JPART2016年度最佳論文大獎-Clerks or Kings? Partisan Alignment and Delegation to the US Bureaucracy

Christine Kelleher Palus & Susan Webb Yackee, Clerks or Kings? Partisan Alignment and Delegation to the US Bureaucracy. Journal of Public Administration Research and Theory, 26 (4), pp. 693-708.

The Beryl Radin Award是頒給發表在PMRA會刊Journal of Public Administration Research and Theory (JPART)（公共管理領域排名第一的期刊）的最佳論文，一年一評。

作者簡介n

Christine Kelleher Palus, 美國維拉紐瓦大學公共管理副教授

• 論文摘要

以往，學者們通常認為公共政策自由裁量權的授權很大程度受到了政黨偏見或政黨派系的影響。簡而言之，民選官員會將政策執行的自由裁量權更多地授權給與自己相同政黨聯盟的行政長官，因為這可以降低未來決策的不確定性。然而，作者借鑒了社會認知理論的觀點，認為在實踐中，這種關係並不存在，反而政黨同盟並未讓行政官員認為其獲得了更多自由裁量權。作者收集了美國50個州政府1978年-2008年的選舉數據以及超過6000個行政長官的數據，驗證自己的研究假設，結果得到證實。

• 獲獎評語

評審委員會認為「Paulus和Yackee開發了一套新穎的測量方法與數據，檢驗一個已經在授權研究文獻中獲得廣泛認為的主張，他們的研究發現卻與現有研究完全不同。新的測量方法完全符合公共管理獨特的學科屬性，即探究實踐（在自由裁量權的測量上使用行政官員的直觀感知而非單純統計規章制度）。通過這種方式，文章重新審視了官僚制度的重要性。」

「Paulus and Yackee developed novel measures and data to test what is probably the strongest and best-supported claim made by the delegation literature, and they produced a finding that is completely at odds with all existing work. The new measure fits with public administration scholarship in nature by asking what happens in practice (the use of perceived discretion) rather than counting formal rules. In this fashion, the paper revisits the importance of the bureaucratic machinery and not just formal statute.」

• 論文全文

An uncomfortable truth about the United States』 system of governance, at least for some Americans, is the crucial role played by public agencies within the policymaking process. Some of this discomfort resides in the fact that the policymaking powers of public administrators are neither clearly spelled out within the US Constitution nor are bureaucrats typically held accountable to the public through the ballot box. Nevertheless, the importance of administrative agencies has grown profoundly since the nation』s founding, with the governance of a complex, modern society largely unthinkable without their contributions to policy formation, implementation, and evaluation ( Rubin 2005 ).

The legal authority of bureaucrats to make such decisions resides in the delegation of policy discretion from elected legislators and executives, such as the president or state governors, to administrative agents. Given this fact, it is not surprising that studying the causes and consequences of policy delegation is a central pursuit of modern American politics and public policy scholarship ( Carpenter 2010 ; Clinton et al. 2012 ; Epstein and O』Halloran 1999 ; Huber and Shipan 2007 ; Meier and O』Toole 2006 ; Rourke 1984 ; Wilson 1989 ). Yet, as Epstein and O』Halloran (1996 , 378) point out, delegation takes many forms; after all, 「[w]hen executive branch officials have significant latitude to implement legislation they will be important policymaking actors in their own right; otherwise, they will be mere clerks carrying out the laws….」 Therefore, key to understanding the importance of delegation within American governance is identifying the degree to which agency officials are either delegated a great deal of policy discretion, or a limited amount, to pursue policy-related decision-making.

In this article, we focus on explaining the perceived amount of policy discretion provided to agency officials. We define policy discretion as the freedom that agency officials have from the influence from political superiors to pursue policy aims or make policy-related decisions. In doing so, we challenge one of the fundamental predictions of modern delegation models: the so-called 「ally principle」 (for a review, see Lavertu and Weimer 2009 ), which typically posits that elected officials delegate more discretion when their partisanship aligns with administrative agents. Such models are often formal in nature and frequently assume that elected officials want to limit the risk of uncertainty over policy outcomes ( Bendor and Meirowitz 2004 ). Delegating policy discretion to like-minded administrative officials is believed to be a primary mechanism for achieving this goal.

In this article, we make two contributions. First, during periods of partisan alignment, we argue that a desire by high-level public officials to conform to the positions of their political party acts to decrease—not increase—perceived levels of policy discretion. Thus, we suggest an anti-ally principle, drawing on insights from social identity theory to suggest why it may result. Second, we empirically test our argument. We employ data collected over a 30-year period by the American State Administrators Project (ASAP). This twice-per-decade survey of state agency heads yields information from all agency types active across the 50 American states. We use data collected in seven surveys from 1978 to 2008, totaling thousands of observations. In particular, we capture measures of policy discretion from the agency administrator』s point of view. In doing so, we advance the literature by arguing that the perceptions of agency officials regarding their policy discretion yield a critical measure of the actual amount of discretion that an agency possesses and are associated with patterns in partisan alignment and misalignment. We supplement the survey information with data drawn from state legislative and gubernatorial elections over the same 30-year time period, which allows us to map the partisan alignment, or misalignment, between political principals and administrative agency leaders.

Using a pooled cross-sectional approach, we employ two conceptualizations of discretion (i.e., two different dependent variables). First, we demonstrate that state agency heads perceive less agency policy discretion on budgets, major policy decisions, and administrative regulations when their partisanship aligns with the governor or with the state legislature. Second, we find that agency leaders observe lower policy discretion—in this case, in terms of their perceived authority over agency program priorities—when their partisanship aligns with executive or legislative political principals. Although the effects of these relationships are small in some specifications, in accordance with our argument, these anti-ally empirical regularities are substantively more impactful when assessed in coordination with the appointment method of the agency head and the amount of contact the agency head has with his/her political principals. Thus, our findings provide consistent evidence challenging the ally principle, and with the caveat of small effect sizes across specifications, these findings offer suggestive support of our theoretical argument, as well.

