Pork Politics: How Earmarks Affect Voter Behavior and Federal Campaigns Travis Braidwood

Home , Bill Nelson, Earmark (politics), Gravina Island, Gravina Island Bridge, Pork barrel

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2013 Pork Politics: How Earmarks Affect Voter Behavior and Federal Campaigns Travis Braidwood

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected] THE FLORIDA STATE UNIVERSITY

COLLEGE OF SOCIAL SCIENCES AND PUBLIC POLICY

PORK POLITICS:

HOW EARMARKS AFFECT VOTER BEHAVIOR AND FEDERAL CAMPAIGNS

TRAVIS BRAIDWOOD

A dissertation submitted to the Department of Political Science in partial fulﬁllment of the requirements of the degree of Doctor of Philosophy.

Degree Awarded: Summer Semester, 2013 Travis Braidwood defended this dissertation on June 18, 2013. The members of the supervisory committee were:

Cherie Maestas

Professor Directing Dissertation

Lance DeHaven-Smith

University Representative

Robert Jackson

Committee Member

Brad Gomez

Committee Member

The Graduate School has veriﬁed and approved the above-named committee members, and certiﬁes that the dissertation has been approved in accordance with university requirements

ii ACKNOWLEDGMENTS

Special thanks to my dissertation committee: Bob Jackson, Cherie Maestas, Brad Gomez and Lance DeHaven-Smith for volunteering to help me through this long process. Thanks as well to my friend and colleague Scott Clifford for his assistance in making this project possible.

I would like to thank the Claude Pepper Center and Jennifer Jerit for their time and resources that permitted me and several other professors and graduate students to conduct experiments in a controlled laboratory setting.

I would like to thank Dale Smith, the Political Science Department at Florida State University, and, again, Cherie Maestas for availing me of opportunities to aid in data collection, including the Research Intensive Baccalaureate Certiﬁcate (RIBC) Program, and the option to teach several unique undergraduate courses. Thanks to these opportunities, I was able to recruit undergraduate students to assist in the data collection process, which saved me countless hours.

Finally, I would like to thank my father, Ken, mother, Patricia, and sister, Jordan, as well as my close friends. Thank you all for always sticking by me and encouraging me to pursue my goals to the fullest.

iii TABLE OF CONTENTS

LIST OF TABLES viii LIST OF FIGURES ix ABSTRACT x

1 INTRODUCTION 1 1.1 Deﬁning Earmarks 3 1.2 Outline 7

2 THEORY AND DESIGN 9 2.1 Pork in the Eyes of Members: Needs and Motivations 9 2.2 Earmarks and Elections 13 2.2.1 The Rewards of the Enlightened 15 2.2.2 New Bridge, What Bridge? The Reality of the Unaware Voter 17 2.2.3 Rewarding Those Who Reward: Earmarks and Campaign Contributions 20

3 MEASURING EARMARKS 23 3.1 Data Accuracy and Sources 24

4 WHO BENEFITS FROM EARMARKS, AND WHY? 28 4.1 Getting Pork: Which Representatives Receive Earmarks? 28 4.2 Getting Pork: Which Senators Receive Earmarks and Why? 33 4.3 Frightened Into Action: Electoral Vulnerability and Effort to Secure Earmarks 37

5 DESIRABLE PORK: DO VOTERS REWARD FOR EARMARK ACQUISITION? 45 5.1 Introduction 45 5.2 Awareness, Credit Claiming, and Desirability 47 5.2.1 Credit Claiming 48 5.2.2 Issue Framing 50 5.2.3 Issue Publics 51 5.3 Hypotheses 52 5.4 Experimental Design 54

iv 5.4.1 Study 1 55 5.4.2 Results of Study 1 56 5.4.3 Study 2 59 5.4.4 Results of Study 2 61 5.4.5 Study 3 64 5.4.6 Results of Study 3 65 5.5 Conclusion 67

6 MYSTERIOUS PORK: THE LACK OF CITIZEN AWARENESS OF EARMARKS 70 6.1 Introduction 70 6.2 Earmarks and Electoral Behavior 71 6.2.1 Self-Interest, Information, and Electoral Rewards 73 6.3 Data Sources 76 6.3.1 Data Accuracy 77 6.4 Methodology 79 6.4.1 Model 1: Project Recall: 2008 CCES 79 6.4.2 Measuring Media Coverage 80 6.5 Results 81 6.5.1 2008 CCES Survey 81 6.6 Approval and Support 86 6.6.1 Pork and Approval 86 6.6.2 Earmarks and Support for the Procuring Incumbent 89 6.7 Conclusion 92

7 PORK AND CAMPAIGNING: FINANCIAL BENEFITS OF EARMARKING 95 7.1 The Entangled Campaign Process 97 7.2 Pork and Campaigning: Financial Beneﬁts of Earmarking 100 7.2.1 Contributions for Projects Theory: Are Earmarks the Result of Campaign Contributions? 103 7.3 Data and Methods 104 7.4 Results 106 7.5 Conclusion 109

8 CONCLUSION 111 8.1 Earmarks and Institutions 112 8.2 Earmarks and Electoral Outcomes 113 8.3 The Future of Research Regarding Pork Politics 115

A SUPPORTING MATERIAL TO CHAPTER 4 117

v B SUPPORTING MATERIAL TO CHAPTER 5 123

C SUPPORTING MATERIAL TO CHAPTER 6 135

D SUPPORTING MATERIAL TO CHAPTER 7 150

BIBLIOGRAPHY 151

BIOGRAPHICAL SKETCH 161

vi LIST OF TABLES

4.1 Representatives Securing Earmarks, Fiscal Years 2008-2010 31 4.2 Senators Securing Earmarks, Fiscal Years 2008-2010 35 4.3 Regression of Past Earmarks on Future Vote, Freshmen Only 39 4.4 2SLS Model Predicting Pork’s Effect on Future Vote Share 42

5.1 The Effect of Positively Framed Pork Treatments on Support for Sen. Bill Nelson (D-FL), Summer 2011 57 5.2 The Effect of Positively Framed Pork Treatments on Support for Rep. Miller (R, FL-1st), Fall 2011 62 5.3 The Effect of Negatively Framed Pork Treatments on Support for Sen. Bill Nelson (D-FL), Spring 2012 66

6.1 Project Recall, CCES 2008 82 6.2 Senator and Representative Approval, CCES 2008 87 6.3 Pork and Incumbent Support, CCES 2008 90

7.1 The Effect of Contributions on Freshmen Representatives Securing Earmarks 107

A.1 Freshmen Representatives Securing Earmarks, Fiscal Years 2008-2010 117 A.2 Heckman Two-Step Estimation of Earmark Acquisition 119 A.3 State Population and Representatives Securing Earmarks, Fiscal Years 2008-2010 120

B.1 The Effect of Pork Treatment on Support for Rep. Miller (R, FL-1st) 123 B.2 Number of Subject Surveys and the Effect of Pork Treatments on Support for Sen. Bill Nelson (D-FL) 124 B.3 Disaggregated Dependent Variables, Study 1 (Summer 2011, Sen. Nelson) 125 B.4 Disaggregated Dependent Variables, Study 2 (Fall 2011, Rep. Jeff Miller) 125 B.5 List of Utilized Local Newspapers 127 B.6 Political Sophistication and Positively Framed Pork Treatments on Support for Sen. Bill Nelson (D-FL), Summer 2011 130 B.7 Disaggregated Dependent Variables, Spring 2012 (Sen. Bill Nelson) 131 B.8 Favorability of the Tea Party on Support for Senator Nelson (FL), Summer 2011 133 B.9 Exclusion of the Party Identiﬁcation Control Variable, All Models 134

vii C.1 Variable Information Utilized in the Analysis of the 2008 CCES 136 C.2 Variable Information Utilized in the Analysis of the 2006 CBS/NYT Poll 137 C.3 The Number of Earmark Projects on Predicting Project Recall 138 C.4 Local News Papers 141 C.5 Pork Dollars and the Ability to Recall Projects, 2008 CCES 143 C.6 Incumbent Approval and Earmark Allocations, CCES 2008 145 C.7 Incumbent Vote, CCES 2008 147

D.1 Tobit Analysis of the Effect of Contributions on Freshmen Representatives Secur- ing Earmark 150

viii LIST OF FIGURES

2.1 Earmark Projects to States and Districts, Fiscal Year 2008 10 2.2 2011 CBS Poll: Identiﬁcation of Earmark’s Share of the National Budget 11 2.3 Previous and Revised Theories of Earmark Attribution 21

3.1 Federal Earmark Allocations, FY2008: Comparing TCS, CAGW and FAADS 26 3.2 Earmarks FY2008-2010: Comparing TCS to CAGW Measurements 27

4.1 Dollar (logged) and Number of Representative Earmarks, FY2008-2010 30 4.2 Dollar (logged) and Number of Senator Earmarks, FY2008-2010 34 4.3 Marginal Effect of Past Earmarks on Future Vote, Freshmen Only 40 4.4 Marginal Effect of Earmarks on Future Vote, 110th and 111th Congresses (2SLS) 43

5.1 2010 CNN Poll: Acceptability of Earmarks 49 5.2 Treatment Effects of Pork on Support for Sen. Nelson (Positively Framed) 58 5.3 Treatment Effects of Pork on Support for Rep. Miller (Positively Framed) 63 5.4 Treatment Effect of Pork on Support for Sen. Bill Nelson 67

6.1 Project Recall and Representative Pork Dollars (logged) 84 6.2 Project Recall and State Newspaper Coverage of Senator Pork 85 6.3 Representative Media Coverage (of Earmarks) on Incumbent Support 92

7.1 The Earmark-Contribution Endogeneity Problem 102 7.2 Marginal Effect of Contributions on Earmark Acquisition, 110th and 111th Con- gresses 108

B.1 Lexis-Nexus Media Search on Earmarks and Congress 132

ix ABSTRACT

This dissertation explores the role of earmarks, also known as pork projects, in several facets of American politics. After reviewing the changing history and various means of measuring earmark projects, I attempt to determine which Members of Congress are most adept at securing earmarks, and whether these projects affect electoral security. Second, this work departs from previous assumptions that pork projects are viewed equally by all recipients, given recipients are made aware of the projects at all. Third, this work challenges existing claims that contend a direct linkage between voter awareness of earmark projects and electoral support for an incumbent; instead, I argue for the role of media dissemination of this information. Finally, this project differentiates itself from the current literature by approaching the impact of earmarks not solely as a means to directly appeal to the majority of voters, but as a quid pro quo to be invoked by Members looking to shore up campaign support. Rather than contend that only voters reward incumbents for project dollars, this paper explores the impact of earmarks on campaign contributions provided by special interest groups.

x CHAPTER 1

INTRODUCTION

Over decades of legislation, Members of Congress (MCs) succeeded in securing billions in earmarked projects. These outlays ranged from highway overpasses, to the construction of advanced aerospace engineering facilities; from stereotypical projects, such as “the bridge to nowhere,” to the construction of food banks and other public service institutions (Taxpayers for Common Sense 2011).

While earmarks have existed as an institutional practice for quite some time in American politics (some scholars would date them to the founding if allocations for road and waterway projects are considered), scholarly work attempting to explain the allocation and utility of earmarks for MCs has been relatively sparse. This deficiency is largely due to the difficulty in collecting data that simultaneously detects earmarks and attributes them to a specific MC, as well as a lack of survey data asking voters their awareness and appreciation (or lack thereof) of pork in their home state or district. This work sheds light on many of these remaining questions regarding earmarks, and attempts to bridge several distinct sub-fields of research: congressional work on why and how members secure earmarks, research on congressional elections and challenger deterrence, work on voter behavior and information processing as it pertains to particularized benefits, and finally, research on campaign contributions and special interest support.

