaska) received $9819 per pupil, while the state with the lowest per pupil funding (Utah) spent 11 The total expenditures increased from $7135 to $8016 over the decade in 20002001, but the difference between the total and current numbers reflects money spent on debt servicing and capital outlays, such as building construction. 25 only $3339. 12 Thus, spending per pupil was over 2.9 times as high in Alaska. The discrepancy between the highest and lowest spending states decreased somewhat during the 1990s, showing that legal and legislative efforts during the decade to provide funding equity might have had some success. For the 1998-1999 school year, the top spending state (New Jersey) spent $10,145 per pupil, while the lowest spending state (still Utah) spent $4210 per pupil. However, with New Jersey spending 2.4 times as much as Utah, this remained a large gap. Even with the decrease in the difference, the states were far from the same in funding at the extremes. Even the data toward the center showed meaningful variation, the interquartile range for the data was $1825 in th 1998-1999—meaning the 75 th than the 25 percentile state spent almost one-third more (32.9%) percentile state (Hurst et al. 2003). States also have a wide variation in their funding levels and mechanisms. Although the average state government spent over five percent of the total personal income of its residents on education spending (Garand and Baudoin 2004), there was considerable variation. Two bordering states, Massachusetts and Vermont had the widest variation with Massachusetts spending only 2.6% of residents’ personal incomes on education spending, while Vermont spent 8.7%. The difference dwarfs the variation in per capita personal income (where Massachusetts was 40% higher, according to the U.S. Department of Commerce) and shows a difference in policy priorities toward education. However, state governments do not represent the full subnational support for education. In 2000, 15 states provided in excess of 60% of all public elementary and 12 The rate in Washington, D.C., was even higher than Alaska, but the District of Columbia has not been considered elsewhere in this dissertation, so it is excluded here. 26 secondary school revenues within their states (including Vermont), while 11 states provided less than 42% with the lowest being Nevada (Wong 2004). In most cases, local school districts are the primary source of education funding in states where statelevel support is relatively low. The primary source of education funding in states with low state government responsibility was the local school district. 13 Because federal money is often tied to specific eligibility requirements and programs, while state money can be used more flexibly, local schools often gain more latitude in policy implementation as state funding increases relative to federal funding. According to Hurst et al. (2003) most of the money that comes from state governments is in the form of general grants to schools based primarily on the number of pupils enrolled. Schools are then given the responsibility for carrying out the educational tasks assigned by the state. Federal money often derives from formulas designed toward funding equalization and aid to those students who live in poverty through the districts that have these students. With this in mind, education funding formulas, in addition to spending levels, clearly represent a significant potential state policymaking variation. Throughout the 1990s, state governments struggled with the need to provide adequate funding. Litigation designed to compel states to provide funding that was more equal between the wealthiest and poorest districts led states to seek innovations (or to resist them) in the area of school finance. 13 Eleven states lost cases in state supreme courts during the 1990s on school funding, including New Hampshire and Vermont whose state spending spiked from the lowest levels for the 1999 school year. As late as the 1998-1999 school year, the New Hampshire state government accounted for less than 10% of the revenue for public schools in the state. 27 Public Dissatisfaction with Educational Performance and Equity By the 1950s, many Americans were already dissatisfied with the quality of the educational system (Timor 1997; Kantor 1991). Much of this dissatisfaction resulted from the sense that unequal educational performance hurt both the moral and economic standing of the United States in Cold War comparisons with the Soviet Union and other European nations (Kantor 1991). Educational inequality was one of the underlying premises for the Brown v. Board of Education 347 U.S. 483 (1954). While the decision recognized that segregation denied equal protection even if the separate schools could be shown to be of the same quality, the inferior nature of schools for African American children in the South was widely understood. Apart from equal protection clauses in most state constitutions, many state constitutions reference a guarantee of efficient and/or adequate educational funding (Moser and Rubenstein 2002). Considerable funding litigation has been based on the need to achieve adequate funding—usually defined in terms of the funding differences between school districts until recent years when a greater focus has been made on outcomes. The question of equity becomes more complex with the addition of other factors making the education of some students more expensive than the education of others. Obvious examples, such as the complexity of funding the education of children with disabilities (Parrish and Chambers 1996) are well noted, but other students, who have different transportation needs, differential internet access, or even access to mentoring (Holland 1996) are less often mentioned in studies of educational finance equity. While solving these problems might be seen as the role of redistributive or 28 remedial federal programs, measures of outcome affect those schools where the most disadvantaged students live. The argument in favor of equality assumed a general awareness among educational specialists and professionals about how to provide quality education, given that sufficient funds were available to carry out the plan. However, the debate over the relationship between funding levels and educational effectiveness is far from settled (Moser and Rubenstein 2002). Certainly, wealthier districts are better able to recruit teachers, provide for facilities, and provide other services to teachers, such as additional training. Wealthy districts would also be able to provide services to students, such as extracurricular activities and tutoring and special education, which poor districts would be unable to provide. In any case, increased funding is generally assumed to be associated with better school districts by those who support educational equity. The 1980s and 1990s saw a shift away from the earlier focus on educational inputs, such as spending, and a movement toward performance outputs, such as scores on standardized tests, graduation rates, and teacher certification requirements (Hurst et al. 2003, vii). This was a recognition that educational equity did not only consist of spending equity but also in the ability to produce students who were learning. A Nation at Risk, a report by President Reagan’s National Commission on Excellence in Education, spurred many of these reforms. The commission blamed failing schools for putting the economic security of the United States in danger, citing high levels of adult illiteracy, students’ performance lagging behind students from other industrialized countries in 19 measures of academic achievement, and students who were not prepared for college courses or to engage in higher level thinking. The report also 29 stated that the trends were toward declining performance (National Commission on Excellence in Education 1983). By focusing on academic standards, A Nation at Risk disputed the previous assumption that public education was working in those places where the financing was adequate. This was a problem not at the extremes of American society but on the average. Unless financing was not adequate in most places—a possible contention—the process was to blame for the failure. While this was a federal report, it spurred action primarily at the state-level initially. Taking up the role suggested by Jewell (1982) and others, state became a laboratory for democracy, engaging in those reforms indicted by their predilections. Other education reformers disputed whether there was any one best way to educate all of America’s youth. Even if the same standards were appropriate everywhere, the process for achieving those standards did not need to be one-size-fitsall (Chubb and Moe 1990). Ladd and Hansen (1999) noted a shift away from plans to equalize school spending to plans designed to require quality education for everyone as measured by scores and standards. This meant a shift away from inputs to outcomes even among those who felt that some students were not receiving their fair share of educational spending. Instead, the argument was that some students were not receiving the quality education to which they had a right, no matter what the government spent. Education Reform: Policy Innovations in the 1990s This project utilizes a 2002 report by the National Center for Education Statistics (NCES) authored by David Hurst, Alexandra Tan, Anne Meek, and Jason Sellers for 30 most of the data included. Additional data on homeschooling came from reports by the Home School Legal Defense Fund. The NCES report, Overview and Inventory of State Education Reforms: 1990 to 2000 is a comprehensive discussion of education reform including numerous tables and charts detailing differences in policies between the states. In many cases, this data is arranged in a useable form, showing which states had adopted different policies at various times through the decade. In other instances, minor transformations were performed on the data, such as determining cut-points or gathering data that had been in list form. Education reform in the 1990s focused on four main areas: 1) standards, assessment, and accountability, 2) school financing reforms, 3) teacher training requirements and school resources, and 4) state support for school choice options (Hurst et al. 2002). From these four broad categories, it is possible to isolate 404 different policy innovations or changes of various degrees of importance. The great majority of these innovations are reporting requirements, assessments, and measures of educational performance in various years and in various subjects. In fact, 204 of the 404 policies fall into the category of standards, assessment, and accountability. Because of the great many measures of standards, assessments, and accountability (and because many of these measures are distinguished only by the year in school in which they were required), it is possible to collapse many of these categories, so that the total number of policy variables considered was only 336, with 136 of these measuring standards, assessments, and accountability. The second category, school financing reforms contained 51 policies and dealt with a range of issues including funding formulas, funding levels, and funding litigation. The third category (and the 31 second largest) was teacher training requirements and state resources. The 117 policies in this area included state funding for Head Start and Pre-K programs, teacher education standards, teacher testing and certification, length of the school day, compulsory attendance, and textbook selection policies. The final category was state support for school choice programs. The 31 policies in this area included open enrollment rules, charter schools, rules for private schools and school vouchers, and laws involving homeschooling. States varied dramatically in their adoption of these policies. The number of policies adopted ranged from states that had a very high level of policy innovation, including Louisiana (200 policies), New Mexico (197), Texas (183), North Carolina (182), and California (179). Five other states adopted fewer than 100 total education reform policies, including Nebraska (66), Iowa (74), Montana (87), Minnesota (95), and Wyoming (96). The mean number of policy innovations in a state was 140. Even among states with a large number of policy innovations it will still be possible to have a great deal of dissimilarity. With a total of 336 recoded policies, none of the states enacted even two-thirds of the possible alternatives. Of course, with this wide lens, it is impossible to determine which states are the most similar in the types of policies implemented. Chapter 4 uses multidimensional scaling (MDS) to look at state education policy similarities. This will be a measure of the proportion of policy innovations that states have in common. Although Hurst et al. (2002) and others might suggest broad categories for education policy as noted above, the MDS will allow each of the recoded categories to have its own influence on state similarity. 32 An Example of 1990s Education Reform: School Choice School choice is one of many major policy innovations states explored during the 1990s. Certainly, school choice programs were not the most commonly adopted of education reforms. However, these programs garnered a great deal of attention from advocates with widely divergent agendas. In addition, many of the debates over the purpose of education reform and the nature of public-funded education itself echoed through the debate on the issue. One result of the move away from assessing educational inputs to assessing outputs was the increasing reference to parents and students as consumers of an educational product (Chubb and Moe 1990). This is in stark contrast to the earlier view that students and their families were treatment subjects to be trained and acculturated (Timar 1997). While focusing on the equality of educational inputs, it remains necessary to find a mechanism for dispensing quality education in large increments. If the focus shifts to individuals and whether they are performing either to their potential or to predetermined standards, equality loses some of it salience. Arguably, students are competing to receive a better education than the next person. The school choice movement was not designed to improve public schools, except in the sense of requiring them to compete with the best schools surviving (Chubb and Moe 1990). Instead, school choice was designed as an opportunity for those who did not want to have their children attend failing schools. While controversial with those who believed that a majority of students would suffer if a select few were allowed to leave, school choice proposals were very popular in some communities (Mintrom 2000). 33 Using the metaphor of markets and choice, the rhetoric of education reform ceased to focus on equal educational opportunity as a civil right for every student. Instead, education could be seen as an economic choice between schools (cf. Henig 1994; Chubb and Moe 1990). This economic approach led to an inevitable application of other economic principles, such as the dangers of monopoly. This, in turn, suggested several methods for eliminating the “monopoly” of having students assigned to public schools. These plans to provide students and parents with educational options were generally called school choice. While school choice was far from the only educational reform advocated during the 1990s, it was probably one of the most controversial. Furthermore, support for school choice divided along political and social lines with the groups who had supported desegregation and educational equality generally being opposed to school choice, while conservatives were much more supportive. The general acceptance that many schools especially in urban school districts provided a poor quality of education led to political pressure for school choice. 14 School choice programs gained support for individual families as an opportunity to escape schools that were perceived to be dangerous and inadequate. Although wealthy parents have always had a great deal of latitude in where they would send their children to school, because they could afford to pay the high price of private schools, parents with more modest incomes have had little choice but to send their children to the school to which they were assigned. This assignment became more controversial and developed a racial element, after the United States Supreme Court began to 14 School choice is a less viable option in small communities both rhetorically and economically. Similarly, it is a policy that is more likely to be adopted where there is an economic infrastructure capable of funding alternative schools. 34 require busing to attain racial desegregation after Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1 (1971). Once court ordered busing programs began students could be assigned to schools outside their neighborhoods, yet those assignments remained mandatory. Many of the early public school choice programs, such as magnet schools were developed to obtain desegregation (or at least to end enforced segregation) in a less controversial manner than busing by allowing students to attend schools with special programs or focuses. These programs were meant to entice students of all races to a school without requiring anyone to choose them (Hurst et al. 2003). School choice programs were not developed as a means for desegregation (or segregation) though. Originally proposed by Milton Friedman in the 1950s, school choice was designed as a market approach to education. Unlike the current system, often called the democratic approach by advocates, school choice plans were supposed to provide parents with alternatives to sending their children to public schools within a defined region (Witte and Rigdon 1993; Mintrom 2000). Instead of providing education directly, under school choice plans the government provided the money it would otherwise allot for the education of a student to whichever school provided that education. Parents would then choose any school that was willing to be paid in that way. Sometimes, school choice plans included only other public schools, but these plans could also extend to charter schools and even private schools (Mintrom 2000). Conservatives had numerous political reasons for promoting these programs. First, school choice was a program with an economic justification that could show conservatives helping students who showed initiative. Second, school choice provided 35 conservatives with the opportunity to blame liberal interest groups, such as the teachers’ unions, for either opposing school reform (with the added benefit of decreasing the power of these groups if school choice was adopted and public schools lost students). Third, this was a policy that transferred educational funding to wherever students attended and did not require new outlays. The fundamental argument over school choice is whether market forces or democratic forces provide students with better schools (Witte and Rigdon 1993). As with most debates over public goods, the question became the best interests of individuals versus the best interest of the system as a whole. Even if a few students might be better off by leaving (a point not conceded by opponents of school choice), the question remained whether other students who remained in the old school would be disadvantaged. If these students who were left behind were minorities, the debate would also invoke questions about de facto segregation (Tedin and Weiher 2004). Because the parents who were most likely to utilize school choice for their children were also the most motivated parents when it came to maximizing their children’s potential, and because public schools were often left to educate the children who were less desired by schools outside the home district, it is difficult to measure the success of the school choice. However, it was initially popular with many parents and other reform supporters. Three types of reformers supported school choice. First, a large number of school choice supporters echoed Chubb and Moe (1990), indicating that when parents had the opportunity to choose their children’s schools, it would force schools to engage in activities designed to entice students. Schools would be forced to compete to keep 36 prices down and/or quality up. Additionally, parents would be able to choose schools with curricula consistent with their educational goals for their children. These goals might vary. Chubb and Moe (1990) argued that no single plan could fit the needs of everyone. A second type of reformer primarily considered the concerns of parents in failing schools. Parents in failing schools cared less about the market forces behind school choice. Instead, the parents sought to get their children out of dangerous or ineffective schools. These advocates also had more success in gaining favorable attention, because they often included minorities, especially in the case of cities such as Milwaukee and Cleveland, who provided cover against charges that school choice was a covert segregation plan. Finally, a third group advocating school choice wanted to have government pay to send their children to private schools or to allow home schooling (Mintrom 2000). These groups did not seek to leave schools because they were failing in their primary educational purpose, but because of religious or political opposition to public schooling. School choice found itself at the center of the debate between those who advocated educational equity and those whose focus was on education excellence. Many researchers and school choice opponents claimed that school choice would lead to more segregation and failing schools would be abandoned to their fate. While this seemed to be the conclusion of professional educators who observed parents, most parents claimed to place the greatest concern on academic performance (Tedin and Weiher 2004). However, if parents used school choice as an opportunity to select schools based on race, America would return to an era of parallel school systems where minorities and the poor would have less access to education. While Tedin and Weiher 37 (2004) found little evidence of segregation as an objective of school choice, other researchers (Smith and Meier 1995) argued that parents were consciously seeking to move their children to schools without minorities. Tedin and Weiher (2004) also found that academic performance was more important to parents of all races than the racial composition of the school. Many advocates of school choice appear to have misread the point at which the competition between school choice and the democratic system would begin. In a market system, as envisioned by Chubb and Moe (1990), the debate would be over which schools provided what parents desired. However, opponents of school choice did not compete against other schools in the market; they competed against school choice in the political (democratic) arena. Whereas Chubb and Moe (1990) and other reform advocates scarcely mention the interest groups involved in school choice policy, most of the major interest groups, such as the NEA and the AFT quickly developed policy positions opposed to school choice. Moe (1995) focused more on the ability of the bureaucracy as being able to limit the scope of change. Of course, the consensus of policy experts (Baumgartner and Jones 1993; Kingdon 2003; Lindblom 1959; Pressman and Wildavsky 1988; Sabatier and Jenkins-Smith 1993; Wildavsky 1964) sees incremental reform as being the normal manner of policy change. 15 15 Clearly, any While Baumgartner and Jones (1993); Kingdon (2003); and Sabatier and JenkinsSmith (1993) saw the opportunities for dramatic change rare, there approach is much more optimistic toward large-scale reform than incrementalists such as Lindblom (1959) and Wildavsky (1964) would predict. Pressman and Wildavsky (1988) suggests that adopting a policy is only the very first step in an evolutionary process that changes both the situation the policy is meant to address and the policy itself. This project does not address implementation. 38 dramatic policy change would occur not because of the market or bureaucratic support but because people within the democratic system decided to push for it. Essentially there are four types of school choice programs, as utilized by the various American states. The first of these is open enrollment. Open enrollment programs do the least to reform schools, inasmuch as they do not change the institutional framework of the school system. With open enrollment, parents can choose between participating public schools. Of the 32 states with open enrollment laws, the rules vary from requiring all schools to accept students who seek to transfer (up to some limit such as being full) in return for the state educational money assigned to that student, to laws that allow schools to participate in open enrollment based on a local decision. The second type of school choice program is the charter school. While the laws in the 36 states with charter schools varied significantly, the essential feature is that private groups, parents, educational institutions, or teachers are allowed to form a school with a degree of autonomy from the local public school system. While the degree of autonomy from local and state school boards varies dramatically, charter schools are allowed to develop curricula and themes that are different from traditional schools. In a sense, this is simply an extension of the magnet school idea, except that in most cases these schools are not developed directly by people within the school district but by outside groups. Individual charter schools, if not the laws that regulate them, are creatures of market force to a degree larger than public school choice and magnet schools are. The third type of school choice program, while only seeming to be incrementally different from the first two, is a change in kind. This plan, private school choice, 39 involves vouchers or some other program designed to send public money to private schools in order to pay all or some of the tuition for students to attend private schools. Programs, such as these, are extremely controversial, because more than threequarters of all private schools have a religious affiliation (Hurst et al. 2003; Broughman and Colacielo 1999). Additionally, these programs provide the greatest degree of independence for parents. As a result of opposition within the educational establishment and perceived constitutional problems with transferring public money to sectarian organizations, only three states have enacted voucher programs. This does not mean that government money is not spent to promote private education in other states. As Hurst et al. (2003) notes, most states allow funding for transportation, textbooks, and other educational supplies in sectarian schools. However, the three states—Wisconsin, Ohio, and Florida—were the only ones to help pay tuition (Hurst et al. 2003). A final type of school choice policy is homeschooling. All fifty states allow parents to educate their children at home under some circumstances. According to Hurst et al. (2003), 1.7 percent of all children in the United States were homeschooled. Surprisingly, the vast majority of states (41) have very few educational requirements for parents in order to school their children at home (Hurst et al. 2003). Additionally, many states provide help for parents who are attempting to educate their children at home, including educational supplies, the opportunity to participate in school activities, and the ability to take some classes within the school setting and other outside of it. The level of approval and supervision of parents required for homeschooling varies dramatically from state to state, but in all cases, homeschooling is seen as the most independent of 40 the school choice options. Most likely, it is the lack of required financing that allows parents so much autonomy. Mintrom (2000) provides evidence that the changes in school choice policy were, while not revolutionary, significant. Hurst et al. (2003) suggests that the direction throughout the 1990s was toward more educational choice for parents with more than half of all states enacting a reform during the decade that made it easier for parents to choose their children’s school. Thirty-six states passed charter school legislation between 1991 and 1999. According to Mintrom (2000), this majority showed that even incremental moves in the direction of school choice programs can lead to dramatic change, even though very few adopted voucher programs designed to pay public money for private school tuition. It also showed the limits of this type of reform. According to Mintrom (2000), initial incremental change led to further advancement, although this did not ultimately lead to a spread of private school voucher plans and unlimited home schooling as some advocates hoped. Certainly, parents have more choices where to send their children, even though the public system has not been transitioned into a market. This is more consistent with the goals of the parents’ reform groups, than the Friedman plan. The contrast between the rapid spread of charter schools during the 1990s and the school voucher movement provides some insight into policy diffusion. The most obvious element of the adoption of charter schools is the rapid nature of the diffusion. Within a decade, 36 states had adopted policies, while none had previously. The first state to implement a charter school plan during the 1990s was Minnesota. California was the next state to adopt a policy in 1992. Afterward, the policy was implemented 41 primarily by other states within the regions of those two states: the Great Lakes area and the West. Although other states, particularly in the South eventually adopted the charter schools, the regions with the highest percentage of states adopting the policy were the same as the regions of the initial adopters. Conversely, only three states chose private school voucher programs. Notably, Wisconsin took the lead. Mintrom (2000) credits this innovation to pro-voucher policy entrepreneurs. They brought together disparate groups and implemented a policy. Ohio followed with a program designed to mimic the Wisconsin plan—focused around Milwaukee—in hopes of helping students in the failing Cleveland School District. However, this program became bogged down in litigation surrounding the Establishment Clause of the First Amendment. By the time the Ohio program prevailed in the U.S. Supreme Court (Zelman v. Simmons-Harris, 536 U.S. 639, [2002]), the momentum for reform was passed. Conclusion Education policy innovation in the 1990s might best be divided into the four categories suggested by Hurst, Tan, Meek, and Sellers (2003): 1) standards, assessment, and accountability, 2) school finance reforms, 3) teacher training and school resources, and 4) state support for school choice options. Within these categories, there are dozens of different measures with a total of 404 identified. Although it is not possible to give every one of these measures the same detailed look as school choice, statistical research in later chapters will provide an overview into the states that engaged in the most similar policy innovations. By looking at the policies as 42 a group, it was possible to find two statewide characteristics that correlated highly with the number and types of state policy reforms. These similarities were racial diversity and state policy ideology. Given the historical review above, as well as the discussion of funding and school choice, it is not surprising that similarities in these two characteristics seem to be reflected in the similarities in state education policymaking. State governments have a long history of being the primary providers of education policy. While the federal government has taken a more active role in recent years, states have remained important players in formulating and implementing their own education policies. Because of this responsibility, education policy appears to be an area where states are most likely to reveal their differences. While the federal role in education funding has increased a small amount in recent years, during the 1990s, the federal government comprised only seven percent of public school revenues. State governments contributed 49% with most of the rest coming from local governments. State governments play a major role in establishing standards and curriculum, evaluating teachers and setting criteria for becoming an education professional. Thus, it should not be surprising that states would have a crucial role in education reform when these were the major topics of education reform in the 1990s. States were created by the differences between residents of the various communities within the colonies and later territories on the North American continent. Those states institutionalized their differences upon becoming states with boundaries and laws. While formed for reasons largely exogenous to education policies, these initial characteristics as well as elements that came later (such as most of their populations) have had a dramatic effect on the types of political communities in these states. 43 In the 1990s, many Americans grew dissatisfied with the perceived inadequacy of education policy. Some of this dissatisfaction came from those who saw a vast difference between the funding levels and educational attainment levels of students in some districts versus those levels in other districts. 44 Chapter 3 State Policy Innovation in the Context of Prior Research State policy communities provided the primary arena for the competition between education reform ideas in the 1990s. This competition was between two very different approaches. One of these approaches was an effort to promote equal educational funding and opportunity. The other focused on increased educational standards within public schools, along with standards and accountability for schools and teachers. States were introduced to a wide range of options. Of course, states also had the option of not engaging in reform, if they either did not find a set of viable alternatives or did not see a problem worth addressing. In addition, states were not required to sample from only one approach. 16 A few state governments, such as Louisiana and California, seemed to choose an all-of-the-above approach. This chapter looks to consider this policy innovation within the context of those state policy communities. While not unique, periods of widespread policy innovation are not the norm within policy communities. Previous research into policymaking has shown a general bias in favor of the status quo with only incremental change (Kingdon 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993). The selection of primary and secondary education policy in the 1990s was intentional—in order to examine state-led change—but this should not be construed as the typical condition of American politics. 