Moving forward, we first review the literature, paying specific attention to potential measurement and data issues in existing work. We then investigate the empirical patterns in our data. We conclude that partisan alignment, which is often thought to free agency officials to make politically like-minded policy decisions, often has the opposite effect. Our results suggest that aligned agents may feel more like 「clerks」 carrying out the wishes of their political principals, rather than 「kings」 who are relatively free to exercise their own policymaking authority. From a normative perspective, these findings paint an intriguing picture: politically aligned actors may feel constrained in their decision-making; however, they are constrained by a political orientation of their own choosing.

Theoretical Foundations

Delegation is common in government, and as Bendor and Meirowitz (2004) note, the practice of political delegation dates back to antiquity. Yet it is only within the last several decades that scholars have wrestled with the causes and consequences, as well as the associated normative implications, of policy delegation to the public bureaucracy. This work has, to some degree, been driven by massive societal changes that make the public bureaucracy a much larger contributor to government policymaking today than even a century ago ( Rubin 2005 ). As Lavertu and Weimer (2009) point out, in modern society, elected officials often need to delegate policy authority and autonomy to administrative agencies to realize the full impact of legislative and executive public policy decision-making.

Political Influence and the Ally Principle

Government bureaucracies in the United States routinely make instrumental decisions regarding public policy formation, for instance via the generation of regulations, policy statements, and guidance documents, as well as take important actions to implement policy, often in the form of adjudication orders and enforcement decisions. Few people today would argue that such policy-related decisions are unimportant; however, many debate the degree to which these decisions are made under the close direction of the bureaucracy』s overseers, including legislatures and elected executives, or whether these decisions are generated with little oversight. This is a crucial debate. As Lewis (2009 ; see also, Meier and O』Toole 2006 ) concludes, delegation and the political control of administrative agencies have rightly developed into a major focus of political science scholarship.

Scholars traditionally focus on the conditions under which elected officials choose to delegate more or less policy discretion to administrative agencies as a way of understanding one of the primary methods of political influence on agencies ( Calvert, McCubbins, and Weingast 1989 ; Epstein and O』Halloran 1999 ; McCubbins, Noll, and Weingast 1987 , 1989 ; Volden 2002a ). As a result, much of the literature focuses either explicitly or implicitly on a Political Influence-Agency Discretion Trade-Off . This trade-off suggests that as political influence (or what some may call 「political control」) over the decisions of public bureaucrats by their political principals increases, then bureaucratic policy discretion decreases (and vice versa). Or, as one scholar summarizes ( Lewis 2007 , 1074), 「[a] widely cited literature」 suggests 「legislatures generally reduce agency discretion by writing more specific statutes, strengthening administrative procedures, and monitoring more carefully.」 Thus, the amount of policy discretion provided to agencies is seen as a fundamental way of understanding legislative-bureaucratic power relations. Discretion, in particular, matters because it establishes—and thereby legitimizes and/or constrains—the range of permissible policy choices that can be made by the bureaucracy ( Epstein and O』Halloran 1996 ). As Duffy (1996) suggests, elected executives, such as presidents and governors, specifically seek to limit discretion as a mechanism of political control.

Lavertu and Weimer (2009) identify two fundamental predictions of modern delegation models. First, political principals delegate more discretion as the alignment of preferences between elected and administrative officials increases. This is the so-called 「ally principle.」 Second, more discretion is provided when the need for expertise within the bureaucracy increases. As Bawn (1995) suggests, there may be a willingness by elected officials to trade some political control of agency behavior for expertise.

In this article, we focus on the ally principle, which has been called an 「ancient rule of political delegation」 ( Bendor and Meirowitz 2004 , 308). Two review essays that focus on delegation models generally conclude that the ally principle is a foundational concept for scholarly thinking with regard to policy delegation (see Huber and Shipan 2007 ; Lavertu and Weimer 2009 ). And although Huber and Shipan (2007) qualify their support, they too conclude that the formal and more limited quantitative evidence assessing the ally principle all point in a supportive direction.

At first blush, Epstein and O』Halloran』s (1996 , 1999 ) formal results and predictions, as well as their empirical research on trade legislation and the regulatory laws passed by Congress from 1947 to 1992, provide just this type of evidence. Epstein and O』Halloran measure the delegation of policy discretion by focusing on the content of congressionally passed legislation, and in particular, they code the procedural provisions imposed on bureaucratic agencies in statutes. In their empirical results, Epstein and O』Halloran (1999) find that when the preferences of Congress and the president are farther apart, Congress typically delegates less discretion to the bureaucracy. Given that they find that a partisan alignment between Congress and the president (i.e., unified government) yields more discretion for agencies, whereas a misalignment (i.e., divided government) results in less delegated authority, Epstein and O』Halloran』s work is frequently pointed to as evidence of the ally principle.

Huber and Shipan (2002 ; see also Huber, Shipan, and Pfahler 2001 ) also focus on the level of agency policy discretion provided in legislation, and in the US context, they study this phenomenon with data drawn from select health laws across the American states. Overall, Huber and Shipan find that more prescriptive and detailed legislation passes state legislatures during times of divided rather than unified government, which suggests that legislatures provide agencies less discretion when there is a partisan misalignment between the legislature and governor (and therefore, presumably, agencies), and more discretion when there is alignment. Despite this overall finding, however, Huber and Shipan (2002) caution that their results are not robust in states with low legislative professionalism nor do they hold when there is a split state legislature. Similarly, Volden (2002a) finds some support for the ally principle in his research on delegated authority to welfare boards. Yet, as Huber and Shipan (2007 , 262) write about Volden』s work, 「Like many others, he [Volden] begins by positing that policy conflict between legislator and bureaucrat will be highest during divided government (i.e. when the governor is of a different party from the legislature), but he finds that in actuality the effects of conflict are more subtle than that.」 In particular, Volden (2002a) concludes that such delegation is often related to other factors, including the appointment authority for bureaucratic officials.

Four Rationales for a Theoretical and Empirical Reassessment of the Ally Principle

What is striking to us, when reviewing this important and well-cited literature, is how ingrained the ally principle is within scholarly thinking, yet simultaneously, how unsettled the arguments and evidence are regarding its existence. Below, we raise four specific concerns, which when taken together, we believe suggest the need to reassess the thinking behind—as well as the empirical evidence in support of—the ally principle.

