The Variable Politics of the Policy Process: Issue Area Differences and Comparative Networks Matt Grossmann Assistant Professor of Political Science, Michigan State University 303 S. Kedzie Hall East Lansing, MI 48823 (517) 355-7655 matt@mattg.org Abstract The politics of policy issue areas differ in multiple ways, including the venues where policies are enacted, the frequency and type of policy development, the relative importance of different circumstantial factors in policy change, the composition of participants in policymaking, and the structure of issue networks. The differences cannot be summarized by typologies because each issue area differs substantially from the norm on only a few distinct characteristics. To understand these commonalities and differences, I aggregate information from 231 books and 37 articles that review the history of American domestic policy in 14 issue areas from 1945-2004. The histories collectively uncover 790 notable policy enactments and credit 1,306 actors for their role in policy development. The politics of each issue area stand out in a few important but unrelated aspects. The Variable Politics of the Policy Process 1 Scholars seek to understand how the political system produces public policy, but the answers may differ across issue areas.1 Though cognizant of these likely differences, scholars rarely consider them systematically. Across issue areas, does American national policymaking take place in the same venues with the same frequency? Is the relative importance of different political circumstances similar? Are the composition of participants in policymaking and the structure of the networks that connect them similar? If differences are widespread, can they be easily summarized by a typology? This paper reviews American federal policymaking in 14 broad domestic issue areas since World War II: agriculture, civil rights & liberties, criminal justice, education, energy, the environment, finance & commerce, health, housing & development, labor & immigration, macroeconomics, science & technology, social welfare, and transportation. Across these issue areas, neither the causal factors in the policy process nor the composition and structure of issue networks are universal. Each issue area is distinct from the others on a few characteristics, but typical in most respects. Separable types of policymaking do not follow from issue area categorizations. As a result, investigations of policymaking are likely to focus on particular aspects of the policy process based on issue area case selection decisions, even though they seek generalized knowledge. The relevant circumstances and actors change with the issue territory, as do the relationships among actors and the relevant political circumstances. Rather than assuming universality in the policy process, relying on typologies, or creating unique theories for each issue area, scholars should be attentive to the few ways that each issue area differs from the others. 1 An online appendix at www.journals.cambridge.org/jop contains a full list of sources used in the content analysis described in this article as well as relevant codebook instructions. Data to reproduce all numerical results and network diagrams will be posted at www.mattg.org on publication. The Variable Politics of the Policy Process 2 I address these variations using historical studies of policymaking. First, I compare general theories of the policy process, policy typologies, and studies of issue networks. Second, I argue that issue area differences are best conceptualized as issue-specific exceptions to general patterns, rather than categorical distinctions based on underlying dimensions. Third, I explain the method, which relies on a content analysis of 231 books and 37 articles that review policy history. Fourth, I review the record of significant policy enactments in each issue area and the explanations for policy change found in these sources. Fifth, I analyze the networks associated with each policy area, relying on information about the actors credited with policy enactments by historians. Sixth, I search for underlying dimensions of issue area differences as well as clusters of issue types. Finally, I provide descriptions of the unique features of each issue area to guide future scholarship. Issue Area Politics and the Policy Process Many theories of the policy process largely sidestep the question of differences across issue areas and are meant to apply to many domains. Punctuated-equilibrium (PE) accounts (Baumgartner and Jones 1993) argue that significant policy change is unlikely without a large increase in consideration of a problem. The multiple streams (MS) account emphasizes the multiple, largely independent, streams of problem definition, politics, and policy (Kingdon 2003). The advocacy coalition framework (ACF) focuses on the ideas and beliefs developed by interest group and government proponents of policy change (Sabatier and Jenkins-Smith 1993).2 Although these theories are all applied flexibly to different issues, their applications tend to concentrate in particular areas. Studies using PE focus more on budgets (Jones and Baumgartner 2012), the MS account 2 The ACF was developed to apply to issue areas that involve scientific disputes and high degrees of belief conflict. Its application is thus somewhat more restricted than other theories. The Variable Politics of the Policy Process 3 draws more from transportation and health (Zahariadis 2007), and nearly 64% of applications of the ACF focus on environmental or energy policy (Weible, Sabatier, and McQueen 2009).3 Issue area differences could help reconcile accounts of policymaking from different theoretical perspectives. For example, PE accounts imply that significant policy change is driven by episodic agendas; other incremental policy changes are thought to be less important. In contrast, historical approaches to policy change (Pierson 2004) argue that most significant policymaking is developmental; it relies on a path dependent process where early decisions constrain later decisions. Alternatively, some issue areas may be more episodic and others more path dependent. Some theories of the policy process explicitly analyze issue area differences. They tend to involve issue categorization schemes that focus on one or two dimensions of variation associated with clear types. Theodore Lowi (1964) proposes a three-part typology: redistributive, distributive, and regulatory. The idea is that scholars should expect to find differences in the politics of each issue area based on the kind of policy under debate and who has something to gain or lose from policy action. Similarly, James Q. Wilson (1980) argues that policy issues can be divided into types based on whether the costs and benefits of policy action in the area are concentrated or dispersed: interest group politics where both are narrow, entrepreneurial politics where only costs are concentrated, client politics where only benefits are concentrated, and majoritarian politics where both are broad. These typologies have been difficult for scholars to follow, since most policy areas have elements of multiple types.4 They have not proved especially fruitful in understanding policy area 3 A content analysis of applications of the ACF (Weible, Sabatier, and McQueen 2009) found few studies