16 A few states, such as Iowa and Nebraska engaged in very little education reform in the 1990s. Others, most notably California, chose many of the reform options. Chapter 4 will consider state variation in greater detail using statistical analysis. 45 Determinants of Policymaking Kingdon (2003) began his study of policymaking by asking, “what makes people in and around government attend…to some subjects and not to others?” (1). This becomes the fundamental question of public policy, when one considers that no “authoritative allocation of values” (Easton 1953), can occur unless the policy community recognizes certain subjects as meaningful concerns for government and places them on the public (governmental) agenda (Kingdon 2003). When looking at the national government, Kingdon (2003) found policymaking to be an interaction between problems, alternatives, and politics, each with unique considerations of timing and importance at a given time. When these three distinct streams came together, Kingdon (2003) saw an opportunity for policy change. 17 As shown in the previous chapter, those streams came together for primary and secondary education policy in the 1990s with many varied interests supporting different policy alternatives. Research into state political systems necessarily expands Kingdon’s (2003) question to explore why some political communities attend to certain policy problems or attend to them more often or more comprehensively, while others do not appear to pay much attention at all. This comparative research ought to focus on the characteristics shared by the states with the most similar innovations. While the spread of ideas might be a plausible explanation, it would seem likely that state governments have similar access to the knowledge. More likely, they vary in the political power accrued by the various actors, including politicians, interest groups, as well as other governmental and non-governmental institutions that influence policymaking within the state. Taken 17 Kingdon (1984; 2003) called these streams: problems, solutions, and politics. 46 together, this defines the state policy community. While policy communities emphasize the role of interest groups, the influence of organized interest groups is seen to vary from state-to-state. Therefore, it is reasonable to investigate policy innovation in a way similar to prior research into state policymaking. Walker (1969) argued that geographical proximity is the crucial mechanism through which policy innovation is transmitted. However, this ignores the ability of government officials, interest groups, and even private citizens to learn either actively or passively about policy alternatives available in other states or simply from policy experts when they recognize a problem. Instead, it is likely that the political feasibility of any potential policy innovation will be determined by support among those actors and groups with connections to state policymaking. If state policy innovations are selected due to the characteristics of state political communities, rather than through a process of learning from neighbors, then innovation can be explained in much the same way as other measures of policy output. While it is possible to debate the advantages of examining authorization versus appropriation, the great advantage of looking into policy innovation is the ability to diminish the influence of past decisions. Prior research has presented many explanations for policymaking decisions within states. Whether state policy research considers popular control and public opinion (Erikson, Wright, and McIver 1993), diversity and policy outcomes (Hero 1998), spending priorities (Jacoby and Schneider 2001), interest group communities (Gray and Lowery 1996), or policy innovation (Berry and Berry 1990; 1992), the study of state-level policymaking has generally recognized that governments in the American states choose 47 very divergent policies. Clearly, the implication of research on states is that states make distinct policy choices and the actions of policymakers indicate both different preferences and resources. This implication allows for a comparative look at state governments and their policy communities with the opportunity to improve understanding into what factors influence the consideration and adoption of policies. Defining Policy Innovation Literature on policy innovation distinguishes between innovation and invention. Walker (1969) defined a policy innovation as any policy that was new in a particular state, regardless of whether the policy had been adopted by any other state or how old the idea was. States were not required to create entirely new policies to be engaged in innovation. Instead, they needed only to adopt a policy. This approach was useful for Walker (1969), because it allowed for the study of diffusion, or the spread of policy innovation as a contagion where ideas are seen as caught through contact. 18 Walker’s definition of innovation appears to be generally accepted (ex. Berry and Berry 1990; 1992; Mintrom 2000; Gray 1973; Eyestone 1977), even by those who saw considerable importance for the independent effects of the internal characteristics of states in policymaking. Therefore, when a state or some other policymaking body implements a policy, and it has never implemented the policy before, then it is an innovation as far as the particular state is concerned. Any policy alternative, which has not been previously 18 This dissertation does not dispute the possibility that the spread of knowledge is a key component of innovation. However, the diffusion of political knowledge is sufficient across the country, and it would seem as likely that a legislator in any state would be aware of efforts in other states and not just the nearest. Instead, the focus of this project is on the selection among alternatives that are viewed as widely available. 48 adopted in a state, becomes an innovation upon adoption, even if it is the last of the fifty states to do so. In this way, policy innovation is studied for the cause of its spread. Building on Walker’s (1969) simple definition of policy innovation facilitates an operationalization of innovation as any identified policy alternative adopted by a state government. 19 Policymakers within state governments should be expected to be aware of many of the actions of other state governments. National organizations ranging from the National Governor’s Association and the National Conference of State Legislatures to unions and associations either representing government employees or lobbying state governments all act as venues for the transmittal of information. Additionally, the national government, national political parties, private citizens, and the news media might all act to bring the activities of one state government to the attention of other states. In a sense this can be seen as the environment in which state policymaking occurs, and the decision of legislators to choose particular states to emulate is internal to those legislators—with the aggregate decision of the state policymakers a function of the internal power structure within each state. If it is true that state policymakers are generally aware of the policy alternatives that are available in other states, as suggested by the many links between state political and bureaucratic leaders, then it is reasonable that all states share political knowledge. Instead, the real cause of diffusion is the shared pressure to innovate throughout the policymaking environment caused by the pro-reform environment of education policy in 19 If one assumes at least one potential policymaker within a state is aware of a given policy alternative, the fact that the policy alternative has not been adopted provides insight into state policymaking—if only to measure the non-centrality of the potential policymaker, but also the political feasibility of the alternative. 49 the 1990s. Rogers (1962), and others who developed the concept of the diffusion of innovation would not have been surprised that a community of 50 would be capable of quickly disseminating information given their resources. Additionally, two major problems exist when looking at regionalism as an explanation for policy innovation. First, the regional classification of states is largely subjective. Many would doubt a cluster analysis based simply on the geographic midpoints of the several states as a basis for assigning regions. Any governmental assignment of states into regions is done already aware of the histories of the various states and other similarities between the states. However, a Post Hoc creation of regions is bound to find areas of the country that share similar policy innovations, but it will be important to ensure that this is not the result of other factors. Second, states within a region are also the states most likely to share common characteristics from demographics to history. The implication of Walker (1969) and others is that states with little else in common would be more likely to adopt similar policies, because they are forced to learn and compete. Instead, regional explanations might reflect the likelihood that nearby states are the most likely to share relevant characteristics. Berry and Berry (1990, 1992) found exactly this. Characteristics of states are important over and above proximity in the decision to adopt outside innovations. Berry and Berry (1990), Mintrom (2000), and others suggested a linkage between internal and external factors. Essentially, states seek to learn from the alternatives selected by other states they perceive to be similar, with problems they perceive to be similar, or with whom the state policymakers believe their state is in competition (Berry and Berry 1990; 1992). Additionally, geographic proximity cannot explain the extent of policy innovation 50 or why states have different policies at all—provided all patterns are not just midtransition snapshots of what will eventually be universal adoption. Inasmuch, as all states except Alaska and Hawaii border another state, independent differences between states must either eventually stop the tide (making diffusion less important than those differences) or any policy adopted anywhere must eventually spread across the country (making diffusion patterns a relatively unimportant listing of the order). 20 Furthermore, the modern political and economic world in which the states exist argues against assuming that state policymakers learn only from what they can see across borders or learn from a short drive. States compete nationally and globally for economic advantage. Interest groups, legislatures, administrators, political leaders, and media all have nationwide as well as regional conferences and affiliations from which it is possible to obtain political knowledge. Therefore, it appears unreasonable to believe that state policymakers who seek solutions do not know about a wide range of alternatives designed to solve policy problems. 21 If it is assumed that policymakers find policy alternatives that were developed outside the region in which a state exists, then the question of why policies are adopted is not one of knowledge but of politics. Even the perception against which state (or states) any given state competes against or by which 20 I refer to the importance within the context of this dissertation. Research into which states are the most innovative or most likely to try something new, or the most diffusive, or most likely to get others to follow, would be interesting research topics in their own right. 21 The assumption of knowledgeable policymakers is not meant to imply an acceptance of the rationalist approaches of Lindblom (1959) and Simon (1986) over organized anarchy (Cohen, March, and Olsen 1972). Instead, the implication should be that whichever model is superior; the policymaker is not more likely to find an alternative locally than nationally. If the policymaker is more likely to select a local alternative, it is because of political factors and not increased awareness. 51 it is influenced becomes a question not of geographic determination but of constructed belief. Explanations for State Policy Innovations Rather than focusing on those passing outside pressures that lead states to adopt new policies, this study of education reform will concentrate on the underlying statewide characteristics that make different states open to the same types of policy changes as other states. In other words, what statewide characteristics are consistent with the choice of certain groups of states to make similar policy choices, while other states make very different choices? This study will endeavor to show these characteristics as substantively meaningful dimensions within the structure of state policy innovations. While such a study might reasonably focus on governmental characteristics, instead the emphasis will be on the citizenry. This examination builds on the considerable previous research into state policymaking. Previous research has argued in favor of several competing explanations for state policymaking. While innovation has been explained primarily in terms of competition and learning between states (Walker 1969; Berry and Berry 1990; 1992; 1994), this is only a tiny subset of all of the explanations for state policymaking. In order to link the study of innovation to the rest of state policy literature, it is useful to compare common explanations such as regionalism (Walker 1969; Berry and Berry 1990), and state wealth and competitiveness (Gray 1973; Mohr 1969), to explanations for state policymaking in general. 52 In this dissertation, the focus will be on three competing models. These include partisanship and ideology (Erikson, Wright, and McIver 1993; Berry, Ringquist, Fording, and Hanson 1998 for ideology) 22 , social diversity (Hero 1998), interest group communities (Gray and Lowery 1996). A closely related institutional effect is legislative professionalism (Squire 2000; Mooney 1995). However, legislative professionalism is closely associated with ideology and partisanship. Recently, Richardson, Konisky, and Milyo (2012) have shown that previous research has not done enough to show the linkage between ideology and legislative professionalism, particularly as it relates to citizen attitudes toward state legislatures and term limits. Erikson, Wright, and McIver (1993) provide a compelling approach to policymaking by associating government action with electoral pressures as measured by statewide partisanship and ideology. In essence, Erikson, Wright, and McIver (1993) suggest that the public achieves its policy ends through voting for its preferences over the long term. This is the most optimistic approach for democratic policymaking on the state level, inasmuch as policy outcomes relate to public desires (although partisanship would appear to be a distortion of public will provided people desire outcomes to 22 Berry et al. (1998) provides an excellent measure of ideology. It measures actual voting behavior of citizens and elites instead of self identification. Additionally, it allows for the selection of specific years or periods of time, such as all of those under study here, and it differentiates between citizen and institutional ideology. The results in this dissertation find citizen ideology to have more impact on policy innovation than institutional ideology and similar impact on interest group communities. Because the results for citizen ideology from Berry et al. (1998) are essentially the same as Erikson, Wright, and McIver (1993), for policy innovation, and because Erikson, Wright, and McIver (1993) create an ideology measure complimentary to their partisanship measure, it is useful to use Erikson, Wright, and McIver (1993). 53 building political institutions such as parties). 23 The major problem that exists with partisanship is the historical distinctiveness of the southern states. Because of their association with Democratic Party one-party regimes during the period from the end of reconstruction through at least the 1940s, historically Democratic states are often conservative and southern, while in other states the Democratic Party was usually the more liberal party. Therefore, it is reasonable to consider partisanship as not the cause of state policy outcomes, but (at least in the South) a result of efforts to maintain certain policy outcomes (Key 1949; 1956). While the parties in 1993 did not reflect a tactic to maintain the political status quo to the degree they had in the 1950s and 1960s, they were remnants created for past struggles and not the cause. While political parties are institutions, the structure of Erikson, Wright, and McIver (1993) focuses primarily on democratic theory and representation—the very antithesis of the argument that institutions, primarily organized interest groups, tend to obtain relatively favorable treatment as compared to diffuse interests (Edelman 1960; Shattschneider 1960). Hero (1998) presented a different approach to explaining state policymaking by suggesting a strong role for social diversity or the influence of the size of black and Hispanic minorities on policymaking. Hero (1998) suggested that the increasing minority population leads to bifurcation or a competition between the white majority and 23 For those who study public opinion, the term ideology is problematic as it relates to the general public with the widespread belief that most people do not possess the political sophistication necessary to have true ideologies (Campbell, Converse, Miller, and Stokes 1960) and fall short of being informed on political issues (Berelson, Lazersfeld, and McPhee 1954). Erikson, Wright, and McIver (1993) address the linkage between ideological self-identification and policy outcome and find a strong correlation. It might be best to assume some cancellation of errors in self-identification, but it might be more correct to assume the citizens are (as the authors propose the link to be) voting for the candidates they assume—correctly or incorrectly—to represent their partisan and ideological interests. 54 minorities (usually to the detriment of the minorities) up to the point where minorities become a large enough portion of the political community to force inclusion in political decision-making, usually in coalition with white ethnics. 24 This approach suggests that partisan and ideological competition is actually a manifestation of racial competition. While Hero (1998) does not consider the role of ideology, it stands to reason that the competing groups in each state would have particular policy preferences. Those groupcentered differences would account for differences between the most diverse states and those states where the influence of African Americans and Latinos was negligible. Thus, Hero (1998) suggested that racial and ethnic group interactions within the mass public of a state influences policymaking by providing three possible structures. First, policymaking might occur in the absence of diversity with an expectation of a shared agenda. Second, policymaking might occur in a bifurcated community, where conflict might occur between dominant and minority interests with the majority setting policy. Third, policymaking might occur within a heterogeneous environment where minority groups might form coalitions and have an input in policymaking. If the level and types of diversity in the United States have a geographic component, it should be expected that what might seem to be regional differences will really be differences in access to power common to a part of the country. Gray and Lowery (1996) examined the assumption that state policymaking was a competition between interest groups. Diverse (pluralistic) interest group communities are expected to have outcomes that represent wider interests and therefore lead to 24 A measure of minority diversity from the 2000 census including African Americans and Latinos proved to have a much stronger correlation with education policy innovations and was used instead of the Hero (1998) data. 55 more representative policy outcomes than communities dominated by a single interest. Interest groups are often seen as purveyors of political information (knowledge) and receptacles of political power. Because organized interests have been shown to have more influence than diffuse interests, only those interests represented by interest groups are seen to have significant influence (see also Schattschneider 1960; Edelman 1960). Gray and Lowery (1996) did not find strong evidence for the influence of interest group communities, but it is an explanation with enough support to consider. A different mechanism for investigating policy communities is the advocacy coalition framework (ACF) developed by Sabatier and Jenkins-Smith (1993). ACF does not model interest groups as existing in a state of nature with the strength of the community defined by its diversity. Instead, ACF sees interests groups as having specialized interests but existing within broad coalitions of like-minded groups. These groups cooperate when they share a common interest and rarely work against each other because they share certain values. One problem with utilizing ACF in studying state governments is the relative paucity of state-level interest groups as compared to the federal government. However, ACF might explain the mechanism for how interest groups that do not share all of the same policy objectives might work toward their mutual interest: for instance, it might explain why both school choice advocates and advocates of standardized testing would work together if they share common values. Both might want students to be tested: testing advocates in order to grade schools; school choice advocates in hopes that if public schools perform badly it will be a catalyst to give students the opportunity to opt out of the public school system. 56 Gray (1973) suggested that policymaking and innovativeness were tied closely to state wealth and electoral competitiveness. Similarly, size and wealth have been associated with the formation of complex interest group communities (Lowery and Gray 1998). This explanation would appear to link several of the approaches listed above. However, a few problems exist. Electoral competitiveness would appear to have either a smaller or distorted influence on policymaking during the 1990s for two related reasons. First, the Republican Party was considerably stronger during the 1990s than during the historical period before the 1990s including gains in places where the Democrats had been more successful (Bibby and Holbrook 2004). This might indicate that competitiveness would be in a state of flux. Second, a considerable part of this change occurred in 1994 and afterward—the period of the study. Because Gray (1973) studied a period when party control was seen as static, it is unclear what the effect of a major change would be. Nevertheless, the next chapter will take this explanation into account. On the other hand, state wealth should be related to the ability of states to fund education. Several conclusions can be drawn from the discussion above. First, correlations between the spatial map and the percentage of the population that is black or Hispanic will provide support for the social diversity explanation presented by Hero (1998). Second, ideology as discussed in Erikson, Wright, and McIver (1993) is not contradictory to the social diversity model (as an explanation for policy outcomes supportive of what the majority desires). 25 Third, the influence of the size of the interest 25 This majority is either white citizens of a state or the coalition of whites and minorities in those heterogeneous states where minority desires must be taken into account. 57 group community should be more influential as it is more diverse, though not simply when it increases in size. Fortunately, primary and secondary education policy during the 1990s allows for an investigation of these theories of state policymaking. Policy change occurred rapidly and the different possible directions of that change allowed for a great deal of divergence. When looking at a policy during a time when most states are active in policymaking, it is possible to look at both the decision to make specific reforms or not make reforms as a willingness or unwillingness to innovate in similar ways. 26 Thus, this is a study of why states choose specific solutions when problems are generally recognized, rather than during times of incremental change. In the context of Kingdon (2003), this is an examination of where the problem stream and the solutions (or alternatives) stream have converged, and the research focuses on where the certain solutions allow for the politics stream to converge as well. Leading researchers, such as Kingdon (2003) and Baumgartner and Jones (1993), suggest that much of all change that occurs within a policy area happens during these rare opportunities for reform. During a period of change, even the lack of innovation is important, as it will be reflect on the absence of action in areas where reform has created new policy options and in the continued presence of policy options that reform has replaced. This means that change was not politically feasible in those states where change did not occur, even though it was feasible elsewhere. Those states with a resistance to change in the 26 Conversely, a relatively static policy area would bias toward previous conditions at the time of the adoption of the policies. If little is changing, it should not say much about a state that it is doing nothing. 58 political environment of 1990s education policy are just as interest as those with openness to reform. Competitive Federalism and State Preferences While the primary focus of this dissertation is the evaluation of the research listed above, several other explanations of policy change are compelling. Mooney (2001) and Volden (2006) suggest a complex and dynamic relationship between information and innovation that should not be oversimplified into an assumption that learning leads inevitably to innovation. Mooney (2001) found evidence of anti-innovation effects in late adopting states when information on policies from nearby states is negative. This makes sense when it is considered that late adopting states begin as fence sitters and not all feedback is positive. Negative information would be expected to have a profound effect on the undecided and should bias against adoption, if learning is posited as the explanation for spread of innovation. Therefore, the failure of states to adopt policy should not be seen as a failure of information to diffuse. Volden (2006) showed that states with successful Children’s Health Insurance Programs are more likely to be emulated, while those with less successful programs are not. State actors do not just learn of the existence of policy alternatives, but also from the mistakes and problems of actors in other states. It follows that a state would make an effort to improve on past innovations. This would explain why all policies are not universally adopted eventually as well as providing evidence that states pay considerable attention to what other states are doing. 59 Additional evidence from the area of competitive federalism suggests that national and regional factors are more important explanations for policy outcomes than local explanations. This is because states compete for productive residents and seek to repel unproductive ones (Lieberman and Shaw 2000; Peterson 1981; 1995). Peterson (1981) suggested that local governments have limited resources and this necessitated an effort to expend most of their resources on the most productive elements of the community. As with municipalities, state-level policy provides perverse incentives within some areas whenever the national government does not mandate uniform standards (Peterson 1995; Lieberman and Shaw 2000). In those areas of public policy where there is no national policy, states are not forced to provide services to those who are less productive. This is because states with inadequate services for the poor, especially in the area of welfare, are able to export their poor citizens to those states that provide better programs. Some policies might attract productive citizens and capital through increased expenditure, but others have the effect of increasing poverty in a state through increased aid (Lieberman and Shaw 2000; Peterson 1981; 1995). This provides an alternative explanation for policy diffusion. Just as Berry and Berry (1990; 1992; 1994) showed evidence of diffusion in cases where the policy innovation was to increase revenue—and thereby to punish investment—only when other states were willing to do so, Lieberman and Shaw (2000) provide evidence that the policy of diminishing benefits should spread when other states are trying to repel the poor. The ramifications of research into competitive federalism are that states might not be learning to provide superior policy. Instead, states might feel the need to produce inferior policy from what they have learned from other states. Although this is a 60 problem most closely associated with the study of welfare policy and the so-called race to the bottom, it has been shown to influence areas such as inner-city public schools (Peterson 1995). Conversely, a reputation for good education policy provides some opportunity for states to draw productive citizens making education a more complex issue. 27 Of course, the productivity of citizens might be a subjective measure. Particular groups and communities might be disadvantaged due to public perceptions that are not supported by evidence (Gilens 1999; Jacoby 2000). This presents two difficulties for any research designed to uncover policymaking differences between states. First, the policy chosen should not be unduly influenced by the race to the bottom, but it should take into account the possibility of this explanation for policymaking. Education policy carries both the potential for elite support and the potential to reflect a concern against providing too many benefits to the poor and to minorities. Second, it is necessary to separate the influence of competition from the influence of information about policies. One way to do this is to choose a policy area within which success would tend to attract productive residents (even if it equally attracts others). Another option is to attempt to account for a potential attempt to repel the less fortunate. 27 Primary and secondary education policy provides a useful test for Hero’s social diversity approach (1998), because attitudes toward education policy might be influenced by whether those empowered by the state’s power structure to make policy view school children who need help from state education policy as members of groups with access to power. 61 Interest Groups and Innovation If interest groups do not influence policymaking, then a great many people have spent considerable time and money in a very misguided effort. Advocates have joined interest groups because of the perception that they make a difference, and researchers have studied them for the same reason. For many researchers, the answer to the question of why government attends to some subjects and not to others has been that interest groups are major players in setting the public agenda (Shattschneider 1960; Edelman 1960; Lowi 1979; Kingdon 1984; 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993). Chapter 5 will examine similarities within states for types of interest group diversity as well. Chapter 6 looks at whether interest group diversity is explained by similar characteristics to those that influence policymaking. In addition to their role as disseminators of information, a great deal of research has pointed to interest group influence in the formulation of policy alternatives. Heclo (1978) spoke of an issue network, a loose web within which interest groups and government strive for influence over policymaking. In this sense, interest groups are members of the same policy community as legislators and bureaucrats who are official representatives of the state. In this approach, policy alternatives should not be expected to form without the input or at least the reaction of these organizations. Interest groups are seen to have a powerful role in policymaking, because they help communicate policy alternatives to states, and because they are often active across state boundaries. Once a critical mass of interest group attention has been achieved, at least some states should be expected to consider the policies advanced by those interest groups with ties to their policymaking community. This critical mass might be a 62 few powerful interest groups or a broad cross-section of all possible groups, but policymaking should not be expected without the participation (or at least awareness) of interest groups. Interest groups should be expected to form in order to fulfill a demand created within the structure of state policymaking. Truman (1951) and Dahl (1961) proposed that interest groups were natural manifestations of the groups they represented with people joining to have their voices heard in an increasingly complex political world. 28 Those who suggested interest groups were unlikely to reflect group interests attacked this view of the interest group community as representative. Olson (1965) suggested that these groups were more likely to represent individual interests, because group interests suffered from the collective action problem of diffuse benefits resulting from concentrated efforts. Others (Shattschneider 1960; Edelman 1960; Lowi 1979) found that organized interests working to benefit specific, usually private, interests were effective in overcoming mass interests when they came into conflict—something they believed happened often. The effectiveness of interest groups in obtaining goals when they remain intact, whatever the result of that effectiveness, is widely attested by both those who advocate and decry interest group influence (Truman 1951; Dahl 1961; Edelman 1960; Olson 1965; Lowi 1979). A general complaint about the nature of interest groups is that private interests generally trump those of broad mass interests, because it is difficult to maintain general interests due to collective action problems (Olson 1965). Therefore, any research into innovation should not forget the importance 28 Thus, states with the most complex political systems should be expected to have the largest interest group communities. Population size and gross state product (gsp) can be viewed as either proxies for this complexity or the cause. 63 of understanding how interest groups come to be. Gray and Lowery (1993; 1996; 1998) have illustrated the wide variation in the size and diversity of state interest group communities. In some states, interest groups represent a complex network of groups, as suggested was possible within the policymaking community by Heclo (1978). In others, there are very few active interest groups, especially with respect to particular policies. 