First, both formal and quantitative studies of the ally principle tend to equate the preferences of public agency leaders with the preferences of the president or state governor. Put differently, key delegation studies typically assume, or derive within a formal setting, that the policy preferences of the agency head are likely to be aligned with the elected executive. Epstein and O』Halloran』s (1996) formal model, for instance, assumes that presidents and agency heads occupy the same ideal position. This assumption is critical because it then allows the concepts of divided and unified government to be used as proxy measures for partisan misalignment and alignment, respectively, in both quantitative and formal models. As Volden (2002b , 111) summarizes, 「In these models, given uncertainty over policy and executive control over the agency, the legislature is only willing to delegate decisions to the agency when the executive』s preferences are similar to those of the legislature, as under unified government.」

Yet, a growing body of work takes aim at this simplifying assumption. In particular, the broader scholarship devoted to measuring directly the preferences of agency officials—as well as modeling agency preferences in tandem with their political principals—suggests that agency preferences often differ from elected officials ( Bertelli and Grose 2011 ; Clinton and Lewis 2008 ; Clinton et al. 2012 ; Nixon 2004 ). Bertelli and Grose (2011) , for instance, find a pattern of ideological divergence between cabinet secretaries and the president, whereas Cohen (1986) suggests that different agency types may be either closer, or farther way, in terms of preference congruence with the president. Such findings lead Clinton and Lewis (2008) to conclude that while in the past it was common for a researcher to use the larger political environment as a proxy measurement for agency preferences, this approach is now seen as problematic. Broadly, this literature suggests the need to reevaluate the assumptions and empirical measurement made within many delegation models—particularly the use of divided and unified government to measure misalignment and alignment, respectively. Put differently, this literature suggests that past efforts to assess the ally principle may have been hampered by an inability to compare the preferences of bureaucrats directly to other important political actors.

Second, existing studies of the ally principle typically measure delegation via the actions of political principals—such as measuring the content or length of legislative statutes. However, a good deal of empirical work suggests that measuring policy discretion from the recipient』s point of view instead may be a better match with the realities of modern administrative decision-making and policymaking. Measuring delegation by the actions of principals is problematic, for instance, because the on-the-ground evidence demonstrating that legislatively imposed administrative procedures, such as those evaluated in Epstein and O』Halloran』s work, actually constrain the behavior of agency officials is mixed, at best ( Lavertu and Weimer 2009 ). Instead, numerous studies suggest that agency officials often work around these administrative procedures (see broadly, Yackee and Yackee 2010 ). For example, Hamilton (1996 ; see also Hamilton and Schroeder 1994 ) explores the bureaucracy』s ability to avoid 「compulsory」 administrative procedures when making policy, and he finds such avoidance commonplace. He studies the Environmental Protection Agency』s use of policy statements and other informal policy pronouncements and concludes, 「[R]egulators should be viewed as strategic as Congress members with respect to the constraints」 ( Hamilton 1996 , 151).

Similarly, Spence (1999) examines a set of procedural requirements imposed on the Federal Energy Regulatory Commission (FERC). He demonstrates that FERC was able to use its perceived discretion to avoid making unwanted policy changes, while also remaining consistent with the agency』s statutory requirements. Yackee (2012) finds that agency use of procedural requirements, such as Advance Notices of Proposed Rulemaking, are often self-initiated and therefore are not seen by agencies, at least not seen exclusively, as limits to agency discretion. Additionally, West and Raso (2013) suggest that only a portion of federal agencies』 rulemaking agenda is directly attributable to specific congressional actions; indeed, a large percentage of the legally binding regulations written by agencies are self-initiated. Consequently, an agency』s understanding of its formal and informal discretion to write new policies and make policy decisions is important and is consequential to agency policy decision-making. Given this scholarship, the amount of delegated discretion available to an agency leader to make policy formation and implementation decisions may be best understood from the eye of the recipient of the delegation (i.e., the agency official), instead of the sender (i.e., the legislature or elected executive).

Third, although the ally principle is a property of many studies of delegation, there is select work that suggests arguments and conditions under which the principle may not hold. For instance, Boehmke, Gailmard, and Patty』s (2006) formal model (see also Gailmard and Patty 2013 ) highlights how interest group politics can interrupt information sharing processes and may disrupt the ally principle. Huber and McCarty (2004) suggest that the capacity of public bureaucrats may generate conditions when allied agency officials are not granted high levels of discretion; this occurs because low capacity and/or low competence levels may make delegation to politically aligned agency officials too risky or too costly for the principal. Similarly, Gailmard and Patty (2014) find that, depending on the transparency of information present between a political supervisor and an agent, there can be scenarios where the ally principle may not hold, whereas Bertelli and Grose (2011 ; see also McCarty 2004 ) suggest that the appointment power of an elected executive may complicate the predictions of the ally principle.

Fourth and finally, the majority of formal models and empirical studies of delegation focus on the policy discretion delegated by the legislature to agencies, leaving any delegation decisions within the executive branch relatively less explored (for an exception, see Duffy 1996 ). This is problematic given that we know that elected executives frequently devolve policymaking authority to administrative agencies. Thus, understanding devolved authority from presidents and state governors, as well as the factors driving agency discretion, are opportunities for inquiry.

Toward an Anti-Ally Argument

Above, we suggest several theoretical and empirical rationales to question the ally principle. We now offer one possible rationale for why, and under what circumstances, an opposite relationship, or an anti-ally principle, may materialize. A key piece to our contribution is that we focus theoretical attention on those agency officials charged with implementing public policies. This perspective differs from most published work. For instance, neither Epstein and O』Halloran (1999) nor Huber and Shipan (2002) collect information on what bureaucrats think or do after a legislature adopts a statute. Although these important books may capture what legislators (or at least the majority of those legislators voting for a statute) intended to do in terms of the amount of policy discretion provided to agencies in statute, they do not address the degree to which these legislative actions are connected to the amount of policy autonomy perceived by public agency officials in the workplace. After all, agency officials may interpret, perceive, and/or act on statutory authorizations or executive mandates in ways that were not intended and may have relatively large latitude in which to do so (see broadly, Hamilton 1996 ; Spence 1999 ; West and Raso 2013 ).