While previous research on earmarks had been stymied by information gathering hurdles, recent reforms in Congressional requirements have led to an inﬂux in scholarly work on the subject. Armed with new, highly accurate data, scholars have endeavored to better explain why pork exists,

1 and how it might work to benefit Members electorally. This work aims to contribute to the growing literature on earmarks in three fundamental ways. First, this study attempts to ascertain the utility of pork as a tool. Specifically, I explore an assumption that has been long held by the overwhelming majority of politicians: that pork dollars delivered to a constituency are always a good thing. Utilizing experimental data, scholars can begin to shed light on this reward hypothesis. Second, after developing a theory to explain pork’s potential effectiveness, this study reexamines the most fundamental assumption regarding pork and elections: pork dollars can help a candidate succeed. Finally, this work investigates the electoral impact of earmarks, not via direct effects on voters that culminate in electoral support, but through indirect effects that alter the ability of a Member to campaign effectively. In other words, given the nebulous nature of voter awareness, might Members actually be reaping benefits from the most ardent, wealthy, and attentive of supporters: special interest groups?

As mentioned, this work departs from previous research on earmarks in several fundamental ways. First, it departs from several existing theories that contend a direct linkage between voter awareness of earmark projects and electoral support for an incumbent. It rejects the claim that earmarks alone will culminate in electoral benefits from appreciative voters. Such propositions require voters undertake the Herculean task of first becoming aware of projects, differentiating those projects from normal spending (i.e. distinguishing particularized benefits), and then correctly attributing the project to the appropriate federal official. For decades scholars have worked under the assumption that earmarks, and other direct efforts, have an electoral impact; this assumption that has mixed empirical support. Second, this work departs from the previous assumption that pork projects are viewed equally by all recipients, given recipients are made aware of the projects at all. Traditionally, earmarks have been treated as a local benefit that is viewed as universally good by constituents, however this overlooks project desirability. Even if we assume voters reward for pork, previous work has not considered the value of particular projects that vary by issue area. For example, college student voters may wish for earmarks to benefit higher education, while military personnel may increasingly support a Member who brings home defense dollars. Finally, this

2 project differentiates itself from the current literature by approaching the impact of earmarks not solely as a means to directly appeal to the majority of voters, but as a quid pro quo to be invoked by Members looking to shore up campaign support. Rather than contend earmarks beneﬁt MCs solely through voter appreciation for projects, this paper explores the role special interest groups play in persuading MCs to procure earmarks in exchange for campaign contributions.

1.1 Deﬁning Earmarks

The term earmark is derived from its original use by herdsmen who would mark notches in the ears of their livestock in order to denote ownership in a manner similar to branding (Porter & Walsh 2006). In modern use, however, the term is meant to denote a differentiation between traditional legislation and particularized projects directed at targeted recipients.

The primary purposes of these projects are twofold: inclusion in legislation as a means to facilitate political deal making, and as a venue to bring beneﬁts to an individual constituency (Ferejohn 1974). The electoral motivations for Members to seek out federal dollars are readily apparent: federal outlays to a district or state allow Members to credit claim, while simultaneously showing would-be challengers that they remain politically adept.

Unlike normal legislation, which has a sponsor(s), and typically various cosponsors, earmarks are not as easily defined or tied to a single Member.In fact, various terms and definitions of legislative earmarks exist. For example, the Congressional Research Service provides several differing specifications to describe the same concept: congressional earmarks, congressional directives, presidential earmarks, and congressionally directed spending items (Brass et al. 2008). We are left with no clear definition; it is possible to view earmarks as any budgetary resource that is directed at a specific entity, geographic region, or allocation created for a solitary purpose. The difficulty lies in drawing a distinction between federal funds inserted into legislation originating

3 from Congress, typically during the committee stage, versus requests directed at federal agencies from the executive, or by speciﬁc members of Congress.

Earmarks originating from Congress consist of projects that are inserted into appropriations, authorization, and/or revenue bills at the behest of Members (Brass et al. 2008). Many of these earmarks ﬁnd their way into report language and joint explanatory statements in the form of expressed intent directed at agency behavior. In other words, Members are often explicit about the intended use of the appropriated funds. While these earmarks are included separate from the agency budget, the budget allocations have the force of law upon passage of the legislation.

Earmarks may also originate from the executive. The President may use a Member of Congress as an intermediary to propose a project, the President can direct Members to alter agency spending to ensure funds are devoted to a specific purpose, the executive may request an agency devote funds to a specific grant or initiative, or the executive may rely on agency discretion to ensure specific localities receive particularized benefits (Brass et al. 2008). Moreover, the executive may rely on consultation with both Members of Congress and agency heads to ensure budget allocation and proper use of allocated funds.

Because of the varied origins of earmarks, it can be quite difficult to arrive at a single definition. Scholars and members of the media add to this confusion with their use the terms “pork” and earmarks,” which can assume a variety of definitions. For the purposes of this work both terms will be used interchangeably to describe inserted projects such as those previously described. However, this does not answer where researchers should turn to arrive at a standard measure. For this, it is best to rely on citizen groups, such as Citizens Against Government Waste (CAGW) and Taxpayers for Common Sense (TCS), which have offered guidance. Additionally, the federal government itself, via the Congressional Research Service (CRS), also offers limited insight into how pork should be defined.

Looking first at Citizens Against Government Waste, CAGW we find that considers seven criteria for a project to be considered a pork project: (1) requested by only one chamber, (2) not specifically authorized, (3) not competitively awarded, (4) not requested by the President,

4 (5) greatly exceeds the President’s budget request from the previous year, (6) not the subject of a congressional hearing(s), and (7) only serves a local or special interest (Finnigan 2007). A review of these requirements raises several concerns; first is the fact that CAGW does not include Presidentially requested earmarks in their calculations. However, this omission may be desirable, depending on whether researchers are primarily concerned with attribution of earmarks to specific MCs, versus total dollars brought into a district or state. Additionally, only one of the seven criteria need be met for an allocation to be considered an earmark, which would not be troublesome if it was not for CAGW’s reliance upon vague classifications, such as “greatly exceeds,” when measuring changes in spending. Nonetheless, the fact that CAGW has been collecting data on pork spending since 1991 means the resulting data are amongst the richest temporally.1

Taxpayers for Common Sense, another citizen advocacy group opposed to earmarks, took advantage of recent policy changes in 2007, which allow them to accurately attribute pork to specific members with a high degree of accuracy. This was made possible by the fact that the House and the Senate enacted changes that required elected officials from both chambers to disclose all earmarks inserted into legislation. TCS defines an earmark as the following:

[L]egislative provisions that set aside funds within an account for a speciﬁc program, project, activity, institution, or location; [t]hese measures normally circumvent merit- based or competitive allocation processes and appear in spending, authorization, tax, and tariff bills (Taxpayers for Common Sense 2008).

TCS has worked to gather this data from individual sources (eg. the Member’s websites). Since TCS can rely on MCs for earmark disclosure, they need not rely solely on a strict deﬁnition.

Finally, the Congressional Research Service (CRS) has also attempted to lend clarity in their reports to Members of Congress. Speciﬁcally, the CRS deﬁnes earmarks as:

1 State-level measures are not available until ﬁscal year 1995, and district-level measures until ﬁscal year 2008.

5 [Provisions] included primarily at the request of a Member [...] authorizing or recommending a speciﬁc amount of discretionary budget authority [...] targeted to a speciﬁc State, locality or congressional district, other than through a statutory or administrative formula driven or competitive award process (Lynch 2008, CRS2).

The House and the Senate, which share a similar deﬁnition, appear to agree with the deﬁnition offered by the CRS. The House sees earmarks as the following:

[Provisions or report language] primarily at the request of a Member [...] providing, authorizing, or recommending a speciﬁc amount of discretionary budget authority, credit authority, or other spending authority for a contract, loan, loan guarantee, grant, loan authority [...] targeted to a speciﬁc State, locality or Congressional district (House Rule XXI, Cl. 9(d); 110th Congress).

The Senate definition is nearly identical (Senate Rule XLIV, Cl. 5(a); 110th Congress). Further- more, both the Office of Management and Budget (OMB) and the President seem to agree with these definitions (Brass et al. 2008). In 2007, both the House and the Senate updated their rules regarding earmarks and further clarified these terms. Namely, the House now considers an earmark to be the following:

A provision or report language included primarily at the request of a Member [...] providing, authorizing or recommending a speciﬁc amount of discretionary budget authority [...] targeted to a speciﬁc State, locality or Congressional district, other than through a statutory or administrative formula-driven or competitive award process (House Rule XXI, Cl. 9(e); 112th Congress).

This deﬁnition is strikingly similar to that provided by the CRS. This latest incarnation has expanded to include earmarks that appear in the bill report language, and now differentiate the request from normal spending by highlighting the fact that it is removed from the normal “competitive reward process.”

While there exists obvious variation in the conceptualization and measurement of earmarks, all three deﬁnitions share commonalities. First, all three agree that earmarks are federal dollars

6 spent on specific projects, or allocated to specific entities. In other words, these are not broad, abstract topics, or generalized agency allocations. Second, by their nature, earmarks are not a part of the usual competitive process that features year-to-year battles over agency funding. Finally, all agree that allocations that appear in statutory text, which are usually inserted by the authorizing committee, are earmarks. What is less clear, however, are committee and conference committee reports that contain agency directives and Member-agency requests with specific spending instruc- tions. While these written requests do not have the force of law, they nonetheless account for the vast majority of specified spending.2 These MC requests function as de facto requirements due to the nature of the requests; in other words, were an agency to ignore these requests, they may face a reduced budget in the future. While Members retain the option of raising points of order against projects, this is frequently waived for political reasons (Finnigan 2007).

1.2 Outline

This project explores some very broad questions pertaining to earmarks, such as who gets pork, and does it matter? To achieve this end, the work has been divided into chapters, each devoted to an individual exploration of pork’s influence. Chapter 2 begins the analysis by providing an exploration of the various theories at issue. As mentioned, this project attempts to bring together several fields of study, including Congressional behavior, elections, voter behavior, and campaigns. The opening chapter attempts to unite these topics in the consideration of the role earmarks play in American politics. Armed with this, Chapter 3 will begin the analysis by introducing earmarks as an empirical measurement, answering both general and descriptive questions. Next, Chapter 4 asks who in Congress benefits from pork, in other words, which Members succeed in securing

2For example, Porter & Walsh (2006, pg 7., n18) point to a 2006 Congressional Research Report that suggests that more than 95% of all earmarks in ﬁscal year 2005 were contained in committee reports and legislative directives, rather than being included directly in the statutory language.

7 projects? Additionally, this chapter will consider the impact of projects on electoral vulnerability. Chapter 5 ventures to remove confounding factors surrounding the impact of pork by analyzing its effect in a controlled setting: the laboratory. When given direct information about earmarks in a respondent’s district or state, do subjects view their MCs more favorably? Chapter 6 asks if the recipients of pork, namely constituents, are aware of the benefits they receive. Only a fraction of citizens claim to have knowledge of pork projects in their state or district, so the question is this: is there any validity to existing academic claims that voters appreciate localized benefits? Finally, Chapter 7 wades into the treacherous world of campaigns and electoral outcomes. Specifically, the chapter proposes a unique approach to the study of elections to determine if contributions prompt Members to bring home more pork.