29 For the most part, the debate in the area of interest group influence has not been over whether interest groups have an influence, but over the impact and mechanism of the influence. Researchers who study national-level policymaking have differed on the implications of interest group influence. If interest groups are broadly representative of the general public, then they exist only as a positive force linking people to a government they might not have the resources with which to interact otherwise. However, many researchers have expressed concern that interest groups are not representative of the mass public, either in their membership or their purpose. Those who have expressed the most concern about the influence of organized interests have either worried that interest groups disproportionately represent the powerful at the expense of the weak (Schattschneider 1960; Edelman 1960) or that organized interests are able to co-opt government, transforming the subject of regulation into a client of the regulatory body (Lowi 1979). Because the benefit in actual policy outcomes for organized interests is widely recognized, it is necessary to determine the degree to which organized interests are out of phase with popular interests. One way to 29 Education policy also shows a lack of diversity within interest groups with local school districts and teachers unions representing a considerable segment of the entire community. 64 determine this would be to see whether the characteristics of states that lead to interest group formation are the same as those that lead to policymaking. If they are the same and work toward the same purpose, then there is little concern about the power of interest groups. If the characteristics are different, then it is crucial to determine whether immutable characteristics make organized interests the opponent of the interests of mass publics or whether parallel forces provide similar outcomes. Under most conditions, policymaking tends toward equilibrium with very little policy change occurring within defined scopes of conflict. Baumgartner and Jones (1993) suggested this was because of the constraining influence of interest groups who have defined relationships with government. So long as the accepted definitions of the debate over policy are accepted little changes. Similarly, Sabatier and Jenkins-Smith (1993) saw the policy community as a long-standing arrangement where different actors have assets to apply to policy problems, but rarely do circumstances allow for dramatic change. Both approaches share the underlying finding that interest group activity creates a policy community involving at least interest groups and government. Furthermore, every approach suggests an outcome that is different when interest groups are involved from the hypothetical case where they are not. Although interest groups are part of the policymaking community, they operate outside the structure of the government. Given the widely accepted result of the debate between pluralism and elitism, interest groups should not be expected to reflect characteristics of the states (or nations) in which they exist (Moe 1981; McFarland 1987). Instead, interest groups within a state are best modeled ecologically as creatures of a state, living within an environment, and nurtured by characteristics of the 65 state (Gray and Lowery 1993; 1996; 1998). An understanding of the interest group environment is seen as crucial to policymaking, and thus its structure is crucial to policymaking research (Gray and Lowery 1996). This environment is not just a product of happenstance. Interest groups should not be seen as existing as a characteristic of a state but because of the characteristics that enable formation. When researchers look at the state level, it is possible to see variation from stateto-state. This variation could be related to the strength and diversity of the interest group community in a state (Gray and Lowery 1996). However, Erikson, Wright, and McIver (1993) found evidence that a strong relationship exists between long-term public opinion within states and policymaking in state governments. Although these represent diametrically opposed strategies for understanding public policy, it might not be surprising to find that both work as explanations of policymaking. Forces leading to citizen mobilization into group membership among a subset of the entire population might be very similar to those that result in attachments to political parties, selfidentification with ideological positions, and a realization of where policy alternatives fit within the framework of those attachments and self-placements. In a sense, this reflects the debate over whether interest groups are manifestations of previously unorganized interests reacting to an increasingly complex government and specific societal problems as pluralists would suggest, with the interest group serving as a vanguard of broader political thought (Truman 1951). This approach would dispute Olson (1965) by suggesting that those groups that exist notwithstanding the difficulty in group formation and maintenance represent many who do not formally belong to the organized group but still hold positions similar to those held by members and are 66 influenced by the forces that led to the group’s formation. Conversely, interest groups might exist as at least partially exogenous creations of policy entrepreneurs (Nownes and Neeley 1996). Evidence that policymaking and interest group formation are caused by the same underlying processes would provide support for the pluralist view that interest groups result from the demand for their services rather than a supply of interested benefactors. In addition, there are several other institutional explanations of state policymaking, evaluating actors within state government as potential interests. Schneider and Jacoby (1996) present evidence that state bureaucracies can often ignore or circumvent the policy initiatives of elected officials and public opinion. Instead, bureaucratic administrators are influenced primarily by environmental conditions within a state, interest groups, and structural characteristics. Research on bureaucratic discretion in general has found a relationship between the similarity of goals of those legislators overseeing the work of government agencies and the agencies themselves (Volden 2002). Independent agencies have more discretion when government is unable to control them, such as conditions of divided government (Volden 2002; Epstein and O’Halloran 1994; 1996; 1999). On the state level, Volden (2002) found a relationship between bureaucratic discretion and legislative professionalism, in addition to preference alignment. The role of bureaucracies within the structure of state government is important, because it suggests an explanation for state-level policymaking that is a common explanation of national-level policymaking. Interest groups can influence governmental institutions differently, because these institutions do not always have the same agenda. 67 From the Schneider and Jacoby (1996) model, it is clear that bureaucratic administrators have a great deal of influence over policy outcomes. Therefore, interest groups appear to have influence not only on the elective, but also the bureaucratic elements of a state government. This means that effects commonly associated with interest groups in states might actually result from bureaucratic power. Bureaucratic autonomy should be expected to have some relationship with interest group manifestation as well as other explanations. Therefore, the research will take into account characteristics of the bureaucracy and the structure of the state government, in order to account for its receptiveness to interest group pressure and its insulation from the public. The nature of the scaling procedure used requires no prior assumptions, but an expectation of strong institutional effects would follow from past research. In education policy, state policy communities would be expected to vary considerably with respect to their interest groups. Some states would be expected to have large communities that provide support for a pluralist conception of policymaking. These states would be expected to have interest groups supported by non-institutional actors (primarily state teachers’ unions and local school boards. 30 On the other hand, some states had very few total interest groups, especially when local school boards were excluded. The few interest groups that were not related to school boards, employee unions and pension funds, and companies with business interests related to education, tended to advocate specific policy changes, such as charter school advocates (and charter schools themselves), or to support educational funding and 30 Research for Chapter 5 indicates that the largest state-to-state variation in interest groups stems from whether local school boards had registered to lobby state governments. State teachers’ unions, while undoubtedly with the potential influential, tended to act as a unified statewide body, rather than to register each local affiliate. 68 quality generally. A look at the specific groups is somewhat dispiriting for pluralism, suggesting strongly that many stakeholders are not represented within the policy community. The lack of non-institutional interest groups is consistent with research into interest groups that argue a collective action problem exists in interest group formation. Expected Influences on Policy Innovation and Interest Groups Many characteristics that are internal to the state policymaking and political community have been suggested as explanations for state policymaking decisions. First, Erikson, Wright, and McIver (1993) suggested that state policymaking is a function of mass ideology and partisanship. While others have debated this democratic approach, a linkage between policymaking and democratic governance would seem a positive development. More to the point, the mass interest might remain even if there are distortions caused by other influences. Hero (1998) and others have posited an explanation based on social diversity. Because race and ethnicity have been longstanding concerns in American society, it should not be surprising if these demographic characteristics help determine policy choices. If racial and ethnic groups do not always mobilize into organized interests, the potential for internal and external forces leading to recognized group interests remains. Both of these explanations are largely noninstitutional, with diffuse interests providing the predominant explanation. Additionally, numerous other similarities—from the structure of state government to economic conditions—have the potential to impact policymaking. These influences should include internal influences, such as the characteristics of the state and the structure of its government; as well as external pressures, such as the influence and 69 volume of organized interests. Factors related to legislative size and the powers of the Speaker, education spending, state employees, and the power of the governor, are all institutional explanations proposed by previous research but somewhat outside the scope of this examination. The density and diversity of the state interest group community is a possible explanation for policymaking as well. Because so much research has suggested a link between the size and structure of the interest group community and the types of policies that a government adopts, Chapter 5 will examine state characteristics associated with similarities in the structure of state interest group communities. If similarities in state interest group communities really influence similarities in state policymaking, then both the underlying structure of those communities and state characteristics related to that structure should be congruent (should correlate) between policy innovation and the interest group structure. Therefore, the research into interest groups should begin with many of the same characteristics. However, there will be a special focus on population and gross state product, because of the findings in Gray and Lowery (1996). 70 Chapter 4 An Empirical Look at State Primary and Secondary Education Policy Innovation Research into policy innovation has focused on both the characteristics of states and regionalism as an explanation for the spread of policymaking. With so many creditable explanations for the decisions of state governments, it will prove useful to look for an underlying structure to policymaking in the states. This chapter will extend previous research by looking at the structure of policy innovation in the area of primary and secondary education policy in the 1990s. By looking for the structure, it will be possible to examine the similarities between states in their decisions to engage in policy innovation. Provided that there is meaningful structure within the policy space for primary and secondary education, this structure will help illuminate the influences on policymaking by state governments. The 1990s marked a period of extensive policy change in state education policy, but this reform was not uniform across the United States. Examining the structure of this policy change will also provide additional understanding into how states vary when given extensive policymaking latitude. This chapter explores the similarities between innovations enacted by the states in order to uncover the underlying characteristics that lead to that variation. This chapter examines 404 policies related to primary and secondary education that were in force in at least one state during the 1990s. The policies can be divided into roughly four categories: standards and accountability, school finance reforms, state support and school choice options, and teacher training requirements and school 71 resources. Standards and accountability, the largest of these groups primarily accounts for student assessments and testing. Teacher training requirements and school resources accounts for state expectations for teachers. This includes rules for textbook selection and school hours, a curious amalgam. State support and school choice includes a range of topics from homeschooling to charter schools. Perhaps the clearest of the four measures is school finance reforms. This category includes the level, source, and mechanisms for education funding. The category also includes state support for early childhood education and special education programs. Hurst, Tan, Meek, and Sellers (2003), who gathered much of this research for the National Center for Education Statistics report, Overview and Inventory of State Education Reforms, 2003-020, characterize the differences between the states as considerable, with some states focusing on very different reforms than others. There are many possible reasons for these differences. Table 1 below shows the total number of policies adopted by each state, as well as the number in each category. 72 Table 1. Primary and Secondary Education Policies, 1990-2000, by State and Category State Total Standards School State Teacher Training Policies and Finance Support and Req. and School Adopted Assess. School Resources Choice Alabama 175 93 11 4 67 Alaska 117 59 10 12 36 Arizona 115 53 14 14 34 Arkansas 157 77 13 11 56 California 179 92 15 13 59 Colorado 113 61 11 12 29 Connecticut 150 65 19 12 54 Delaware 159 84 16 12 47 Florida 169 81 11 15 62 Georgia 164 85 12 15 52 Hawaii 111 36 12 10 53 Idaho 132 60 8 14 50 Illinois 149 68 15 12 54 Indiana 164 77 15 9 63 Iowa 74 13 8 10 43 Kansas 132 70 9 11 42 Kentucky 165 83 12 5 65 Louisiana 200 106 15 16 63 Maine 129 54 13 11 51 Maryland 165 96 15 6 48 Massachusetts 149 80 16 14 39 Michigan 125 64 14 14 33 Minnesota 95 33 9 16 37 Mississippi 128 59 8 9 52 Missouri 142 66 16 14 46 Montana 87 26 14 6 41 Nebraska 66 21 14 8 23 Nevada 146 75 13 11 47 New Hampshire 129 51 16 14 48 New Jersey 131 56 18 14 43 New Mexico 197 107 11 16 63 New York 150 80 19 17 34 North Carolina 182 86 11 18 67 North Dakota 106 44 13 10 39 Ohio 157 72 13 17 55 Oklahoma 158 64 9 10 75 Oregon 146 61 9 14 62 Pennsylvania 118 43 16 12 47 Rhode Island 132 51 18 13 50 South Carolina 161 80 5 10 66 73 Table 1. (cont’d) State South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming Total Policies Adopted Standards and Assess. School Finance 98 149 183 152 129 165 127 170 114 96 34 65 87 76 68 78 57 79 51 38 15 12 14 13 14 14 16 11 13 12 State Support and School Choice 8 11 12 11 8 12 12 10 15 11 Teacher Training Req. and School Resources 41 61 70 52 39 61 42 70 35 35 Table 1 illustrates the wide disparity between the number of policies adopted by individual states. A few states were particularly active, such as Louisiana with 200 policies adopted and New Mexico with 197. Nebraska was the least active state with only 66 policies. While the states also varied in each of the categories, school funding presents an interesting comparison. Here states varied between five (South Carolina) and 19 (Connecticut and New York) policies. In the standards and assessments category, the range is even more shocking with Iowa’s 13 policies adopted out of 204 a clear sign that this state largely opted out of this area of reform. Not surprisingly, the states that were most active overall were particularly active in the area of standards and assessments. Later in this chapter, it will be possible to use multidimensional scaling to evaluate each of the states and the similarity of the specific policies they adopted to obtain insight into why they are different. Many possible reasons exist for the differences between states in the adoption of education policies. One of these might be citizen ideology. Erikson, Wright, and McIver (1993) presented evidence that ideology influences political decision-making along a 74 dimension that is at least partially independent of partisanship. These citizens would be likely to utilize their system of beliefs to formulate opinions on major issues. Such that some citizens do not have a developed ideology, this might weaken any correlations found, but it does not preclude the possibility. Ideology might also work through the medium of a closely associated variable. Also, those who identify themselves as liberal might also align with groups they perceive to be liberal, even if a true ideology is not present. The Erikson, Wright, and McIver (1993) model would then suggest that the strength of political parties within a state (partisanship) should provide an explanation. Along with ideology, partisanship would provide a happy picture of state education policy, where citizens would receive the education policy they sought, given their beliefs of how government should provide education. In this case, the priorities would not necessarily be spending priorities, but choices between different potential solutions to policy problems. However, the nature of political parties evolves across time. Both major political parties saw a shift in the regions where they expected to receive the strongest support during the period of this investigation. Republican support declined in many Northeastern and Pacific West states, while it increased dramatically in Southern states where their electoral support prior to the 1960s had been almost non-existent. Arguably, this realignment did not result simply from the addition of voters at the center of state politics, while each party maintained its core support. Instead, this was a reordering of politics with the Southern Politics described by V.O. Key (1949) only a distant relation to the one described by Black and Black (2002). Nevertheless, 75 partisanship is a common explanation for policymaking that needs to be examined in research such as this. The intersection between American history and education policy that best points to a competing explanation for state policy innovation decisions is race and ethnicity. Arguments in favor of school choice, public school financing, and accountability—even when aimed at helping students in failing schools—often occur within the context of the perception that these failing schools are mostly black, inner-city schools (Mintrom 2000). If liberals have sought equity in school financing, it has often been to provide opportunities for minorities. Conservatives framed school choice as an option to help minorities as well (Moe 1995; Mintrom 2000). One indication the perception that schools with large minority populations are the only ones failing is not based entirely on reality is the high percentage of parents who have a favorable view of their local school district (Loveless 1997). However, Mintrom (2000) notes that one of the driving forces behind school reform in American cities is the perception that those schools do not serve their largely minority populations. If race has been proposed as a reason some Americans see a problem (cf. Gilens 1999, relating to Welfare policy), racial diversity has been proposed as a possible solution. Hero (1998) suggested that policymaking outcomes could be explained by the level of social diversity within a state and the type of coalitions that minorities might form to influence state policymaking. While the level of diversity within a state might not be determinative of the precise coalition arrangements derived, it should suggest a mass public that could be mobilized to address policy problems from the prospective of minority groups. Therefore, racial disparity in educational policy or 76 attainment might be perceived as a policy problem, but racial diversity might also be seen as providing a potential for policy innovation. Spatial Analysis using MDS Exploring the structure of primary and secondary education reforms during the 1990s should provide three important pieces of information, 1) the similarities between state policy innovations in this policy area, 2) the meaning of the major dimensions of this policy innovation, and 3) the key characteristics of states with the greatest similarity in policy innovation. As just mentioned, there is a strong rationale to support the belief that ideology and minority diversity constitute major elements of the policy space’s dimensionality. The first step to this process is to create an accurate spatial representation of policy innovation in primary and secondary education policy. In order to increase understanding of state policy innovation decisions, this chapter utilizes a spatial model derived from multidimensional scaling (MDS) using the technique of alternating least squares (Young and Lewyckyj 1996). This spatial analysis will reflect the dissimilarity between states in the types of policy innovations they adopt in terms of distance, while paying special attention to those external variables consistent with an underlying dimensionality of the data observed using the MDS. As states are more dissimilar, they will be shown to be more distant. This will create a map in policy space, while proving some insight into the implications of certain underlying characteristics of the various states. Then the states will be classified, using a cluster analysis to provide information on those states with the most in common in terms of their policymaking decisions. Thus, the spatial analysis will show both the 77 characteristics associated with state variation and those states that are the most similar, creating a picture to show why states made the policy decisions they made. Although measuring the difference between states in terms of policy innovation is of interest simply as a way to categorize state educational policy preferences, the spatial model also makes it possible to measure other characteristics usually associated with policymaking in order to draw conclusions specifically for primary and secondary education policy. Therefore, as Rabinowitz (1978) noted, the MDS model is exploratory, descriptive, and evaluative in nature. For instance, if the distance between two states on the coordinate axis for policy innovation correlates highly with the geographic regions or even geographic space, this will provide information about why states are similar. However, other considerations, such as racial diversity, ideology, or the interest group community might also influence policy innovation space, as has been suggested by previous policy research. Using the data, it will also be possible to look at how these states are similar in ways other than the types of policies they implement. All of the major theories of state policymaking might be examined in this way. This should shed some light on why states make the policy decisions they do, using the results of the study and previous research into state policymaking. Methods MDS is a series of techniques designed to estimate parameters in order to present a spatial representation of proximity data and enhance the understanding of the structure between objects (including possibly people, concepts, outcomes, and decisions), as well as assessing the fit of this proposed structure (Davison 1983). 78 Proximity data measures the distance between two stimuli based on how similar they are. Stimuli (in this case states) that have more in common with each other are measured as closer to each other than objects that are farther away in terms of Euclidean space. Therefore, the positive distance between two objects is represented by a measure of their dissimilarity. Because proximity data is used for MDS, it is necessary in turn to make certain assumptions about the data that describe how the data is interpreted in terms of distance. First, the distance between any object and itself (or some object in exactly the same location) is 0. If two objects are the same, they have no distance between them. For instance, two states that had made the same adoption decision for every policy under consideration would have the same spatial location. Second, the distance between any two objects must be non-negative. They are 0, if in the same location and have increasing distance as they have more differences. Third, the distance from point A to point B is equal to the distance between point B and point A [D(a,b) = D(b,a)], also called the symmetry property. In other words, the distance is treated as the space between Detroit and Chicago where it is possible to go back and forth, and not the distance between Tuesday and Thursday and Thursday and Tuesday. Fourth, D(a,b) plus D(a,c) ≥ D(b,c) , also called the triangle inequality property. 31 MDS provides a solution for the spatial representation of these distances in the minimum number of dimensions necessary to provide a reasonable fit between the data in a matrix of dissimilarities between objects. This is analogous to determining the 31 This is less stringent than the straight line rule that D(a,b) + D(b,c) ≥ D(a,c). Instead, the space is like a triangle, where the distance between any two sides is greater than the distance of the third side assuming all angles are positive. 79 spatial distance between cities by measuring how many miles they are apart, and then further defining their location by measuring the distance from several cities and determining where a city must be located, if it is the given distances from the other cities (Kruskal and Wish 1978). Given the distances provided by the dissimilarity of objects, the goal of MDS is to provide a configuration of data in the minimum number of dimensions, so long as the representation of the data still maintains a close relationship to the observed distances. Because the goal of MDS is spatial representation, the utility of the model declines as the number of dimensions becomes more difficult to represent graphically (usually no larger than two or three dimensions). As noted in Bartholomew, Steele, Moustaki, and Galbraith (2002), it is always possible to find a perfect fit for a matrix with n objects in (n –1) dimensions. This would simply be the distance of every object to every other object. However, this type of perfect fit is not the goal of the process, because it provides no more information than the original measure of distance. MDS provides the most information when a balance is struck between the closeness of the approximation of the representation and the minimization of dimensions. Then, a researcher can look to directions or dimensions within the space that represent properties or characteristics of data and/or clusters of data points could represent a set of common characteristics. In principle, MDS is done by constructing a table of distances, then applying a formula designed to provide a result closely approximating those distances in a defined number of dimensions, utilizing a measure of goodness-of-fit to show the degree to which the results are like the original table (Bartholomew et al 2002). This involves adjusting the locations of the stimulus points to minimize the difference between the 80 distances as shown on the graphical representation and the input matrix. The map shows the configuration of the stimulus points in a manner designed to maximize the 2 goodness-of-fit, while the measure of goodness-of-fit (usually R and SSTRESS) 32 reveals the correspondence between the mapped locations of the stimuli and the original input data matrix. The program used to conduct the MDS in this study (ALSCAL) utilizes an alternating least squares procedure to fit the model to the data (Young and Lewyckyj 1996). Additionally, it is important to note several limitations to the fixed nature of the MDS results. All of these are a result of the arbitrary nature of where the configuration is placed in space. First, a reflected or mirror image of the coordinate points produced by the MDS will represent the matrix just as well as the original result. Second, adding a constant to every point (moving the entire configuration in any direction) provides an equally good representation of the proximity data. Third, multiplying each value by a constant (increasing the distance between the objects, but not the relative distance) does not result in a worse representation. Fourth, the data may be rotated without changing the quality of the representation (Bartholomew et al 2002). In other words, the configuration of points can be seen as a connected whole like a child’s toy made from interlocking pieces that might be moved from place to place in order to be better observed but with its structure unconnected to any fixed point. Therefore, the MDS output is standardized and represented by a varimax rotation of the points to maximize the variance of the coordinates within the space created by the coordinate axes. 32 2 R measures the proportion of the variance explained by the model. SSTRESS is actually a badness of fit measure, so as it decreases, goodness-of-fit increases. 81 Even though the coordinate axes do not represent any particular substantive dimension, it is possible to uncover data about hypothesized relationships. One approach used in this study is to regress external variables on the point coordinates and using the results to embed a meaningful dimension into the data. Because the discussion above points to several explanations of policymaking, it will be possible to look at these substantive dimensions to see their correlations with the point coordinates (as suggested by Rabinowitz 1978). A second approach is to conduct a hierarchical cluster analysis to group similar stimuli. This process begins by considering each stimulus (or state) as an individual group. In principle, the next step would be to join the two states that are the most similar to define a new group with the midpoint of the two stimuli defining the location of the group. This process continues with continued iterations of the two nearest groups joining to create new groups at their weighted midpoints until only a pre-determined number of groups remain. This process can be illustrated using a dendrogram that measures the relative similarities between stimulus objects. The goal of this process is to find groups that have meaningful similarities. Both of these steps will be employed in this analysis. Data This paper requires the use of three types of data. First, the policy innovation data is derived primarily from Overview and Inventory of State Education Reforms, 2003-020 with additional information from the Home School Legal Defense Association (2006). Because nearly half of the provisions concern different standards and assessments (often reflecting different years where student performance is assessed), 82 this data is recoded. As a result, the MDS reflects 336 policy innovations, and is a 50 X 50 matrix of policies each state shares with each other state. States with a greater number of shared policies (similarities) are deemed to be more alike and closer spatially. Once the MDS provides a spatial representation, it will be possible to examine the meaning of substantive dimensions within the policy space illustrated in the MDS. Figure 1 will show the pattern in policy adoptions that most closely aligns with the policy space. Figure 2 will look for other state characteristics that form an underlying dimension with the MDS result. This search requires data related to several key theories for state policymaking mentioned earlier. A few of these variables were seen originally in Politics in the American States (Gray and Hanson ed. 2004) and checked against their source data. While measures such as legislative professionalism (Squire 2000) and fulltime state employees (which figured heavily in Jacoby and Schneider [2001]) are included, the primary focus was on data related to partisanship and ideology: two measures of a democratic linkage between public policy and the preferences of the mass public. Therefore, the data from Erikson, Wright, and McIver (1993) for state partisanship and ideology is used. However, the measure of minority diversity comes from taking the proportion of African Americans plus Hispanics from the 1990 U.S. Census. This measure is similar in intent to the minority diversity approach envisioned by Hero (1998). Regional and population data also use U.S. Census data and are designed to divide the states into roughly equal-sized historical regions. The paper will also consider measures of Gross State Product from the Bureau of Economic Analysis, though this will figure more heavily in Chapter 5. The next step will be a 83 regression of the major variables examined in Chapter 4 as independent variables to explain each dimension. Finally, the chapter will also report a hierarchical cluster analysis using the coordinates from the MDS. So the results of the MDS will become data as well. Spatial Results of the MDS The variety of explanations offered for state-level policymaking in the United States would seem to suggest a possibility of a high dimensionality for the MDS solution. If each explanation has merit, then there must either be a correlation between those explanations or many dimensions, each with some explanatory power on a dependent variable. However, the scree plot in Figure 1, on the next page, provides evidence that the most appropriate number of dimensions was two. 33 The improvement in the SSTRESS scores is over two times greater going from one dimension to two than going two dimensions to three. In two dimensions, the results for 2 the SSTRESS measure of fit provided by the ALSCAL program was .15131. The R was .89. Because of the many explanations for state policymaking, a two-dimensional solution provides some interesting information by itself. In fact, it might suggest interrelatedness for the many competing explanations. For instance, the competing explanations of race and partisanship might have an underlying relationship. Other explanations might share characteristics with ideology. 