Thus, we take as a starting point that administrator-based perceptions of discretion are important to the policy decisions made in public sector agencies and provide a critical measure of the actual amount of discretion that an agency possesses. It is in this context that we aim to explore the relationship between partisan alignment and policy discretion. As stated above, we know that the literature typically posits that elected officials delegate more discretion when their partisan orientation aligns with administrative agents. The implication is, then, that administrative agents will perceive a higher level of discretion when they are of the same party as their political principal (be it the legislature or an elective executive). We explore this implication by suggesting that instead of an ally principle, the opposite relationship may materialize in practice. Specifically, we hypothesize that agency officials, on average, will perceive a lower level of discretion when there is partisan alignment with a political principal then when they are misaligned.

We draw on insights from the social and political psychology literatures to offer one potential explanation for what may seem, at first, to be a counterintuitive hypothesis, but that we ultimately suggest fits within modern theoretical conceptualizations of group behavior. Specifically, we focus on social identity theory, which is a prominent argument developed primarily by Tajfel and Turner (1979 , 1986 ) focused on the psychological processes linked with being a member of a particular group. As noted by Ashforth and Mael (1989) , people often group themselves, as well as others, into social categories which include gender, religion, and other types of organizational membership, and these self-groupings form the underpinnings for social identity theory. Essentially, the argument posits that a central tendency of humans is to sort and to make judgments about others based on whether other individuals are like, or un-like, themselves. As noted by Curley (2009 , 654), 「Individuals can and likely have differing strengths of identities within the group, but being a member of the group has an effect on the individual』s cognition and behavior.」

The main prediction derived from social identity theory is that people frequently simplify their environment by making 「in-group」 and 「out-group」 distinctions ( Brewer 2007 ; Huddy 2001 ; Judd and Park 1988 ; Kinder and Kam 2009 ; Park and Judd 1990 ). As Nicholson (2012 , 52) explains, 「In so doing, they seek to align their views with an in-group and differentiate themselves from out-groups.」 Indeed, the argument that individuals 「conform to their own membership groups (in-groups) over groups to which they do not belong (out-groups) is among the most well-established phenomena in social psychology」 ( Brewer 2007 , 729).

In the realm of modern American politics, the Republican and Democratic parties play a critical role in generating key in- and out-groups for political discourse and decision-making. As noted by Greene (1999 , 394), 「[s]ocial identity presents a powerful paradigm through which to examine party identification in the American context,」 and as Green, Palmquist, and Schickler (2002 , 3) broadly conclude, partisanship in the United States has 「persistence and motive power.」 1 Research demonstrates, for example, that a person』s partisan identification can affect how one sees the political world, including shaping a person』s opinion on policy issues and of political leaders (see broadly, Jacoby 1988 ; Kam 2005 ). And, oftentimes, 「[w]hen people feel a sense of belonging to a given social group, they absorb the doctrinal positions that the group advocates」 ( Green, Palmquist, and Schickler 2002 , 4). Kelly』s (1988) research also suggests that there is a positive relationship between the strength of a person』s partisan identification and the degree to which he or she views in- and out-group distinctions as important, whereas the work of Brewer (2007) suggests that in-group conformity pressures may be accentuated when there is a perceived threat against the person』s group. Furthermore, and of particular importance, is that a person』s group membership—in this case, his or her partisan identification—often affects attitudes, actions, and behaviors.

Thus, we build on Nicholson』s (2012 , 52) conclusion that 「[i]n the political world, parties serve as the relevant in- and out-groups…」 and we do so by stretching social identity theory to help us understand the delegation relationship between a political principal (be it the legislature or an elective executive) and administrative agents. It is important to acknowledge that this represents a theoretical departure from most political science applications of social identity theory, which have largely been applied in the context of mass citizen behavior. However, some scholarship has also used it as a lens to explain larger issues in international relations and politics (e.g., see Curley 2009 ; Theiler 2003 ). More broadly, we see no reason to suspect that in-group allegiances and biases would be any less salient to administrative agents—particularly high-level administrative agents—who often work in political charged and public policy relevant contexts.

With this backdrop, we argue that conformity to in-group views serves as a theoretical mechanism driving our hypothesized anti-ally principle relationship. According to this line of thinking, we theorize that, as party representatives and political elites in their own right, high-level administrative officials in public sector agencies, either consciously or subconsciously, feel a heightened need to advance the priorities and goals of their in-group (which during times of alignment is represented by the political principals of their same party). Thus, the need to conform to the political positions of his or her party may act as a constraint on an agency official』s perceptions of policymaking autonomy—hemming in an agency official』s perceived policy discretion and thereby serving as a mechanism to protect and to further the in-group. Put in different words, 「social identification leads to activities that are congruent with the identity」 and 「support for institutions that embody that identity…」 ( Ashforth and Mael 1989 , 20). Given these pressures, we would expect administrative agency officials with policy-related portfolios to believe themselves to be more constrained in their bureaucratic policy and regulatory decision-making when their political partisanship aligns with their elected superiors』 (and conversely, less obligated when they are misaligned).

To flesh this out more fully, one can also think about the circumstances that may exacerbate these anti-ally feelings. If these conformity pressures are driving perceptions, then we would likely see them intensified when there is partisan alignment and, for example, when the agency official is also appointed to their position by the political principal. Under this scenario, there would be an even greater desire to further the policy priorities, orientation, and outcomes of their in-group and its primary leader. This may be connected to a need to protect and further the agenda of the party over one』s personal agenda or the agenda of the agency. Additionally, one would expect that more regular contact between the political principal and agency official would further these tendencies, causing an official to perceive even less policy discretion when there is partisan alignment because the relevant group attachments would be regularly triggered. Finally, when there is a greater sense of perceived threat against the person』s group, one would expect larger effects. This would be the case during times of divided government, for instance, when one political principal aligns with an agent but when there are different parties leading the legislature and executive branches.