8 CHAPTER 2

THEORY AND DESIGN

2.1 Pork in the Eyes of Members: Needs and Motivations

In the limited research regarding earmarks, answers as to why Members pursue pork has been the subject of some consideration. Shepsle & Weingast (1981) and Ferejohn (1974) contend that the reason pork is accepted, often by overwhelming majorities rather than minimum winning coalitions (Riker 1962), is due to the uncertainty legislators feel in regards to the size of winning coalitions. This uncertainty prompts Members to rely upon the norm of universalism, which results in coalitions approaching unanimous size. Stein & Bickers (1995, 143), despite this finding, contend that universalism would suggest that every district should receive “at least some benefit, albeit at different levels.” However, as Figure 2.1 reveals, even a cursory glace at recently released data on earmarks projects reveals that this conclusion is incorrect. Clearly, the vast majority of districts (96%), and all states, received earmark projects in fiscal year 2008. Balla et al. (2002) attempt to reconcile these differing theories by suggesting that those seeking earmarks are ever cautious of exploiting the minority party for fear of being categorized as wasteful. However, if the majority party allows the minority to secure earmarks as well, no such charges of majority exploitation

9 could be levied. Thus, the reason we should expect to see earmark distributions like those in Figure 2.1 is because of a need for partisan blame avoidance.1 That is not to say that the minority party must be treated as equal partners, rather that they simply be permitted some indeterminate share of pork projects that is not extremely disproportionate from that of the majority.

Figure 2.1: Earmark Projects to States and Districts, Fiscal Year 2008

While Balla et al. (2002) find support for blame avoidance theory using data on higher education, recent publications (Lazarus 2009, 2010; Lazarus & Steigerwalt 2009; Engstrom & Van- berg 2010) that incorporate all pork categorizations raise doubts as to the breadth of these findings. Specifically, recent work by Jeffrey Lazarus has raised concerns with Balla et al.’s (2002) claim

1Earmarks in Figure 2.1 are counts of state and Representative total projects; the data was gathered by Taxpayers for Common Sense. More on this measure will be discussed below.

10 that political parties stand as the crucial factor in earmark acquisition. As Lazarus (2010) points out, party does not predict the distribution of earmarks, rather Member and institutional factors (ideology, seniority, party leadership positions, and membership on the appropriation committee and subcommittees) are amongst the strongest predictors.2 Moreover, partisan blame avoidance theory relies on two latent assumptions regarding blame: ﬁrst, that the public is attentive enough to respond to a message claiming that one party was exploiting earmarks at the expense of the other (i.e. that partisan blame could be heard). Second, that the public would be able to weigh such a message against the seriousness of the accusation. In other words, that the public is able to distinguish claims of waste resulting from earmarks versus other forms of government waste. This raises a fundamental question: how much knowledge should we expect the average American to possess when it comes to earmarks? As Figure 2.2 reveals, the majority of Americans (41% in this 2011 CBS poll) readily admit they possess no knowledge about how much of the federal budget is devoted to earmarks; moreover, a mere 16% were able to correctly identify the correct portion of the budget (< 5%). Clearly, as suggested by Balla et al., if members are concerned that one party may receive blame as being wasteful, such fears are molliﬁed by the fact that most Americans lack a frame of reference when it comes to pork.

Figure 2.2: 2011 CBS Poll: Identiﬁcation of Earmark’s Share of the National Budget

2That said, my own analysis (Chapter 4) ﬁnds that party is a signiﬁcant predictor in the House context, but that other factors related to institutional position are stronger.

11 The evolution of explanations surrounding the causal mechanism of earmark acquisition has left scholars in a state of ambiguity when it comes to predictive theory. This work attempts to lend clarity to this debate by reconciling several differences that have been raised over the past thirty years of research. It is clear that universalism cannot be rejected outright, seemingly all members have access to pork if they choose; it is equally clear that such a theory fails to account for institutional factors. How then do we explain which members secure pork? The answer lies in a combination of institutional factors (i.e. the ability of a given member to secure pork given his or her position in the institution), and the need of an individual member. Members of Congress are self-interested actors (Mayhew 1974b), ever-seeking to reduce electoral threats, while simultaneously working to augment electoral security. Complicating this task is the fact that members are surrounded by uncertainty in regards to reelection, and therefore choose to act strategically (Jacobson 2009). Therefore, we should expect that the Members themselves are the key determinant of the need for pork where party merely functions as an enabler. In other words, Balla et al. (2002) are correct in saying that the parties tolerate pork from their counterparts for fear of the shadow of the future (that the majority party will one day be the minority), but the key pork determinants are not the parties but Members looking to minimize risk by maximizing constituent satisfaction.3

Consistent with a theory of universalism, risk minimization through acquisition does not propose that securing earmarks is solely conditional on the party of a Member. Readers may question the lack of consideration of the role of parties; however, given that earmarks are primarily an individual affair, such a proposal is not revolutionary. Additionally, MCs face a series of constraints; these constraints are political in nature (e.g. reaching group consensus and working with opposite-minded colleagues), and practical (e.g. time and effort). Those wishing to secure pork are no doubt aware of the effort that will be required to ensure a project of any sizable quantity makes its way into a bill. This will require close attention throughout the legislative process. Ulti-

3As the next section will detail, this does not imply that all Members need be active earmark seekers. Rather, this merely contends that Members each possess an individual calculus for maximizing their chances for reelection given their perceived constituency desires.

12 mately, MCs must engage in a cost-beneﬁt analysis whereby they weigh the costs of effort against the perceived beneﬁt to their constituents.

The question then becomes: what can MCs do to ensure their earmark efforts are successful? The clearest answer lies in committee service. Committee assignment serve two fundamental functions when it comes to earmarks; first, once an MC is in a position of power that Member can use his or her vote as leverage to secure more projects. Given the size of earmarks when compared to the federal budget, earmark requests from an MC’s legislative peers is essentially costless. The cost of delaying or voting down provisions of a bill, on the other hand, can be significantly more costly. Second, earmarks are secured in committee and subcommittee at the request of individual Members, and it is the party (via the individual assignment request) that determines committee assignment. Earmarks make their way into (sub)committee reports, which expresses legislative intent.4 Therefore, MCs appointed to positions of power within (sub)committees charged with appropriations (e.g. Members of the Appropriations Committee, or chairs and ranking members of the appropriations subcommittees) should prove the most adept at securing earmarks. While we might expect party to influence assignment, such party-based effects should be reflected in committee assignment.

2.2 Earmarks and Elections

A number of scholars have contended that earmarks bolster constituent support (see Fer- ejohn 1974; Baron 1990; Shepsle & Weingast 1981; Jacobson 2009; Stein & Bickers 1995 for examples), however such arguments make several assumptions about the electoral process. It is the contention of this work that such a direct relationship does not exist; rather such “reward”

4This is why earmarks are also referred to as “Congressionally-directed spending requests.” See Taxpayers for Common Sense for more on this process < www.taxpayer.net/library/article/earmarks-and-earmarking-frequently- asked-questions >

13 effects can be more accurately explained by elite support. Specifically, when earmarks are considered to have an electoral affect, the underlying assumptions are threefold. First, voters are assumed to be aware of projects, or the change in number of projects, from the previous year. In order to reap benefits from projects brought home, Members must assume that the recipients are minimally aware. Second, directly beneficial pork requires that voters differentiate earmarked dollars from normal federal spending. While it is obvious that normal federal spending may influence voter behavior, it is largely out of the control of individual Members of Congress.5 Finally, voters must be able to accurately attribute federal earmark projects to the appropriate member of Congress, or, at minimum, to all incumbents. Alterations in pork-barrel spending are meaningless if voters are not able to electorally reward the procurers.

It is the contention of this work that the aforementioned heroic assumptions are unrealistic and not actualized. Following a discussion of earmark acquisition, this work addresses the outliers: Members who are able to inculcate their constituents with knowledge of their pork-barrel efforts, and what scholars should expect the impact of earmarks to be on an enlightened citizenry. Next, I detail how this work differs from numerous others in its contention that earmark projects go largely unnoticed or unattributed by most voters. Further, this section proposes a theory to explain why past studies (Stein & Bickers 1994, 1995) have arrived at differing conclusions: namely, that by failing to account for media exposure, scholars have incorrectly attributed citizen awareness to possession of political information. Finally, this project considers an alternative causal story regarding earmarks: namely, that pork may play an additional role in regards to reelection via Member inﬂuences on special interest support.

Chapter 4 relies upon the theory proposed above to assess the role of institutional position on pork acquisition across several years of earmarks. Additionally, while a number of studies have claimed pork aids MCs in their electoral goals (Baron 1990; Jacobson 2009) only one attempts to test whether this relationship exists (Stein & Bickers 1994), and that test found no relationship. Af- ter discussing earmark distribution and acquisition, Chapter 4 reexamines this fundamental ques-

5For example, Medicare, Social Security, and Medicaid are in part or wholly federal programs, but the allocations are determined by formula.

14 tion. Speciﬁcally, this work proposes to explore the endogenous relationship between electoral vote and earmarks using an instrumental variables approach. This approach helps to overcome some of the shortcomings in prior studies, which failed to account for the reciprocal relationship.

2.2.1 The Rewards of the Enlightened

There are two obvious, fundamental critiques of the notion that earmarks provide a direct electoral benefit to candidates: first, such a benefit assumes voters are grateful for federal dollars brought home; second, in order to reward or punish, voters must be aware of particularized benefits derived from pork projects. In other words, voters must be able to successfully connect the actions of elected officials to specific attributable benefits, which requires “both the knowledge and the beliefs of the voters” (Popkin 1991, 96). This argument, and indeed all attribution arguments, rely on traditional notions of voter support: the idea that voters champion their member but despise the institution (Fenno 1978; Mayhew 1974b; Hibbing & Theiss-Morse 2002). In regards to pork, this raises a fundamental question: in spite of the fact that most voters are unaware of the earmark efforts of the Members, does that necessarily imply that pork could never benefit incumbents? It is the contention of this work that the answer to that question is a resounding no. However, this is conditional on the information readily available to a given voter.

In all likelihood, most voters are not cognizant of most information regarding pork, or at the very least, are only aware of projects that are seen as personally beneﬁcial and observable. However, assuming for a moment that voters could be made aware of the projects coming to their district or state, and are self-interested (Downs 1957; Riker & Ordeshook 1968; Arrow 1951),6 these enlightened citizens should view the responsible member more favorably. Moreover, recipients will be more grateful for rewards that are salient and of personal importance. In other words “[t]eachers might be expected to be more aware of new grant awards for education, while envi- ronmentalists should be attuned to the ﬂow of monies to the district [or state] for environmental

6This merely implies that voters are able to act in a self-interested way; it does not necessitate any of the accompa- nying assumptions associated with rational behavior.

15 protection” (Stein & Bickers 1994, 380-81). Earmarks to a particular locality that are not personally beneﬁcial are also expected to increase support, but less so than direct beneﬁts.

Readers may ask if such voter attentiveness only exists in an idealized world, what is gained by such a hypothetical? The answer to this question is twofold. First, Members actively attempt to make their earmark efforts known to the public. As Gimmer et al. (2012) recently showed, this is largely accomplished via press releases offered by the Members themselves; further, this credit-claiming activity seems to transcend ideological and party lines. For example, Jeff Miller (Republican, FL-1st), one of the House’s most consistently conservative Members, prominently displays several press releases touting money he has returned to his district, such as $215,236 to the “Literacy and School Libraries Program for Holmes County.”7 The second reason why such a question is important lies in the assumptions found in the literature. Stein and Bickers (1994) identify several propositions that must prove true if earmarks are to have an impact on elections. Amongst these is the idea that “voters reward their representatives on election day for securing district beneﬁts” (1994, 383). However, this presupposes that a majority of voters in a given constituency view received beneﬁts in a positive light, yet such a relationship has never been shown. It is entirely possible that voters view all earmarks as undesirable; this proposition is not totally unreasonable given a lack of public support for aggregate pork spending. This study attempts to shed light on these uninvestigated questions.