of economic policy, social welfare policy, agriculture, criminal justice, or housing. Although there is no equivalent content analysis of the other two frameworks, I noticed few applications of PE to civil rights, science, or transportation and few applications of MS to education or housing. The Variable Politics of the Policy Process 4 differences, but new typologies have nonetheless proliferated (Smith 2002). The continued interest in typologies highlights the need to understand variation across issue areas in the venues where policymaking takes place and the factors responsible for policy change. I investigate this variation, focusing on several categories of explanations for policy referenced in both issue area histories and the general literature on public policy: media coverage, public opinion, interest groups, international factors, state and local factors, research, events, and path dependence.5 If differences across issue areas produce distinct politics, scholars should also observe different kinds of networks emerging in different areas. In the classic formulation of “issue networks,” Hugh Heclo (1978) argues that experts form relationships based on reputations for issuespecific knowledge. Other scholars analyze these relationships, finding a “hollow core” with no central player arbitrating conflict in many issue domains (Heinz et al. 1993). Yet comparative analysis of issue networks is rare. In addition, some scholars argue that not all policy communities are large and broad enough to merit the label of issue networks (Marsh and Rhodes 2004). Others argue that policymaking in some areas may instead resemble iron triangles involving a set of client interest groups, an executive agency, and relevant congressional committees 4 Applying Lowi’s typology might categorize criminal justice and energy as regulatory, transportation and health as distributive, and social welfare and housing as redistributive. Apply Wilson’s typology might categorize energy as interest group politics, housing and labor as client politics, environment and criminal justice as entrepreneurial politics, and macroeconomics as majoritarian politics. 5 This article considers the factors external to government institutions that are the focus of policy process theory, rather than negotiations within the three branches. These external factors are cited as causes of policy change, however, rather than contextual factors driving other determinants. The Variable Politics of the Policy Process 5 (Berry 1989).6 To investigate variation across networks, I examine the composition of actors involved in each issue area and the configuration of their relationships. Issue Area Differences as Exceptions to General Patterns Extant research has not uncovered typologies that successfully explain how either the politics of policymaking or the character of networks differ across issue areas. Kevin Smith (2002, 381) advocates a move from typologies to taxonomies, classifying items “on the basis of empirically observable and measurable characteristics.” This paper generally takes this approach, addressing two fundamental problems of policy typologies. First, typologies assume that differences in the politics of issue areas can be distilled into only a few important dimensions. Second, they assume that most issue areas will fall in a clear zone along these dimensions, enabling scholars to place them in boxes. Both assumptions may be false. Issue areas may have broadly similar policy processes and each issue area may stand out in only a few important aspects. This perspective should apply to both the institutions and circumstances that make policy change possible (the focus of the policy process literature) and the actors responsible for policy change and their relationships (the focus of the issue networks literature). Whether scholars are looking at where and how often policy change occurs, the role of circumstantial factors in driving policy development, or the people and organizations that jointly bring it about, they should not expect issue area differences to conform to any typology.7 6 Issue areas like agriculture, energy, housing, labor, science, and transportation have these three institutions and are sometimes considered candidates for networks that resemble iron triangles. 7 Interactions between context and political factors also help produce policy change, but historians do not discuss interactions with enough consistency to enable incorporation into content analysis. The Variable Politics of the Policy Process 6 Issue area differences manifest themselves in both obvious and subtle ways. It should be no surprise that criminal justice policy change happens more often in the courts compared to other areas; after all, a large proportion of court proceedings confront related issues. Learning that energy policy is less likely to be affected by public opinion than other areas, in contrast, may elicit more surprise. These differences are unlikely to be reducible to a few categories. Categorizing criminal justice as a court-centered issue area, for example, would miss all of the ways that it is similar to other issue areas while highlighting only one of its features. Similarly, categorizing energy policy as immune from public opinion would also put too much emphasis on a single aspect of its politics. Issue network differences are also unlikely to allow categorization into separable types. In particular, the composition of networks (such as the partisan or institutional affiliations of its members) may vary independently of their structure. I find that a large issue network bridging two branches of government determines macroeconomic policy, for example, but this may not correspond to a category that any other issue network fits well within. Issue area differences are thus unlikely to correspond to the characteristics that make typologies useful. Scholars should instead specify the differences between issue areas, even if they only amount to a series of exceptions to the typical policy process and the common features of issue networks. Compiling Policy Area Histories Specifying the differences across issue areas requires comparative studies of many different policy processes. To make that possible, I rely on secondary sources of policy history. Policy specialists often review extensive case evidence on the political process, attempting to explain how, when, and why public policy changes. These authors, who I call policy historians, identify important policy enactments in all branches of government and produce in-depth narrative accounts of policy development. David Mayhew (2005, 245-252) used policy histories to produce his list of landmark The Variable Politics of the Policy Process 7 laws; he found them more conscious of the effects of public policy and less swept up by hype from political leaders than contemporary scholarly or journalistic judgments. The analysis here relies on 268 books and articles that review policy history since 1945. I compile published accounts of federal policy change in 14 issue areas, each corresponding to a category from the Policy Agendas Project (PAP).8 I exclude the foreign policy areas of defense, 8 The agriculture category, category 4 in the PAP, covers issues related to farm subsidies and the food supply. The civil rights & liberties policy area, category 