Assessing the Argument

We evaluate the argument with quantitative data collected from state political elections, as well as in-depth survey information drawn from state administrative agency heads. Using these data in a series of pooled, cross-sectional analyses, we present evidence illustrating the empirical reality of the anti-ally principle with respect to the delegation of policymaking authority to bureaucrats. In the sections below, we first explain the data and methods used to conduct our analyses. We then present the empirical results and a corresponding discussion of their implications.

Data and Methods

The ASAP provides the primary data used in all subsequent analyses. Beginning in the 1960s, this survey collected information at two junctures during each decade from all US state agency heads in administrative agencies within the 50 states. Several questions were added in 1978 to the survey that form the basis for measuring agency policy discretion. Thus, we use the survey years of 1978, 1984, 1988, 1994, 1998, 2004, and 2008. The response rates, varying from 21% to 54%, have been confirmed as representative of the full population of leaders contacted by ASAP ( Bowling 2010 ; Brudney and Wright 2002 ; Cho and Wright 2001 ). 2 In the end, these data represent the full range of administrative agencies active across the 50 states over the past three decades—from social services to economic development to corrections. Moreover, given the high turnover in these executive-level agency positions, the findings also represent a broad swath of opinions from political elites who are often cycling in and out of government, and cycling through different administrative agencies and government positions.

We specifically focus on the perceptions of agency heads because of their decisive role within state politics and state agencies ( Barrilleaux 1999 ; Dometrius 2002 ). As Krause and Joseph O』Connell (2011 , 2) note, agency leaders 「often enjoy disproportional power over policy administration」 (see also Kaufman 1981 ; Wilson 1989 ), whereas Edwards (2001 , 14) echoes this sentiment, stating that existing 「laws often permit agency heads wide latitude in making public policy.」 Schneider, Jacoby, and Coggburn (1997 , 240) specifically validate our attention to state administrators, noting that 「[a]dministrators in state agencies have wide latitude to make vital decisions in important public programs. And yet, surprisingly little is known about this process.」 Bertelli et al. (2013) comment on the importance of agency administrator perceptions: although all agency officials are subjected to some form of political accountability, Bertelli et al. (2013 , 4) write, 「bureaucrats must in practice develop subjective expectations of the autonomy they enjoy in relation to a particular policymaking effort.」 Moreover, these agency-driven perceptions are believed to play an essential role in shaping bureaucratic actions and decision-making ( Bertelli et al. 2013 ; Brehm and Gates 1997 ).

Thus, although we recognize that the informed assessments and attitudes of agency heads do not explain all facets of political reality, we also suggest that such perceptions are important to properly understanding the administrative policymaking context. As Hoffmann (1967 , 57) writes, 「[P]erceptions are more than a part of political reality: they mold it, insofar as they are the springs and fuel of action」 (see also Weick 1979 , 1995 , 2001 ). Indeed, a variety of scholars have similarly relied on the perceptions of agency heads to evaluate such topics as interest group influence, government contracting, and principal-agent relationships ( Bowling and Ferguson 2001 ; Brudney, Hebert, and Wright 2000 ; Brudney et al. 2005 ; Jacobson, Palus, and Bowling 2010 ; Kelleher and Yackee 2006 ; Palus and Yackee 2009 ). Of specific interest to this article, perceptions are uniquely situated to understanding and to uncovering the theoretical issues at play; after all, public agency officials act on their perception of the legal and policy discretion available to them to make decisions.

To assess the argument, we use the ASAP survey data in combination with state legislative and gubernatorial elections drawn from the same 30-year time period. The electoral data allow us to map the partisan alignment, or misalignment, between political principals and administrative agency heads. 3We employ a pooled, cross-sectional approach to data analyses. To control for time effects, we include survey-year fixed effects in all model specifications. We also control for state-level fixed effects in keeping with the literature, which suggests potential clustering around difficult to measure state-specific commonalities. Finally, we include fixed effects to control for type of agency included.

Dependent Variables

Discretion is defined the freedom of agency officials from the influence from political superiors to pursue policy aims or make policy-related decisions, and our operationalization of this construct, which serves as our dependent variable, uses two different measurement conceptualizations. We describe these variables below. Moreover, for each measure of Discretion , we separately tap: (1) perceived discretion from the governor and (2) perceived discretion from the legislature. Consequently, in subsequent analyses, we estimate a total of four separate models.

The first indicator, Discretion-Policy , is constructed from a series of questions on the ASAP surveys that measure state administrators』 perceptions of influence on agency policy decision-making. Thus, this indicator makes use of the Political Influence-Agency Discretion Trade-Off , which posits that as the influence of political principals on the decisions of public bureaucrats increases, then bureaucratic policy discretion correspondingly decreases (and vice versa). The specific wording of the ASAP survey question follows: In making agency decisions it is usually possible to identify and weigh several major sources of influence. Among these are: (1) governor, (2) legislators, (3) clientele groups. And then: Please indicate below (by circling) the degree of influence each source has on decisions your agency makes in the following decision areas. Respondents evaluated the influence of these actors on four particular decisions—Total Agency Budgets Level, Budgets for Specific Programs, Major Policy Changes, and Agency Rules/Regulations—using four choice categories [None (0), Slight (1), Moderate (2), and High (3) degree of influence].

For the dependent variable Discretion Policy: Governor , we added all of the responses for the evaluations of the governor together, and then subtracted that score from 12 (the maximum possible value). Thus, lower levels of political influence results in higher levels of agency discretion (and vice versa) within these key agency policymaking tools, and the resulting value is an indicator of perceived policy discretion from political principals by administrative agency heads. We similarly created two additional variables for the Legislature and Groups , which serve as an additional dependent and/or control variable, respectively, in the multivariate analyses. The alpha values for all are appropriate (between 0.70 and 0.80), indicating the reliability and the suitability of the scale construction.