A considerable amount of work has been conducted on the ability of the news media and political elites to inﬂuence public attitudes to adopt a given policy position (Iyengar 1991; Nelson et al. 1997)—the primary concern being the legitimacy and stability of public opinion if it can be so easily manipulated. This manipulation is thought to depend on which message is repeated most often (Zaller 1992; Cappella & Jamieson 1997; Nabi 2003; Chong & Druckman 2007), versus the strength of a frame conditional on the source of the information (Brewer 2001; Druckman 2001a). The processing of the information provided by frames assumes that considerations be stored in memory (i.e. are available), that the information be readily accessible, and this information is able

7“MILLER ANNOUNCES FUNDING FOR SCHOOL LIBRARIES AND LITERACY IN HOLMES COUNTY” < http://jeffmiller.house.gov/News/DocumentSingle.aspx?DocumentID=137600 > Accessed: October 21, 2011

16 to be appropriately applied to a given context (Chong & Druckman 2007, 639-40). Once exposed, the likelihood that a frame will shape an individual’s opinion depends on the quality of the proposed argument, as well as the source (Eagly & Chaiken 1993; O’Keefe 2002). In the laboratory setting we can control for the confounding effects related to the legitimacy of the source, and instead focus solely on the strength of the frame via the breadth and depth of the information.

To examine the degree of voter gratitude for earmarks, this work turns to the growing area of research that asks how well informed the public is in regards to political information. This branch of research has approached the topic in two differing ways. One method has been to rely on statistical models that multiply impute expected response values that would exist if the respondents were fully informed (see Althaus 1998; Bartels 1996; Delli Carpini & Keeter 1996; Gilens 2001). The alternative route involves the experimental setting: supplying respondents with varying amounts of political information, and measuring whether their responses vary (see Fishkin 1997). Chapter 5 of this work employs the latter approach. Obviously, a natural situation would be difficult to capture in the real world, however, such a treatment could be observed in a controlled setting. This could be accomplished in two ways: the first involves reminding voters of the policy process by simply defining earmarks.8 The alternative entails providing respondents with a cue containing specific information that allows them to make inferences based on that information.9 As Druckman et al. (2010) point out, these are distinctly different treatments. For the purposes of this work, we are interested not in whether respondents know what pork is (the first approach), but in whether attributing a project to a specific Member alters evaluations; in other words, gauging the effect of information akin to that provided by the news environment.

2.2.2 New Bridge, What Bridge? The Reality of the Unaware Voter

Intuitively, the notion that a member can increase voter support by providing tangible beneﬁts seems plausible. Given that politicians are self-interested (Mayhew 1974b), why else take ad-

8See Druckman 2001a,b; Wood 2000; Price & Tewksbury 1997 for more on the signiﬁcance of framing effects. 9See Eagly & Chaiken 1993; Romero 1996; McLeod & Shah 2008; Sniderman et al. 1991 for more on cues.

17 vantage of the privileges elected office brings if not for personal gain? To be sure, scholars have asserted a connection between service and the personal vote, be that in the form of pork,10 the franking privilege, or constituent service (Cain et al. 1984; Stein & Bickers 1994; Cover & Brum- berg 1982; Johannes 1984; Serra 1994; Serra & Cover 1992; Romero 1996). Furthermore, while many have written on district marginality and incumbency advantage (Mayhew 1974a; Jacobson 2009; Fiorina 1977), as King and Gelman (1991) point out, only a third of the variance can be explained by incumbency advantage alone. Therefore, it is not surprising some scholars have looked to direct benefits provided via earmarks to account for this discrepancy. The difficulty with this proposed causal connection, however, lies in the information costs imposed on voters.

The requisite cost of processing information regarding earmarks is staggering. In order to argue earmarks directly alter electoral outcomes, we must assume that voters are supremely attentive and engage in rational cost-beneﬁt analysis that appropriately weighs the value of projects brought home by Members, and that voters engage in retrospective analysis that permits them to compare previous acquisitions to more recent projects. Moreover, government ﬁnancing of state and district projects is complex. This fact is not lost on voters. Evidence suggesting that public opinion is coherent on matters of government attribution is mixed (Arceneaux 2005; Schneider & Jacoby 2003; Thompson & Elling 1999, 298). The concept of representation at the federal level is grounded on the assumption that citizens make attributions regarding policy outcomes. This requires that correct opinions be formed about government, that citizens can evaluate government performance, and that citizen’s evaluations provide guidance for policy alterations, or maintenance of the status quo (Arceneaux 2005, 300). That said, citizens need not directly analyze the political process to achieve this end; heuristics, as well as cues and frames, all work to ease the imposed mental tax. When asking how much information the average citizen possesses concerning earmarks, researchers must consider the distinction between civil knowledge versus generalized heuristics (Delli Carpini & Keeter 1996, 52). Citizens only have the option of learning about earmarks if they are presented with an opportunity and possess motivation (Luskin 1990). At some

10Current theories or pork attribution (Stein & Bickers 1995, 127) suggest that voters assess pork retrospectively, making “comparisons about the state of affairs in their district” as compared to the previous (ﬁscal) year.

18 point, absent even modest understanding, low information rationality (i.e. excessive reliance on heuristics) can lead to erroneous conclusions, especially when the heuristics are overly simpliﬁed (Mondak 1994).

In the present case, Figure 2.2 revealed that that most Americans have opted not to attempt to make an inference on earmark spending. This suggests such a flagrant lack of information that most are willing to admit their ignorance rather than offer speculation. And while it may appear pessimistic, hypothesizing that the American public, on the whole, is largely disinterested in politics is hardly revolutionary. Eliasoph (1998) found that of those we might expect to be the most engaged in discourse, namely those in civic organizations, wanted nothing more than to avoid discussions about politics. Hibbing & Theiss-Morse (2002) found that most Americans are more content leaving the political processes to elites, since facing the realities of political debate and compromise is unpalatable. Even the largely optimistic findings of Jacobs et al. (2009, 4) offer a sobering conclusion: “public deliberation does not reach the high expectations of its proponents,” in fact, there is a notable absence of “rational communication and outcomes that generate agreement and politically efficacious citizens.”

Given this general lack of attentiveness, how then can we explain why previous research has reached the opposite conclusion that earmarks directly affect electoral outcomes,11 given that most citizens lack information about pork spending and attribution? This study advances that self- reported claims of citizen awareness are actually the product of press releases (Gimmer et al. 2012) and media exposure, rather than the pork itself. Specifically, those claiming to have knowledge of their Member’s efforts are actually recalling the amount of media coverage of such behavior, rather than the behavior itself. Indeed, previous research on media exposure confirms that the breadth and depth of media coverage increases policy-specific knowledge (Barabas & Jerit 2009; Friske & Taylor 1991). And while previous work has confirmed that MCs rely on press releases to disseminate their legislative accolades (Gimmer et al. 2012), scholars have yet to confirm the role of the media. Similarly, while previous findings have suggested that the causal link between

11Including the ﬁndings that will be presented in Chapter 4 conﬁrming this effect.

19 earmarks and electoral gain is dependent on the amount of pork a Member secures, these previous results are more accurately described as the ability of an MC to tout his or her procurements. Given the imperceivable nature of earmarks, it is unrealistic to assume that the average voter is persuaded by pet projects. Rather, I contend that voters will reward credit-claiming activities that are perceived to beneﬁt them. In other words, voters reward incumbents when they are made aware of local gains. The theoretical contention that media coverage, rather than reality, shapes opinions on earmarks can be empirically tested via appropriately derived hypotheses. Chapter 6 fully explores this theory by testing the reality of earmark acquisition against the reports of survey respondents. If correct, claims of knowledge about earmarks should be largely explained not by the volume of projects, but by the degree of media discussion about the efforts of a respondent’s Members.

2.2.3 Rewarding Those Who Reward: Earmarks and Campaign Contribu-

tions

Early attempts to predict the effect of campaign contributions on electoral outcomes found that the more challengers spent, the better they did on election day; yet the more incumbents spent, the worse they did (Jacobson 1980). This, of course, was due to the fact that incumbents spend more as the electoral threat grows. After controlling for this reciprocal effect, it was found that challenger spending was strongly related to the outcome, while incumbent spending had little or no effect (Jacobson 2009; Ansolabehere & Gerber 1994; Green & Krasno 1988; Bartels 1991). In short, we can expect that “heavy spending by [an] incumbent is a sign of electoral weakness” (Jacobson 2006, 196).

Much like the literature on campaign contributions and electoral outcomes, this work investigates which candidates secure campaign contributions from special interest groups; also like the campaigns contributions literature, the potential for endogeneity is painfully clear. However, here I wish to investigate the role of earmarks and their indirect effect on voters as a means to

20 bolster campaign contributions. While there has been a notable amount of recent research looking to the origin of this funding (Herrnson 1992; Green & Krasno 1988, 1990; Krasno et al. 1994),12 what has not been deeply explored is the possibility of donations with expectations of reciprocity.

Figure 2.3: Previous and Revised Theories of Earmark Attribution

This work investigates the electoral impact of pork-barrel spending in a way fundamentally different from that of past research. This difference is detailed in Figure 2.3. Previous proposals have suggested a theory of direct causal connection between earmarks and electoral outcomes (Fig- ure 2.3, top). Fundamental to this theory is the assumption that earmarks have a substantive impact upon the behavior of legislators, or in the words of Shepsle and Weingast (1981, 110), “pork, in various forms, will always serve as a part of the legislators’ response to his voters’ retrospective question, What have you done for me lately?” A clear underlying assumption is that Members seek pork for the purpose of credit claiming, since earmarks would be direct evidence of a Member’s individual success. Baron (1990) too suggests a connection in his study between pork spent by House Members and electoral consequences of funding for Amtrak. He considers “constituent preferences” as a possible explanation for distributive resources (Baron 1990, 886), proposing that the recipients of beneﬁts are able to attribute federal dollars to a particular individual. In yet another example, Jacobson (2009, 235) argued the following:

12Research on this topic must, by its nature, be recent because rules mandating disclosure and record keeping by the FEC did not begin until the Federal Election Campaign Act (FECA) was amended in 1974.

21 Electoral logic inspires members to promote narrowly targeted programs, projects, and tax breaks for constituents and supporting groups without worrying about their impact on spending and revenues. Recipients notice and appreciate such specific and identifiable benefits and show their gratitude to the legislator responsible at election time.

Clearly there is a shared consensus that earmarks hold the potential for altering elections, and are therefore a viable tool for incumbents.

This work departs from the aforementioned examples by proposing an alternative, although not mutually exclusive, causal structure (Figure 2.3, bottom). Rather than contending that earmarks provide a direct benefit to Members via increased voter support, this work investigates the possibility that incumbent gains from earmarks are also due to secured outlays that benefit large campaign donors. Given the reciprocal nature of this theory, two testable hypotheses result: first, that changes in the amount of pork a Member is able to secure will result in corresponding changes in campaign contributions. This hypothesis has seen some exploration in recent literature from Rocca & Gor- don (2013) who found that defense earmarks did indeed result in increased contributions. Second, the reverse relationship also exists: that changes in the amount of contributions to a member results in that Member securing more earmarks. The implications of the latter of the two relationships is more fully explored in Chapter 7.

22 CHAPTER 3

MEASURING EARMARKS

Those attempting to study earmarks are presented with three sources of data, each with varying strengths and weaknesses: Citizens Against Government Waste’s Pig Book, Taxpayers for Com- mon Sense (both of which have been previously discussed), and the Federal Assistance Awards Data System (FAADS). Before attempting to determine the impact of earmarks on any facet of politics we must ﬁrst determine the accuracy of the measurements.

Beginning ﬁrst with CAGW’s data collection efforts: CAGW has been collecting information on pork spending since 1991, and has archived its records in a searchable format on its website.1 As previously mentioned, CAGW relies on their own seven-point scheme for determining which projects are considered pork; while this process does have its limitations, it is nonetheless consistent. While CAGW has data on pork spending in the aggregate beginning in 1991, it is not until 2008 that the site allows users to attribute pork projects to speciﬁc Members. Conse- quently, CAGW data only allows researchers to observe national year-to-year trends, and provides a resource for comparison amongst states.