33 Instead of eigenvalues as in a factor analysis, a scree plot for MDS shows decreases in the value of the stress measure of fit (Kruskal and Wish 1978). When interpreting a scree plot, it is necessary to look for the “elbow” or point where the dimension to dimension decline in stress values decreases dramatically, and then levels off. 84 Figure 1. Scree Plot of SSTRESS for Policy Innovation .3 y .2 .1 1 2 3 4 Dimensions Figure 2 (on page 87) shows the two-dimensional configuration for state policy innovation in primary and secondary education. 34 34 The dimensions shown provide an The initial results reported in ALSCAL are corrected to require the size of both dimensions to be the same. This does not change the correlations between the points on the coordinate axis and the substantive variables, but it adjusts the slope of the substantive variables regressed on the stimulus points. 85 interpretation of the coordinate locations of each of the states related directly to the policies they adopted. The results have been rotated to maximize their variance on the axis and adjusted so the range of the data on each dimension is the same. They offer an opportunity to analyze the data based entirely on the policy similarity between the states. While all states are represented by open-faced circles, each state has been highlighted for reference purposes. The first thing that seems clear is that the three largest and most diverse states (California, Texas, and New York) all appear in the bottom-right of the configuration, while smaller, more homogenous states (Montana, Iowa, and Utah) are in the top left. Iowa and Montana are among the least active reform states, although Utah was toward the high end of the middle of the pack. 35 Similarly, Iowa and Nebraska adopted few policies and are near the extreme left of the map. In general, it is easy to see that the states on the left-hand side of the policy innovation map are less diverse than those on the right. The solid line in Figure 2 has a slope of -0.086 and correlates at .918 with the total number of policies adopted by each state. States on the right-side of the map are also more widely dispersed than those on the left. The policies associated with the second dimension (the Y-axis) are less clear. The dashed line represents the strongest of these dimension—the number of school funding policies adopted by each state. The slope of the dimension is 4.874 and its correlation with the state coordinates is .597. The lower-right of the map is the most widely dispersed area. 35 Perhaps, Utah’s location is evidence that the upper-left is not simply a measure of non-innovators. However, it is close to the center on the left-right dimension. 86 Figure 2. Stimulus Points with Dimensions related to State Policy Innovation 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Note: The solid line represents number of policies (r = .918). The dashed line represents school funding policies (r = .597) The next step is to determine appropriate substantive variables to regress on the stimulus points to uncover meaningful relationships. Table 2 shows the correlations between the stimulus points and several substantive variables proposed and described in the previous chapters. Significant correlations (at p < .05) are shown in boldface. As 87 might be expected, given the many explanations proposed in previous research for the causes of state differences in policymaking several variables show a moderate to strong correlation. Aesthetically, partisanship and ideology, the two components suggested by Erikson, Wright, and McIver (1993) are both significant and provide dimensions that diverge enough to show two different effects. However, minority diversity provides the strongest correlation. Additionally, many researchers see the issue of race as the cause of the historical one-party system in the South (Key 1956; 1949; Black and Black 2002). Because racial politics pre-exists partisan politics, and partisan politics has been suggested to be an institutionalization of racial politics (Key 1949; Hero 1998), replacing partisanship with minority diversity as a substantive dimension in policymaking would be reasonable even if the correlation for minority diversity were not very much stronger. Table 2. Correlations between Substantive Variables and Stimulus Points Variable Minority Diversity Interest Groups Population Per Capita GSP State Emp./10,000 N Correlation (r) Variable Partisanship .703 Ideology Gubernatorial Power .366 Speaker Power .558 Total Size of Legis. .590 .261 Legis. Professionalism 50 Correlation (r) .562 .499 .325 .269 .253 .496 Table 2 shows the correlations between the substantive variables. While all of these variables have some correlation the policy space for state primary and secondary education policy innovation, some of these are much stronger than others. For the theoretical reasons shown earlier, racial and ethnic diversity makes sense as a variable. Although legislative professionalism and per capita gross state product (gsp) are strong 88 variables, ideology stands as a variable that is more closely related to a fundamental characteristic. As noted by Richardson, Konisky, and Milyo (2012), legislative professionalism is closely associated with ideology. While the authors merely suggested that ideology influences attitudes toward professional legislatures, ideology is a more basic state characteristic and ostensibly a temporally prior one. In fact, Table 3 shows a regression model suggesting the strong relationship between ideology and legislative professionalism in the data used for this analysis. Table 3. Regression Model to Examine Legislative Professionalism Scores Variable Ideology Partisanship Minority Diversity Interest Groups Population Per Capita GSP Constant N = 48 Coefficient -0.629 -0.013 -0.010 -0.017 -0.001 -0.134 4.213 Standard Error 0.240 0.165 0.172 0.075 0.001 0.242 0.766 T-Score -2.620 -0.076 -0.056 -0.229 -2.663 -0.582 5.499 P-Value .012 .940 .955 .820 .011 .622 .000 2 R = .560 There is also a strong correlation between population and legislative professionalism (Squire 2000). Increased population and legislative professionalism could suggest increased influence for interest groups, but the measure of the interest group community shows only a moderate to weak relationship between interest groups and policy innovations in primary and secondary education (Gray and Lowery 1996). As part of the interest group explanation, interest groups, population, and legislative professionalism will be treated in more detail in Chapter 5. At this point it is useful to 89 focus primarily on the relationship between policy innovation in primary and secondary education and ideology and minority diversity with the understanding that the formation and effect of interest groups will be considered at a later point. Figure 3. Stimulus Points with Dimensions related to Minority Diversity and Ideology 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Note: The solid line represents Minority Diversity (r = .703). The dashed line represents Ideology (r = .499) 90 Figure 3 shows the configuration introduced in Figure 2 with the axes for policy innovations replaced by regressing minority diversity and ideology, in turn, on the stimulus points. Minority diversity has a slope of –0.242 and ideology has a slope of 3.964. Given that the stimulus points were determined prior to the slope, it is interesting to see how and whether the characteristics of individual states are reflected by these OLS lines. The correlation between minority diversity and the stimulus points is .703, while the correlation between ideology and the stimulus points is .499. By highlighting several of the states, it quickly becomes clear that the dimensions make sense in terms of what they are explaining. However, a final step will serve to validate the minority diversity and ideology dimensions. For each dimension, an OLS regression with the dimension using the other highly correlated variables from Table 2 will test the explanatory power of minority diversity and ideology. Owing to the regression result in Table 3, legislative professionalism will be excluded. Table 4 shows the regression for minority diversity while Table 5 shows the regression for ideology. Table 4. Regression examining variables associated with Dimension 1 of Policy Innovation Variable Partisanship Ideology Minority Diversity Interest Groups Population Per Capita GSP Leg. Profess. Constant N = 48 Coefficient 0.016 -0.006 0.022 -0.003 4.77 E-05 -0.003 0.012 -0.975 Standard Error 0.006 0.010 0.006 0.003 2.11 E-05 0.009 0.006 0.493 2 R = .630 91 T-Score 2.595 -0.582 3.504 -1.001 2.261 -0.307 2.070 -1.975 P-Value .013 .564 .001 .323 .029 .761 .045 .055 While there are many significant variables in Table 4, minority diversity is clearly significant even with the strong and similar partisanship variable included. States such as California and Texas with a great deal of diversity are spatially distant from those such as Montana and North Dakota with very little diversity. Three southern states, Georgia, North Carolina, and Texas, which had seemed to fall within the same region on the map as California and New York, now fall within a group that has many other southern states. With only a few exceptions, the results show that states with a great deal of minority diversity enact education policy innovations at a much higher rate than those with little diversity. Table 5. Regression examining variables associated with Dimension 2 of Policy Innovation Variable Ideology Partisanship Minority Diversity Interest Groups Population Constant N = 48 Coefficient -0.028 -0.145 -0.002 0.002 -3.600 E-05 -0.145 Standard Error 0.009 0.179 0.007 0.003 2.170 E-05 0.179 T-Score -3.188 -0.812 -0.244 0.499 -1.646 -0.812 P-Value .003 .421 .808 .620 .107 .421 2 R = .580 However, states that differ in terms of ideology also have very different agendas for education policy innovation. As the low correlation between minority diversity and ideology (and the width of the angle between the two OLS lines) suggests, there is little or no ideological effect on states from increased diversity. The substantive dimensions also emphasize the wide dispersion of states within the area with both a great deal of diversity and a liberal mass ideology. 92 Even though it is possible to make general substantive conclusions about minority diversity and ideology from the maps of policy space, a few exceptions serve as reminders that the coordinates are a measures of shared policy innovations. While Mississippi is diverse enough to expect policies similar to its western neighbor, Louisiana, Mississippi is much nearer Tennessee, Utah, and Idaho. Indiana, a state sandwiched between Michigan, Illinois, and Ohio, has adopted policies similar to California and Louisiana, rather than its neighbors, who have fairly similar policies. Nevertheless, these are exceptions and it might be possible to explain each of these: perhaps, conservatism is more central in Mississippi than race (Hero [1998] might argue that minorities in the state have little influence); perhaps, the influence of nearby Chicago in the minds of Indianans leads to greater concern over minority diversity than the states’ population would indicate. While investigating the substantive dimensions has provided considerable information about the policy space, a more systematic grouping of the states could provide a better analysis for individual comparisons. Results from the Cluster Analysis The next step is to look at the results of the hierarchical cluster analysis to determine whether the results indicated by the dimensions make sense in terms of how the individual stimulus points group together. The best method is to utilize a dendrogram to indicate the cluster arrangement along with the iterative steps in the clustering algorithm where states grouped together. The results show the order in which the states were clustered with each other, showing the iteration in which they were linked. The most similar states are those whose lines connect the soonest (the 93 furthest left). It should be noted that any branch could be rotated, so the proximity of non-connected branches does not indicate closeness. It is possible to choose any number of clusters in a hierarchical cluster analysis by moving from the right to the left on the dendrogram; however, selecting four clusters allows for three larger groupings, while creating an interesting though smaller cluster. Table 6 shows the members of each cluster. As can be seen, three of these clusters are reasonably large, while Cluster #3 is small in terms of the number of states. Cluster #3, however, includes two of the largest and most diverse states (California and Texas). Table 6. Cluster results from Hierarchical Cluster Analysis of MDS Stimulus points Cluster #1 Alabama Arkansas Florida Georgia Kansas Kentucky Maryland Mississippi Nevada North Carolina Oklahoma Oregon South Carolina Tennessee Utah Virginia West Virginia Cluster #2 Alaska Arizona Colorado Hawaii Idaho Iowa Maine Minnesota Missouri Montana Nebraska North Dakota Rhode Island South Dakota Vermont Washington Wisconsin Wyoming Cluster #3 California Indiana Louisiana New Mexico Texas Cluster #4 Connecticut Delaware Illinois Massachusetts Michigan New Hampshire New Jersey New York Ohio Pennsylvania Cluster #1 primarily includes southern and border states. These states are located in the upper-right quarter of policy space primarily indicating support for 94 conservative policies—though the states to the extreme right figure to have been busy policy innovators generally. Most of these states are diverse, although there are a few exceptions. Cluster #2 primarily contains states from the Midwest and West, although it stretches as far east as Maine, Rhode Island and Vermont along the upper tier of the United States. For the most part, these are much less active policy innovators. Cluster #3, including California, Texas, New Mexico, and Louisiana (in addition to Indiana) were all active policy innovators in both conservative and liberal policies. The first four states mentioned are among the most diverse in the United States. New Mexico and Louisiana also border Texas. Cluster #4 contains mostly Midwestern states bordering the Great Lakes and Northeastern states. These states primarily enacted liberal policies. Interestingly, there is a high correlation between membership in this group and states with a priority toward what Jacoby and Schneider (2001) particularized benefit policies. Nine of the ten states in this cluster range 1st and 14th on their scale of particularized benefits to collective goods, including the top five (553). While Cluster #1 and Cluster #2 show some difference with Cluster #2 the most favorable toward collective goods of any of the clusters, the difference between the others and Cluster #4 is stark. Cluster #3 is slightly more slanted toward particularized benefits than Cluster #1 and Cluster #2, but the difference is far less stark. Next, Table 7 shows descriptive statistics for the characteristics of each cluster. This should provide insight into how the states in each cluster are different from the other clusters. In a few clusters (such as Cluster #2) the states vary considerably on some variables. To the degree the cluster reflects similarity in policy innovation; this 95 suggests a lesser role for the given characteristic in determining primary and secondary education policy in the state. Table 7. Descriptive statistics for the Characteristics of each cluster (Medians) Minority Ideology Partisanship Predominant Region Diversity (%) Cluster #1 21.59 -16.80 12.83 South Cluster #2 8.74 -14.96 0.43 West and Midwest Cluster #3 34.68 -17.02 13.78 CA, TX, NM, MS, IN Cluster #4 18.65 -7.63 4.75 Northeast and Midwest All States 17.69 -14.30 7.11 Note: Minority Diversity is the percentage of the population comprised of African Americans and Latinos. Ideology is % liberals minus % conservatives (all regions had more conservatives). Partisanship is % Democrats minus % Republicans (all regions had more Democrats). Predominant Region shows the region where states are located or the individual states when this is more informative. The next step is to look at each of the clusters in turn, applying what is observed in Table 7 and elsewhere to draw conclusions about the clusters and state policy innovation in education. Cluster #1 is comprised of 17 states, thirteen of which are in the South Census Region. In addition, Kansas borders Oklahoma (which is in the South Region), and the final three states: Utah, Nevada, and Oregon border each other. This includes all of the states in the South Region except Louisiana and Texas. The membership of this cluster suggests that the unique nature of the South in American politics (as suggested by Key 1949 and others) continued through the 1990s. None of the other large clusters is dominating this much by states in a single region. In general, these are states with a relatively large amount of minority diversity, and were notable for being controlled by the Democratic Party at least until the 1990s despite being ideologically conservative. While this dominance was passing away in the 1990s, 96 Georgia did not elect its first post-reconstruction Republican Governor until 2002 and several state legislatures remained in Democratic Party hands until after the period of this research. Cluster #2 is comprised of 18 states divided mainly between the West Census Region and the Midwest Region (but including three New England states from the Northeast). However, twelve of these states are in a contiguous swath across the northern tier of the United States starting in Wisconsin and running to Washington State (dipping as low as Missouri) These states include the great plains states (called the “West North Central” division by the Census Bureau) except Kansas. In addition, the Cluster includes Arizona, three New England states: Maine, Rhode Island, and Vermont, and Alaska and Hawaii. These states have several things in common. First, they are all primarily rural (with the exception of Rhode Island). On average they were the most Republican in their partisanship, though they varied greatly in their ideologies. The only states that do not fit one of these definitions are Hawaii, Maine, Vermont, and Rhode Island. These states also had the least minority diversity. However, they were typical of the country as a whole in terms of ideology on average with a great deal of variation among them. Cluster #3 is a cluster of five states. Four of these states are in the far south of the United States and have the largest proportion of minority diversity. The fifth state, Indiana, does not appear on the surface to be very similar to the other four states. Two of these states, Texas and California, are both populous and extremely large geographically. Louisiana and New Mexico border Texas and have extremely diverse 97 populations. 36 These states vary dramatically in terms of partisanship and ideology, but they all have at least some significant minority diversity, with the states other than Indiana being more diverse than any other cluster. These states were the most active innovators. Cluster #4 is a cluster of ten states. Three of these states are in the Midwest and border the Great Lakes: Illinois, Michigan, and Ohio, while seven others are in the Northeast. In general, they were more liberal than the average for the country, slightly more Republican, and their minority diversity was average. In years past, they would have been termed typical states, comprising what had been the industrial heartland of the country: New York, New Jersey, Pennsylvania, Ohio, Michigan, and Illinois. As would be expected, these states, having slightly liberal to liberal mass ideologies and a percentage of minority diversity that is typical of the United States produce primary and secondary education policies very different from more conservative states with greater diversity. While more liberal and more diverse than Cluster #2, these states are in many ways more similar to Cluster #2 then Cluster #1. These states engaged in more innovation than Cluster #2 states, but in general the policies were more liberal. This is further illustrated by the very strong tendency of these states to finance what Jacoby and Schneider (2001) called particularized benefits such as welfare, hospitals, health care, corrections, transportation, and employment security. These are policies that affect people on an individual level. Notably, when education is conceptualized as a particularized benefit, it entails providing access for individual children to good schools. 36 It is unclear to what degree the influence of French dialects and other Caribbean immigration in Louisiana leads to the same outcomes as Spanish and Central American immigration in Texas, New Mexico, and California. Another possibility is the location of two small states near a state the size of Texas just leads to Texas-like policies. 98 When conceptualized as a collective good, it involves improving the quality for the benefit of society. Taken together, the cluster analysis results provide support for the influence of minority diversity and ideology, while introducing regionalism as an influence. This is especially true as it relates to the South and those large rural areas from the Mississippi River to the far side of the Rocky Mountains. Implications This chapter has sought to investigate the policy space related to state-level primary and secondary education in order to find what might influence the structure of this policy. By using an MDS map, it has been possible to suggest that policy innovation is structured along two dimensions: minority diversity and ideology. Because they provided a weaker theoretical explanation or did not correlate as strongly with the policy innovation data, other explanations of state-level policymaking provided less insight in the area of education policy. By looking at the nature of the clusters derived from a hierarchical cluster analysis, it was also apparent that regional explanations remain strong—though this regional explanation might not exist solely along Census Bureau defined regions, because one of the clusters stretched primarily across the Great Plains all the way to the Pacific Ocean along the Northern tier of the United States. The hierarchical cluster analysis showed that states with similar policy innovations typically had the predicted ideologies and levels of minority diversity. Extreme levels of ideology or minority diversity appeared to overcome the effects of the other variable (as was the case in Cluster #2 and Cluster #4). However, this analysis 99 ignored several substantive variables that appear to have a relationship with interest group formation. These variables were also correlated at least moderately with policy innovation, and it will be the primary focus of Chapters 5 and 6. This chapter argues that the dimensionality between minority diversity and policy innovation is strong, it existed prior to any of the other explanations of policy innovation (as well obviously as any of the policy innovations of the 1990s), and led to one of the other variables with a strong correlation with policy innovation, partisanship. In other parts of the country, the presence of minorities (measured by minority diversity) might not have been as important as other factors. During the 1990s, states engaged in many policy innovations designed to improve primary and secondary education. However, the direction of this reform appears to have been set by state characteristics that had existed for many years prior to the reform period. The cluster analysis provided considerable insight into the value of ideology as an explanation. The states in Cluster #4 were considerably more likely to also be the states that have spending priorities toward particularized benefits as measured by Jacoby and Schneider (2001). All of the states in Cluster #4 were on the liberal side of the policy innovation space, but they varied dramatically in the number of policies adopted. However, the Jacoby and Schneider (2001) data clearly illustrate a difference in the types of policies to which these states usually give their appropriations. Nine of the states within the cluster support are among the 14 most favorable toward particularized benefits. Only Delaware, the smallest and most diverse of these states is outside the top 14 states in particularized benefits. Meanwhile the states in Cluster #1 appear to be the most conservative states, and have a slight bias toward adopting 100 policies aimed at standards and accountability. The final two clusters differ between those who have adopted the “none of the above” approach (Cluster #2) and the “all of the above” approach (Cluster #4). While this chapter does not consider interest group diversity, it provides considerable insight into state policy innovation. The similarity between the dimensions for state policymaking activity and minority diversity is a strong finding with a correlation of .918. Ideology also provides an insight into policymaking, particularly when it comes to school funding policies, although this might be through an intervening variable such as legislative professionalism. Nevertheless, state mass ideologies are most likely older than state political institutions. Chapter 2 argued that state borders trapped residents within political communities. Although the citizenry has renewed many times, this chapter argues that citizen characteristics remain a powerful influence over time. 101 Chapter 5 An Empirical Look at State Primary and Secondary Education Interest Group Communities Kingdon (2003) began his study of policymaking by asking, “what makes people in and around government attend…to some subjects and not to others?” (1). In a democracy, the simple, reassuring answer would be that policymakers attend to subjects salient to their constituencies, either as a noble public service designed to achieve citizen’s preferences or in the effort to avoid losing a future election. However, three obvious factors limit the ability of the populace to influence particular decisions made by policymakers. First, constituents often have unformed opinions about issues and little knowledge of which policymakers are responsible for them (cf. Campbell, Converse, Miller, and Stokes 1960). Second, policymakers often do not have clear ideas as to the desires of their constituents; much less what they would find minimally acceptable (Schattschneider 1960). Third, Gray and Lowery (1996) note that the electoral connection is limited by the fairly long length of time between elections (1). Therefore, the question arises, what is the mechanism (or system) for maintaining a permanent connection between policymakers and constituent interests? This chapter explores the size and structure of the interest group community for primary and secondary education, looking at the characteristics of states that are associated with education interest group formation. Although it will be tempting to draw conclusions about the broader interest group community from this look at the groups influencing education policy, this chapter will focus entirely on those interest groups with 102 an interest in education policy. 37 This exploration will look both at the number of registered education policy interest groups within a state and the similarity between the different types of interest groups shared between states. Thus, the chapter will examine two characteristics of interest group communities—termed density and diversity respectively by those who look at the communities ecologically (Gray and Lowery 1996). 38 Finding similarities between states for density and diversity will provide insight into which states have the most similar interest group communities. Furthermore, this information will test the association between interest groups and policymaking. In order to increase understanding of state interest group communities, this chapter represents the density and diversity of each community. This will require two models. The first will be an ordinary least squares regression model showing those characteristics of states that lead to large (or dense) interest group communities. The second will be derived from multidimensional scaling (MDS) using the technique of alternating least squares (Young and Lewyckyj 1996). This spatial analysis will reflect the dissimilarity between states in the types of primary and secondary education interest groups engaged in lobbying the state government of each state. The dissimilarities will be represented in terms of distance, while paying special attention to those external variables consistent with an underlying dimensionality of the data observed using the 37 Major theories of interest group influence from perspectives as diverse as Lowi (1979) to Heclo (1978) to Sabatier (1993) envision the connection to occur between groups taking an interest in a specific policy or policy area. This argues against a claim that the non-education policy interest group community affects education policymaking. 38 Although density and diversity are conceptually different, the paucity of active interest groups within some states will inevitably lead to a lack of diversity. The number of education interest groups in some states was less than ten. Not surprisingly, the diversity in those states is often low as compared to states with a large number of interest groups. 103 MDS. As states are more dissimilar, they will be shown to be more distant. The advantage of using the MDS will be the ability to uncover not only diversity but to compare the types of diversity between states. 39 The MDS will create a map in interest group space, while providing some insight into the implications of certain underlying characteristics of the various states. 40 Next, the use of external variables will attempt to determine the cause of the dissimilarities between states. Finally, the states will be classified, using a cluster analysis to provide information on those states with the most in common in terms of the types of interest groups. Thus, this chapter will show both the characteristics associated with variation and the most similar states, creating a picture of the interest group community. Determinants of Interest Group Formation and Diversity The debate concerning the influence of organized interests is one of the most resilient in political science. This debate was fueled by the large increase in the number and organizational resources of interest groups in the period after roughly 1950 (Nownes and Freeman 1998; Gray and Lowery 1996). While the importance of specific interest groups, especially in national policymaking is widely recognized, the influence and the structure of the interest group community has been widely debated (Truman 39 Instead of focusing only on whether an interest group community is diverse, this will allow for a look at the structure of that diversity. Borrowing from the metaphor of population ecology, this will not just determine the number of species that live in a region but also their type. 40 Density might have been explored in this way, except that by its nature, the model would have had only one dimension, because the distance would be determined solely by the size of the community. The number more or less of interest groups formed in each state would determine the distance. 104 1951; Schattschneider 1960; Edelman 1960; Dahl 1961; Olson 1965; Heinz et al. 1993). Dahl (1961) was one of the first to recognize the role of interest groups in local politics, but the premise has been the same on the local, state, and national level—interest groups are designed to provide the policymaking community with access to a point of view. Interest group formation is typically seen as a result of community resources more than as a result of the number of possible interests (Bischoff 2003; Gray and Lowery 1996). Gray and Lowery (1996) attribute interest group formation to a state’s economic capacity as measured by gross state product (GSP)—analogous to gross domestic product on the state level. This measurement is useful for two reasons. First, GSP measures the total output in goods and services for a state in a given year. Therefore, it provides the standard measure of the resources available to support interest group formation as well as all other activities within a state. Second, simple algebra demonstrates that it is a measure of average individual wealth (as measured in per capita GSP) times the population of a state. Therefore, GSP serves as an interaction term between average individual resources and population size. If economic development is related to interest group formation, it should be seen in a correlation between GSP and interest group formation. However, it should be noted that the bivariate correlation between population and GSP was .993 in 1998, while the correlation between per capita GSP and aggregate GSP was only .217. This would suggest the influence of population greatly exceeds the influence of wealth. 41 41 This should not be surprising, given that the largest state (California) was over 65 times more populous than the smallest (Wyoming), while the wealthiest state (in terms of per capita personal income (again California) was just less than twice the lowest 105 In principle, population should have an influence on the diversity of interest groups, because it influences both the number of points of view needing representation and the difficulty (or expense) for any one individual who seeks access to a representative. However, Gross State Product should have a greater influence on interest group density (using a population ecology model), because it measures the ability to support a large community. A leading explanation for interest group formation is the mobilization of both the general populace and specific interests, when government is acting within a policy area (Leech et al. 2005). Primary and secondary education saw an increase government action during the 1990s, so it should be expected that the number of groups should have increased. This presents a problem for establishing the causal relationship between interest groups and policy innovation, as was briefly considered in the previous chapter. However, prior activity by states in terms of education reform should suggest reasons for differences between states in interest group density. Similarly, the countermobilization explanation suggests a reason for increased diversity when states are engaged in periods of reform (Truman 1951; Berry 1997; Lowery et al. 2005). The proportion of a state’s economic resources expended on education policy is also a measure of pre-existing government activity and the saliency of the policy (and thus the ability and need to establish countervailing interest groups after the distortion caused by early interest groups is first noticed). If interest groups form primarily to protect the economic interests of specific groups affected government, rather than the diffuse (Mississippi). Therefore, GSP can be seen as only a slight modification on state population. 