2, includes issues related to discrimination, voting rights, speech, and privacy. The criminal justice area, category 12, includes policies related to crime, drugs, weapons, courts, and prisons. Education policy, category 6, includes all levels and types of education. The energy issue area, category 7, includes all types of energy production. The environment issue area, category 8, includes air and water pollution, waste management, and conservation. The finance & commerce area, category 15, includes banking, business regulation, and consumer protection. Health policy, category 3, includes issues related to health insurance, the medical industry, and health benefits. Housing & community development, category 14, includes housing programs, the mortgage market, and aid directed toward cities. Labor & immigration, category 5, covers employment law and wages as well as immigrant and refugee issues. The macroeconomics area, category 1, includes all types of tax changes and budget reforms. Science & technology, category 17, includes policies related to space, media regulation, the computer industry, and research. Social welfare, category 13, includes anti-poverty programs, social services, and assistance to the elderly and the disabled. The transportation area, category 10, includes policies related to highways, airports, railroads, and boating. The Variable Politics of the Policy Process 8 trade, and foreign affairs, but cover nearly the entire spectrum of domestic policy areas.9 For each issue area, I search multiple book catalogs and article databases using keywords from the topic lists and subcategories available at policyagendas.org. To find additional sources, I use bibliographies and literature reviews. Rather than sample, I construct a population of sources based on several exclusion criteria. To focus on broad historical reviews of the policy process, I exclude sources that do not identify the most important enactments, those that focus on advocating policies or explaining the content of current policy, and those that cover fewer than ten years of policymaking. I also exclude sources that analyze the politics of the policy process from a single theoretical orientation without a broad narrative review of policy history.10 The full list of sources, categorized by issue area, is available in the supporting materials on the journal’s website. With the help of research assistants, I read each text and identified significant policy enactments. I include policy enactments when any author indicated that the change was important and attempted to explain how or why it occurred. The relevant portions of the codebook and instructions are available in the supporting materials. For each enactment, I code whether it was an 9 The PAP divides policymaking into 19 categories. Three categories cover foreign policy and two categories do not have an associated separable policy history literature (government operations and public lands). Foreign policy may be subject to different dynamics than those studied here and is typically reviewed in international relations scholarship, rather than policy history. 10 At least 80 policy studies using the ACF (Weible, Sabatier, & McQueen 2009) and at least 35 policy studies undertaken to study PE (listed at policyagendas.org) are excluded. This guards against a search for confirming evidence, where scholars emphasize factors that are central to theory. Policy historians also share biases, but their collective judgment serves as a useful comparison to theoretically driven research. For a comparison of the advantages of each, see Grossmann (2012). The Variable Politics of the Policy Process 9 act of Congress, the President, an administrative agency or department, or a court. I also categorize it by issue area based on the PAP issue area codebook.11 I code all policy histories for the factors that each author judged significant in each policy enactment. To capture their explanations, I have coders ask themselves 61 questions about each author’s explanation of each enactment from a codebook. Based on these questions, I record dichotomous indicators of whether each author’s explanation included each factor for every significant change in policy that they analyze. The supporting materials include the relevant factors included in the content analysis classified into the categories used here. Coders of the same volume reach agreement on more than 95% of all codes.12 In the results below, I aggregate explanations across all authors, considering a factor relevant when any source considered it part of the explanation for an enactment.13 Most authors rely on their own qualitative research strategies to identify significant actors and circumstances. For example, the books that I use quote first-hand interviews, media reports, reviews by government agencies, and secondary sources. I rely on the judgments of experts in each 11 For the list of policy enactments, an assistant reassessed codes for policymaking venue and issue area and, where available, compared our codes to those in the PAP database. The Krippendorff’s Alpha reliability score was .903 for the venue analysis and .848 for the issue area analysis. 12 Percent agreement is the only acceptable inter-coder reliability measure for many different coders analyzing a single case; other measures are undefined due to lack of variation across cases. 13 I use this minimal standard because many policy histories did not include substantial explanatory material on some of the policy enactments that they viewed as significant. Some authors consistently list more explanatory factors than others. Further analysis revealed that author differences do not substantially change the relative frequency of the explanatory factors associated with each issue area. The Variable Politics of the Policy Process 10 policy area, who have already searched the most relevant available evidence, rather than impose one standard of evidence across all cases and independently conduct my own analysis that is less sensitive to the context of each policy debate. Collectively, however, policy historians likely still have biases that are reflected in their focus.14 These biases, if similar across scholars, also inevitably color the aggregate content analysis. The policy histories therefore offer a useful comparison to current theoretically focused research on the policy process, but not a definitive test of their claims. I also record every individual and organization that was credited by policy historians with bringing about policy change. For each policy enactment mentioned by each author, I catalogue all mentions of credited actors (proponents of policy change that were seen as partially responsible for the enactment). I then combine explanations for the same policy enactments, aggregating the actors that were associated with policy enactments across all authors.15 The typical explanation credits the few actors most responsible for each policy change. Coders of the same volume reach agreement on more than 95% of actors mentioned as responsible for each enactment. I also count the number of 14 Grossmann (2012) reviews these likely biases. Historians may be less likely to notice policymaking in administrative agencies and lower courts compared to laws passed by Congress. No measure of differences in author research method, scholarly discipline, or time period explain the differences across issue areas reported here, but unexplained variation in author judgments remains. There is variation in the emphasis authors put on each factor that they mention, but not enough similarity in the language they use to incorporate it in the analysis. I also assume equality across policy changes of very different scope by using frequencies from the population of all significant policy enactments. 