Taxpayers for Common Sense relies on changes to institutional rules that required Members to disclose their earmarks. In 2007, both the House and Senate enacted changes that required elected officials from both chambers to disclose all earmarks inserted into legislation. Initially this self-inflicted regulation was met with fevered reluctance; indeed, many Members still make the task of finding their earmark disclosures quite difficult.2 Easing the task are the efforts of citizen

1See general website at < www.cagw.org >. 2There is no ofﬁcial repository for earmarks, rather it is left to Members to disclose their awards. This has led some

23 groups, such as Taxpayers’ for Common Sense, which were quick to exploit this newly-disclosed data, making such information readily available on-line.3

The final source of data is the aforementioned Federal Assistance Awards Data System, FAADS. FAADS is overseen by the U.S. Census Bureau (Department of Commerce), in conjunction with the Office of Management and Budget (OMB), which provides policy oversight. FAADS was begun in 1981 via Congressional statue; it functions as a repository for all federal financial assistance, which is updated quarterly (Management and Budget, Office of 2009). Federal assistance comes in a variety of forms, from direct loans to action-by-action grants, and is the product of thirty-three federal departments and agencies, all of which report their spending to FAADS. The data are sorted according to location of initial receipt (this money could then be reallocated) on an action-by-action basis, or as an aggregate county total. This means that in many instances the data may be attributed down to the state, district, or city location, depending on how the funding allocation is targeted.

3.1 Data Accuracy and Sources

Prior to 2007, there were two alternatives available to scholars who wished to attribute earmarks to specific Members of Congress: the first was tediously combing through every bill in search of pet projects, the alternative was turning to FAADS data to attempt to make inferences regarding who was responsible for the projects. Selecting the first route only marked the beginning of the difficulty, since attempting to tie an earmark to a specific Member of Congress necessitated tracking the life of a bill. Once an earmark was discovered, researchers were required to pin-down the commit-

Members to delay reporting, to bury reports on personal websites, or to make the information available only to those willing to contact the Member’s office. 3Information on the number, value, responsible Member, district or state benefiting, and details of the specific earmark can be found at < http://www.taxpayer.net/search_by _category.php?action = search_by_category &category=Earmarks >

24 tee or sub-committee responsible for the amended bill. Once the (sub-)committee was identified it would be necessary to once again comb through revisions of the legislation (or consult committee minutes) to find the original sponsor. The alternative route streamlined the data collection process since the Census Bureau keeps detailed yearly records on all federal funds that are dispersed to states and localities. What the FAADS data are lacking, however, is any meaningful way to attribute a given state or local project to a Congressperson. Without this information, scholars have been unable to differentiate if a project was the handiwork of a Representative, or a Senator, the result of a nationally applied program, or an executive request to a specific agency. Furthermore, FAADS data do not attempt to catalog earmark projects, rather they are a record of all federal direct spending.4 This poses a problem for scholars investigating earmarks, as it requires either relying on the data in their current form, which would introduce a great degree of error, or attempting to filter out earmark projects from normal spending. While Bickers and Stein (Bickers & Stein 1996; Stein & Bickers 1994, 1995) attempt to avoid this problem by focusing on the number of new awards rather than the dollar amount, this raises concerns regarding accuracy.5 In addition, the FAADS data itself must be considered suspect since differentiating pork-barrel projects is near impossible.

For example, Bickers & Stein (1991) collect FAADS data from 1983-1990 looking only at non-automatic spending.6 They conclude that in the first three quarters of fiscal year 1990,7 the federal government spent $573.9 billion on direct outlays. This is a staggering amount considering that for fiscal year 1991 CAGW reported a mere $3.1 billion went to earmarks. We see similar glaring inconsistencies when we compare Stein and Bicker’s (1994, 396) report of over 19,000 new projects during the 100th Congress (1987-1989), yet there were only a modest 546 earmark projects in 1991, according to CAGW. Figure 3.1 reveals this discrepancy applied to more recent

4For example, FAADS data also include detailed receipts of spending by the Social Security Administration, De- partment of Defense, and the Department of Homeland Security. 5More will be said on this later. In short, by relying on the number of projects without regard for the value associated, the authors impose the assumption that all projects are essentially equal, regardless of value. 6This includes federal money to state governments, counties, cities and towns, special districts, schools and universities, Indian tribes, non-profits, private universities, small businesses, individuals and for profit business. 7The final quarter was not available at the time their book was published.

25 data. While the FAADS data remains correlated with both TCS data (.67, p <.001) and CAGW data (.62, p <.0001), there remains doubts as to accuracy.

Figure 3.1: Federal Earmark Allocations, FY2008: Comparing TCS, CAGW and FAADS

Comparison of the TCS and CAGW measures, on the other hand, reveal a high degree of correlation. As Figure 3.2 reveals, there is a consistent relationship between the two measures over

fiscal years 2008-2010 (FY2008 : r = .97; FY2009 : r = .91; FY2010 : r = .98, all p <.001). This is good news for researching earmarks. Since TCS is the most accurate measure of earmarks that exists, its association with CAGW data (assuming this accuracy holds for previous years) means scholars may be able to rely on CAGW data, which is temporally more expansive than TCS. However, given there is no TCS data before fiscal year 2008, the certainty of this relationship is far from certain. Additionally, no group can provide Member-attributable or district-level data prior to fiscal year 2008.

26 Figure 3.2: Earmarks FY2008-2010: Comparing TCS to CAGW Measurements

27 CHAPTER 4

WHO BENEFITS FROM EARMARKS, AND WHY?

This chapter seeks to reﬁne, expand, and correct past efforts that have explored who gets earmark projects and why (Lazarus 2009, 2010; Lazarus & Steigerwalt 2009; Bickers & Stein 1996; Stein & Bickers 1994; Ferejohn 1974; Frisch 1998; Engstrom & Vanberg 2010). Speciﬁcally, this chapter will focus on earmark procurement by asking three questions: which Representatives excel at securing earmarks, which Senators excel at securing earmarks, and what role does electoral vulnerability play in these efforts? This is achieved by exploring the theoretical expectations derived from the theory introduced in Chapter 2—namely, that Members wish to minimize electoral risk while maximizing constituent satisfaction. However, when it comes to earmarks, such desires are largely determined by institutional factors. By testing theories against recent data that are more expansive and of higher quality, I am able to revise existing hypotheses in lieu of the theory presented in Chapter 2.

4.1 Getting Pork: Which Representatives Receive Earmarks?

The ﬁrst section addresses the ability of House members to secure projects by expanding upon the recent research conducted by Lazarus (2009; 2010), as well as nearly identical work by Engstrom & Vanberg (2010). The studies of both Lazarus and Engstrom and Vanberg rely on ﬁscal year 2008

28 data when testing earmark acquisition models;1 this section will expand upon that research. This will be accomplished both by considering more recent data from the same source (Taxpayers for Common Sense) across ﬁscal years 2008, 2009, and 2010. Of primary concern are institutional factors that may advance or inhibit an MC’s ability to secure pork. As Chapter 2 described, the ability of MCs to secure earmarks are largely a result of institutional position, which is achieved with the aid of the party. Moreover, while any MC can seek earmarks, the projects must nonetheless be inserted into appropriations, authorization, or revenue bills (Brass et al. 2008). Consequently, this section will focus on pivotal institutional factors: being a committee chair, MC tenure, party, party leadership position, membership on the appropriations committee, or being a cardinal (appropriations subcommittee chair or ranking minority member).2 As predicted, institutional position in the House is fundamental to understanding earmark acquisition.

As previously discussed, earmarks are primarily measured in two ways: the value of the projects attributed to each Member, and the number of outlays secured by each MC. Figure 4.1 displays both the value (logged) and number of earmarks for the U.S. House over fiscal years 2008, 2009, and 2010. The figure reveals that the value of the projects brought home to a Member’s district is stable over time, with slightly more members opting to forgo earmarks ($0 logged dollars) in recent years. The number of earmarks shows a similar downward trend, with increasing numbers of Representatives going without any projects. Interestingly, this trend appears to have started in fiscal year 2009, one year after public disclosure of earmarks began. This suggests that Members may have changed their behavior as a result of increased exposure.

1Both sets of authors raise methodological concerns in their model speciﬁcation; for example, Lazarus fails to log the dollar totals attributed to a Representative, which raises issues of normality given that the uncorrected earmark data (ﬁscals year 2008) has a skewness of 6.3 and a kurtosis of 56.404. This error was corrected in his 2010 adaptation. Rather than address the censorship caused by MCs with 0 earmarks, Engstrom & Vanberg (2010) simply omit these MCs from their analysis. This leads to highly skewed results. 2Ideology, as measured by the Representative’s DW-NOMINATE score, was also considered, but it was highly correlated with party (.942)

29 Figure 4.1: Dollar (logged) and Number of Representative Earmarks, FY2008-2010

While the above ﬁgure reveals a great degree of variance in project acquisition, it does not predict which MCs are better able to secure projects. To shed light on this basic question, an analysis of earmarks (both value and number) was conducted. In a manner similar to that of Lazarus (2009), I constructed a predictive model with the individual Representative as the unit of analysis. However, unlike Lazarus, the following model looks at earmark procurement over several years with greater attention paid to institutional position.3 Table 4.1 displays the results of two estimations: the ﬁrst captures the ability of Representatives to secure earmarks as measured

3Namely, I rely on time series cross sectional models to gain insight into the temporal stability of previous ﬁndings.

30 by their dollar value (logged), while the second (right) measures the number of projects.

Table 4.1: Representatives Securing Earmarks, Fiscal Years 2008-2010 Tobit Negative Binomial Negative Binomial Variable Coeff. (std. err.) Coeff. (std. err.) Incidence Rates Democrat 2.784*** 0.364*** 1.439 (0.475) (0.068) Appropriations Comm. 2.503*** 1.142*** 3.134 (0.444) (0.072) Majority Cardinal 0.815* 0.137 1.147 (0.370) (0.084) Minority Cardinal 0.223 0.130 1.138 (0.788) (0.113) Party Leader 0.231 0.627*** 1.872 (1.298) (0.186) Committee Chair -0.509 0.112 1.118 (0.738) (0.119) Ranking Min. Member 2.205** 0.407*** 1.503 (0.786) (0.113) Tenure 0.053* 0.014** 1.014 (0.023) (0.005) Marginal 0.792+ 0.361*** 1.435 (0.448) (0.071) FY2009 Dummy -1.315*** -0.042 0.959 (0.264) (0.031) FY2010 Dummy -2.005*** -0.281*** 0.755 (0.341) (0.043) Constant 11.582*** 1.448*** 4.253 (0.518) (0.075) σ 4.929*** (0.257) α -0.785*** (0.093) pseudo-R2 0.023 N 1,307 1,307 Tobit regression’s censor point at 0 with the dependent variable as logged earmark dollars. Negative Binomial dependent variable is count of individual earmarks. Both two-tailed standard errors, clustered by Congressional district. ***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10

Table 4.1 (left column) displays the results of a time series cross-sectional (TSCS) model correcting for censorship at $0.4 The results are similar to those of Lazarus (2009, 1056), but

4Censorship exists for members who make a conscious effort to avoid sponsoring or cosponsoring legislation

31 reveal important differences with the inclusion of multiple years. First, institutional position appears to play a strong role in the dollar amount of earmarks secured; appropriations committee members, majority cardinals, and ranking minority members all secured more valuable earmarks. Additionally, those with greater experience (tenure), and greater electoral uncertainty (marginal Representatives) also secured more earmarks. Finally, across the three ﬁscal years considered, Democrats secured more earmarks than their Republican counterparts. This is not especially surprising as Democrats held the majority of House seats over all three ﬁscal years; moreover, it is consistent with conventional wisdom of Democrats favoring greater governmental spending.