106 interests of student, then the portion of GSP spent on education is a good judge of the stakes of education policymaking. Berkman (2001) suggested that interest groups do not form independently of their political system, as he argues that Gray and Lowery (1996) do. Specifically, professional legislatures tend to depress the number of interest groups, because fulltime legislators with sufficient resources have less need to obtain information from lobbyists. Therefore, interest groups recognize that they are less effective and subsequently leave competitive communities. Berkman (2001) finds that research into professional legislatures supports this argument, finding a strong difference between legislatures based on their level of professionalization. For instance, Squire (1993; 1988) and Berry, Berkman, and Schneiderman (2000) indicated that professional legislators spend more time with their constituents and are more insulated from electoral pressures than less professional legislators. Conversely, less professional legislatures and legislatures with a large number of members should provide a greater access for interest groups (and a greater need to seek information from groups in order to inform members). However, the size of the interest group community remains dependent on the population of the state and other factors, because the meaning of crowded is relative to state resources. Whatever the need, the resources to form interest groups limit growth. With this evidence, it should not be surprising that interest group membership or supporters see less advantage to continued existence in states with small, professional legislatures (Berkman 2001). Another important consideration is the size and professionalization of the bureaucracy. When those who work for the state represent a large segment of the 107 population, they act as gatherers of information and advocates for policies much in the same way staffers for a professional legislature do. Others (Jacoby and Schneider 2001; Gormley 1996) suggest the role of bureaucrats as lobbyists on their own. Professional bureaucrats tend to remain in the government, even as elected and appointive officials leave government. Senior bureaucrats represent institutional memory within government, having both expertise and the ability to look beyond shortterm political concerns. In addition, they are the primary implementers of policy. 42 This should lead to a decreased advantage to interest group formation and maintenance. Specifically within the area of primary and secondary education policy, a large percentage of those directly affected by education policy are government employees (mostly professional educators, school districts, and school administrators being influenced by legislatures and bureaucracies). 43 In addition to the factors and conditions mentioned above, which have been proposed to explain the differences between state interest group communities, it is important to explore the characteristics of state governments associated with policy innovation. Perhaps, this will uncover a mechanism by which primary and secondary education interest groups influence primary and secondary education policy. Among others McKenzie (2004) and Hayward (2003) noted differences between racial groups in terms of the institutions commonly expected to mobilize political activity with churches 42 A policy implementer would seem to be a more influential actor during periods of incremental change. Models of bureaucratic control typically focus more on whether policymakers implement directives from political branches than what those branches initially pass (Hammond and Knott 1996). 43 Local school districts, alone represent 27.6% (327 of 1183) of all education interest groups that lobby state governments. Teachers unions represent the second largest group at 6.9% (82 of 1183). 108 having a greater role. While Hill and Leighley (1994) found that mobilizing institutions were less important in elections than socioeconomic factors. Huckfeldt and Sprague (1992) saw the preeminent role of political parties. If these factors influence other aspects of political activity, they should at least be evaluated for interest group formation. Taken together with the influence of ideology, partisanship, and race and ethnicity for policy innovation shown in the previous chapter, the need to account for the influence of these characteristics on interest group formation should be implicit. This will be useful, because it will rule out the possibility that interest groups form for the same reasons policy innovations are considered. Educational Interest Groups in the American States This chapter requires the use of three types of data. First, the MDS analyses and the cluster analysis utilize proximity data for state primary and secondary education policy interest group communities in 1998. The raw data from which the proximities were derived is the list of primary and secondary education policy interest groups as compiled by the David Lowery and Virginia Gray updated through 1998 (see Gray and Lowery 1996; 1998; Lowery and Gray 1998). The Lowery and Gray (1998) data list all interest groups lobbying state governments and was categorized into policy areas and organization types with the name for each group active in the state. For the purpose of this analysis, only groups engaged in lobbying for primary and secondary education were used. This meant excluding non-education groups and those groups representing colleges and universities, museums, etc. 109 For the measure of interest group community size, the next step is to total the number of interest groups and measure similarities based on the proximity of the total number. In order to measure diversity, a list of 51 interest group categories was derived from the 1183 groups active in lobbying state governments in 1998. By far, local school boards and educational unions represented the largest categories. The next step was to divide the number of interest groups within each category by the total number of interest groups within a state in order to show diversity. The final data transformation for the MDS is to develop a matrix of interest group category proportion matches between each combination of two states for each model. This is a square 50 X 50 matrix of the states where the number in each cell represents the proportion of the entire primary and secondary education interest group community states represented by the same categories of interest groups. In MDS terms, this is called a similarity matrix. States with a greater number of similarities are deemed to be more alike and closer spatially. 44 The second step was to collect data in order to find the substantive dimensions within the coordinate data produced by the MDS. This search required data on several of the key theories for state interest group formation and policymaking mentioned in the previous chapter. In some cases, this research used data derived for prior policymaking and interest group research. This included the measure of legislative professionalism found in Squire (2000) and state partisanship and ideology found in Erikson, Wright, 44 While it would have been possible to focus on the number of interest groups in each category, instead of that category’s proportion of the entire community, the result would have been extremely similar to the measure of interest group density. The effort was to find a measure that had the potential to diverge from interest group density, if it were measuring something unique. 110 and McIver (1993). A few of these variables were seen originally in Politics in the American States (Gray and Hanson ed. 2004), and then checked against their source data. This would include the total number of fulltime state employees and the institutional powers of legislative speakers and state governors. A measure of state employees per 10,000 state residents in the year 2000 was found through The 2012 Statistical Abstract (U.S. Census Bureau 2012, Table 466). This chapter used a measure of minority diversity derived from taking the proportion of African Americans plus Latinos from the 1990 US Census. This measure of minority diversity was similar in intent to the minority diversity approach envisioned by Hero (1998). Population and Regional variables (designed to create four roughly equal sized regions) use 2000 US Census data. Similarly Gross State Product and Per Capita State Product use data from the Bureau of Economic Analysis, a part of the US Department of Commerce from 1998. The data collected to measure density was also be used for the regression used to measure influences on density. The density of an interest group community is simply the total number of policies in a state. In an OLS regression, this total is used as the dependent variable and the variables listed above to determine the structure related to the interest group diversity are used as independent variables. In so doing, it is possible to make some determinations regarding interest group density. Finally, the cluster analysis will require the coordinates from the MDS. These results will represent the data in the number of coordinates found in the dimensional solution from the MDS. The two coordinates will allow each of the states to be represented spatially. 111 Methodology The chapter will utilize MDS to create a spatial representation of the similarities between state interest group communities. Chapter 4 presents a discussion of MDS. For this chapter, it is important to note that instead of looking at dissimilarities in individual policies, this chapter is looking at dissimilarities in the proportion of each type of interest group represented within a state. Therefore, two states that have local school boards as one-third of all interest groups would be deemed similar even if 10 local school boards engaged in interest group activity in one state and only 5 in another. In this way, the MDS is meant to represent interest group diversity and not just density. Regression Results for Density The theory and data suggest several variables to examine for their relationship with the number of interest groups within a state. As noted above and in Chapter 3, these include population, gross state product (GSP), legislative professionalism, the size of the state bureaucracy, the size of the legislature, education spending, ideology, and region (particularly the western states). In addition, the research will examine minority diversity, a leading cause of variance in education policymaking according to the previous chapter. In keeping with the other sections of this dissertation, the three goals of this analysis will be exploration, description, and evaluation (see Rabinowitz 1978 with respect to MDS). OLS regression will allow for an exploration of the leading characteristics associated with the number of interest groups within a state. The regression line, along with an associated scatterplot will provide both a mathematical 112 and graphic representation of the data. Finally, the results will allow for an evaluation of theories for the density of interest group communities. Previous research points to several explanations for the size of state interest group communities. The strongest of these, as suggested by Gray and Lowery (1996), is gross state product (GSP). Unfortunately, there is a problem with GSP. It is highly correlated with population (.993) while not necessarily making a compelling argument for its superiority. Both Population and GSP can be used as measures of a state’s capacity to support interest groups, and absent a strong relationship between per capita GSP and interest group density, one is essentially the other in terms of their effect on either a regression or MDS. The correlation between interest group density in primary and secondary education and per capita GSP is .026. Therefore, there is no particular reason to use one variable instead of the other. Certainly, the demand by the public to have representation could be related to population. This is particularly important, because of the large number of primary and secondary education interest groups related directly to local school boards and their employees. A school district does not need to represent only wealthy people to have access to state decision-makers. 45 A second variable with the potential to be related to interest group formation is the number of fulltime state employees. Where the state population considers demand for interest groups to represent a point of view and the capacity of a community to provide for an interest group, the number of fulltime state employees is suggestive of a 45 This is not to suggest that local school boards are perfect representations of community interest. Within a community, access to power might determine whether someone can influence a school board to care about a particular concern. However, any school might be sufficiently influential and greater wealth within a community might not afford them a second school board. 113 supply of access points. However, state employees might also be the ones who are the most likely to form interest groups. In many respects, state employees have the greatest stake in state policymaking. In education policy, teachers and other school employees are represented by interest groups related both to labor relations through their unions and to their professional disciplines through professional associations, associations related to their academic discipline, and boards and commissions related to their field. Chapter 3 suggests several other variables for consideration. Each of these has garnered support in previous literature. It should be noted, that legislative professionalism and state employees as a proportion of the state population have moderate negative correlations with interest group density. This is the relationship predicted by prior research (Squire 1988; 1993; Berry, Berkman, and Schneiderman (2000); and Berkman (2001) for legislative professionalism; Gormley 1996 and Jacoby and Schneider (2001) for size of bureaucracy). However, it does suggest that the positive correlation between fulltime state employees and interest group density might also owe to the large population differences between states. In fact, population might be driving any difference that does not control for population. In some respects, the strong positive relationship between fulltime state employees and interest group formation should be seen as counter-intuitive, if a bureaucrat is seen as similar to a professional state legislator (who should depress interest group formation). However, it should be noted that many state employees are public school teachers. Therefore, instead of having a role primarily as regulators of outside contractors, many government employees are actually workers in the field of 114 education. Therefore, there number reflects a demand on the part of employees to have representation. Counter mobilization should be seen clearly in terms of local governments (school boards) also seeking to be represented. As can be seen in Table 8, these variables work at cross-purposes. As the population increases, the number of interest groups increases. However, as the number of fulltime state employees increases—the population remaining the same, the more likely a state is to have fewer interest groups. This is consistent with previous research that sees a bureaucrat as a type of public sector lobbyist, gathering and disseminating information. This supports the contention in Berkman (2001) that interest group activity is not independent of the political context in which it exists. Rather, large professional bureaucracies tend to suppress interest group community size, even as increases in population make interest groups more likely. The influence of state wealth, as measured by GSP, and not used as a proxy for state size, is not supported by the results of this research. 46 Table 8. Regression Models explaining the Density of Interest Group Communities Variable Model #1 Population Fulltime State Employees West .110 (.014) ** -.664 (.112) ** --.7731 2 Model #2 (Region Included) .104 (.013) ** -.606 (.113) ** 11.902 (3.717) ** .8145 R N (Number of States) 50 50 A single asterisk indicates a t-statistic with a p = (α < .05) for a one-tail test. Two asterisks indicate p = (α < .01) for a one-tail test. 46 Other variables that are predicted to influence interest group density were tested, including legislative professionalism, per capita state employees, minority diversity (because of its influence on policymaking), size of state legislature, ideology, education spending, and region. None of these variables had an influence independent of the included variables when tested. 115 Spatial Results of MDS The first step in conducting the MDS was to find the appropriate number of dimensions. The goal of MDS is to find the minimum number of dimensions that provides a meaningful and appropriate view of the data. For presentation purposes, it is ideal for this number of dimensions to equal two. However, a three or more dimension spatial representation is possible in MDS with a perfect fit with a total number of dimensions equal to N minus one. As noted above, this is an issue of trading fit for interpretability. Figure 4 (on the next page) shows a scree plot for the MDS in 1, 2, 3, and 4 dimensions. The scree plot provides evidence the appropriate number of dimensions is two, using the approach of looking for an “elbow” in the SSTRESS data. 2 In two dimensions the SSTRESS measure of fit result was .14223. The R was .95252. 116 Figure 4. Scree Plot of SSTRESS for Primary and Secondary Education Interest Group Diversity .2 y .15 .1 .05 1 3 2 4 Dimensions Figure 5 shows the two-dimensional configuration for state primary and secondary education interest group community diversity. At this point, the coordinates have no substantive meaning, as the results have been rotated to maximize their variance on the axis and adjusted so the range of the data on each dimension is the same. 117 Figure 5. Diversity Two-Dimensional Configuration of Stimulus Points for Interest Group 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Even without substantive dimensions, Figure 5 provides several insights. The most obvious of these is the distance of Hawaii from the other states’ primary and secondary policy community. While Hawaii is unique in several ways: it is not located on the North American continent as the other states are, whites do not constitute a plurality, and it is the last of the fifty states to join the United States, perhaps it is most 118 notable in this research for being the only state without local school boards. The state school board runs all of the public schools in the state directly. The state is also one of the smallest. Therefore, Hawaii does not have the characteristics that seem to drive interest group formation in public school policy. Local school boards, which represent over a quarter of all interest groups (27.6%), do not exist. Population, which is strongly correlated with density, is small. Hawaii also has the most fulltime state employees per capita in the United States and the tenth most professionalized legislature. This would appear to predict a very low interest group density. Not surprisingly, Hawaii had only one interest group. This was the Hawaii State Teachers’ Association (Gray and Lowery 1996). This was a statewide union that also served as the representative of all teachers in the state. California, the state with the most primary and secondary education interest groups (184), is on the far right side of the interest group space. While th California has a highly professional legislature, it was 49 in per capita state employees and is a state with many interest groups. Another characteristic of the interest group space that is clear from Figure 6 is the seemingly linear arrangement of states heading down and to the right in the upperright of the space. The cause of this arrangement is less clear than the differences between Hawaii and California. However, it is possible to look for characteristics of state government where the correlation of the data and other substantive variables are high. The next step is to determine appropriate substantive variables to regress on the stimulus points to uncover meaningful relationships. The variables examined will be those discussed previously in this chapter. Because past research (Gray and Lowery 119 1996) has suggested density and diversity as unique characteristics of interest group communities, an important substantive variable is density (or the influence of having a large number of interest groups on diversity). This is because density could either result from either the repetition of existing types of groups (for instance, many local school boards and teachers’ unions) or the existence of new groups. In addition to interest group density, Table 9 shows the correlations (or the strength of the straight line relationship) between the stimulus points and the possible underlying dimensions for interest group diversity proposed and described in the previous sections. Significant correlations (at p < .05) are shown in boldface. As expected, there is a strong relationship between the density of the interest group community and the diversity. Given the explanatory power of population and its close correlation with GSP, it is also not surprising that a significant but fairly weak relationship exists between those two variables and interest group diversity. On reflection, it is surprising that it was not stronger. Of more interest is the relationship between both the number of state employees per 10,000 state residents and legislative size with interest group diversity. The other suggested variables are not significant. Table 9. Correlations between Substantive Variables and Stimulus points Variable Correlation (r) Variable Interest Group Density .624 Partisanship Ideology (Berry Inst.) .344 Ideology (Berry Citizen) Minority Diversity .236 Gubernatorial Power Competitiveness .104 Fulltime Employees Population Speaker Power .357 GSP Total Size of Legis. .367 State Emp./10,000 Legis. Professionalism .589 N = 50 Highlighted Correlations are significant at p < .05 120 Correlation (r) .210 .360 .076 .316 .160 .399 .327 The results above suggest that the best explanation for similarities in interest group diversity is similarities in the density of interest groups and the number of state employees per 10,000 residents. When measuring interest group density, the total number of state employees was important, but the number of employees as a proportion of the total population was considerably less so. In fact, interest group diversity appears to increase as the number of per capita fulltime state employees increases, while density declined. The next strongest correlation is between legislative size and interest group diversity. However, the large gap in the relative sizes of the correlations suggests a much weaker role. Thus, the strongest influences on interest group diversity are the relative size of the bureaucracy given the size of the state and the size of the interest group community. The next step is examine the influence of interest group density and state employees per 10,000 on interest group diversity by replacing the coordinate axis in Figure 2 with substantive dimensions created by regressing these variables with the coordinate points from the MDS. Figure 6 shows those dimensions. 121 Figure 6. Regression Lines Showing Dimensions for Interest Group Density and State Employees per 10,000 State Residents 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Interest group density has a slope of -.411 and the proportion of state employees to state residents has a slope of .605. This means that as the number of state employees per 10,000 increases, there is an expected move toward the top right on the map, and as the number of primary and secondary education interest groups in a state increases there is an expected move toward the bottom right at the slope indicated. As with Figure 6, Figure 7 shows the tight clustering of most of the results, because the 122 data points have not moved. Thus, the negative sloping line suggested in the analysis of Figure 5 reveals itself as a dimension associated with interest group density. The tight cluster is not surprising, given the nature of the substantive dimensions. In particular, it is notable that interest group density is a variable heavily skewed to the right with a mean of 23.66 and a median of 18. Two-thirds of the states fall below the mean, providing less ability for distinctness than in the case of the others. While excluding California minimizes the effect, mean continues to exceed the median. For most states, the number of primary and secondary education interest groups is smaller than the possible number of categories. The number of state employees per 10,000 residents has a much less pronounced skew to the right with a mean of 178.92 and a median of 170.5. One of the most notable results is that, as expected, the states with few active interest groups have less diversity. They are similar in the sense that most categories of primary and secondary education interest groups do not exist within their states. Similarly, there is a tendency toward less variety, because certain interest groups, such as teachers unions are active in essentially all states, decreasing the opportunity for variety even more in those states. With few exceptions, states with more interest group density show a tendency for more activity by a wide variety of groups. This will be explored more thoroughly in the next section of the chapter. Cluster Analysis Building on the analysis of the dimensions, it is possible to gain additional insight into the manner in which the individual stimulus points group together by looking at the 123 results of a hierarchical cluster analysis. It is possible to choose any number of clusters; but the data suggests particular attention should be paid to two clusters (representing 41 states) from the region in the top-right corner of the graph to uncover what makes these to clusters different. Additionally, it will be useful to look at four clusters of states that are more widely separated (representing the other nine states). The extreme position of Hawaii tends to suggest less separation between the other states than would be apparent, if Hawaii were excluded. However, the extreme score for Hawaii is consistent with the difference between the interest group community in Hawaii and other states. It should be noted that other states that fall closest to Hawaii also have very small interest group communities. So although these clusters are not as closely aligned with each other as the two main clusters, examining their similarities might prove useful. Table 10 shows the members of each cluster, while Figure 7 shows the locations. As can be seen, the two clusters are both large and close to each other, while the others range across the remainder of the available primary and secondary education interest group space. With Cluster #4 being of interest because it is entirely comprised of large interest group communities. 124 Table 10. Cluster results from Cluster Analysis of MDS Stimulus points Cluster #1 Cluster #2 Cluster #3 Alaska Colorado Florida Georgia Illinois Indiana Iowa Kansas Louisiana Maine Minnesota Missouri Montana Nebraska Nevada New Mexico New York Ohio Oklahoma S. Dakota Virginia Washington Wisconsin Wyoming Alabama Maryland Arkansas W. Virginia Connecticut Delaware Idaho Kentucky Massachusetts Mississippi N. Hampshire New Jersey North Carolina North Dakota Pennsylvania South Carolina Tennessee Texas Vermont 125 Cluster #4 Cluster #5 Arizona California Michigan Oregon Rhode Island Utah Cluster #6 Hawaii Figure 7. Relative Cluster locations with each State’s Cluster Indicated 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Note: Cluster 1 is represented by hollow circles, Cluster 2 by hollow squares, Cluster 3 by plus signs, Cluster 4 by triangles, Cluster 5 by filled squares, and Cluster 6 by diamonds. Next, Table 11 shows descriptive statistics for the characteristics of each cluster. This table provides insight into how the states in each cluster are different from the other clusters. Table 11 provides clear evidence that the greatest cause of clustering is 126 the density of the interest group community within a state. While states vary in terms of the other variables that correlate significantly, the most pronounced differences are reflected in the number of interest groups within the policy area. Within the two large clusters, only two of the 17 states in Cluster #2 have as many primary and secondary education interest groups as the average for Cluster #1. Seven states in Cluster #2 have fewer interest groups than any in Cluster #1. Meanwhile, 13 of the 15 states with the largest interest group density (17 of 21) were in Cluster #1. Meanwhile, the first, second, sixth, and twelfth most dense interest group communities comprise the four states in Cluster #4. The remaining five states (comprising the entire bottom left of the spatial model) range between 1 and 12 primary and secondary education interest groups. Table 11. Descriptive Statistics for the Characteristics of each Cluster (Means) Cluster #1 Cluster #2 Cluster #3 Cluster #4 Cluster #5 Cluster #6 All 50 States Interest Group Density 25.08 15.25 9.5 76.25 6.00 1 23.66 State Employees per 10,000 171.67 192.81 175.00 132.50 204.00 453 178.92 Citizen Ideology (Berry et al. 1998) 45.45 52.32 64.65 52.44 51.62 74.3 48.89 Size of Legislature 141.92 182.38 161.00 112.00 127.00 76 148.48 The clusters also re-enforce the relationship between interest group diversity and the number of state employees per 10,000 residents. The states in Cluster #1 have fewer state employees given their population than the states in Cluster #2, and although the average is higher than for Cluster #4 (of four states), seven of the eleven states with 127 the smallest proportion of state employees are from Cluster #1. There is less clarity for the other variables, but the states in Cluster #1 also are more conservative and have smaller legislatures. There does not appear to be a strong relationship between either size or region and interest group diversity. Discussion The spatial and cluster analyses have shown that increased interest group density in primary and secondary education policy results in more interest diversity than exists in states with less density. This is consistent with the conclusions of Gray and Lowery (1996; 1998) that interest groups diversity is related to density. However, Gray and Lowery (1996) found that the most powerful interests remain central even when density increases. The natural question is what types of primary and secondary education interest groups exist in those states with a large number of interest groups that do not exist when there are fewer interest groups. In order to examine this question, it is necessary to look again at the types of primary and secondary education interest groups that are shown in the data from Gray and Lowery (1996; 1998) as updated. As noted in Chapter 2, 27.6% of all primary and secondary education interest groups are local school districts (327 of 1183). However, the percentage of local school districts represented from state-to-state varies considerably. Cluster #4 consists of the states with the most local school boards lobbying the state government and in three of these states; local school boards represent a majority of the primary and secondary 128 education interest groups. 47 Other than local school boards, the next four largest categories of interest groups are: state teachers unions (82), associations of school administrators, business administrators, and superintendents (71), advocacy groups for teaching specific courses (65), and school board associations (46). Together, these interest groups represent 22.3% of the interest groups within each state. Because the modal score for each of these groups is one, the diversity level of interest group communities with low densities are more strongly influenced by these groups. This is seen starkly with teachers’ unions. Most states (even the very large ones) had either one or two state teachers unions registered, while only Nebraska and New Jersey had no state teacher’s union registered as a lobbyist with the state government. This is most significant with Hawaii, which had only one interest group—the State Teachers’ Association. In most cases, states with few interest groups in the policy area have fewer local school boards lobbying them, while those with the highest proportion of school boards tend to cluster together (and often have the densest communities). 48 Taken together, primary and secondary education policy within state interest group communities might best be described in terms of labor/management negotiation with little outside influence by groups that are neither associated with schools nor their employees. 47 The three states, California, Arizona, and Oregon also have the first, second, and twelfth largest interest group communities respectively. The fourth state in the cluster, Michigan, had 43.9% (18 of 41) of its community comprised of local school boards. These states also rank from first through fourth in terms of local school boards as a proportion of the entire community. 48 A dilemma arises as to whether school boards are seen as similar to a corporate board, in which case they represent a powerful interest, or as representatives of the mass public which elects them, in which case they represent a diffuse interest. 