15 I use this standard because some policy histories did not include any actors associated with some policy enactments that they viewed as significant. Further analysis using different standards (such as majority rule across authors) revealed similar results. The Variable Politics of the Policy Process 11 members of Congress, interest groups, and government organizations credited with policy enactments in each issue area and categorize the actors ideologically, based on whether they were Democrats (or liberal organizations) or Republicans (or conservative organizations).16 I use affiliation networks to understand the structure of relationships in each issue area. These networks include all of the actors that were partially credited with a policy enactment in each issue area, with undirected ties based on actors that were jointly credited with the same policy enactments.17 This does not necessarily indicate that the actors actively worked together, but that they were both on the winning side of a significant policy enactment and that a policy historian thought they each deserved some credit. The affiliation network ties are valued as integer counts of the number of shared policy enactments between every pair of actors.18 To assess the extent to which differences across issue areas can be easily summarized, I use nonmetric multidimensional scaling and k-means cluster analysis (see Everitt et al. 2011). First, I construct a dissimilarities matrix between all pairs of issue areas based on their differences on the number of enactments in each venue, the percentage of enactments associated with each causal 16 Actors that could not be easily categorized were put in separate unidentified categories. 17 The networks are not made up of all of the participants in policy communities. Instead, they include people credited with policy enactments in each issue area. All of the issue areas analyzed here therefore have some network of actors jointly credited with influencing policy. Because not all ties in the networks convey political collaboration, however, the results are not directly comparable to past research that analyzes networks of working relationships or coalitions. 18 If an individual and an organization of which they were a part were both credited, they are treated as separate actors in the network. These instances accounted for a small minority of all connections. The Variable Politics of the Policy Process 12 factor, and the characteristics of each issue network. I then place the issue areas in dimensional space and clusters. The goal is to see whether typologies can account for issue area differences. Policy Enactments Across Issue Areas Policy is enacted in every branch of government and issue area, though hardly with equal frequency. Figure 1 depicts the number of significant policy enactments in each issue area in each venue since 1945, separating laws passed by Congress from executive orders by the president, administrative agency rules, and court decisions. Unsurprisingly, Congress dominates policymaking in most issue areas. Nevertheless, a few issue areas stand out for the extent to which policy enactments occur in other branches of government. In civil rights, criminal justice, and finance & commerce, policymaking occurs disproportionately in the judiciary. Enactments in the energy and science & technology domains are more likely to come from administrative agencies. [Insert Figure 1 Here] Policymaking in each issue area also differs dramatically in its frequency: health and the environment are associated with more policy enactments. This is partially a consequence of their consistent prominence on the government agenda. The correlation between the total policy enactments in each issue area and the number of congressional hearings over the entire period is .49. Transportation, however, is regularly on the agenda without producing many policy enactments. Issue areas also differ substantially in the extent to which their policymaking is path dependent or episodic. Figure 2 compares the percentage of policy enactments where policy historians referred to factors related to path dependence with the percentage of enactments where they pointed to particular events driving policy change. This does not indicate that the historians used any language related to the theoretical concepts of path dependence or focusing events; most did not. Instead, explanations involving path dependence included any statement that the enactment The Variable Politics of the Policy Process 13 was an extension of an earlier policy, that an earlier choice made the enactment more likely, or that an earlier choice eliminated a potential alternative policy. Explanations involving events pointed to the effects of war, economic downturn, a government financial problem, a focusing event such as a school shooting, or a case highlighting problems in a previous policy. [Insert Figure 2 Here] The results show that policy enactments in agriculture, energy, housing, and labor are most likely to be path dependent. Enactments in energy and macroeconomics are most likely to be associated with events, especially nuclear disasters and economic downturns. These two potential sets of explanatory factors do not directly trade-off with one another. Some policy changes were not associated with either category of factors. Others were associated with both past policy choices and focusing events, such as reauthorization of an environmental statute in response to a natural disaster. Nevertheless, more episodic policy areas were associated with more congressional hearings. The number of hearings in each issue area is correlated at .46 with the difference between the percentage of enactments that were episodic and path dependent. Analyzing the policy agenda may thus track episodic issue areas while missing significant enactments in areas with more path dependence. Reported Circumstances Responsible for Policy Enactments Policy histories also point to somewhat different types of circumstances in explaining policy change in each area. Table 1 reports the percentage of policy enactments in each issue area associated with six categories of causal factors. These categories are not mutually exclusive or exhaustive. They were the external circumstances mentioned by policy historians most often and are common components of theories of the policy process. Explanations involving media coverage point to general attention or specific articles. In the public opinion category, I include references to public views, issues raised in an election campaign or by constituents, or a public protest. The Variable Politics of the Policy Process 14 Explanations involving interest groups include advocacy by non-governmental organizations, business interests, professional associations, or unions. Those involving international factors include references to foreign