Table 4.1 (right column) displays the results of a TSCS count model, with the number of earmarks over fiscal years 2008, 2009, and 2010 as the dependent variable. As the table reveals, many of the same variables that were significant in the continuous model are also significant in the negative binomial model, the strongest being membership in the appropriation committee, which garners a member 3 additional projects, ceteris paribus. As in the alternative specification, being a Democrat, a party leader, a ranking minority member, years of service, and election from a marginal district all lead to an increased number of projects.

An alternative specification to those considered above is the possibility that the likelihood of an MC choosing to forgo earmarks is not the same data generating process that determines when a Member has $0 in earmarks. In other words, perhaps there exists unobserved factors that predetermine if a MC is able to securing earmarks in the first place.5 If such situation existed, a modified version of the Tobit model, a two-part selection model (also known as a hurdle model),

that features pork projects. This is due to the fact that Members are restricted to a minimum of $0 in sponsoring and cosponsoring, regardless of whether a Representative truly desires a lower (negative) dollar value of earmarks. Sigelman & Zeng (1999) clarify the relationship being proposed here; they use the example of PAC contributions, since it could be “imagined to include negative as well as positive values, with $0 representing a censored negative observation” when a “PAC might wish that it could make a “negative contribution” to a disliked candidate by taking dollars away” (170). For a model considering the allocation of pork, we can imagine an instance where a Member may desire to have sponsored less than zero dollars, as this may make the member appear especially opposed to the utilization of earmarks, but such a value is not possible. The assumption is that some Members are not satisﬁed with $0 in earmarks attributed to their name, and wish to reduce or eliminate earmarks awarded to other Members. In fact, this may occur at the committee and sub-committee levels when a Member works to strike a piece of pork from proposed legislation; however, we cannot observe this action. 5The Tobit model assumes that the same processes that results in some MCs having $0 and others having more than $0 are the same (Cameron & Trivedi 2010, 553).

32 might be more appropriate. A test of the correlation of the error terms between MCs securing earmarks compared to those who did not (e.g. MCs securing $0 in earmarks), and second step model predicting the amount of earmarks secured, suggested the two processes are not independent. In other words, a tobit will likely suffice. Nonetheless, the alternative specification using a two-step Heckman model is featured in the appendix; the variables found to be statistically significant in the tobit model remained so in the Heckman specification.

Ultimately the results of the analysis reveal that many of Lazarus’ (2009; 2010) findings change when several years of earmarks are considered. In contrast to Lazarus’ findings, being a Democrat was a strong predictor of earmark acquisition; moreover, being a member of the Ap- propriations Committee or a minority cardinal were shown to have a considerably stronger effect in the analysis above. Finally, being electorally vulnerable, which was not featured in previous models, also evidenced to be a strong predictor of an MC securing more earmarks. In conclusion, institutional position matters a great deal in the House when it comes to securing pork projects. The next section asks if these findings are consistent in the context of the Senate.

4.2 Getting Pork: Which Senators Receive Earmarks and

Why?

Section two continues the analysis into the Senate by building on the work of Lazarus (Lazarus & Steigerwalt 2009) both expanding the time frame and improving the analysis. For example, Lazarus & Steigerwalt (2009) depart from their previous analysis by opting to look solely at the number of projects both Senators and House members secured in ﬁscal year 2008, rather than the dollar amount Senators were able to bring home to their state. While such codiﬁcation is not unprecedented (both Stein & Bickers (1994) and Bickers & Stein (1996) rely upon the change in the number of new awards), it raises questions regarding the purpose and role of earmarks.

33 Namely, by looking only at a count, the authors discard any variance in the potential impact of a project. For example, the $5 million spent on California’s Golden Gate National Recreation Area would be treated the same as the $38,000 spent on the Hatboro (Pennsylvania) Union Library Restoration project since both are a single earmark project. Moreover, the distribution of the number of earmarks is considerably less uniform than the dollar values, even before the dollar values are logged. Figure 4.2 features both the logged dollar value and the number of Senator earmarks across ﬁscal years 2008, 2009, and 2010. The ﬁgure suggests that the distribution of earmarks remains relatively stable over time, with a wide degree of variance between Senators.

Figure 4.2: Dollar (logged) and Number of Senator Earmarks, FY2008-2010

Given the variation in value from project to project, it seems plausible that the value of

34 the earmarks will play a signiﬁcant role, especially in regards to electoral outcomes (which is the concern of Stein & Bickers (1994), Bickers & Stein (1996) and Lazarus & Steigerwalt (2009)). Consequently, this section will consider differing measurement schemes in order to ensure the validity of the ﬁndings of past research.

Table 4.2 features a predictive model of Senator earmark acquisition across three ﬁscal years (2008, 2009, and 2010) and two Congresses (110th and 111th). Following in a vein similar to Lazarus & Steigerwalt (2009), a number of controls are included regarding institutional position: appropriations committee membership, being a chair or ranking minority member of an appropriations subcommittee (a cardinal), a party leader, and being a standing committee chair or ranking minority member. Additionally controls were added for party, tenure, and electoral marginality (considered marginal if previous electoral victory was less than 60%).6 Finally, given the time series cross-sectional nature of the data, year covariates were added, with ﬁscal year 2008 as the base year.7

Table 4.2: Senators Securing Earmarks, Fiscal Years 2008-2010 Tobit Negative Binomial Negative Binomial Variable Coeff. (std. err.) Coeff. (std. err.) Incidence Rates Democrat -3.321+ 0.014 1.014 (1.721) (0.273) Appropriations Comm. 6.949*** 1.596*** 4.931 (1.434) (0.283) Majority Cardinal 0.874 0.112 1.118 (1.688) (0.256) Minority Cardinal -1.759 0.223 1.250 (1.518) (0.321) Party Leader 1.093 0.784* 2.189 (1.466) (0.337) Committee Chair 2.422 0.108 1.114 (1.543) (0.251)

6Again, ideology was going to be included, but it was highly correlated with party (.922). 7As in the previous section, an alternative two-step estimation was considered (see appendix). Not only were the same variables signiﬁcant, but there was stronger evidence that that the errors in a Senate hurdle model were not different from 0, strongly suggesting there was no independent process at work between members securing pork in the ﬁrst place, and those securing greater values of earmarks.

35 Table 4.2 - continued

Tobit Negative Binomial Negative Binomial Variable Coeff. (std. err.) Coeff. (std. err.) Incidence Rates Ranking Min. Member 1.045 0.343 1.409 (1.606) (0.267) Tenure -0.017 0.012+ 1.012 (0.053) (0.007) Marginal 0.208 0.293* 1.341 (0.825) (0.146) FY2009 Dummy -0.828 -0.059 0.943 (0.504) (0.070) FY2010 Dummy -2.309* -0.467*** 0.627 (0.976) (0.105) Constant 12.734*** 1.613*** 5.020 (1.433) (0.230) σ 6.863*** (0.656) α 0.141 (0.190) pseudo-R2 0.0377 N 296 296 Tobit regression’s censor point at 0 with the dependent variable as logged earmark dollars. Negative Binomial dependent variable is count of individual earmarks. Both two-tailed standard errors, clustered by state. ***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10

Table 4.2 (left column) begins with a TSCS consideration of the logged amount of earmarks each Senator has secured, while accounting for the censoring effect of $0.8 The results reveal that being a member of the appropriations committee remains a strong predictor of earmarks, being a Democrat actually lessens the amount of pork a Senator secures, albeit this coefficient is barely significant at conventional levels. Aside from these two variables, none of the related predictors were revealed to be statistically significant.

Utilization of a count model allows for the comparison between the model presented here (right column) and previous results, namely, those of Lazarus & Steigerwalt’s (2009, 358). As the results reveal, there are several differences. First, while the effect of a Senator being on the appropriations committee remains positive, this effect is notably greater across several years (Lazarus & Steigerwalt ﬁnd membership increases the number of projects by 1.38; my model ﬁnds this number

8See footnote 4 for an explanation as to why a tobit model was employed.

36 is closer to 5 projects, ceteris paribus). Additionally, unlike the ﬁndings of Lazarus & Steigerwalt (2009), being a Democrat is not a strong, positive predictor of acquisition. In fact, the continuous model (left column) found being a Democrat is negatively associated with project dollars; this suggests temporal variation not captured by previous models. Unlike previous research, a TSCS count model consideration reveals that party leaders secure 2 more projects than non-leaders, ceteris paribus, where previous research found no effect. Finally, electoral uncertainty, which was absent from Lazarus & Steigerwalt’s (2009) model, also increased the number of projects.

The analysis above reveals two things: first, that the Senate, as one might expect, is notably different from the House when it comes to securing pork, with few variables proving meaningful predictors in the Senate context. Second, scholars will draw quite different conclusions regarding earmarks when looking at the number of projects secured versus their dollar value. As table 4.2 revealed, while a number of institutional and electoral factors contribute to the number of projects a Senator secures, the value of those projects is less easy to predict. If a Senator wishes to rely on earmarks to ensure electoral support, he or she would need to actively disseminate information about the many projects secured rather than rely solely on large, valuable projects that let money do the talking. On the other hand, if the purpose of pork is not solely for the benefit of constituents, but rather to benefit specialized interests (as will be discussed in Chapter 7), the grander projects may indeed be the most desirable.

4.3 Frightened Into Action: Electoral Vulnerability and Effort

to Secure Earmarks

This section will address an issue that is invariably raised when scholars attempt determine if earmarks affect electoral outcomes: how can we investigate this relationship while avoiding inherent

37 endogeneity? Before attempting to tackle questions surrounding whether earmarks may function as a means to improve electoral support, scholars must consider the role vulnerability plays in member efforts to bring home projects. Stein & Bickers (1994) attempt to parse this out by looking at a Member’s pork acquisitions in the previous year (t-1) as a predictor of electoral marginality in the following election (t). They found no relationship between success securing earmarks and electoral vulnerability; however, the authors make no attempt to address the endogeneity that exists between electoral vulnerability and the ability of a Member to secure earmarks in the first place.9 Attempting to determine the role of earmarks on electoral margins is made especially difficult by the fact that previous electoral results may have influenced Member behavior regarding earmarks. In other words, it is easy to imagine a Member redoubling his or her efforts to reduce electoral uncertainty, including securing more pork, in an effort to improve future electoral prospects. Since MCs can be expected to secure pork in proportion to their perceived risk, pork-barrel projects are not randomly assigned, and therefore not exogenous. This section proposes an expanded and revised analysis of this question using more robust and temporally expansive data, as well as an alternative method that avoids the causal pitfalls.

There are two means of gaining insight into the problem of causation. First, in addition to relying on recent, accurate earmark data, this section investigates a subsample of the population: freshmen Representatives. Having never been able to secure pork in the past, freshmen MCs, while able to secure earmarks following their ﬁrst election, have not yet learned the effect of earmarks on their reelection prospects. Freshmen MCs give us a unique look at the short term effects of earmarks (the amount procured over one election cycle), without the potential for long- term, downstream effects of earmarks affecting voter behavior. Moreover, given their institutional position, freshmen are less likely to be able to secure earmarks.10 While this doesn’t eliminate the endogeneity problem, it does offer insight into the short-term effect of pork while controlling

9Despite this, Bickers & Stein (1996) seem to ignore their previous findings and test a model of quality challenger emergence that utilizes the change in the number of new awards as the key independent variable. Not surprisingly, they find it significant. 10The assumption that freshmen are less adept at securing earmarks is empirically tested in the Appendix. Not surprisingly, freshmen Representatives are notably less able to acquire pork dollars.

38 for any past earmark efforts affecting the electoral outcome. This provides insight into whether pursuing an instrumental variables approach is worthwhile. Second, and most importantly, it is possible to rely on a statistical solution to endogeneity via the instrumental variable approach, assuming there is a viable instrument available. Unlike Stein and Bickers (1994), I hypothesize that earmark acquisition has a greater effect on electoral success than previously discovered.