The second indicator of discretion, Discretion-Initiatives , is constructed from a series of questions that ask the administrators to assess whether shifts in program priorities were the result of policy initiatives or actions originated by a series of actors, including legislators, the governor, and clientele groups. Here, we see that less discretion is present when an administrator observes one of these actors as initiating new policies or actions for their agency because the freedom of the agency official from political influence to pursue policy-related aims is constrained. For example, if the respondent noted that the governor was a source of new initiatives, then the variable, Discretion-Initiatives: Governor was coded as a 「0.」 If he or she was not identified as a source of new initiatives, then the variable was coded as a 「1.」 We similarly created variables for the Legislature and Groups , to be used as additional dependent and/or control variables.

Independent Variables

The key explanatory variable of interest in subsequent analyses taps whether or not the individual administrator is of the same political party as either the governor or legislature in their respective state in the survey year. In the realm of modern American politics, the Republican and Democratic parties often play a critical role in generating the relevant in-group and out-group attachments for elite political discourse and political decision-making. Thus, to construct the Partisan Alignment variable, we utilized a survey question in which administrators were asked the following: 「Generally speaking, do you consider yourself to be a Democrat; Republican; or an Independent?」 If they responded 「Independent,」 then a follow-up was asked: 「Are you closer to the Democratic Party, Republican Party, or Neither?」 From these responses, we constructed a five-point indicator for the partisan self-identification of the administrator. We then matched each administrator』s party ID with data drawn from state legislative and gubernatorial elections over the same 30-year time period, thus mapping the partisan alignment, or misalignment, between political principals and administrative agency heads. In the models explaining discretion from the governor, Partisan Alignment focuses on the administrator-governor partisan match; coded as a 「1」 for a party match and a 「0」 for a non-match. Similarly, in the legislative models, the focus is on the administrator and the legislature. A legislative-administrator match is present when the administrator is of the same political party as the majority in both houses of the state legislature; a zero is recorded otherwise (which includes split-branch legislature scenarios).

Our remaining explanatory variables fall into three categories: (1) Agency Characteristics, (2) State Context, and (3) Administrator Expertise. 4 The three variables included under 「Agency Characteristics」 are all drawn from the ASAP data. The variables Number of Employees and Agency Budget measure characteristics related to the size of the agency. Number of Employees is an ordinal variable with six categories, ranging from under 25 to over 5,000 employees. Agency Budget is measured in millions and is inflation adjusted to 2008 dollars. 5 Larger agencies may hold more discretion from political principals. We also include the indicator of Proportion of Agency Budget from Federal Funds because if an agency receives more of its budget from federal funds, then it may be less politically dependent on the legislature and/or executive.

The second group of variables explores broader contextual factors at work in the states, and we specifically include variables tapping the overall climate of perceived discretion. In the models explaining discretion from the governor, for example, we include as controls perceptions regarding discretion from the legislature and groups. In the models focused on the legislature, we include as controls perceptions regarding discretion from the governor and groups. The final two contextual variables measure the presence or absence of Divided Government , and whether or not the state is located in the American South . The 「Administrator Expertise」 variables tap varied dimensions of educational and professional experience that an agency head brings to his or her position—including Education , Years in Current Position , Previous Experience in Agency , and Previous Experience in Another State Government. Greater expertise may be related to a greater degree of perceived discretion.

Finally, to assess the circumstances that are theorized to exacerbate the anti-ally feelings of high-level agency administrators, we include two additional substantive variables. First, we add an indicator of whether or not an administrator came to his or her position through Governor Appointment. Agency heads appointed by the governor, as opposed to another appointment method, such as via an independent board or commission, may affect levels of discretion. Second, in an additional set of models, we assess the degree of contact between a political principal and agency officials. We do so by using a series of ASAP survey questions that ask administrators to assess the frequency of personal contacts (either by phone or face-to-face) with their political principals. The respondents evaluated the frequency of contacts using the following scale: Daily, Weekly, Monthly, Less than Monthly, or Never. We combine the impact of the governor and governor』s staff ( Governor Contacts ), as well as legislators and staff ( Legislator Contacts ). We consider less frequent contacts to be 「Never」 and 「Less than Monthly,」 which are coded as 「0,」 whereas the more frequent contacts—「Daily,」 「Weekly,」 and 「Monthly」—are coded as 「1.」

Results

As highlighted above, we assess the argument using two different conceptualizations of perceived discretion—each measured from vantage point of the agency official—for a total of four dependent variables. We use a fixed effects ordinary least squares regression technique to estimate the models associated with the first dependent variable, Discretion-Policy . For Discretion-Initiatives, we use fixed effects logistic regression estimation, as the dependent variables are dichotomous. The first set of results is found in table 1 .

Across all measures of discretion, we find preliminary evidence challenging the ally principle and suggestive support for our theorized relationship. This is evidenced in table 1 by the Partisan Alignment variable』s statistically significant ( p < .05), negatively signed coefficient in all model specifications. Thus, when agency heads are politically aligned with either the governor or the state legislature, state agency heads perceive less—not more—policy discretion on budgets, major policy decisions, and administrative regulations ( Discretion-Policy ). Agency leaders also observe lower policy discretion in terms of their perceived authority over agency program priorities ( Discretion-Initiatives ) when their partisanship aligns with legislative or executive political principals.

These results suggest, across two distinct measures of perceived discretion, that agency heads consistently feel more constrained when their partisanship matches that of their political principals. Thus, we uncover anti-ally principle findings. These results are all the more intriguing given the data』s range: they emerge in data that capture all types of state administrative agencies, in all 50 states, across multiple decades. To get a better sense of the magnitude of these results, we calculated the predicted values (or probabilities, in the case of the logit models) for the dependent variables. To generate these values, we shifted the main predictor variable, Partisan Alignment , from 「Misaligned」 to 「Aligned,」 while holding the other variables constant at their mean value. The results for Model 1, for instance, demonstrate a 4% decrease in discretion in the presence of partisan alignment. 6 We want to emphasize that this result is small in its effect size, suggesting a shift in the dependent variable of less than one point on the 12-point scale when moving from 「Misaligned」 to 「Aligned.」 This implies that administrators are certainly not acting as kings during periods of partisan alignment; however, the evidence that they are mere clerks is only weak in its magnitude in this specification. The results for Model 3 demonstrate a 7% decrease in the predicted probability of shifts in program priorities.