examples and international pressure or agreements. Explanations involving state or local factors include references to state or local actions that preceded federal action or reports from state or local officials. For explanations involving research, I include references to data or research findings, think tank or academic involvement, or research reports. [Insert Table 1 Here] Media coverage was most commonly associated with policymaking in macroeconomics, the environment, social welfare and transportation. Reports of pollution, poverty, and dilapidated infrastructure all play roles in policy development. Public opinion was a commonly reported cause of enactments in macroeconomics, civil rights, and labor and significantly less common in energy, finance, and science. Public concern over economic conditions, for instance, was regularly credited with macroeconomic policy change. Interest group influence was quite common in most issue areas, but was significantly more common in agriculture, transportation, the environment, and civil rights. Historians regularly credit lobbying by industry groups in agriculture and transportation as well as advocacy by public interest groups for civil rights and environmental protection. Science & technology policy registered the highest rate of international influence, with the Soviet launch of Sputnik serving as the most prominent example. State or local influence on policy enactments was most common in the areas of housing and civil rights but was significantly less common in science and agriculture. According to policy historians, factors related to research were commonly associated with policy change in most issue areas, with the exception of civil rights. Policy historians regularly cite new data as well as summary reports from government agencies as factors in policy change. The Diversity of Issue Networks The Variable Politics of the Policy Process 15 Policy histories also credit particular individuals and organizations with bringing about policy change in each area. The networks that I analyze enable a visualization of the relationships among these actors. Figure 3 depicts a sample of four issue networks. Nodes are actors credited with enactments; links connect actors credited with the same enactments. Black nodes are Democrats or liberal organizations; white nodes represent Republicans or conservative organizations; others are grey. Actors in the legislative branch are represented as circles; actors in the executive branch are squares; diamonds represent those in the judicial branch and triangles are non-governmental actors. [Insert Figure 3 Here] There is remarkable variation in the composition and structure of networks across issue areas, though none resembles a hollow core. None of the issue areas have a clearly bifurcated network polarized by ideology, though there are differences in degree. There is also substantial cross-branch interaction in most, but not all, issue areas. Table 2 reports several characteristics of the composition of each issue area’s network. Members of Congress dominate half of the networks and interest groups dominate two of the networks; others have a mix of central players. Organizations like executive agencies are central in the transportation network. [Insert Table 2 Here] Table 3 reports common characteristics of the structure of each issue area’s network.19 Size is the number of actors. Density is the average number of ties between all pairs of actors. In this case, the interpretation is the average number of policy enactments for which each pair of actors in the network shared credit. Paul Hallacher (2005) uses equivalent notions of size and interaction to 19 Although there are fit statistics proposed for some of these measures (beyond the fitness measure for the core-periphery model), they generally require simulated data and are not usefully compared to familiar statistics (Wasserman and Faust 1994). The Variable Politics of the Policy Process 16 compare subgovernments and issue networks, suggesting that issue networks have high values for each. The core-periphery model in Table 3 compares each network to an ideal type in which a central group of actors is closely tied to one another and surrounded by a periphery of less connected actors. The fitness statistic reports the extent to which the network fits this ideal type; the core density statistic reports the density within the group of actors identified as the core of the network (a categorical distinction). High values here would indicate that a network is far removed from the “hollow core” that characterized some previous networks (see Heinz et al. 1993). [Insert Table 3 Here] Degree Centralization measures the extent to which all ties in the network are to a single actor. The Clustering Coefficient measures the extent to which actors that are tied to one another are also tied to the same other actors.20 Table 3 also reports two versions of the E-I (externalinternal) index to track cross-branch and cross-party ties. The index measures the extent to which ties are disproportionately across groups (positive) or within groups (negative). The groups for the first index are actors in Congress and those in the executive branch; for the second, the groups are Republicans or conservative organizations and Democrats or liberal organizations.21 20 Higher clustering coefficients indicate that, if actors share ties to other actors, they are more likely to be tied. To determine whether actors divide into clustered neighborhoods throughout the network, it is useful to compare the clustering coefficient to the overall density in each network. 21 The indexes are calculated as (number of ties between actors in the two different groups - number of ties between actors in the same group)/(sum of both of these types of ties). These indexes may be measures of the extent of conflict (negative) or cooperation (positive) across groups; more conflict is usually seen as an indicator of issue networks rather than subgovernments (Hallacher 2005). The Variable Politics of the Policy Process 17 Issue networks differ in their structure across issue areas. Some networks are large and dense like transportation, whereas others are small and dense like agriculture. The health network is large but sparse but science is small and sparse. The issue networks that most resemble the core-periphery structure are civil rights and the environment. The most centralized networks are housing (around the U.S. Conference of Mayors) and labor (around the AFL-CIO). Clustering is most evident in the environment and least evident in energy. The networks most polarized by partisanship are the environment, science, and civil rights. The networks with most ties between the legislative and executive branch are housing, finance, and macroeconomics; civil rights and science had the least cross-branch oriented networks. Dimensions of Issue Area Politics To evaluate the number of underlying dimensions of issue area politics and to see where issue areas sit relative to one another on these dimensions, I use nonmetric multidimensional scaling. This provides two kinds of output: information about how well a model with each number of dimensions fits the data and a scale score for each case on each dimension. A typology that successfully made sense of the differences in politics across