Table 4.3: Regression of Past Earmarks on Future Vote, Freshmen Only Variable Coeff (std. err)

Ln(Porkt−1) 0.005+ (0.003) Votet 1.059*** (0.107) Democrat -0.085*** (0.021) 110th Cong. Dummy 0.004 (0.041) Constant 0.028 (0.0654) R2 0.457 N 118 OLS Regression; freshmen only. Two-tailed standard errors, clustered by state and district. ***p < 0.001, **p < 0.01, *p < 0.05, + p < 0.10

As mentioned above, freshmen MCs are unable to secure earmarks prior to their ﬁrst election. In other words, freshmen may only begin attempting to alter their electoral uncertainty during their ﬁrst reelection bid. The model below treats previous earmark dollars (Ln(Pork)t−1) as an exogenous predictor of vote share in the next election. Table 4.3 features the results of a time series cross-sectional model over the 110th and 111th Congresses (for a total of 118 freshmen MCs).11 As the table reveals, past earmarks are a positive predictor of future electoral support. This connection is weak, however. A one unit increase in logged pork dollars results in a .005% increase in electoral support, ceteris paribus.

11Only a limited number of controls are needed since the model considers a sub-sample of the House (freshmen). Moreover, controls for institutional position are largely meaningless since no freshmen in the sample were majority/minority chairs or cardinals.

39 Figure 4.3: Marginal Effect of Past Earmarks on Future Vote, Freshmen Only

Figure 4.3 depicts this relationship visually. Holding all other continuous variables at their mean, and binary variables at their mode, the figure features the marginal effect of moving from minimum (0) to the maximum (16.8) value of earmark dollars (logged). As the figure depicts, this results in a change of approximately 54% to 62%. Contrary to previous research, the results suggest that earmark allocations can have a rather large effect on district marginality. That said, given the small sample size and limited time frame, it is not surprising the earmarks variable was on the cusp of a conventional level of statistical significance. Moreover, while freshmen may be less likely than their experienced counterparts to be able to turn to earmarks as a means to reduce electoral uncertainty (i.e. being a freshmen means the assignment of earmarks is closer to being random than non-freshmen), this does not remove potential for endogeneity. Consequently, an additional analysis will be considered that relies on a methodological solution.

While the results above are encouraging, they rely on a subsample of the population that still may suffer from the endogeneity present in previous research. An alternative solution to the

40 endogeneity problem is methodological: an instrumental variable approach. The difficult task is finding an instrument that predicts earmark allocations, but is unrelated to electoral outcomes. While trying to predict if PAC contributions from the defense industry lead to greater defense earmarks, Rocca & Gordon (2013) relied on the square root of all earmarks as an instrument to solve the endogeneity problem. However, while that instrument worked mathematically, it lacked “validity [based] on a persuasive argument” (Cameron & Trivedi 2010, 181). I propose an alternative instrument: state population. Population is typically not considered a meaningful predictor of earmarks in the House context, but previous work has considered land area (Lazarus 2010) and population (Lazarus & Steigerwalt 2009) in the context of the Senate. Moreover, the use of state populations as an instrument has found support in related contexts, such as campaign spending (Gerber 1998). The reason for its absence as a control in House procurements is obvious: populations do not vary much from district to district. That said, as an instrument state population need only be a modest predictor of earmark acquisition that is unrelated the endogenous dependent variable (in this case, future vote). In the electoral context, state population is theoretically unrelated to electoral outcomes, but it may affect earmark allocations since a larger state population means a greater cadre legislators from the same state. Assuming Representatives from the same state work towards common goals, state population should function as a means to capture larger, more powerful legislative contingents, which also would mean an increased likelihood of passing of legislation. To test such a contention, models similar to those featured in table 4.1 were created with the addition of a state population variable. In both the logged dollar value and count specifications, the state population was a statistically significant predictor of earmark acquisition (see the appendix for full model specifications). Additionally, a number of robustness checks were conducted post-estimation to confirm the presence of endogeneity, and state population’s role as an adequate instrument (see appendix); these results confirm both. Armed with a viable instrument, it is now possible to reconsider the findings of previous analyzes.

41 Table 4.4: 2SLS Model Predicting Pork’s Effect on Future Vote Share 2SLS, First-Stage 2SLS, Second-Stage

Dependent Variable Ln(Porkt−1) Votet+1 Variable Coeff. (std. err.) Coeff. (std. err.)

Votet 5.048*** 0.236 (0.000) (0.201) Ln(Porkt−1) 0.068* (0.031) Tenure 0.162*** -0.013* (0.000) (0.006) Democrat 1.294*** -0.079 (0.001) (0.051) 110th Cong. Dummy 3.072*** -0.152 (0.000 ) (0.097) State Pop. (in millions) 0.039** (0.005) Constant -0.168 (0.175) R2 0.209 N 871 871 Two-Stage Least Squares regression. Two-tailed standard errors, clustered by state and district. ***p < 0.001, **p < 0.01, *p < 0.05

Relying on state population as an instrument, table 4.4 estimates the effect of previously securing earmarks on electoral vote share. The left-most column features the reduced form model (first stage) of the Two-Stage Least Squares (2SLS) regression, and the right column features the results of the fully specified model. As table 4.4 reveals, not only is the previous pork variable a significant predictor, but the 2SLS model suggests it has a stronger effect than eluded to in the freshmen only specification (table 4.3). The substantive interpretation of the coefficient is made easier by looking at the marginal effect of earmarks across the range of values.

42 Figure 4.4: Marginal Effect of Earmarks on Future Vote, 110th and 111th Congresses (2SLS)

Figure 4.4 shows the effect of earmarks (logged) on vote share over the range of earmarked values for the 110th Congress (top) and 111th Congress (bottom) while holding all other variables at their mean (continuous) or mode (discrete) values. Figure 4.4 (top) reveals that moving from the

43 smallest value of earmarks (about 11.5 logged dollars) to the maximum (about 18.9 logged dollars) results in a change in expected vote share of 40% to 89% for the 110th Congress.12 That said, as the confidence intervals around the estimate reveal, for most Members this change is closer to the mean value (approximately a 13% increase when moving from 14 to 16 logged dollars). Figure 4.4 (bottom) visualizes the same results over the 111th Congress. Moving from the minimum positive value (11.4) to the maximum (18.6) results in a change in support from 54% to 100%. Obviously the confidence intervals around these estimates are quite large at high values, nonetheless, the message is clear: earmarks are significantly boost electoral security.

What remains unclear from this analysis is whether earmarks are the direct causal mechanism responsible for the boost in electoral support. In other words, while voters do appear to reward MCs for their past efforts, it is unclear from this straightforward analysis how the public is being made aware of earmarks. Are Members directly informing the public of their Congres- sional success, or might MCs need to rely on press releases (see Gimmer et al. 2012) or media sources to dissemination this information? These concerns will be explored in depth in Chapter 6. Moreover, while the results of this section have revealed pork’s effect on electoral outcomes, the causal mechanism behind constituent appreciation for such projects is unclear. A positive response from recipients is a prerequisite if MCs ever hope to exploit the earmarking process. The notion of constituent appreciation for particularized beneﬁts is the topic of discussion in the next chapter.

12The smallest value in any ﬁscal year is $0, but for visual ease the ﬁgures start at the smallest substantive value.

44 CHAPTER 5

DESIRABLE PORK: DO VOTERS REWARD FOR EARMARK ACQUISITION?

5.1 Introduction

Scholars and Members of Congress alike have long assumed that money brought back to a district or state in the form of earmarked projects would garner appreciation from constituents. Several polls over a number of years have revealed that, in the aggregate, earmarks do not enjoy popular support. This raises a question: does pork behave similarly to notions of individual Member evaluation; that is, just as constituents love their Member but despise the institution of Congress, so too do they cherish their own pork projects, while despising allocations in the aggregate? While such a connection seems intuitive, it has never been empirically veriﬁed. This chapter attempts to shed light on this question by turning to experimental data to ascertain the impact of information regarding earmarks on Member support.

The ability of incumbents to alter reelection consequences has been given much academic speculation; as Mayhew (1974b, 28) stated: “left unanswered is the question of whether congressmen in search of reelection are in a position to do anything about it.” Facing electoral uncertainty, Members turn to credit claiming: generating the belief in constituents “that one is personally responsible for causing government [via] pieces of governmental accomplishment” (Mayhew 1974, 53). This can take several forms, such as casework, utilization of the franking privilege, frequent

45 trips home, and producing earmarked projects for a state or district. While all of these efforts leave open the question of effectiveness given low political awareness, incumbents nonetheless actively utilize such beneﬁts.

The issue of budget allocations for pet projects has often been the subject of public scrutiny. For example, during a debate in the latest race for the Republican Presidential primary Mitt Rom- ney remarked to his then-rival Rick Santorum: “[w]hen I was ﬁghting to save the Olympics, you were ﬁghting to save the bridge to nowhere.”1 Santorum quickly rebuffed such accusations. How- ever, in the same debate candidate Newt Gingrich commented that there are both good and bad earmarks. Indeed, on other occasions Santorum has made statements akin to those of Gingrich, such as his remark during an interview that “I’ve defended my earmarks in the sense that I’m proud of the money that I did set aside for priorities in my state instead of having bureaucrats do that.”2 Clearly Members and candidates tailor their narrative regarding earmarks to their audience, demonizing the projects in some instances, while touting them to recipients. What remains unclear are the beliefs of those receiving such messages.

This chapter is concerned with the role earmarks play in relation to the ability of office- holders to influence opinion via direct spending. Voters must be able to successfully connect the actions of elected officials to specific attributable benefits, which requires “both the knowledge and the beliefs of the voters” (Popkin 1991, 96). Survey evidence tells us that citizens are remarkably uninformed about particularized spending.3 Nonetheless, many scholars have maintained, implicitly or explicitly, that earmarks breed appreciative constituents and safer elections (Stein & Bickers 1994; 1995; Bickers & Stein 1996; Alvarez & Saving 1997; Herron & Shotts 2006; Jacobson 2009). That said, this conclusion is not universally shared. For example, Sellers (1997) argues that electoral benefits from pork are conditional on Member-constituent ideological congruence.

1“Candidates Hammer Out Differences Over Earmarks at Debate.” February 22, 2012. The New York Times Blogs (The Caucus). 2“Rick Santorum Defends Spending on Pet Projects.” Dec. 29, 2011. CBSNews: Political HotSheet. < http://www.cbsnews.com/8301-503544_162-57349917-503544/rick-santorum-defends-spending-on-pet-projects/ >. 3For example, the 2008 Cooperative Congressional Election Study (CCES) found that when respondents were asked if they could “recall any speciﬁc projects that your members of Congress brought back to your area,” only 17% answered in the afﬁrmative.

46 While Lee (2003) and Feldman and Jondrow (1984) contend the opposite: earmarked dollars do not lead to actualized electoral gains. While many researchers have relied on presupposed effects, none have attempted to ascertain the connection between pork and opinion. This work attempts to ﬁll that gap.

This chapter proceeds as follows. In the first section, I provide a brief overview of the theoretical underpinnings surrounding pork and representation. This involves both reviewing arguments asserting a connection either exists or does not, and proposing how such connections can be tested. Next, I discuss the methodology of the experiments employed to test how information regarding earmarks affects opinion, as well as detail hypotheses concerning how different frames of pork (positive and negative) should affect opinion. Finally, the chapter presents the findings of independent experiments. Ultimately this work finds that pork does not always guarantee increased support for the incumbent. When particularized benefits are framed in a positive light, and especially when they are personally relevant to the recipient, Members can indeed expect an increase in support. However, even modest changes in the framing of the earmarked projects can greatly reduce or eliminate these gains.