Despite the small magnitude of these effects, it is important to note that these table 1 findings are observed while controlling for several factors that have been hypothesized by others to drive the amount of delegated authority available to US administrative agencies, including the presence of Divided Government . Recall that prior literature typically relied on measures of divided government to empirically assess the ally principle. In our measurement strategy, in contrast, we can directly operationalize the alignment between agency administrators and political principals, whereas Divided Government is included to capture circumstances where the governor and legislature are of opposing political parties. Across all model specifications, Divided Government is an insignificant predictor of perceived discretion. We also are able to consider the discretion from other political influences on the agency head in our models. For example, as evidenced in Model 2, respondents report higher levels of discretion from the legislature when they also perceive higher levels of autonomy from the governor. The same pattern is observed for the governor』s models.

In all tables, we control for other potential alternative explanations, including the size or 「power」 of an agency, as measured by the Number of Employees and the Agency Budget . Number of Employees generally produces a negative relationship with discretion, with a statistically significant result observed in three of the four models, suggesting that administrators in larger agencies feel more constrained than those in smaller agencies. These results suggest that larger agencies may receive, at times, more attention from political principals due to the scope of the agency』s programing reach or potentially a larger and more important agency policy agenda. The variable tapping the Proportion of Agency Budget from Federal Funds is negative and statistically significant in all of the models for discretion from the governor, yet positive and significant in the legislative models. Consequently, when more of an agency』s budget comes from the federal government, an agency head perceives less discretion from the governor, yet more from the legislature. Because agencies typically derive most of their funds from legislative appropriations, it is not surprising that agencies with greater flexibility in funding streams perceive more autonomy from legislatures.

The models also control for administrator expertise—another important potential alternative explanation. 7 Here, varied results are returned, suggesting these constructs are not consistent drivers of perceived policy discretion. The only common finding is that Years in Current Position is statistically significant and positive across all models explaining discretion from the governor. This implies that expertise gained as an administrative head yields greater autonomy from the elected executive. All in all, and despite the fact that these variables are similar to those used by other research to gauge administrator expertise (e.g., Lewis 2007 ), we find little evidence that administrator expertise is consistently driving perceptions of delegated policy discretion by agency heads. The inclusion of the variables remains important, however, given Huber and McCarty』s (2004) finding that bureaucratic capacity—and particularly low capacity—may impact the delegation of discretion from political principal to agency official.

In our argument, we suggest that anti-ally perceptions are anticipated due to pressures within the broader political and partisan context in which administrative decision-making is taking place. Although the models across table 1 control for that political environment, we now move toward assessing this testable implication more directly. For instance, we have argued that a type of 「partisan threat」 across in-group and out-group distinctions (see broadly, Brewer 2007 ) in the decision-making environment may exacerbate anti-ally sentiments and help to drive these patterns. According to this line of thinking, as political elites, agency heads feel—either consciously or subconsciously—a heightened need to advance the priorities and goals of their political party, which is represented by the political leaders of their same party, especially when under threat. As applied here, such a 「partisan threat」 would manifest itself during period of divided government because as an agency head feels more obliged to act on the party』s behalf when politically aligned partisans occupy a leadership role in one elected branch of government but not the other. To assess this possibility, we removed Divided Government from the list of control variables in the model specifications in table 1 . We then ran a series of conditional models to explore the table 1 relationships in the presence or absence of divided government. We demonstrate, in sensitivity analyses not shown here, that the obligation felt by agency heads over policy decision-making increases in size during times of divided government (and remains statistically significant and negative) and decreases in size (or became statistically insignificant) during times of unified government across both conceptualizations of discretion. These findings are returned for both governor and legislative model specifications, which suggests that conflict within the broader political and partisan context in which state administrative decision-making is taking place may indeed affect perceived policy discretion and authority. 8

In table 2 , we add Governor Appointment in its various forms to the model specifications. Given our argument, we would expect that the coefficients to be statistically significant and negative, and also that they would be larger in their substantive impact than those reported in table 1 . To demonstrate this, we unpack the potential joint effect of Governor Appointment and Partisan Alignment by estimating the models explaining discretion from the governor using a series of dummy variables to capture the nature of the interactions. The baseline category (not included in the estimations) is 「Misaligned, Not Appointed.」 The remaining three categories are: 「Aligned, Appointed,」 「Misaligned, Appointed,」 and 「Aligned, Not Appointed.」 The results, as shown in table 2 , demonstrate continued evidence of an anti-ally principle; they are uniformly negative and significant. For instance, when an administrator is both politically aligned with the governor and was appointed by the governor, he or she perceives the least amount of policymaking discretion.

These results suggest such administrators, whom one might expect to have greater trust and therefore policy autonomy from that governor, perceive less freedom in their choices and in their execution of major initiatives and program priorities. Another way to convey these relationships is to calculate the predicted values (or probabilities, in the case of the logit models) for the dependent variables shown in table 1 while setting the Partisan Alignment and the Governor Appointment variables to 1 and while holding the remaining model variables constant. These actions return results that, in the governor』s models for instance, demonstrate an almost 10% decrease in policy discretion (Model 5) and a 22% decrease regarding major policy initiatives (Model 6) when the governor appoints the state agency head and that administrator is aligned with the governor』s partisanship. These results are statistically significant and imply evidence that is in keeping with our argument that pressures to conform to their political party』s positions and its leader may drive lower perceptions of policy discretion.

Finally, we further explore the theoretical argument by adding the Governor Contacts or Legislator Contacts variables, respectively, into the full model specifications. We expect that a higher level of contact between a political principal (or his/her representatives) and the agency head will augment the anti-ally leanings uncovered thus far in the data. We find support for this testable implication in table 3 . Indeed, as expected, the direct effect of all three variables, Partisan Alignment , Contacts , and Governor Appointment , is negative and significant in the governor』s models (Models 7 and 9), whereas, as expected, Partisan Alignment and Contacts are negative and Governor Appointment is positive in the legislature models (Models 8 and 10). To assess the combined substantive effect of these expectations, we calculated the predicted values (or probabilities) for the four dependent variables while simultaneously varying Governor Appointment , Partisan Alignment , and Contacts (either governor or legislator) accordingly and setting all other constructs to their means. We find a 13% change in Model 7, 9% shift in Model 8, 33% change in Model 9, and a 24% adjustment in Model 10. These changes suggest that agency officials perceive even less policymaking discretion under these joint conditions.