issue areas would require differences to be summarized by a small number of dimensions where the cases clearly separated into clusters. Figure 4 depicts a multidimensional analysis of issue area dissimilarities, using the characteristics of policy change reported in Figures 1 and 2, the reported circumstances associated with policy change from Table 1, and the characteristics of issue networks from Tables 2 and 3.22 22 It uses a dissimilarity matrix of the Euclidian distances between pairs of issue areas for all variables in the three tables and two figures. I also created an alternative dissimilarity matrix based on The Variable Politics of the Policy Process 18 Although the model is depicted in two dimensions, the results suggest no clean break in the number of dimensions. The measure of Stress_1, a fit statistic where lower numbers indicate a better fit, is .075 for a one-dimension solution, .037 for a two-dimension solution, .016 for a three-dimension solution, and .003 for a four-dimension solution. [Insert Figure 4] I also use k-means cluster analysis to divide the issue areas into clusters based on the same dissimilarities. The Calinski-Harabasz and Duda-Hart procedures (see Everitt et al. 2011) for determining the appropriate number of clusters were inconclusive, with the former producing improved fit statistics with 3 and 13 clusters and the latter producing improved fit statistics with 4 and 5 clusters. A two-cluster solution divides agriculture, crime, energy, finance, and science into one cluster (the issue areas on the right side of Figure 4). One interpretation is that these issue areas are associated with less popular mobilization whereas the others involve broader economic and social policy agendas. An alternative three-cluster solution separates crime into its own cluster from this group. A four-cluster solution has a separate cluster for agriculture and for civil rights and labor. Beyond this number, the algorithm begins separating each issue area into a unique cluster. Rather than clean categories, distinctions among issue areas are best seen as continuous differences. The results suggest that there is no easy way to summarize the differences across the political processes surrounding each issue area and their associated issue networks. A two-by-two typology, for example, would have an especially poor fit with the data. We cannot be sure that two dimensions best account for issue area differences, that the issue areas divide cleanly along those dimensions, or that four clusters would be the most appropriate division. standardized versions of all of these variables. A low-dimensional solution based on the standardized measures does not fit the data as well and fails to produce differentiated clusters. The Variable Politics of the Policy Process 19 I also analyze several reformulations of the data to ensure that the results were not a product of methodological decisions. First, I divide the data across time into subsets of 15-year periods and searched for an underlying structure. Second, I categorize the issue areas into a larger number of subtopics, each covering a smaller territory. Third, I analyze only the policy enactments considered significant by most authors in each area and limit the explanatory factors considered to only those with consensus across authors. Fourth, I construct distinct sets of dissimilarities based on network characteristics and based on the factors reportedly driving policy change. Issue area differences did not produce clear clustering on a few underlying dimensions in any of these analyses. Some readers may interpret the findings as evidence that there are large differences among the scholars studying these issue areas, rather than the issue areas themselves, but the evidence does not point in this direction. First, authors covering policy enactments outside of their area of focus (such as health policy historians explaining the political process behind general tax laws) reached most of the same conclusions about who was involved and what circumstances were relevant as specialist historians. Second, there were few consistent differences in the types of actors credited and few differences in relevant circumstantial factors reported based on whether the authors used interviews, quantitative data, or archival research, whether the authors came from political science, policy, sociology, economics, history, or other departments, or how long after the events took place the sources were written. There were idiosyncratic differences across authors, but they did not produce the differences across issue areas. Making Sense of Issue Area Similarities and Differences There are some seemingly universal features of the policy process, but there are also important differences in each issue area’s politics. The 14 domestic policy areas analyzed here differ in their frequency of policymaking, their common venues, the circumstantial factors enabling policy The Variable Politics of the Policy Process 20 change, the actors responsible for enactments, and the structure of networks. On each dimension of issue area differences, most issue areas fall in the middle rather than at the extremes. Policy typologies are unlikely to isolate the different features of policymaking in each issue area. Indeed, the available policy area typologies are not predictive of the differences analyzed here. Theoretical distinctions made by Lowi (1964) and Wilson (1980) are not helpful in distinguishing among the types of politics present in each issue area. Distinctions between iron triangles, issue networks, and policy communities are equally unhelpful. This is a sign that the theories that produce typologies should be subjected to more scrutiny. The project of creating minimalist models of the policy process based on ideal types may not be helpful. The iron triangle ideal type, for example, mixes three independently varying dimensions of network structure: a strong core, cross-institutional links, and bipartisan links. Instead of assuming that issues will fall clearly into boxes, scholars should acknowledge that issue area differences are widespread but not very amenable to categorization. Table 4 lists descriptions of the features of each issue area that stand out when compared to the others, including the type of policymaking, the circumstances associated with policy enactments, and the composition and structure of the governing network. All policy areas stand out in some ways in comparison to the others. This comparative analysis should enable authors of case studies to check whether their findings are likely to apply only to a few issue areas or generalize to the policy process as a whole. [Insert Table 4 Here] There are also important similarities in policymaking that are reportedly common across issue areas, even though I rely on 14 distinct literatures on a broad spectrum of domestic policy. First, Congress is the most frequent maker of significant policy in nearly all issue areas and it is responsible for the bulk of policymaking in most areas. Second, all policy areas have some policy enactments where path dependent explanations are apt and others where event-related explanations The Variable Politics of the Policy Process 21 are apt. Third, interest groups and research reportedly play a common role in policymaking in most issue