129 The cluster analysis shows differentiation between the states in terms of interest group density. This is the primary explanation for the differences in the type of groups in each state and is very consistent with Gray and Lowery (1996; 1998). All four of the states in Cluster #4 not only have high interest group densities, but they also have more local school boards lobbying state governments than the average state. In fact, two of these states have more local school boards as lobbyists than the total number of primary and secondary education interest groups in a majority of states. As compared to the extremes, 41 states show less to differentiate the interest group communities in education. None are especially large when compared to California, which has a very large and diverse community with numerous professional organizations related to occupations and fields of study in addition to groups advocating for policies from the outside. Even so, a majority of California’s primary and secondary education interest groups were local school boards (96 of 184). In contrast, Cluster #1 averaged 15 interest groups and Cluster #2 averaged 25. Implications This chapter explored interest groups that lobby state governments in primary and secondary education, using MDS to map the similarities between states in the policies they adopted to measure diversity, as well as an OLS regression to explore causes of population density. In some respects it is surprising and disappointing that population and size of the state bureaucracy provide so much of the explanation for interest group density, while interest group diversity appears to be an extension of that density. In many states there are so few interest groups that it is not really surprising 130 that the community would not be diverse. In each state except Nebraska and New Jersey, at least one teachers’ union was represented among the interest groups. Therefore, a state, such as Hawaii, that has only one interest group shares in common with other states that have very few interest groups, a community with a state teachers’ union interest group as a powerful figure. Many of the states with large interest group communities owe that large figure to the great number of local school boards and school districts involved in lobbying the state government. Unlike the unions that typically had a single state organization involved in lobbying, many school boards tended to seek influence on their own. The large percentage (27.6%) of the primary and secondary education interest groups represented underscores this. It should also be noted that this strategy is not necessarily superior to the strategy of teachers’ unions and other statewide unions for school employees that had more central control. Because primary and secondary education policy depends so much on people within the educational field for interest group activity, it would seem to support the notions of Chubb and Moe (1990) or even Friedman (1962) that the policy community in education exists primarily for its own benefit. However, this is not fully consistent with the policies adopted during the 1990s. If it can be assumed that education reform presented a challenge to schools and their employees, and then it would follow that in education policy, the interest groups at the state level were not the primary actors in this policy area. 131 Chapter 6 A Configuration Comparison of State Policy Innovation and State Interest Group Communities As noted in Chapter 5, the measure of interest group diversity used in this dissertation is somewhat different than the measure used by Gray and Lowery (1996). Gray and Lowery (1996; 1998) examined whether the interest groups were institutions, membership organizations, or associations (as well as the policy area with which they were concerned). Because Chapter 5 only considered a subset of educational interest groups (those focused on primary and secondary education), it was necessary to create sub-categories related to their place within the primary and secondary education policy community. The similarities then were based on the proportion of like interest groups each state shared. Chapter 5 considered what caused states to share interest group types. This chapter considers whether states with similar dispersions of interest group communities are likely to adopt similar policy innovations. Although the substantive variables are not generally the same for the policy innovation dimensions and the interest group diversity dimensions (within primary and secondary education), the structure of the two has not yet been compared directly. Whereas the evaluation of population density is straightforward, because the key attribute of density is simply the number of interest groups, an evaluation of population diversity needs to account for not only the amount but also the type of interest groups active within the policy community for primary and secondary education in each state. Despite the lack of strong support for a correlation between interest group diversity and 132 policy innovation from the other associated variables it is important to compare the two, because the question of their relationship goes to the heart of the widely-accepted argument that interest groups advantage the groups they represent. If interest group diversity has little or no influence, then it matters little whether a given group is represented in a state as compared to mass support represented by ideology and the interaction of racial and ethnic groups within the general population. This comparison will provide a measure of the correlation (congruence) between the policy innovation and interest group diversity structures obtained through MDS, as well as further insight into the possible importance of similarities in interest group diversity on the similarities in policy innovations occurring in states. However, if interest group diversity provides an explanation for policy innovation, a strong argument can be made for the centrality of the pluralist/elitist debate regarding interest group influence. Comparing Configurations (Congruence) The purpose of Chapter 4 and Chapter 5 was to explore the set of points each derived from the effect of a single set of stimuli (policy innovation or interest group diversity, in primary and secondary education policy) on 50 objects (the states). In principle, MDS is done by constructing a table of distances, then applying a formula designed to provide a result closely approximating those distances in a defined number of dimensions (also called attributes), utilizing a measure of goodness-of-fit to show the degree to which the results are like the original table (Bartholomew et al 2002). This involves adjusting the locations of the stimulus points to minimize the difference between the distances as shown on the graphical representation and the input matrix. 133 The program used to conduct the MDS in this study (ALSCAL) utilizes an alternating least squares procedure to fit the model to the data (Young and Lewyckyj 1996). This chapter compares the two matrices, in order to determine whether the structure of interest group diversity and policy innovation is similar. In essence, the stimulus points are compared to determine whether the states are arranged similarly with respect to policy innovation and interest group diversity. A high correlation between the structures would suggest states with similar diversity in interest group communities engage in similar types of policy innovation. Several characteristics of MDS both allow and make it necessary to transform and rotate the configurations in order to maximize the correlations (Davison 1983). First, because the structure of the MDS results is derived entirely from internal stimuli, MDS results have no fixed locations relative to external stimulus points. All of these are a result of the arbitrary nature of where the configuration is placed in space. Second, the original results are rotated using varimax, in order to show differences clearly. However, varimax rotations are designed to account for the data without regard to external considerations. Therefore, a rotation that maximizes the correlation between two independent matrices would be purely accidental. To compare the configurations, it is necessary to make several transformations of the data, in order to best represent their similarity. The principle behind configuration comparison is to find a rotation designed to maximize the correlation between an obtained stimulus configuration and some target configuration. Because any rotation of an MDS solution represents the same stimulus configuration (though from a different perspective), and any rotated and un-rotated solution is equally arbitrary with respect to its specific location except relative to other 134 points in the solution, either or both MDS coordinate matrices may be rotated to facilitate comparison (Davison 1983, 22-28). Davison (1983) also notes the target configuration for the comparison may either be derived from theory or from a prior stimulus configuration (195). For a comparison of policy innovation and interest group diversity, either configuration might be deemed the target configuration. Because both derive from MDS results and not theory neither configuration is determined to necessarily exist prior to the other one and the measurement of their correlation is designed simply to determine how alike the two configurations are. Once the configurations are rotated, the comparison of the two matrices allows an evaluation of the strength of a hypothesized relationship between the two configurations (i.e. how alike they are). If the two configurations were precisely the same, they would be congruent. The degree of their correlation is the measure of their congruence (Davison 1983, 190-201). If there is a strong association between interest group diversity and policy innovation in primary and secondary education, it would be expected that the configurations, when rotated to minimize divergence would have a great deal of proximity between an object in the target configuration and the same object in the obtained configuration. 49 This has the potential to provide many different insights. For instance, MDS results for the effect of a stimulus on a set of objects from two different periods of time could be compared with their similarity pointing to a continued strength of an effect. If congruent, the effect would not have changed over time. In this chapter, 49 As noted elsewhere, it is an assumption of MDS that the distance between an object and itself is zero. Therefore, if the two configurations were perfect replicas, the distance between the target and the rotated obtained configuration would be zero for every object. As the divergence increases (measured as the sum of squared divergence), and thus the correlation decreases, the assumption that these are similar configurations, representing similar concepts becomes less tenable. 135 the level of congruence will determine whether similarity in state interest group community diversity is associated with similarity in the types of primary and secondary education reform policies. Because MDS results represent the data even when they have been reflected, rotated, multiplied by a constant, or a constant has been added, it is possible to make several transformations of obtained stimulus configuration designed to maximize its similarity to the target stimulus without diminishing its ability to represent the data. 50 Once the mean and range of the values of the stimulus points are standardized for each matrix, it is possible to calculate a multiplicative constant that minimizes the (sum of squared) discrepancies between the target matrix and the other matrix, once rotated to minimize those discrepancies (Schonemann and Carroll 1970; Gower 1971; 1975; Davison 1983, 197-201). For this chapter, a program developed by William G. Jacoby (2007) performs the task. Conceptually, this is the equivalent of inflating and deflating the space like a balloon so it is the same size before arranging the points to minimize their differences. Once this is done, it becomes possible to calculate a correlation between the target configuration (matrix) and the rotated obtained configuration (matrix). This correlation is termed the congruence coefficient (Davison 1983, 198). As the value of the congruence coefficient (correlation) increases, the more similar the two configurations are. 50 Although the description of each of these transformations is geometric, each results from a transformation of the matrix that does not change the relative relationships between the individual entries. 136 Interest Group Diversity and Policy Innovation Although many leading explorations of interest group influence in the United States have been conducted on the national level (Shattschneider 1960; Edelman 1960; Olson 1965; Heclo 1978; Lowi 1979; Kingdon 1984; 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993), much of the policymaking affecting the lives of Americans occurs on the state or local level. This is especially true in primary and secondary education. Therefore, the conclusion by Gray and Lowery (1996) that an interest group community is limited in the size of its growth by the ability of the political environment to support (carry) it is extremely important (254). Chapter 5 suggests that this ability to carry interest groups is determined largely by the size and complexity of state governments. The influence of the business groups versus not-for-profits dichotomy within primary and secondary education policy figures to be muted with only 33 (of 1183) interest groups representing business (school manufacturers/builders and book publishers), business/education partnerships, education consultants, and education television—not including private, Montessori, and charter schools. However, government agencies and associations of government agencies comprise a large percentage of all state interest groups (43.6%) with local school boards alone representing 27.6%. For this reason, it should not be surprising if primary and secondary education behaves differently from other policy areas. However, education policy—as one of the most salient state-level policies—is of independent interest. Chapter 5 examined both interest group density and interest group diversity. Because 48 of the 50 states had at least one interest group lobbying the state government on behalf of state teachers’ unions (for a total of 82), the effect of increased 137 interest group density was to lessen the proportion of all groups in a state represented by the one or two teachers’ union groups in each state. Not surprisingly, the strongest factor in interest group diversity was density. The next strongest influence was the proportion of state employees per state resident. 51 In addition to examining the level of diversity, the MDS allowed for a look at many different ways states could be dissimilar simultaneously. Although Chapter 4 presented a model for policy innovation in primary and secondary education that was not especially favorable for any explanation involving interest group diversity, a few factors might point toward the models being similar. Arguing against the similarity is the substantive dimensions proposed for policy innovation. These were the proportion of minority diversity in a state—a component of the social diversity approach suggested in Hero (1998)—and citizen ideology—as proposed by Erikson, Wright, and McIver (1993) and Berry et al. (1998). The correlation between these variables and policy innovation was .703 for minority diversity and .499 for the Erikson, Wright, and McIver (1993) ideology measure. 52 These variables are consistent with two common explanations for policymaking generally—social diversity 51 Interest group density was examined in a regression in Chapter Five, and population and the number of full-time government employees were found to be associated with density. Because of the similarity between these two independent variables and the seeming association between proportion of government employees to state population for diversity, it calls into question the independence of the very concept of diversity from density. 52 The correlations between interest group diversity and these variables were .236 for minority diversity and .132 for Erikson et al. (1993) ideology. Whereas shifting from the Erikson et al. (1993) measure to the Berry et al. (1998) citizen measure increases the correlation between ideology and policy innovation only slightly (.499 to .516), this change increases the correlation between ideology and interest group diversity from .132 to .360 (making it a significant, though still a weak to moderate correlation). 138 (Hero, 1998) and democratic representation (as described in Erikson, Wright, and McIver, 1993). However, the MDS policy innovation dimension attributed to minority diversity in Chapter 4 could be replaced by dimensions associated with several more weakly correlated but significant variables. 53 These variables are per capita gross state product (r = .590), population (r = .558), and partisanship (r = .562). While partisanship is not significantly correlated with interest group diversity, population (r = .357) and per capita GSP (r = .367) have weak to moderate correlations. This does not suggest a high congruence between the two configurations, but there could be some similarity. Another useful result of a comparison will be the opportunity to examine the results to see whether interest group diversity might be explanatory for a subset of the states. For instance, interest group density varies dramatically with a mean of 23.66 and a standard deviation of 25.27. With such a large range (1-172), it is reasonable to assume that the influence of interest groups varies from state to state. Perhaps, diversity will matter in those states with enough interest groups for them to have an influence—or conversely, interest groups might only matter when they have a virtual monopoly (this would be consistent with the iron triangle model). If some states appear to have the types of policy innovations expected given their interest group communities while other states do not, it might be possible to uncover a pattern. This might provide insight for further research, even if it is impossible to make any determinations at this point. 53 Each of these dimensions has a stronger correlation with the stimulus coordinates for policy innovation than either measure of ideology. However, their relationship (as measured by the regression slope) with the stimulus coordinates is closely related to minority diversity (and as such cannot account for the added variation of a second dimension). 139 Results for the Configuration Comparison The differences between the spatial models for policy innovation from Chapter 4 (Figure 8) and for interest group diversity from Chapter 5 (Figure 9) provide considerable insight into the results that are to be expected even when interest group diversity is rotated and (in this case) deflated in order to optimize the fit between the two configurations. Figure 8 (on the next page) shows the policy innovation configuration to have a relatively uniform dispersion of stimulus points across the entirety of the policy innovation space with few of the states ranging far from their neighbors. At the same time, Figure 9 shows a few states to be outliers far away from any of the other states, with the remaining states tightly bunched as compared to those outlier states in a pattern highly suggestive of a linear relationship between the coordinates of the many states congregated in the upper-right of the chart. The shapes of these two configurations appear very different, with the policy innovation space more evenly dispersed. 140 Figure 8. Two Dimensional Configuration of Stimulus Points for Policy Innovation 1 y .5 0 -.5 -1 -1 -.5 0 .5 x 141 1 Figure 9. Diversity Two Dimensional Configuration of Stimulus Points for Interest Group 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Because congruence is a correlation measure defined by the covariance and variance of the two matrices, there will be a tendency in optimal fitting procedure to minimize the greatest differences between items in one matrix to the corresponding items in the other—even at the expense of increasing the size of the smaller differences (Davison 1983, 196-197). For this reason, optimizing the fit between interest group diversity and policy innovation through transformations in the interest group diversity 142 matrix will necessarily involve deflating the interest group matrix, so its most extreme items are nearer the scores for the corresponding items for policy innovation. The necessary amount of this deflation is directly related to whether the extreme scores for interest group diversity are in the same states as extreme scores for policy innovation (and if this extremity is always in the same direction) 54 . Figure 10 shows the rotation and deflation of the interest groups configuration from the original solution to the optimal configuration. As can be seen, the deflation is dramatic, while the rotation has been in excess of 45 degrees. Because of the deflation, very little difference remains between the great majority of the state interest group diversity scores. In order to minimize the differences of the most extreme scores, it was necessary to push the remaining data very close together. A clear interpretation of this result is that there are not sufficient differences in the diversities of state interest group communities to account for the differences in state policy innovation. This is especially clear in Figure 11 (on page 145), where the states with the most extreme scores for interest group diversity are not shown to have extreme policy innovation scores. Instead, the scores of these states seem consistent with the minority diversity and ideology of each state. 55 54 Because of the possibility of rotating the interest group configuration without decreasing its ability to accurately represent the solution, it does not matter where on the policy innovation configuration the states with extreme scores for interest group diversity would be so long as they are close together and extreme relative to other state policy innovation scores. 55 Remembering that minority diversity is a measure of the proportion of African Americans and Latinos, and the large Pacific Islander and Asian populations of Hawaii are not included. 143 Figure 10. Optimally Rotated Interest Group Diversity Configuration and the Original Interest Group Diversity Solution 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Note: The Optimally Rotated Interest Group Diversity Configuration is represented by open circles, clustered near zero) and the Original Interest Group Diversity solution is represented by plus signs, widely dispersed). 144 As mentioned previously, it is possible to calculate a correlation between the matrices for the two configurations. Although the correlation is said to be larger as the value approaches 1, Davison (1983, 198) and others have been reluctant to suggest a specific result that indicates a line of demarcation between a satisfactory and unsatisfactory level of correlation for interpretation purposes. Davison’s (1983, 198) rough rule of .7 for a good to fair match might seem stringent, but any standard is somewhat arbitrary. In Chapter 4 and Chapter 5, a correlation of .4 has been used as a target for a moderate correlation. 145 Figure 11. Innovation Optimally Rotated Interest Group Diversity Configuration and Policy 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Note: The Optimally Rotated Interest Group Diversity Configuration is represented by open circles and is mostly clustered near zero. The Configuration for Policy Innovation is represented by plus signs and is widely dispersed). Hawaii is included in graph. 146 The correlation between interest group density and the coordinates for policy innovation was .366. Because interest group density is the strongest correlation (.624) between any variable and the interest group diversity measure, it would be surprising if interest group diversity and policy innovation had a high correlation score in any case. In fact, the effort to set a boundary point is somewhat moot, given that the measure of correlation between the two configurations is .110. By any standard, this is a weak correlation suggesting a lack of similarity between the concepts, at least as it concerns every interest group community taken together. While this does not preclude an influence of interest groups within some of the state interest group communities, it presents a considerable challenge. Additionally, Figure 11 shows the relatively wide dispersion of the scores for policy innovation relative to interest group diversity (when optimally fitted). The close clustering of scores near (0, 0) suggests that interest group diversity provides little added information about policy innovation. However, one final attempt to find some association between interest group diversity and policy innovation is possible; using a comparison between the clusters for interest group diversity in Chapter 5 and the location of the policy innovation coordinates in Chapter 4. Perhaps, some types or levels of interest group diversity lead to some common outcomes for policy innovation. This is quite possible, if interest group communities need to meet certain thresholds to have any influence (and the states that do not meet this threshold introduce a considerable degree of random noise). If policy innovation and interest group diversity are similar concepts in some way under certain conditions that are not clear from looking at the entire United States, then looking at subsets of the interest group 147 community to see whether at least some of these have adopted similar policies should provide additional insight. The hierarchical cluster analysis of interest group diversity in Chapter 5 suggested six clusters. However, two of these clusters accounted for 41 of the state interest group communities. Another cluster (Cluster #4 in Chapter 5) accounted for four other states, including California, with extremely large interest group communities. The other five states might reasonably be termed atypical. Essentially, their interest group communities were extremely small. This chapter will look at Clusters #1 and #2 from Chapter 5, as well as a third spatial model showing the remaining states. Table 12 shows the clusters as they are examined in this chapter. As expected from the discussion of maximizing the correlation between the two configurations, it is the most extreme scores that appear to have the best fit. Possibly, this indicates that extremely large interest group communities lead to better representation, while the small interest group communities provide representation to only a few groups. The logic of this argument would suggest that, for states, this dichotomy is only clear at the extremes. However, this might be an artifact of the optimization and not a sign that extreme scores for the presence or absence of interest group density leads to an influence on policy innovation. 148 Table 12. Hierarchical Cluster Analysis of MDS Stimulus Points for Interest Groups Cluster #1 Alaska Colorado Florida Georgia Illinois Indiana Iowa Kansas Louisiana Maine Minnesota Missouri Montana Nebraska Nevada New Mexico New York Ohio Oklahoma South Dakota Virginia Washington Wisconsin Wyoming Cluster #2 Alabama Arkansas Connecticut Delaware Idaho Kentucky Massachusetts Mississippi New Hampshire New Jersey North Carolina North Dakota Pennsylvania South Carolina Tennessee Texas Vermont Remaining States Arizona California Hawaii Maryland Michigan Oregon Rhode Island Utah West Virginia The first step will be to determine whether the states in these clusters have similarities in their policy innovation locations. Chapter 5 has already shown that the dimensions on which these states have the highest correlations have been the level of interest group community density and the proportion of state employees. Therefore, it is not surprising that these groups are defined by these characteristics. The states in Cluster #1 had (on average) slightly larger interest group communities and slightly smaller proportions of their population working for their state governments. The states 149 in Cluster #2 had slightly smaller interest group communities and slightly larger proportions of their populations working for state governments. However, Chapter 5 did not examine policy innovation in the states within these clusters. This will be examined in Figures 12 through 14. Figure 12. Interest Group Cluster #1 and Policy Innovation 1 y .5 0 -.5 -1 -1 -.5 0 .5 x 150 1 Figure 12 shows interest group diversity Cluster #1 as adjusted for the configuration comparison done in this chapter. The first conclusion to be drawn is that the states included in this cluster come from across the entirety of policy innovation space. In fact, every policy innovation cluster is represented in this interest group diversity cluster. Therefore, almost any set of policy innovations is possible, when a state has a large diversity of interest groups. Policymaking is certainly not determined by whether states have large interest group diversity. However, there might be an overrepresentation of policy innovation clusters #2 and #3 (twelve of 18 and three of five respectively). Only nine of the other 27 states are represented. The states in policy innovation Cluster #2 were typically in the Midwest (west of the Great Lakes) and across the Northwest of the United States, and were often less diverse and more Republican than average (in the 1990s). However, the largely diverse policy innovation Cluster #3 was also well represented. In short, interest group diversity Cluster #1 provides little insight, except possibly to suggest small Midwestern states seem to be capable of large interest group communities in primary and secondary education policy. 56 This does not explain policy innovation however, because both states where policies related to equity and states focused on standards and accountability are wellrepresented within the states in Cluster #1. 56 As noted previously the main cause for larger (denser) interest group communities in primary and secondary education is having a large number of local school boards among the lobbying interest groups. If these states have more local control, then this might explain the interest group density here. 151 Figure 13. Interest Group Cluster #2 and Policy Innovation 1 y .5 0 -.5 -1 -1 -.5 0 .5 1 x Figure 13 shows interest group diversity Cluster #2 as it was adjusted for the configuration comparison in this chapter. The most obvious conclusion to be drawn is the under-representation of policy innovation Cluster #2 (three of 18). As noted in the previous chapter, these under-represented states are the ones with the fewest policy innovations. The eleven states with the fewest policy innovations were in this policy innovation cluster. In addition, every one of the states in the missing cluster, except 152 Missouri, was in the bottom half of the states in the total number of policy innovations. Of course these states have other characteristics in common. They are typically in the Great Plains through the Rocky Mountain West and generally small in population. Their governmental structures vary dramatically from legislative professionalism (Squire 2000) to the size of the state’s population. Both liberal and conservative states are represented in this cluster, as are states ranging in population size from the very small to the very large. Chapter 4 also suggested that these states were influenced by their low level of diversity. Ideologically, these states were typical of the country as a whole, but they were more Republican than most. Hero (1998) would have suggested that states with low levels of diversity (homogenous states) would be more likely to see reform efforts as help for the entire community as opposed to a conflict between groups within society. Instead, the low level of education reform might be seen as an example of greater satisfaction with the schools in these states. With these states absent from the cluster of states with more interest groups, it suggests that interest group density leads to a certain policymaking density. However, this result is very tenuous, because states with small interest group densities do not appear to show any pattern. 57 57 There also appears to be no correlation between proportion of the public who are state government employees and the number of policy innovation in education. 153 Figure 14. Interest Group Diversity for Remaining States and Policy Innovation 1 y .5 0 -.5 -1 -1 -.5 0 .5 x 154 1 Figure 14 does not show a single cluster but includes each of the remaining states that did not fit into the large clusters for interest group diversity. For the most part, these states do not have extreme scores for policy innovation (as many had for interest group diversity). The states with extremely small and non-diverse interest group communities comprise five of these states. None of these states stand out as atypical in terms of the types of policies innovations enacted. Gray and Lowery (1996, 254) suggested that only extremely high concentrations (and/or rapid growth) of interest groups should be expected to result in a distortion of policymaking. The position of California as both the state with the largest interest group community (and most diverse) and with extreme scores for policy innovation provides some evidence of this. However, California is on the extremes of many trends among the American states from being the largest to being one of the most diverse. Of the four states with large primary and secondary education interest group communities with a large proportion of these interest groups being local school boards, only California and Michigan have scores that are at least somewhat extreme. Unfortunately for the belief in interest group influence (at least as measured by its ability to distort outcomes predicted by other causes), the types of innovations adopted by these two states were very different—as well as predicted by moderate diversity and strong citizen liberalism in the case of Michigan, and by massive diversity and more moderate citizen liberalism on the part of California’s public. 155 Conclusions This chapter suggests that there is little similarity between the structure of the diversity of a primary and secondary education interest group community in a state and the structure of a state’s primary and secondary education policy innovation. In fact, the correlation (congruence) between the two spatial representations is very low (.110). While there is some indication that states with few interest groups adopt fewer policies, this does not translate into a similarity between states with similar interest group diversity patterns adopting a similar pattern of policy innovations. These, instead, are two measures of density: how many policy innovations and how many interest groups. The effort to find that certain types of interest groups are correlated with certain types of policy innovations did not uncover evidence of such congruence. Instead, the results from Chapter 4 provide more support for minority diversity and ideology as explanations of the types of policy innovation selected. Given the low correlation between policy innovation and interest group diversity, it is more reasonable to assume both are caused by different influences. Previous chapters have suggested that interest group diversity within primary and secondary education is associated with interest group density and the proportion of the state population working for the state government. Interest group density in turn is related to both full-time government employment and population (with some regional bias toward more interest groups in the western states). Thus, education interest groups appear to be products of the need to link people to their government more than specific policy preferences, while the most profound societal divisions determine education policy innovation preferences. 