The implication of our empirical results thus far is relatively clear: high-level agency officials believe themselves to be more constrained during bureaucratic policy and regulatory decision-making when their political partisanship aligns with their elected superiors. Although the size of the substantive effects vary across the model specifications, the results are important because they consistently imply a refutation of the traditional conceptualization of the ally principle, as well as suggest a need to reevaluate the politics of partisan alignment as applied to the critical governance topics of delegation and agency policy decision-making. Thus, it seems that an agency leader appointed by the governor and who shares the governor』s political persuasion is more—not less—likely to feel constrained in his/her abilities to set agency program priorities, as well as feels more constrained in his/her ability to pursue agency budget, policy, and/or regulatory aims. Substantively, the appointment type of the agency head, as well as their level of contact with their political principals, affects these relationships.

It is worth emphasizing here that feelings of constraint due to political influence from elected officials are not, normatively speaking, a 「bad outcome」 for society. Indeed, the hallmark of a responsive public bureaucracy resides in the degree of political accountability present within the administrative state. However, what is interesting about this argument is not the constraint felt per se, but who feels the constraint more or less acutely. Indeed, the in-group and out-group distinctions, as applied to this specific case, paint an intriguing picture: politically aligned administrative heads may feel more constrained in their agency』s policy decision-making.

It is also worth noting that we have provided only one possible explanation for why these findings materialize in the data. Much more additional work is needed to fully understand the theoretical mechanisms at play. That is because, although our results are preliminary and suggestive that in-group and out-group pressures may drive some portion of the perceptions attached to policymaking autonomy, they are not confirmatory. Indeed, other explanations are also possible. For instance, it may be that politically aligned co-partisans enter public service during times of alignment expecting more discretion and then react more strongly when there is interference by a governor or the legislature of the same party. We cannot rule out this possibility in our data. What we can do (and, indeed, have done), however, is to control for set of other important variables in our analyses, including the length of service of agency officials (and many are long serving), which ought to mitigate this potential explanation.

Discussion

One of the central questions within the system of American public policymaking and management is: Are bureaucratic agents—who are tasked with implementing the nation』s policies and the writing of its regulations—clerks or kings? Conventional wisdom holds that partisanship plays a role in explaining when political principals will delegate greater policymaking discretion to their administrative agents. Indeed, traditional conceptualizations of the so-called 「ally principle」 go one step further by suggesting that an alignment in partisanship between a political principal and his or her agent will encourage elected officials to delegate authority because it limits uncertainty over policy outcomes. Presumably, such delegation provides agency officials considerable latitude in their policymaking decisions—in essence, it allows the freedom to behave more like autonomous kings than mere administrative clerks.

In contrast, our findings demonstrate an anti-ally principle, suggesting that when agency officials are politically aligned with their elected principals, such administrators feel more—and not less—constrained in their policy-related decision-making. Using over 6,000 survey observations drawn from three decades in all 50 states and encompassing all policy areas addressed by state administrative agencies, we find that a shared political affiliation between the governor or the legislature and an agency head restricts feelings of administrator autonomy. Although the substantive effect of this finding is often small, it is significant and robust across different conceptualizations of discretion related to policymaking authority and policy priorities. This result is also demonstrated while controlling for a variety of other rival explanations in the literature used to predict administrator discretion.

Our explorations into the explanations for these findings support the conclusion that administrative agents may desire to further the policy preferences expressed by leaders within their political party due to in-group conformity pressures, feelings that may be exacerbated by a sense of 「partisan threat」 present during periods of divided government. Moreover, we also uncover evidence that the appointment of a bureaucrat by the governor further limits feelings of administrator autonomy over major policy decisions, as does numerous contacts between agents and their aligned political principals. When taken as a whole, these results suggest the important role that partisanship plays in limiting—not expanding—the perceived autonomy of aligned agency officials. Put differently, our findings suggest that greater partisan alignment does not allow administrators, in the words of Lynn and Hill (2008) , to feel like liberated policy 「creators,」 but instead, it appears to remind such administrators that they have obligations and constraints in place that limit their policy choices.

These anti-ally principle results are striking because they largely contradict the extant theoretical and empirical delegation literatures. To some degree, however, the measurement advancements made in this article may explain part of this discrepancy. For instance, as we mentioned earlier, instead of assuming the partisanship of agency heads matches that of the elected executive—a limiting assumption that is used by most of the literature—we generate a new measure that matches the stated partisan affiliations of agency heads to the partisan affiliations of their political principals. Another measurement difference is our focus on the perceived discretion available to agency heads to make bureaucratic policy decisions, whereas other related research, most notably, Epstein and O』Halloran』s (1999) and Huber and Shipan』s (2002) work, focuses on the text of legislative statutes. As we state earlier, we see our focus on perceptions as a major contribution of this study; after all, agency heads rely on their perceptions of the actual amount of delegated authority available to their agency when making public policy decisions.

In closing, although our findings offer provocative insights into the nature of the relationship between administrative agency leadership and their elected political principals, continued attention to this important public management topic is critical. For example, although our models control for personal characteristics associated with administrator expertise, future work should gather interview data from agency administrators and elected officials to explore the interplay between partisanship, expertise, and perceptions of discretion and autonomy. Additionally, research that builds on our patterns uncovered here ought to take the next logical step of linking perceptions of discretion within agencies to tangible actions taken by agencies. For example, scholars may link performance and output measures, on such topics as the writing of agency rules and regulations and/or adjudicative orders, to perceived statutory authority to make such decisions, or may build on and extend our application of social identity theory to drill down even more on the theoretical mechanisms driving these intriguing empirical regularities.

參考文獻：略

推薦閱讀：