areas whereas public opinion and media coverage play less frequent roles. Fourth, all issue areas are associated with networks of actors credited with policy change, including members of Congress, executive branch agencies, and interest groups. These similarities are generally consistent with the textbook treatments of federal policymaking that focus on institutions. The issue area differences, however, have important implications for the generalizability of research findings. Issue area case selection decisions make large differences in likely findings. For example, Kingdon’s (2003) study of the policy process is based on case studies of health and transportation. If he had instead chosen to study education and labor, he might have shown more influence for public opinion and international factors. Likewise, because a great deal of scholarship using the ACF focuses on environmental policy (Weible, Sabatier, and McQueen 2009), scholars may be more likely to find influence by interest groups and research. The results may also show where each of the policy process frameworks could be productively applied. The research and coalition focus of the ACF, for example, might be useful in studies of agriculture and housing. The focus on episodic determinants of policy change in PE studies has been useful in studies of nuclear energy and budgets (Jones and Baumgartner 2012). This comparative research shows that energy and macroeconomic policies more broadly are the most likely to be associated with episodic causes of policy change. Theories of path dependent policymaking, in contrast, may be most appropriate in agriculture, education, and housing. The findings also have implications for normative discussions of the policy process. Scholars who sought to divide policymaking into categories (e.g. Wilson 1980, Lowi 1964) associated their typologies with judgments about the relationship between the policy process and democratic values. If issue area politics cannot be easily predicted based on whether policies tend to benefit majorities or minorities, general claims about where policymaking is likely to be more or less democratic may The Variable Politics of the Policy Process 22 not hold up to scrutiny. Similarly, politics in every issue area may be labeled “interest group politics” to some degree, even though groups are not the only important actors in any area. Claims about iron triangles and issue networks were also meant to raise concerns that policymaking did not live up to America’s founding principles. Iron triangles supposedly involved domination of policymaking by political insiders. Heclo (1978) also viewed issue networks with concern, arguing that they came with a “democratic deficit” because they empowered technocratic elites. Disproportionate involvement by administrators and scientists, however, is only one source of difficulty in matching our expectations of wide participation with the reality of the policy process. Disinterested citizens representative of the nation as a whole do not make up any issue networks. Instead, each issue area is associated with distinct distributions of political elites. Despite differences across issue areas, there is potential ground for general theories of the policy process and associated critiques of the relationship between democracy in theory and practice. Across the issue areas analyzed here, all issue areas involved multiple institutions, interest groups, and diverse policymakers. They incorporated several circumstantial factors and responded to both past policy development and current events. There are many factors in the policy process but some are much more frequently influential than others. Issue area differences are decipherable and should be emphasized, but the similarities in the relative importance of each component of policymaking across issue areas are just as important. The American national government has both a general policy process and some unique variants for each issue area. The Variable Politics of the Policy Process 23 References Baumgartner, Frank R. and Bryan D. Jones. 1993. Agendas and Instability in American Politics. Chicago: University of Chicago Press. Berry, Jeffrey. M. 1989. The Interest Group Society, 5th ed. New York: HarperCollins Publishers. Everitt, Brian S., Sabine Landau, Morven Leese, and Daniel Stahl. 2011. Cluster Analysis, 5th ed. Hoboken, NJ: Wiley. Grossmann, Matt. 2012. “Interest Group Influence on US Policy Change: An Assessment Based on Policy History.” Interest Groups & Advocacy 1 (2): 1-22. Hallacher, Paul M. 2005. Why Polity Issue Networks Matter: The Advanced Technology Program and the Manufacturing Extension Partnership. Lanham, MD: Rowman & Littlefield Publishers. Heclo, Hugh. 1978. “Issue Networks and the Executive Establishment.” In The New American Political System, ed. Anthony King. Washington: American Enterprise Institute. Heinz, John P., Edward O. Laumann, Robert L. Nelson, and Robert H. Salisbury. 1993. The Hollow Core: Private Interests in National Policy Making. Cambridge: Harvard University Press. Jones, Bryan D. and Frank R. Baumgartner. 2012. “From There to Here: Punctuated Equilibrium to the General Punctuation Thesis to a Theory of Government Information Processing.” Policy Studies Journal 40 (1): 1-19. Lowi, Theodore. 1964. “American Business, Public Policy, Case-Studies, and Political Theory.” World Politics 16 (4): 677-715. Kingdon, John W. 2003. Agendas, Alternatives and Public Policies, 2nd Ed. New York: Addison-Wesley. Marsh, David and R. A. W. Rhodes. 2004. “Policy Communities and Issue Networks: Beyond Typology,” In Social Networks: Critical Concepts in Sociology, ed. John Scott. London: Routledge. Mayhew, David R. 2005. Divided We Govern: Party Control, Lawmaking, and Investigations, 1946-2002. New Haven: Yale University Press. The Variable Politics of the Policy Process 24 Pierson, Paul. 2004. Politics in Time: History, Institutions, and Social Analysis. Princeton: Princeton University Press. Sabatier, Paul A. and Hank C. Jenkins-Smith. 1993. Policy Change and Learning: An Advocacy Coalition Approach. Boulder, CO: Westview Press. Smith, Kevin B. 2002. “Typologies, Taxonomies, and the Benefits of Policy Classification.” Policy Studies Journal 30 (3): 379-395. Wasserman, Stanley and Katherine Faust. 1994. Social Network Analysis: Methods and Applications. Cambridge: Cambridge University Press. Weible, Christopher M., Paul Sabatier, and Kelly McQueen. 2009. “Themes and Variations: Taking Stock of the Advocacy Coalition Framework.” Policy Studies Journal 37(1): 121-40. Wilson, James Q. 1980. “The Politics of Regulation.” In The Politics of Regulation, ed. James Q. Wilson. New York: Basic Books, 357-94. Zahariadis, Nikolaos. 