156 These results suggest two possible sets of conclusions—one set if the relationship between education policy and education interest groups can be generalized to all policy areas and another set if it is unique. First, if education policy is typical of other policy areas, then it is clear that interest group diversity is far less important than interest group density in policymaking. Interest group density is the driving force in diversity with diversity impossible without a threshold level of density. Furthermore, interest group density, unlike interest group diversity has a significant and weak to moderate association (.366) with policy innovation space, while diversity is correlated at only .110. However, this might be mitigated by the essentially constant nature of state teachers’ unions as interest groups within a state. They can never represent the largest share of the interest group community except in the smallest interest group communities. Local school districts represent the largest amount of the variation, so density becomes a proxy for the representation of school district employers versus teachers’ union employees. The conclusion offered is generally supported by other research into interest groups and policymaking (Shattschneider 1960; Edelman 1960; Olson 1965; Lowi 1979; Kingdon 1984; 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993) that a few interests dominate policymaking. Local school boards might represent different interest groups without representing different interests (as measured by viewpoint). The counterargument is that even interest group density is far from the strongest influence on policy innovation. Minority diversity and citizen ideology can largely account for the differences between states. These approaches are supported in part by two different lines of policymaking literature. Hero (1998) supported a social diversity approach that explained group conflict in terms not of the 157 small, organized interests but the larger societal cleavages. The approach of Erikson, Wright, and McIver (1993) explained policymaking in more optimistic terms of democratic theory, where citizen ideology and partisanship explained policy outcomes. If it is possible to generalize from primary and secondary education policy innovation and the corresponding interest group to the entire interest group community and all policy innovation, then we are left with the conclusion that policy innovation occurs as a result of the mass public pressing its preferences influenced by racial (and ethnic) interests as well as ideological predispositions across a spectrum of policies. However, there are reasons to suspect that primary and secondary education is not like other policy areas. First, no one should be surprised if disaggregating the policy community leads to different preferences for specific policies. For instance, Gilens (1999) and Jacoby (1994; 2000) suggested that some policies (especially welfare) stimulate different attitudes within the public. These attitudes need not be rooted in their surface manifestations. The saliency of minority diversity in this study supports some of the same underlying concepts (though education and welfare are very different policies). In addition, Jacoby and Schneider (2001; 2009) show that states have varying spending priorities for different policies. Jacoby and Schneider (2001) singled out education policy as being different from other policies because of its great importance within state budgeting. No matter what state governments would choose to do, it is considered one of their greatest priorities (while still being considered a collective good policy). The relationship between spending priorities and policy innovation does not need to be direct; however, policy alternatives for primary and education policy reform (as argued previously) were well-known by the public (standards and accountability 158 versus funding equalization). Given that one of the alternatives was closely associated with state spending, it should be expected that spending priorities and innovation would be associated for education policy. If it is not true that spending priorities and policy innovation are similar concepts for all policies, it is further evidence that policy areas are different. If the weak relationship between the interest group community and policy innovation in primary and secondary education policy innovation cannot be generalized to all policy areas, the finding remains useful. First, this provides support the idea that policymaking cannot be studied in whole assuming it is like a fractal (similar in subsets of decreasing size—such as single policy areas representing all policies, single cities representing all like cities, etc.). Second, education policy provides insight into an interest group community where there is no single large business interest but many small public sector employers/service providers (the school districts). This and other differences between education policy and other policy areas might provide information about why education policy is different. In this sense, it is possible to examine education policy comparatively not only from state to state but with other policies. Finally, education policy is one of the most salient issues for state government. The discovery that this is not a policy defined interest group activity but by mass preferences (whether nefarious or reputable origins) is reassuring to advocates of democracy in the states. In the final chapter, I will look to the implications of mass preferences driving public policy. 159 Chapter 7 Mass Preferences as an Explanation for State Policy Innovation This dissertation has argued that state primary and secondary education policy innovation in the 1990s aligned along two dimensions: minority diversity and ideology. While other possible explanations, such as interest group density, state population, and partisanship, correlated with these policy reforms, minority diversity and ideology had both a strong theoretical basis and strong empirical support. Together they present a consistent story of state education policy innovation during this time period. There appears to be little evidence that similarity of interest groups within a state’s interest group community has a direct influence on policymaking. In fact, interest group diversity appears to correlate most strongly with interest group density, but it is the number of groups and not their type that seems to have a possible influence on state education policy. Other than minority diversity and ideology, this research also indicates a regional influence within education policy innovation. Each of these results is consistent with an element of prior research, while suggesting several clear distinctions as well. This concluding chapter will summarize these research findings, place those findings within the context of state policymaking literature, and discuss its implications for state education policy. In many ways, the No Child Left Behind Act of 2001 (NCLB) can be seen as the culmination of the efforts in school reform begun in the 1980s with A Nation at Risk and the education reforms proposed at that time. NCLB answered many of the concerns of President Reagan’s education task force. The law purported to increase accountability 160 on both the state and local levels, provide greater educational choices for parents and their children, and even provide flexibility to states, school districts, and schools in carrying out reforms. President George W. Bush, who viewed the legislation as one of his chief priorities in the early days of his administration, had been governor of one of the most active states in education policy innovation (as measured by Chapter 4), and so this might be seen as a triumph for the view that states had served as policy laboratories. The education reform push at the state level had seemingly triumphed in Washington (Department of Education, NCLB Executive Summary 2004). However, as the culmination, NCLB might also be seen as the ending event in the 1990s era of high state policymaking activity. The focus of the Bush-era reforms had been on school and teacher accountability—the reforms advocated primarily by the conservative and highly diverse (and mostly Southern) states from the 1990s. While President Obama’s advocacy for a new NCLB did not abandon accountability, it focused instead on educational equity and opportunity, encouraging teachers to go “where they are needed most,” and improving assessments. President Obama’s approach was somewhat more consistent with the liberal and diverse states (often from the Northeast) as seen in Chapter 4, but it is also consistent with educational reform efforts in the era before A Nation at Risk (Department of Education, A Blueprint for Reform 2011). Perhaps, the biggest change, though, is that the debate is occurring at the national level instead of at the state level. It is still too early to determine when or whether state educational policy activity will rebound to the levels seen in the 1990s, when states seemed to take the lead as the national government appeared to devolve policymaking power. However, the history of 161 American education suggests that periods of dormancy in state policymaking should not be mistaken for extinction. Similarly, the focus on educational policy has typically been transitory. Whatever the lasting impact of the reforms (or at least the language of reform) from the 1990s: standards, accountability, school choice, teacher training, and flexibility for states and local schools, this period of educational reform provides a great deal of information about the similarities and differences between the choices states make in education policy, when given that choice. Minority Diversity, Ideology, and Region in Education Policy The policy space representing policy innovations in state primary and secondary education policy is highly correlated with minority diversity (.703) in a multidimensional scaling model. With few exceptions, these diverse states were much more active in developing policy innovations during the 1990s. Primarily, these innovations involved such issues as standards and accountability, school choice, and rules for teacher training. However, the most diverse states also engaged in efforts to promote educational equity, suggesting responsiveness to the concerns of the minority populations within those states. The less diverse states were considerably less active in all areas of education reform. While partisanship (.562), state population (.558), and interest group density (.366) all correlate at least moderately with state education innovation along similar proposed dimensions to minority diversity, none are as strong. Additionally, the dissertation argues that partisanship and interest group density are temporally subsequent to the racial and ethnic characteristics of the states. Similarly, the largest 162 states are almost uniformly more diverse than the small states (at least when diversity is measured by the proportion of residents who are African Americans or Latin Americans as in this study). Ideology provides a powerful explanation for policy variation between the states that is independent of the effect of minority diversity. Unlike partisanship (.562) and population (.537) with their very high correlations with minority diversity, the correlation between minority diversity and ideology is (-.019). As compared to other potential variables, ideology offers the most unique dimension when compared to minority diversity. This suggests that the use of both minority diversity and ideology together provides the greatest total insight into the variation between states in primary and secondary education policy. Additionally, ideology corresponds with a simple and standard explanation for differences in policy preferences (Erikson, Wright, and McIver 1993) and attitudes toward racial and ethnic groups. Although this runs into the ongoing concern about whether citizens have discernible or even meaningful ideologies (Campbell, Converse, Miller, and Stokes 1960), the difficulty people have utilizing ideology to make policy choices helps explain why the correlation is not even stronger. Ideology is also strongly associated with legislative professionalism, another variable associated with state policymaking and a possible institutional mechanism for an influence of ideology. On the surface, the strong correlation between minority diversity and state policy innovation provides support for the social diversity approach to explaining policymaking outcomes (Hero 1998; Tolbert and Hero 2001). Certainly, there is a large variation between the states with the most and least diversity. Hero (1998) argued that minority 163 diversity and the racial and ethnic structure of those with political power in a community would explain policy outcomes better than either partisanship or ideology. However, this is just a surface comparison. The results from Chapter 4 show that diverse states, even those without a large enough minority population to argue that a state was “heterogeneous,” tended to implement more policy innovations during the 1990s. While an argument can be made that many of the innovations aimed at standards and accountability could have been aimed at pointing out the problems caused by politically powerless minorities, highly diverse states, such as California, Texas, New Mexico, and Louisiana implemented more policy innovations than the supposedly bifurcated states of the Deep South, not different ones. Many innovations, such as school choice and charter schools that might be seen as the most conservative policies, carried the support of minority communities in large cities (Mintrom 2000). A more nuanced investigation might be able to determine whether the large number of policy innovations in diverse states could result from different challenges in those states. If policymaking is designed to solve problems, then a greater set of problems should lead to more policymaking. Conversely, citizens in a homogenous state might be more likely to view educational failure as problem for children like their own, while citizens in a diverse state are still likely to see education as important to their community. In any case, Mintrom (2000) provides the argument that support existed in many minority communities for the types of policy innovations adopted by the most active states in the 1990s. This dissertation also presented evidence for the independent influence of ideology in policy innovation. The influence of citizen ideology in policymaking was a 164 contribution of Statehouse Democracy (Erikson, Wright, and McIver 1993). The effort to link elections to policy is the unifying theme of Statehouse Democracy. In essence, the authors asserted that without a link between elections and policymaking, democracy did not exist. 58 If mass preferences are to influence policy, then it is still necessary to establish the measure of preference that should be associated with policy. Partisanship is a measure of the political “team” to which they are members. However, in the United States partisanship is not the measure of beliefs. Dahl (1957) termed parties “a team of men seeking to control the governing apparatus by gaining office in a duly constituted election” (25). By this measure, partisanship could either represent a set of shared values or not. As a coherent system of beliefs that constrains decision-making, ideology is a more likely measure of actual beliefs. Erikson, Wright, and McIver (1993) point to a very low correlation between ideology and partisanship of .08 (17). Generally, liberal states were less active in policy innovations during the 1990s. Most likely, this is because fewer unique liberal policies were presented in the mix during this time. One area where the differences were clear was in education funding. The large number of options for innovations in the area of standards and accountability provided more policies for conservative states (though many of these might not have been as significant—such as the number of different years where students were required to take proficiency tests). However, the minority diversity dimension comes much closer to representing a measure of how many total policies were adopted. 58 Statehouse Democracy (Erikson, Wright, and McIver 1993) begins with a quote from V. O. Key (1961) “Unless mass views have some place in the shaping of policy, all the talk about democracy is nonsense.” So clearly, Erikson, Wright, and McIver (1993) are not the initiators of the premise that democracy IS the matching of mass preference with government action. In essence, though, the authors make a compelling case that this democracy is existent within the states, rather than just a principle or goal. 165 With the moderate correlation between ideology and policy innovation, coupled with the moderate correlation between partisanship and policy innovation, the results of this research are not inconsistent with the Statehouse Democracy model. The MDS model provides a meaningful map of the data and clearly indicates that the role of minority diversity should not be ignored. This is especially true, because of the regional nature of data that becomes visible in the cluster analysis. In the 1990s, the states in the American South were at once the most conservative and the most dominated by the Democratic Party. This appeared to be an anomaly in the post-WWII United States, where the Republican Party was widely seen as the more conservative of the two parties in states where both parties were strong. In recent years, the partisan affiliations of elected officials has come to more closely approximate the outcomes predicted by state mass ideology with the Republican Party doing very well in the South, while the Democratic Party has strengthened in much of the Northeast. Generally, this has happened over an extended period of time, beginning around the time of V. O. Key’s (1949) Southern Politics in State and Nation. At the time of Southern Politics, Key noticed a few Republican breakthroughs in Border States (and considerable disordered political competition within the Democratic Party), but very little evidence of a two-party South. After the passage of the Civil Rights Act, and in stages where Republican gains extended increasingly to the local level, the Democratic Party became less powerful in the South. In the 1990s, this process was still ongoing. The states that clustered together to represent those that had been the most active policy innovators, but also the most conservative policy innovators in 1990, were 166 primarily Southern states. Thirteen of the 17 states were either Southern or Border states. While it is in itself interesting that the propensity to implement primary and secondary education policies in the 1990s had so much to do with whether slavery had been legal in that state in 1865, it should also remind the reader that political party identification was also an artifact of the past. While racial diversity might be seen as a characteristic that helped create political party identification in the South, this diversity has remained a characteristic, while partisanship has continued to evolve. This would seem to indicate that minority diversity would stand to remain an important characteristic to explain policy innovation, even as partisanship might eventually be subsumed into ideology. 59 A final consideration is the role of regional similarities in policymaking. Regional explanations of policymaking typically focus on either the spread of information (diffusion) or of policy pressures related to the decisions of other states (Walker 1969; Berry and Berry 1990; 1992). However, this spread of information has often been tested as it relates to the transfers of particularly mobile assets such as money. Berry and Berry (1990; 1992; 1994) looked at lotteries and tax policy. In each of these cases, these policies tend to maximize state competition, because of the view that money spent out of state or paid out of state is lost. However, education policy is distinguishable from virtually every other type of state policy. Certainly, some would see 59 Erikson, Wright, and McIver (1993) noted that the correlation between partisanship and ideology in non-Southern states was .48. While this is a moderate correlation, it would not be enough to argue that the concepts of ideology and partisanship were the same even if it were a nationwide phenomenon. However, major political parties in many countries are avowedly ideological. The shift in the South (and somewhat in the Northeast) illustrates the danger in associating support for one of Downs’ (1957) “teams” with a population characteristic. 167 it as what Jacoby and Schneider (2001) termed a “particularized benefit,” aimed at individuals who need a special service (as distinguished from those who would be able to pay for the policy themselves. Yet, this does not represent most of society. In fact, Jacoby and Schneider (2001) noted that education policy was quite unlike any other policy provided by states—and is viewed as nearer to a collective good without being a pure example of such (554). This would indicate that states have both some of the positive pressure to provide the collective good, as they would for roads or police protection, and (if the race-to-the-bottom approach is correct) at least some of the pressure not to provide services that are too good and encourage an influx of needy people. If economic competition is not the cause for the pressure to behave as other nearby states do, what is? The answer is likely as simple as who lives within a state. In fact, the level of minority diversity is one of the primary regional differences between states. The two clusters that were found to have large numbers of states bordering each other also were very similar in regional patterns of diversity. Not only is the southern United States the most diverse portion of the country, there is also a stretch of the country along its northern tier and through the Midwest (west of the Mississippi River) that had very little diversity as the 1990s began. These are also states where there was very little policy innovation during the 1990s. Additionally, there is regional variation in essentially every characteristic that might be used to examine state policy innovation. For these reasons, it is most useful to examine why regions are different from each other, rather than to assume Southern-ness or some other name for an amalgam of states explains their difference. 168 The Nature of State Interest Group Communities The data from Chapter 4 show a weak to moderate relationship between policymaking and the number of interest groups in a state interest group community. However, this is only one characteristic of an interest group community. Gray and Lowery (1996) presented the characteristics of an interest group community in two ways. First, the total number of interest groups was defined as interest group density. The breadth of the community, the number of types of groups, was defined as the interest group diversity. Chapter 5 investigated both the density and diversity of interest groups within primary and secondary education policy on the state level. Because density is simply the number of groups, the best representation for density proved to be a linear regression model. The two characteristics of states that were best associated with the density of interest group communities were state population and the total number of full time state employees. These two variables worked at cross-purposes with population leading to more density, while the number of full time state employees led to a decrease. The results of the regression are consistent with earlier research. Gormley (1996) and Jacoby and Schneider (2001) predicted a decrease in the density of interest group communities associated with the number of full time state employees. In the area of education policy, the role of state employees is enhanced because many of the people most directly subject to policymaking in the area of primary and secondary education are state employees: both the teachers and administrators in public schools. Population is another crucial variable, because interest groups provide both a needed linkage between government and the people as population size increases, but more 169 significantly, because as states increase in size the resources with which groups might form increases. Western states are also shown to have more interest group density than other states—largely but not entirely through the influence of California with its extremely large interest group community in primary and secondary education. Given previous research, the causes of interest group density were not surprising. However, interest group diversity presented a different challenge. When evaluating diversity, it is necessary to consider all of the ways in which an interest group community can vary. Primary and secondary education is already a small subset of the wide variety of interest groups active in a state. However, it is a particular purpose of this research to determine whether a specific interest group community can influence policy in its policy area. Gray and Lowery (1996) showed that diversity does not determine policy across policymaking across all policies, but this research is looking at a specific segment and in greater detail. This dissertation used multidimensional scaling to model spatially the dissimilarities between interest group communities in each state. In order to create this model, it was necessary to categorize the 1183 interest groups active in lobbying state governments in the area of primary and secondary education policy into 51 different categories. By far, the most common of these groups were local school boards and teachers’ unions. In many respects, these interest groups represent the management/labor conflict carried into the state policy arena. However, these groups have many interests in common, because state governments represent such a large source of funding for primary and secondary education. 170 The results for diversity show a clear dimension for interest group diversity of interest group density. While the great concentration of interest groups in just the two areas of school boards and teachers’ unions is supportive of Gray and Lowery’s (1996) concern that diversity does not keep pace with increases in density, it is certainly true that without density, very little diversity seems to exist within state primary and secondary education interest group communities. When there are very few interest groups, almost all of them are local school boards and teachers’ unions. When a state had many interest groups, the greatest proportion of them remained local school boards and teachers’ unions. Often the increased number of local school boards represented all but a tiny fraction of the difference between the states with many and those with only a few. Even in California, the Jupiter of states, with 184 of 1183 interest groups (15.6% of all interest groups), half of the interest groups were school districts. Conversely, Hawaii had only one primary and secondary education interest group—perhaps owing to its unique state school system without local school districts. In addition to interest group density, the only other variable to correlate fairly strongly with interest group diversity was the number of state employees per 10,000 people. This rate of state employees correlated at .589, while interest group density correlated at .624. The rate of state employees, literally total state employees divided by population, suggests a further similarity between the concept of diversity and density when it comes to state education policy. Senior bureaucrats might be the most important lobbyists for the interests of their own agencies, making the effort to develop relationships with policymakers, and utilizing their expertise and long-term focus to help solve policy problems in ways favorable to their agency. However, it is also true that a 171 fairly high percentage of state government employees are not bureaucrats in the way commonly conceptualized, but instead, teachers. In 2000, almost 3 million of the 18 million state government employees were public school teachers (US Census Bureau, Table 254 and Table 461). Few Americans voluntarily place as much trust in any other governmental official as public school parents place in teachers. In keeping with the mass opinion theme of this research, a large percentage of the population is able to determine from first-hand interactions its view of school effectiveness and needs. Certainly, many Americans have interactions and more intimate interactions with government employees in education than in other policy areas. This is a unique characteristic of education policy. However, it is also possible that education policy simply represents a subset of the entire state policy community and it is necessary to see how statewide organization in education interests compares to other policy areas to see which receives the attention of policymakers. 60 This result for interest group diversity suggests several normative challenges for those who are concerned that interest groups represent a small segment of societal interests. Lowi (1979) was especially concerned with the power of interest groups to capture those agencies who purpose it was to regulate them. This is consistent with other research (Schattschneider 1960; Edelman 1960) that argues that organized interests have undue influence on policymaking. While the results of a configuration 60 Using this approach, education policy would be compared to health care, transportation, criminal justice, and other policies to determine whether it is more strongly represented than the others. If state education policy has more density, it could indicate a diverse general policy area where education interests have disproportionate strength. Thus, the states with the largest portion of their interest group community comprised of elementary and secondary education organized groups should have more interest in education on the part of state lawmakers. 172 comparison did not support the notion that interest group diversity is linked to policymaking in education policy, the seeming concentration of control of access to government in states with dense interest group communities by a very few interest groups types should be an issue for concern. Furthermore, if increasing the number of interest groups, leads to a larger proportion of those groups representing a small segment of interest, but it is also the chief mechanism for increasing diversity, then the solution is unclear. Interest Groups and Policy Innovation The results of this research cast a doubtful eye on the role of interest groups in state policymaking, either as linkages between policymakers and the public, or as malignant forces that skew public preferences. On the national level, the assumption that interest groups dominate the policy arena is widely-accepted. Although the research has not been uniform in its view of the normative nature of interest group, varied explanations of policymaking have come to terms with the role of organized interests (Shattschneider 1960; Edelman 1960; Lowi 1979; Kingdon 1984; 2003; Baumgartner and Jones 1993; Sabatier and Jenkins-Smith 1993). However, the connection between organized interests and state governments has not been as clear. Gray and Lowery (1996) suggested that interest group density had some role, while interest group diversity did not. On the surface, that result is counterintuitive. If diversity does not matter, then policy innovations (or other measures of state policy) are susceptible to an increase in the same types of interest groups but not to the addition of a new voice. 173 The results appear to support the view that increased density leads to increased diversity, just as indicated by Gray and Lowery (1996). Furthermore, the configuration comparison provides considerable evidence that there is little correspondence (correlation) between policy innovation and the types of interest groups in the primary and secondary education policy communities of the various states (.110). This correspondence is much less than the correlation between interest group density and policy innovation (.366). When considering the high degree of correlation between interest group density and interest group diversity (.624), it is surprising just how much less there is between interest group diversity and policy innovation. This would seem to support the conclusions of Gray and Lowery (1996) with respect to the entire state policy community. Conversely, this outcome would appear to support neither those whose expectations represent the fondest hopes of pluralists (Truman 1951; Dahl 1961), nor the greatest fears of those who fear the distortions of interest groups (Olson 1965; Lowi 1979). Although the weak to moderate correlation between interest group density and policy innovation should not be ignored—and might yet represent a significant relationship, if it were tested in such a way, an improved mechanism for the relationship between interest group density and public policy should be found. If as some suggest (Edelman 1960; Schattschneider 1960; Olson 1965; Lowi 1979) interest groups distort policy outcomes away from the outcome expected if the voice of each member of the mass public had equal weight, then it does not stand to reason that additional members of the policy community—all attempting to echo the initial voices—should lead to a new outcome. 174 The cluster analysis for interest group diversity presents several questions that should be of interest, when it is compared to the cluster associated with the fewest policy innovations in Chapter 4. The interest group cluster associated with a fairly large (but not too large) density of interest groups appears to have either very few or a great many policy innovations. Similarly, California, a state with too many interest groups to fit comfortably in either of the two large clusters, is also among the states with the most policy innovations. 61 If this result is indicative of anything, it probably suggests that interest groups help states to accentuate their already existent policy innovation preferences. The states with fewer interest groups would appear to enact policy innovations more closely aligned with their ideologies. Perhaps, this is suggestive of an avenue for further research. Primary and Secondary Education Policy Innovation The 1990s were a period of extensive state activity educational policymaking. As such, research might indicate that this would be a period when small interest group communities that do not represent the entire population would have less influence than typically. Baumgartner and Jones (1993) suggested that public policy is usually mired in periods of incremental change. During those periods, a few powerful interest groups have clear lines of communication within venues that are stable and longstanding. However, Baumgartner and Jones (1993) suggested there are also periods of great 61 Texas is the final state among those with a large number of policy innovations. It does not qualify as a state with a large interest group density. However, its interest group density is relatively high for its region of the country. Nevertheless, it might correctly be seen as atypical in terms of interest group density and supportive of the minority diversity and ideology only approach. 175 change, where these established rules no longer apply, and new policymakers as well as interests begin to pay attention. As noted in Chapter 2, it is common in the area of education reform for reformers to shop venues, attempt to raise the stakes of the conflict by claiming crises, and even to associate education with larger national issues, such as economic security or national defense. If this is what occurred during the 1990s, a diminished role for organized interests would be expected. This would especially be true, if those interests represented only a segment of policymaking thought in education policy. The dominant interest groups in elementary and secondary education policy are local school districts and employee unions. While those interest groups are often in competition with each other over whether resources or potential resources are to be spent on salaries or elsewhere, these two groups have many interests in common. Chief among their shared interests are their efforts to secure additional resources at the federal, state, and local levels. To the degree that acquisition of these revenues is the goal of their industry (in the way that both automakers and auto workers’ unions hope that cars are sold), then both labor and management spend much of their time on the same page—attempting to make their brand of educational provision appear effective but for the need of additional resources. If there is to be significant elementary and secondary education reform, these powerful interest groups must either be divided against each other or steamrollered by events (or a crisis). Mintrom (2000) proposed that school choice advocacy in the 1990s occurred largely through the efforts of policy entrepreneurs. These policy entrepreneurs overcame the difficulty typically encountered in forming large interest groups designed 176 to represent the mass public by having a single individual bear both the financial burden and provide a face to the group desires. Whether the advocacy efforts of a single person are adequately measured as being lobbying in educational policy is unclear. Additionally, interest groups within other policy arenas might moonlight as education policy advocates. These might include universities and other post-secondary education providers seeking to increase the abilities of students they receive, and civil rights organizations who seek to benefit disadvantaged groups through education. Similarly, if the perceived crisis in education led industry groups to lobby on education policy, this might not be clear from the documentary evidence, because they might also have advocated that year on their parochial interests. For instance, efforts by oil companies to bring attention to educational deficits might be seen more as public relations on the part of those corporations, but they suggest to viewers a continuing crisis. They might also be viewed as efforts by corporations to externalize the cost of creating potential employees, if a commercial advocates math and science education particularly. Almost all of the organized interests in education policy have a financial interest in public education. However, their preferences for policy innovations do not always reflect the interests of either the mass public or even every organized interest’s interest in elementary and secondary education policy (Mintrom 2000; cf. Chubb and Moe 1990). Whatever the cause, when the period of dramatic policy change occurs, it would be expected that the mass public’s views on the issue are more salient than during those periods of incremental change. When this occurs, the mass public would be expected to bring its beliefs as well as its biases to the search for policy innovations designed to solve the crisis. In a sense, ideology and minority diversity are highly 177 reflective of the public’s beliefs and biases. Therefore, it might be possible to argue that a period of rapid policy change in elementary and secondary educational is the best time to uncover characteristics of a state’s mass public that lead to innovations within the policy area, while at the same time; it gives a poor indication of the role of those groups who have the most influence during normal times. st The 21 Century has not yet been a time when education reform has caught the attention of the American public. In addition to the passage of No Child Left Behind, the st first decade of the 21 century also marked the beginning of a new crisis for the American government that was not immediately linked to progress in education. Instead, the war on terror led to increased spending on two wars, a fitful decade for economic growth, and a homeland security effort that focused on nationalization. In a sense, this might be thought of as normal times for elementary and secondary education policy. If so, this presents an opportunity to examine whether interest groups have gained influence in education policy. One drawback to the assumption that these are typical times in education is the sluggish economic picture. If state governments do not have the resources to fund education, then this might lead to additional competition between state governments and education unions. Much of this conflict animated news reports in 2011 from states such as Wisconsin and Ohio where policymakers tried to cut teachers’ benefits and union bargaining rights. These efforts would be state reforms of a very different kind than those seen in the 1990s, and a type of policy innovation not even studied for this dissertation. 178 Points for Future Research This research is limited in its focus to elementary and secondary education. While this is one of the most salient issues the American federal system trusts largely to the states, it is a single issue. State governments are key policy players in a range of issues from policing to health care. State policy communities exist in every area states consider. 62 Whether the most effective place to proceed is by comparing education to various collective goods versus particularized benefits or to look at the breadth of policymaking, a look at education is limited to drawing conclusions in a single area. If limiting future research to elementary and secondary education policy, it would be useful to look at the extent to which the 1990s were typical of other periods. While the history of education policy suggests that the national government has acted as a liberalizing force through much of its career as an arbiter of education policy, it would be useful to consider the location of the national government in policy space using the various incarnations of the Elementary and Secondary Education Act. Therefore, the next step would be to look at the states for their variance from the preferences of the national government as well as an examination across time. If the interest groups in elementary and secondary education policy had diminished influence during the 1990s, this could be demonstrated by looking across time to see whether that influence varies. Education policy has been seen as a remedy for many different societal ills, it would be expected that the specific problem education policy innovation has been tasked to solve should point to the group advocating that solution. 62 Jacoby and Schneider (2001) consider 15 policies including veterans’ benefits, housing and community development, parks and recreation, inspections, police and law enforcement, natural resources, government administration, highways, education, welfare, hospitals, health care, corrections, transportation, and employment security. 179 Additionally, this research minimized the consideration of policy spending— focusing almost entirely on policy innovation or the specific policies adopted. However, current economic conditions in education policy in the United States invite the question of whether a policy innovation that has not been fully funded is really a policy change. Edelman (1960) suggested that mass (or diffuse) interests receive words, while organized interests receive policy. On a different level, it is perhaps possible that a somewhat diffuse interest might receive policy during times of significant policy change, while organized interests receive funding. It is clear that future research should take this possibility into account. Lowi (1979) suggested that it is a tactic of interest groups to attempt to remove their needs from political debate. 63 Nevertheless, policy innovations and policy spending priorities appear to be as linked as any policy authorization/appropriations dyad. Chapter 4 suggested strongly that state policy innovation and state spending priorities are linked for at least a few states. While many states, particularly those where there are either very large or very small levels of total policy innovation (and where either very high or very low levels of minority diversity appears to be a salient characteristic of the state), appear to have a less direct connection between innovations and spending, this is not true of the states where ideology is most salient. Where ideology is an important component, and states have engaged in a moderate to high level of total innovation, conservative states appear to innovate less in education and 63 Lowi’s (1979) discussion of agriculture policy points to the effectiveness of both small farmers and large mills use of the symbolic importance of the family farm in American political history to overcome the power of market forces that had consolidated other producers. Public education policy certainly offers another symbolically loaded, yet financially strapped political arena. 180 balance toward collective goods policies (as defined by Jacoby and Schneider 2001), while liberal states innovate somewhat more and balance toward particularized benefits. Finally, further research could focus more on the nature of charter school arrangements from state-to-state. While charter schools appear to have developed as a key surviving element of the school choice movement, the policy is represented by only a few variables. While this is most likely appropriate for research on all policy innovations in elementary and secondary education, a closer look at school choice and charter schools could provide a great deal of insight, especially because it was such a controversial subject when it was first introduced and seemed to cut across political and racial boundaries (cf. Mintrom 2000). Conclusion The period of rapid education policy change driven by state governments during the 1990s has ended. NCLB shifted the focus of the policy to the national government. The war on terror and the difficult economic conditions over the past decade decreased the salience of education reform, as people focused on other issues. However, this did not mark a permanent end to educational innovation in the United States. American history has seen several periods where there has been considerable policy change and others when the pace has been much slower. Additionally, the level of government that has taken the lead in education has also varied across time. Through most of this time, the national government has taken the lead whenever education reformers have been able to convince the public that a crisis has existed. Often this crisis has involved 181 inadequate funding, stark inequality, or a perception that America’s national defense or economic security has been harmed by an inadequate educational system. During the 1990s, states were allowed to innovate in the area of elementary and secondary education, while the national government was unable to affect much change because of divided government and considerable belief that states were the most appropriate venue for education policy. By the time the 1990s ended, states had engaged in hundreds of different policy innovations (404 were measured in this research). The policy innovations selected appear to have aligned along two dimensions: a state’s level of minority and a state’s citizen ideology. Meanwhile, the interest group communities of the states appeared to have little correlation on those policy innovations with the number of interest groups providing a stronger argument than the diversity of those interest groups. The results are highly suggestive that something different is at work for policy innovation in primary and secondary education than for other seemingly similar concepts, such as policy spending or policy outcomes. Nevertheless, the results are consistent with previous research such as Gray and Lowery (1996), who were skeptical of the role of interest group diversity in state policymaking, Erikson, Wright, and McIver (1993), who suggested that state ideology would have an influence on state policymaking, and Hero (1998), who suggested a role for minority diversity. When President George W. Bush signed NCLB, in a large sense he was nationalizing a policy that had been formulated on a state level. Essentially every reform was consistent with a policy that had gained popularity in at least a few of the reform-minded states. Yet these innovations also ran counter to the spirit of the 1990s 182 by requiring (or incentivizing) states to implement policies mandated by the national government (Department of Education. NCLB Executive Summary. 2004). Even now, states are balking at the requirements of NCLB, with state waivers a key element of President Obama’s reform efforts (Department of Education. A Blueprint for Reform. 2011). While this dissertation ends with the observation that the period studied is potentially unlike the present, it should also serve as a cautiously optimistic view of the democratic system. If democracy is a linkage between policy preferences and policy outputs, then it is heartening to see both the beliefs and the biases of the American people in close correlation with the policy innovations enacted during a period of recent history. Additionally, race and ethnicity have been controversial issues in American history, yet it would appear the greatest number of education reforms was enacted in those states where diversity was the highest. This was true not only in states where the diversity was too great for policymakers to ignore the concerns of minorities, but also in those states where diversity was somewhat less (places where Hero [1998] had suggested that a bifurcated polity would ignore the concerns of minorities). Although it is possible to debate the motives behind policies designed to punish failing schools, many of these reforms were the same as those advocated by reformers and not those seeking the demise of minority education (Chubb and Moe 1990; Mintrom 2000). While views toward African Americans and Latinos might have influenced whether Americans thought their schools were failing, the reforms suggested did not simply represent white voters’ efforts to impose these innovations on minorities (cf. Mintrom 2000, for a discussion on the large number of minorities who favored education reform). 183 Nevertheless, this paints a portrait, not without warts, of citizen preference influencing public policy, a measure of democracy. 184 BIBLIOGRAPHY 185 BIBLIOGRAPHY Bartholomew, David J., Fiona Steele, Irini Moustaki, and Jane I. Galbraith. 2002. The Analysis and Interpretation of Multivariate Data for Social Scientists. Texts for Statistical Science. Boca Raton (FL): Chapman and Hall/CRC. Baumgartner, Frank R. and Bryan D. Jones. 1993. Agendas and Instability in American Politics. Chicago: University of Chicago Press. Berelson, Bernard R., Paul F. Lazarsfeld, and William N. Mc Phee. 1954. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press. Berry, Frances Stokes, and William D. Berry. 1990. “State Lottery Adoptions as Policy Innovations: An Event History Analysis” American Political Science Review 84: 395-415. ________. 1992. “Tax Innovation in the States: Capitalizing on Political Opportunity.” American Journal of Political Science. 36: 715-743. ________. 1994. “The Politics of Tax Increases in the States.” American Journal of Political Science. 38: 855-859. Berry, William D., Michael B. Berkman, and Stuart Schneiderman. 2000. “Explaining Incumbency Reelection.” American Political Science Review. 94(4): 859-874. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 19601999.” American Journal of Political Science. 42(1): 327-348. Berkman, Michael B. 2001. “Legislative Professionalism and the Demand for Groups: The Institutional Context of Interest Population Density.” Legislative Studies Quarterly. 26(4): 661-679. Berry, Jeffrey M. 1997. The Interest Group Society. New York: Longman. Bibby, John F. and Thomas M. Holbrook. 2004. “Chapter 3: Parties and Elections.” In Gray, Virginia and Russell L Hanson ed. Politics in the American States: A Comparative Analysis, Eighth Edition. Washington: CQ Press. 62-99. Bischoff, Ivo. 2003. “Determinants of the Increase in the Number of Interest Groups in Western Democracies: Theoretical Considerations and Evidence from 21 OECD Countries.” Public Choice. 114(1/2): 197-218. 186 Black, Earl and Merle Black. 1987. Politics and Society in the South. Cambridge MA: Harvard University Press. ________. 2002. The Rise of Southern Republicans. Cambridge MA: Belknap Press of Harvard University Press. Boehmke, Frederick J. and Richard Witmer. 2004. “Disentangling Diffusion: The Effects of Social Learning and Economic Competition on State Policy Innovation and Expansion.” Political Research Quarterly. 57: 39-51. Brown v. Board of Education, 347 U.S. 483 (1954). Broughman, Stephen P. and Lenore A. Colaciello. 2001. Private School Universe Survey: 1999-2000, NCES 2001-330. Washington: U.S. Department of Education, National Center for Education Statistics. Burns, Nancy. 1994. The Formation of American Local Governments: Private Values in Public Institutions. New York: Oxford University Press. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. New York: John Wiley and Sons. Chubb, John E. and Terry M. Moe. 1990. Politics, Markets, and America’s Schools. Washington: The Brookings Institution. Cohen, Michael D., James G. March, and Johan P. Olsen. 1972. “A Garbage Can Model of Organizational Choice.” Administrative Science Quarterly. 17: 1-25. Cox, Gary W. and Jonathan N. Katz. 2002. Elbridge Gerry’s Salamander: The Electoral Consequences of the Reapportionment Revolution. Cambridge: Cambridge University Press. Dahl, Robert A. 1961. Who Governs?: Democracy and Polwer in an American City. New Haven CT: Yale University Press. Davison, Mark L. 1983. Multidimensional Scaling. Wiley Series in Probability and Mathematical Statistics. New York: John Wiley and Sons. Department of Education. 2004. The No Child Left Behind Act of 2001. Executive Summary. Document Archive. www2.ed.gov/nclb/overview/intro/execsumm.html. Retrieved July 6, 2012. ________. 2011. Priorities in a Blueprint for Reform. www2.ed.gov/policy/elsec/leg/ blueprint/publication_pg3.html#part3. Retrieved July 6, 2012 187 Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row. Dreier, Peter, John Mollenkopf, and Todd Swanstrom. 2004. Place Matters: Metropolitics for the Twenty-First Century. Second Edition, Revised. Lawrence: University of Kansas Press. Eastman, John C. 1998. “When Did Education Become a Civil Right? An Assessment of State Constitutional Provisions for Education, 1776-1900. The American Journal of Legal History. 42(1): 1-34. Easton, David. 1953. The Political System. New York: Alfred Knopf. Edelman, Murray. 1960. “Symbols and Political Quiescence.” American Political Science Review. 54: 695-704. Epstein, David, and Sharyn O’Halloran. 1994. “Administrative Procedures, Information, and Agency Discretion.” American Journal of Political Science. 38: 697-722. ________. 1996. “Divided Government and the Design of Administrative Procedures: A Formal Model and Empirical Test.” Journal of Politics. 58: 373-397. ________. 1999. Delegating Powers: A Transaction Cost Politics Approach to Policy Making Under Separate Powers. Cambridge: Cambridge University Press. Erikson, Robert S., Gerald C. Wright, and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. Cambridge: Cambridge University Press. Eyestone, Robert. 1977. “Confusion, Diffusion, and Innovation.” American Political Science Review. 71: 441-447. Foner, Eric. 1990. A Short History of Reconstruction: 1863-1877. New York: Harper & Row. Frankfort-Nachmias, Chava and David Nachmias. 2000. Research Methods in the Social Sciences. Sixth Edition. New York: Worth. Franklin, John Hope. 1961. Reconstruction After The Civil War. Chicago: University of Chicago Press. Friedman, Milton. 1962. Capitalism and Freedom. Chicago: University of Chicago Press. 188 Garand, James C. and Kyle Baudoin. 2004. “Fiscal Policy in the American States.” In Politics in the American States: A Comparative Analysis. Eighth Edition. Virginia Gray and Russell L. Hanson ed. Washington: CQ Press. Gilens, Martin. 1999. Why Americans Hate Welfare: Race, Media, and the Politics of Antipoverty Policy. Chicago: University of Chicago Press. Gormley, William T., Jr. 1996. “Accountability Battles in State Administration.” In Van Horn, Carl E. The State of the States. 3rd Edition. Washington, DC: Congressional Quarterly Press. Gray, Virginia. 1973. “Innovation in the States: A Diffusion Study.” American Political Science Review. 67: 1174- 1185. Gray, Virginia and Russell L. Hanson ed. 2004. Politics in the American States: A Comparative Analysis. Eighth Edition. Washington: CQ Press. Gray, Virginia and David Lowery. 1993. “The Diversity of State Interest Group Systems.” Political Research Quarterly. 46: 81-97. ________. 1996. The Population Ecology of Interest Group Representation: Lobbying Communities in the American States. Ann Arbor, MI: University of Michigan Press. ________. 1998. “Representational Concentration and Interest Community Size: A Population Ecology Interpretation.” Political Research Quarterly. 51: 919-944. Grodzins, Morton. 1966. “Centralization and Decentralization in the American Federal System.” In Elazar, Daniel J., ed. The American System: A New View of Government in the United States. Chicago: Rand McNally. Hammond, Thomas H. and Jack H. Knott. 1996. “Who Controls the Bureaucracy?: Presidential Power, Congressional Dominance, Legal Constraints, and Bureaucratic Autonomy in a Model of Multi-Institutional Policymaking.” Journal of Law, Economics, and Organization. 12:1 (April) 121-168. Hayward, Clarissa Rile. 2003. “The Difference States Make: Democracy, Identity, and the American City.” American Political Science Review. 97(4): 501-514. Heclo, Hugh. 1978. “Issue Networks and the Washington Establishment.” In Anthony King, ed. The New American Political System. Washington, DC: The American Enterprise Institute, 87-124. Hegreness, Matthew J. 2011. “An Organic Theory of the Fourteenth Amendment: The Northwest Ordinance as the Source of Rights, Privileges, and Immunities.” Yale Law Journal. 120: 1820-1884. 189 Heinz, John P., Edward O. Laumann, Robert L. Nelson, and Robert H. Salisbury. 1993. The Hollow Core. Cambridbe, MA: Harvard University Press. Henig, Jeffrey R. 1994. Rethinking School Choice: Limits of the Market Metaphor. Princeton: Princeton University Press. Hero, Rodney. 1998. Faces of Inequality: Social Diversity in American Politics. New York: Oxford University Press. Hill, Kim Quaile and Jan E. Leighley. 1994. “Mobilizing Institutions and Class Representation in U.S. State Electorates.” Political Research Quarterly. 47(1): 137-150. Holland, Spencer H. 1996. “Project 2000: An Educational Mentoring and Academic Support Model for Inner-City African American Boys.” The Journal of Negro Education. 65(3): Educating Children in a Violent Society, Part I: 315-321. Huckfeldt, Robert and John Sprague. 1992. “Political Parties and Electoral Mobilization: Political Structure, Social Structure, and the Party Canvass.” American Political Science Review. 86(1): 70-86. Hurst, David, Alexandra Tan, Anne Meek, and Jason Sellers. 2003. Overview and Inventory of State Education Reforms: 1990 to 2000, NCES 2003-020. Project Officer: Edith McArthur. Washington: U.S. Department of Education, National Center for Education Statistics. Jacoby, William G. and Saundra K. Schneider. 2001. “Variability in State Policy Priorities: An Empirical Analysis.” Journal of Politics. 63: 544-568. ________. 2009. “A New Measure of Policy Spending Priorities in the American States.” Political Analysis. 17:1-24. Jacoby, William G. 1994. “Public Attitudes Toward Government Spending.” American Journal of Political Science. 38:336-361. ________. 2000. “Issue Framing and Public Opinion on Government Spending.” American Journal of Political Science. ________. 2007. "CONFCOMP: A SAS Macro for Configuration Comparison." Unpublished MS. Jewell, Malcolm E. 1982. “The Neglected World of State Politics.” The Journal of Politics. 44(3): 638-657. Key, V. O., jr. 1949. Southern Politics in State and Nation. New York: Knopf. 190 ________. 1956. American State Politics: An Introduction. New York: Knopf. Kantor, Harvey. 1991. “Education, Social Reform, and the State: ESEA and Federal Education Policy in the 1960s.” American Journal of Education. 100(1): 47-83. Kingdon, John W. 1984. Agendas, Alternatives, and Public Policies. Boston: Little, Brown. ________. 2003. Agendas, Alternatives, and Public Policies. Second Edition. New York: Longman. Kruskal, Joseph B. and Myron Wish. 1978. Multidimensional Scaling Sage Series: Quantitative Applications in the Social Sciences. Beverly Hills (CA): Sage. Ladd, Helen F. and Janet S. Hansen. 1999. Making Money Matter: Financing America’s Schools. Washington: National Academy Press. Leech, Beth L., Frank R. Baumgartner, Timothy M. La Pira, and Nicholas A. Semanko. 2005. “Drawing Lobbyists to Washington: Government Activity and the Demand for Advocacy.” Political Research Quarterly. 58(1): 19-30. Lieberman, Robert C. and Greg M. Shaw. 2000. “Looking Inward, looking Outward: The Politics of State Welfare Innovation under Devolution.” Political Research Quarterly. 53: 215-240. Lindblom, Charles E. 1959. “The Science of ‘Muddling Through.’” Public Administration Review. 39:517-526. ________. 1979. “Still Muddling, Not Yet Through.” Public Administration Review. 39: 517-526. Lowery, David, Virginia Gray, Jennifer Wolack, Erik Godwin, and Whitt Kilburn. 2005. “Reconsidering the Counter-Mobilization Hypothesis: Health Policy and Lobbying in the American States. Political Behavior. 27(2): 99-132. Lowery, David and Virginia Gray. 1998. “The Dominance of Institutions in Interest Group Representation: A Test of Seven Explanations. American Journal of Political Science. 42:231-255. Lowi, Theodore J. 1979. The End of Liberalism: The Second Republic of the United States. New York: Norton. McConnell, Grant. 1966. Private Power and American Democracy. New York: Alfred A Knopf. 191 McFarland, Andrew S. 1987. “Interest Groups and Theories of Power in America.” British Journal of Political Science. 17: 129-147. McKenzie, Brian D. 2004. “Religious Social Networks, Indirect Mobilization, and AfricanAmerican Political Participation.” Political Research Quarterly. 57(4): 621-632. Mintrom, Michael. 2000. Policy Entrepreneurs and Social Choice. Washington: Georgetown University Press. Moe, Terry M. 1995. “Private Vouchers.” In Terry M. Moe, ed. Private Vouchers. Stanford: Hoover Press. ________. 1981. “Toward a Broader View of Interest Groups.” Journal of Politics. 43: 531-543. Mohr, Lawrence B. 1969. “Determinants of Innovation in Organizations.” American Political Science Review. 63: 111-126. Mooney, Christopher Z. 2001. “Modeling Regional Effects on State Policy Diffusion.” Political Research Quarterly. 54:103-124. ________. 1995. “Citizens, Structures, and Sister States: Influences on State Legislative Professionalism.” Legislative Studies Quarterly. 21:57-72. Moser, Michele and Ross Rubenstein. 2002. “The Equality of Public School District Funding in the United States: A National Status Report.” Public Administration Review. 62(1): 63-72. Mossberger, Karen. 2000. The Politics of Ideas and the Spread of Enterprise Zones. Washington, DC: Georgetown University Press. National Commission on Excellence in Education. 1983. A Nation at Risk: The Imperative for Educational Reform. Washington: U.S. Government Printing Office. Nownes, Anthony J., and Grant Neeley. 1996. “Public Interest Group Entrepreneurship and Theories of Group Mobilization.” Political Research Quarterly. 49: 119-146. Nownes, Anthony J. and Patricia Freeman. 1998. “Interest Group Activity in the States.” Journal of Politics. 60(1): 86-112. Obama, Barack. 2011. “Remarks by the President in State of Union Address.” The White House: Office of the Press Secretary. 192 Olneck, Michael R. 1989. “Americanization and the Education of Immigrants, 19001925: An Analysis of Symbolic Action.” American Journal of Education. 97(4): 398-423. Olson, Mancur. 1965. The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge MA: Harvard University Press. Parrish, Thomas B. and Jay G. Chambers. 1996. “Financing Special Education.” The Future of Children 6(1): Special Education for Children with Disabilities: 121-138. Peterson, Paul E. 1981. City Limits. Chicago: University of Chicago Press. ________. 1995. The Price of Federalism. Washington, DC: Brookings. Pressman, Jeffrey L. and Aaron Wildavsky. 1984. Implementation: How Great Expectations in Washington are Dashed in Oakland. Third Edition. Berkley: University of California Press. Rabinowitz, George. 1978. “On the Nature of Political Issues: Insights from a Spatial Analysis. American Journal of Political Science. 22:793-817. Rogers, Everett M. (1962) 1995. Diffusion of Innovations. New York: Free Press. Sabatier, Paul A. 1993. “Policy Change Over a Decade or More.” In Sabatier, Paul A. and Hank C. Jenkins-Smith, ed. 1993. Policy Change and Learning: An Advocacy Coalition Approach. Boulder, CO: Westview Press. Pp. 13-39. Sabatier, Paul A. and Hank C. Jenkins-Smith, ed. 1993. Policy Change and Learning: An Advocacy Coalition Approach. Boulder, CO: Westview Press. Savage, Robert. 1978. “Policy Innovativeness as a Trait of American States.” Journal of Politics. 40: 212-219. Schattschneider, E. E. 1960. The Semisovereign People:A Realist’s View of Democracy in America. New York: Holt, Rinehart, and Winston. Schneider, Saundra K. and William G. Jacoby. 2006. “Citizen Influences on State Policy Priorities: The Interplay of Public Opinion and Interest Groups.” A Paper Delivered at The State Politics and Policy Conference at Texas Tech University. ________. 1996. “Influences on Bureaucratic Policy Initiatives in the American States.” Journal of Public Administration Research and Theory. 6: 495-522. Schram, Sanford F. and Carol S. Weissert. 1999. “The State of American Federalism 1998-1999.” Publius: The Journal of Federalism. 29(2): 1 -34. 193 Sielke, C.C. and T. Holmes. 2002. “Overview of Approaches to State School Funding.” In American Education Finance Association, Public School Finance Programs of the United States and Canada, 1998-1999. American Education Finance Association. Simon, Herbert A. 1986. “Theories of Bounded Rationality.” In C. B. McGuire and Roy Radner. Decision and Organization. Minneapolis: University of Minnesota Press. 161-176. Smith, Kevin B. and Kenneth J. Meier. 1995. “Public Choice in Education: Markets and the Demand for Quality Education.” Political Research Quarterly. 48(3): 461-478. Snyder, Thomas D. and Charlene M. Hoffman. 2003. Digest of Education Statistics, 2002, NCES 2003-60. U.S. Department of Education, National Center for Education Statistics. Squire, Peverill. 2000. “Uncontested Seats in State Legislative Elections.” Legislative Studies Quarterly. 25:131-146. ________. 1993. “Professionalism and Public Opinion of State Legislatures.” Journal of Politics. 55: 479-491. ________. 1988. “Career Opportunities and Membership Stability in Legislatures.” Legislative Studies Quarterly. 13: 65-82. Stein, Mark. 2008. How the States Got Their Shapes. New York: Smithsonian Books/Collins. Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1 (1971). Tedin, Kent L. and Gregory R. Weiher. 2004. “Racial/Ethnic Diversity and Academic Quality as Components of School Choice.” The Journal of Politics. 66(4): 11091133. Timar, Thomas B. 1997. “The Institutional Role of State Education Departments: A Historical Perspective.” American Journal of Education. 105(3): 231-260. Timar, Thomas B. and David L. Kirp. 1988. “State Efforts to Reform Schools: Treading Between a Regulatory Swamp and an English Garden.” Educational Evaluation and Policy Analysis. 10:75-88. Tolbert, Caroline and Rodney Hero. 2001. “Dealing with Diversity: Racial/Ethnic Context and Social Policy Change. Political Research Quarterly. 54: 571-604. Truman, David B. 1951. The Governmental Process, Political Interests, and Public Opinion. New York: Knopf. 194 Tyack, David and Thomas James. 1986. “State Government and American Public Education: Exploring the ‘Primeval Forest.’” History of Education Quarterly. 26(1): 39-69. United States Census Bureau. 2012. The Statistical Abstract. Retrieved October 3, 2012 at http://www.census.gov/compendia/statab. United States Department of Education. 2009. “Elementary and Secondary Education: Improving Basic Programs Operated by Local Education Agencies (Title 1, Part A).” Retrieved March 17, 2009 at http://www.ed.gov/programs/titleiparta/index. html. Volden, Craig. 2002. “A Formal Model of the Politics of Delegation in a Separation of Powers System.” American Journal of Political Science. 46: 111-133. ________. 2006. “States as Policy Laboratories: Emulating Success in the Children's Health Insurance Program.” American Journal of Political Science 50(2): 294312. Walker, Jack. 1969. “The Diffusion of Innovations Among the American States.” American Political Science Review. 63: 880-899. Witte, John F. and Mark E. Rigdon. 1993. “Education Choice Reforms: Will They Change Am erican Schools?” Publius. 23(3): The State of American Federalism, 1992-1993: 95-114. Wildavsky, Aaron. 1964. The Politics of the Budgetary Process. Boston: Little Brown. Wong, Kenneth K. 2004. “The Politics of Education.” In Virginia Gray and Russell L. Hanson, ed. Politics in the American States: A Comparative Analysis. Eighth Edition. Washington: CQ Press. pp. 357-388. ________. Funding Public Schools:Politics and Policy. Lawrence: University Press of Kansas. Young, F. W. and R. Lewyckyj. 1996. ALSCAL User’s Guide. Fifth Edition. L.L. Thurstone Psychometric Laboratory, University of North Carolina, Chapel Hill, NC. Zelman v. Simmons-Harris, 536 U.S. 639 (2002). 195