2007. “The Multiple Streams Framework: Structure, Limitations, Prospects.” In Theories of the Policy Process, ed. Paul Sabatier. Boulder, CO: Westview Press. The Variable Politics of the Policy Process 25 Table 1: Reported Circumstances Associated with Policy Enactments in Issue Area Histories Media Public Interest International Coverage Opinion Groups Agriculture 18.42% 21.05% 63.16% 13.16% Civil Rights & Liberties 21.31% 31.15% 67.21% 11.48% Criminal Justice 25% 30.77% 30.77% 0% Education 12.12% 27.27% 48.48% 12.12% Energy 18.18% 13.64% 36.36% 13.64% Environment 29.9% 25.77% 69.07% 12.37% Finance & Commerce 10.34% 6.9% 36.21% 3.45% Health 10.53% 11.58% 36.84% 7.37% Housing & Development 16.67% 13.89% 58.33% 0% Labor & Immigration 16.07% 30.36% 55.36% 10.71% Macroeconomics 22.92% 41.67% 54.17% 14.58% Science & Technology 7.89% 7.89% 36.84% 23.68% Social Welfare 22.22% 22.22% 38.89% 0% Transportation 22.22% 17.78% 57.78% 0% The table reports the percentage of enactments that involved each factor, according to policy historians. State or Local 0% 24.59% 9.62% 18.18% 18.18% 19.59% 5.17% 9.47% 19.44% 12.5% 8.33% 2.63% 13.89% 6.67% Research 47.37% 22.95% 42.31% 42.42% 31.82% 54.64% 22.41% 38.95% 44.44% 37.5% 41.67% 28.95% 36.11% 31.11% The Variable Politics of the Policy Process 26 Table 2: Issue Area Governing Network Composition Congress Members # Links Agriculture 17 9.9 Ag. Dept, Farm Bureau Congress Civil Rights & Liberties 63 20.4 NAACP, JFK, MLK Jr. Int. Groups Criminal Justice 20 6.7 ACLU, Bar Assoc. Int. Groups Education 54 19.5 Edith Green, NEA Congress Energy 20 5.3 Ford, Ted Kennedy Congress Environment 36 13.1 Ed Muskie, J. Blatnik Mixed Finance & Commerce 19 3.4 Eisenhower, LBJ Congress Health 46 9.7 Truman, Mary Lasker Congress Housing & Development U.S. Conf. of Mayors 39 9.7 Mixed Labor & Immigration 76 14.8 AFL-CIO, Labor Dept. Mixed Macroeconomics 42 11.6 Wilbur Mills, Treasury Congress Science & Technology 12 2.8 FCC, Nixon Mixed Social Welfare 45 11.4 Wilbur Mills, Social Sec Congress Transportation 24 16.1 Ford, Ted Kennedy Gov. Orgs. The table reports characteristics of the actors credited with policy enactments in each issue area. Most Central (Degree) Dominant Type Interest Groups # Links 8 9.9 52 4.9 16 8.4 27 16.1 12 4.1 34 7.2 3 3 31 9.9 29 14.6 63 17.4 12 15.2 12 2.8 24 12.7 29 18.6 Government Orgs. # Links 6 6 16 23.1 11 7.2 15 16.1 9 2.4 33 13.1 7 3.1 26 8.7 22 9.5 24 16.5 16 20.4 13 4.7 18 16.7 30 20.8 The Variable Politics of the Policy Process 27 Table 3: Issue Area Governing Network Structural Characteristics Core-Periphery Core Degree Clustering Size Density Fitness Density Centralization Coefficient Agriculture 49 0.17 0.7 0.88 9.50% 1.02 Civil Rights & Liberties 210 0.1 0.69 1.07 7.21% 1 Criminal Justice 83 0.08 0.52 1 10.52% 1.02 Education 170 0.1 0.47 0.62 9.83% 0.97 Energy 65 0.07 0.59 1 5.96% 0.91 Environment 144 0.09 0.68 4.39 7.74% 1.2 Finance & Commerce 54 0.06 0.48 1.05 4.34% 0.99 Health 141 0.07 0.49 1.17 6.60% 1.01 Housing & Development 119 0.1 0.48 0.87 15.50% 1 Labor & Immigration 211 0.1 0.49 0.84 14.18% 1.14 Macroeconomics 118 0.12 0.67 1.07 9.94% 1.06 Science & Technology 70 0.06 0.29 2 4.38% 0.98 Social Welfare 136 0.1 0.56 1.04 8.06% 1.07 Transportation 127 0.19 0.76 1.29 13.48% 1.12 The table reports structural characteristics of the affiliation networks associated with policy enactments in each issue area. E-I Index CongressBipartisan Admin. -0.11 -0.14 -0.34 -0.28 -0.5 -0.17 -0.14 -0.23 -0.18 -0.28 -0.21 -0.42 -0.03 -0.03 -0.28 -0.01 -0.07 -0.23 -0.06 -0.19 -0.02 -0.01 -0.38 -0.29 -0.13 -0.24 -0.13 -0.15 The Variable Politics of the Policy Process 28 Table 4: The Relative Features of Each Issue Area’s Politics Agriculture Civil Rights & Liberties Criminal Justice Education Energy Environment Finance & Commerce Type of Policymaking Regular, path dependent enactments Multi-branch, frequent, and path dependent Disproportionately judicial enactments Path dependent enactments by Congress Disproportionately administrative, event-driven Many enactments with congressional dominance Split policymaking between Congress and courts Many enactments in Congress Path dependent enactments in Congress Relevant Circumstances High interest group influence High state/local influence but low research influence High research influence but no international influence No dominant influential factors Low public opinion and high state/local influence High media, state/local, and research influence Low research influence Network Composition Small, mostly congressional network Large network with many interest groups Small network with central interest groups Large network dominated by members of Congress Small network, concentrated in Congress Disproportionately Democratic network Small, congressional and presidential network Large network, dominated by members of Congress Network Structure Dense core-periphery network Core-periphery structure but low centralization Executive-congressional divide Core with satellite clusters Low centralization Dense core with high clustering Sparse decentralized network No dominant influential Sparse ties factors Housing & High state/local, low public Centralized network with Large diverse network Development opinion influence partisan divide Labor & High public opinion and Large, diverse network, Sparse network, with high Multi-branch enactments Immigration group influence centralized on AFL-CIO clustering Event-driven, legislative High public opinion and Congress-dominated Dense network with high Macroeconomics enactments research influence network inter-branch/bipartisan ties Science & Disproportionately Low public opinion but Diverse, with many Sparse, disconnected Technology administrative enactments high international influence government organizations network with no core Infrequent, path dependent High media but no Congress-dominated Social Welfare Large divided network enactments international influence network Regular congressional High interest group, no Large diverse network, with Dense core-periphery Transportation enactments international influence government organizations network; high clustering The table reports descriptions of where each issue area stands out among the others, based on the analysis of policy area histories conducted here. Health The Variable Politics of the Policy Process Figure 1: Policy Enactments by Venue and Policy Area The figure depicts the number of policy enactments in each branch of government from 1945-2004, based on policy histories of each issue area. 29 The Variable Politics of the Policy Process Figure 2: Developmental and Episodic Reported Causes of Enactments by Policy Area The figure depicts the percentage of policy enactments in each issue area that were reportedly affected by path dependence (including earlier policy choices that made the enactment more likely or eliminated alternatives) and focusing events (including wars and economic downturns). The reports are based on policy histories in each issue area covering significant policy enactments from 1945-2004. 30 The Variable Politics of the Policy Process 31 Figure 3: A Sample of Issue Networks Education Social Welfare Labor Transportation Nodes are actors credited with policy enactments in each area. Links connect actors credited with the same policy enactments. Democrats are black; Republicans are white; others are grey. Shape represents branch of government; circles are legislative; squares are executive; diamonds are judicial; triangles are non-governmental. The Variable Politics of the Policy Process Figure 4: Nonmetric Multidimensional Scaling of Dissimilarities Among Issue Areas The figure is a two-dimensional plot of nonmetric multidimensional scaling results of dissimilarities based on the reported factors in policy change since 1945 and the characteristics of their issue networks. The dissimilarities are based on the characteristics of issue areas reported in Figures 1-2 and Tables 1-3. 32