5.2 Awareness, Credit Claiming, and Desirability

Despite numerous claims that earmarks yield constituent support, such a connection presupposes that voters are aware of local pork beneﬁts. Numerous scholars have asserted a pork-favorability connection (Shepsle & Weingast 1981; Baron 1990; Ferejohn 1974; Jacobson 2009), while others have explored the possibility of pork bolstering awareness of the projects (Stein & Bickers 1994, 1995). Most recently, Gimmer et al. (2012) explored the role of press releases related to credit claiming for grant acquisition on Member support. However, despite a recent rise in the number of articles on earmarks, no work exists establishing how tone affects constituent appreciation for

47 earmark efforts. This chapter investigates the consequences of bringing federal dollars home in the form of earmarks for those who are made aware of such efforts.

Members of Congress believe that pork matters for reelection, which prompts elected officials to credit claim. For example, Rep. Mike Thompson (CA-1st) was quick to take credit for HUD dollars allocated to his district, saying “I am proud to support these funds so that folks can continue to get the help they need to get back on their feet” in a press release featured on his website.4 Such claims are not rare events; Members are incentivized to remind their constituents of the good deeds done on their behalf. Facing electoral uncertainty Members opt for a scattershot approach to campaigning, taking advantage of numerous resources at their disposal (Erikson & Wright 2009; Jacobson 2009). Therefore, regardless of the actual electoral gains from pork, so long as members assume earmarks garner votes, such behavior is likely to continue. Addition- ally, as Lupia and McCubbins (1998) point out, citizens generally lack factual knowledge about specific policies; however, this does imply they are incapable of learning. Especially proficient Congresspersons possess the ability to alert their constituents given significant effort.

5.2.1 Credit Claiming

In the aggregate, Americans stand in strong opposition to earmarks (Figure 5.1).5 This raises a question: given the lack of popular support, why do members procure such projects? As Fenno (1978) pointed out, Members have an incentive to credit claim when it comes to tangible beneﬁts. While earmarks, much like the institution of Congress itself, lack popular support, “Members of Congress run for Congress by running against Congress” (1978, 168). In other words, it is not surprising to ﬁnd that voters respond differently to aggregate Congressional behavior versus that of their Member. This notion of voter support, that voters champion their member while despising the institution, has been explored by numerous scholars (Stein and Bickers 1994; Fenno

4< http://mikethompson.house.gov/News/DocumentSingle.aspx?DocumentID=274333 > 5Question wording: “Members of Congress sometimes add provisions to legislation that include government spending projects for their own home states and districts, sometimes known as “earmarks.” Do you think this practice is generally acceptable or not acceptable?”

48 Figure 5.1: 2010 CNN Poll: Acceptability of Earmarks

1978; Mayhew1974b; Hibbing & Theiss-Morse 2002). This also coincides with Harbridge and Malhotra’s (2011) ﬁnding that while support for Congressional bipartisanship was high, strong partisans vehemently opposed such behavior from their Member. Moreover, such differentiation has been found to apply to how voters assess their Member (Born 1990; Parker and Davidson 1979). Ultimately, regardless of the aggregate assessment of earmarks, what matters for Members is evaluation of these outlays by constituents.

Despite consistently negative aggregate views, scholars nonetheless have often discussed pork as a universally positive good. For example, Stein and Bickers (1994) identify three assumptions that must be true if earmarks are to have an impact on elections. Amongst these is the notion that “voters reward their representatives on election day for securing district beneﬁts” (1994, 383). However, this presupposes that a majority of voters in a given constituency view beneﬁts in a positive light. After all, earmarks are not wholly framed as a desirable good. For example, John McCain ran as a vehement opponent to pork-barrel spending in the 2008 presidential race, stating “[t]hat kind of thing is going to stop when I’m president of the United States of America” (Kane 2008). As was discussed above, appeals from McCain and his ilk are not without public support.

49 This raises the question: how should we expect recipients to respond to information about localized beneﬁts that are framed negatively? This chapter attempts to shed light on this by considering varying messages regarding these controversial allocations.

5.2.2 Issue Framing

If it were possible to be certain constituents had knowledge of earmarks, would this change their opinion about a given incumbent, and if so, in what direction? Any alteration in opinion would depend largely on the manner in which the subject was informed. For present purposes, given the lack of insight into the expected impact of knowledge of local earmarks, this study will begin by simply enlightening respondents. For that we need look no further than individual behavior. As previously mentioned, assuming that voters are self-interested, we should expect to see that voters reward Members for direct beneﬁts. In other words, information provided to voters about pork brought home to a respondents’ state or district should increase the support for a given member of Congress.

What the aforementioned research has not considered is variation in how pork is presented. The information environment surrounding earmarks is not universally positive. Nonetheless, many scholars have proceeded under the assumption that the message to constituents is one of service. While scholars recognize the potentially muted response in reciprocity for pork because of a lack of awareness, they often fail to consider the varying nature of the message itself.

Finally, when discussing media effects, researchers must carefully consider both the impact of issue framing and priming effects. Claiming that pork projects are universally beneﬁcial presupposes that coverage of those outlays is framed in a positive light. Voluminous literature on the issue (Iyengar 1991; Nelson et al. 1997; Chong & Druckman 2007) has explored the impact of framing effects on individual opinion before and after exposure to a given frame. An auxiliary analysis of media coverage of pork projects suggests that the tone of coverage is often negative, even for local coverage of local projects. For example, an investigation of state media coverage of

50 pork projects in 2007-2008 revealed that 43 percent of the article mentions had negative connotations.6 That said, the amount of media coverage appears to be relatively consistent over time. A Lexis-Nexis search over the past two years of major news sources in the United States reveals an average of 250 hits with a standard deviation of 41 articles per month.7 In short, the varying nature of the message presented to voters necessitates the exploration of the implications of both positive and negative information. This chapter addresses this concern as well.

5.2.3 Issue Publics

Citizens are not uniformly concerned about all political issues. The notion that citizens follow certain issues with varying intensity based on individual preference is an idea stretching back to Converse’s (1964) coining of the term “issue public.” Politics, generally speaking, is a secondary concern for most citizens; given time and resource constraints, it is not surprising that most tend to direct their attention to issues of personal importance. The existence of variable-publics (a series of intersecting issue-speciﬁc publics) has found support in research. Gilens (2001) found that respondents with policy-speciﬁc knowledge were better informed than those with general policy knowledge. In a similar vein, Krosnick (1990, 59) found most Americans are members of a few issue publics “each composed of citizens who are passionately concerned about a single issue.” Finally, Price et al. (2006) found that membership in a variable issue public (health care policy in Price et al.’s study) explained half the variance when predicting opinion strength and political participation.

Member earmarks are not confined to a specific policy; they expand over various issue areas (Lazarus 2009). This gives Members the opportunity to make direct appeals to a wide array of issue publics. As recent work by Gimmer et al. (2012) finds, Republicans and Democrats alike rely on press releases to claim credit for such acquisitions. If members are aware of the desires of their

6This search consisted of all U.S. Senators holding ofﬁce between October 1, 2007 and October 1, 2008. Searches were conducted on the state newspaper with the highest circulation, where available. Additional information is available in the appendix. 7Details on the search process are featured in the appendix.

51 primary and reelection constituencies (Fenno 1978), we should expect that Members tailor their earmarking efforts to these constiuencies’ desires. As Rocca & Gordon (2013) find in their exploration of defense earmarks, district factors (veterans in the district) and Member characteristics (Defense Subcommittee membership, Armed Services Subcommittee membership, and Military Constriction Subcommittee membership) are both significant predictors of defense earmark acquisition. Moreover, Lazarus (2010) confirms that this connection spans across a large array of issue areas. Consequently, investigations about the utility of earmarks should not assume all projects are viewed equally in the eyes of recipients. Rather, we should expect that constituents preferences vary according to their issue public(s), with greater attentiveness to issues of personal concern.

5.3 Hypotheses

The congressional literature suggests that those in recipient districts or states should reward the member responsible for bringing home the bacon. Yet, as the experimental literature makes clear, negative framing of an issue can seriously depress otherwise positive sentiment. Given the discussion in the proceeding sections, a number of empirical questions are raised at the intersection of voter behavior and issue framing. Armed with these underlying theories, it is possible to derive a series of hypotheses:

H1: General information about earmarked dollars being secured for local beneﬁt will increase favorable evaluations of the responsible member of Congress, ceteris paribus.

H2: Particularized information about earmarked dollars that are personally relevant and secured for local beneﬁt will increase evaluations of the responsible member of Congress, ceteris paribus.

52 H3: The effect of information of earmarks that are personally relevant will be greater than the effect of general information about earmarks.

The hypotheses above paint a straightforward picture: that people generally like local pork projects, and especially appreciate projects that are devoted to self-interested causes. In other words, individuals have preference for certain issue areas that are given greater importance than others, which is unsurprising given that “the nation may be conceived of as an amalgamation of issue publics, [or] groups of people with highly important attitudes toward specific policy options” (Krosnick 1990, pg. 81). It should be noted that, while the aforementioned hypotheses predict that projects targeted to specific groups will win the admiration of recipients, they remains agnostic as to the impact of pork dollars for those not in the self-interested group. In other words, there are no a priori expectations for subjects who are not members of an issue public that would benefit from the outlays.

These are not the only questions raised, however. Not all discussions of pork are benign, yet it is difﬁcult to predict respondent evaluation following a negatively framed treatment. While the allocations are nonetheless beneﬁcial, aggregate studies of earmark evaluations suggest the potential to view such projects with disfavor (see Figure 5.1). While it is expected that negative frames will depress support for the earmark allocations, the degree of the effect is unclear; hence, the following two hypotheses:

H4: Information about earmarks framed as wasteful spending will attenuate the positive effect of pork outlays from the responsible member of Congress, ceteris paribus.

H5: Information about earmarks that are personally relevant that are framed as wasteful spending will attenuate the positive effect of directly beneﬁcial pork outlays, ceteris paribus.

53 Hypotheses four and ﬁve attest to the expectation that the tone of the coverage of earmarks is pivotal to ascertaining citizen response, while making no prediction as to the strength of attenuation. To put it another way, I expect negative framing will drive down support, however no prediction is made as to the degree of that effect.

5.4 Experimental Design

The experiments here seek to conclude if pork projects are viewed positively by those receiving them, and assess the role self-interest plays in member appreciation for earmarks. To determine this, a series of three experiments were conducted in the Summer of 2011, Fall of 2011, and Spring of 2012 in the Claude Pepper Center on the Florida State University’s campus. Students were recruited from various political science courses by offering them extra credit in one course in exchange for their participation. The recruitment efforts resulted in student samples of 250, 438, and 507, respectively.

One potential area of concern with this type of experiment is the reliance on a student sample. Sears (1986) suggested that relying on college students inherently introduced uncertainty as to the generalizability of any experimental ﬁndings. However, Druckman and Kam’s (2011, 53) investigation into this claim concluded that in regards to the external validity of student samples “there is nothing inherent to the use of student subjects that reduced experimental realism.”8 Ad- ditionally, there is the concern that the repeated use of students may lead to contamination of the participant pool. A variable capturing the number of political science experiments a student has taken part in was recorded for each study; the variable was not statistically signiﬁcant. Neverthe-

8That said, Druckman and Kam (2011) did ﬁnd a limited number of student characteristics differed from the general population, amongst which was the amount of political information (students had more than the general population). A political information index variable was tested in the Summer 2011 experiment, however it proved insigniﬁcant. The inclusion of the measure did not substantively alter the results. Such estimates are available in the appendix

54 less, this alternative speciﬁcation is featured in the appendix.

As mentioned, the experiment was conducted in three studies, each with different treatments and expectations. The first looked at pork awareness framed in a beneficial way as it pertains to a U.S. Senator. The second study verified the findings of the first as they pertain to U.S. House members, as well as investigating the desirability of pork across multiple issue areas. The final study again relied on assessment of Senator pork; however, in this instance the pork was framed in a negative context.

5.4.1 Study 1

In the ﬁrst study, respondents were asked to evaluate Senator Bill Nelson (D-FL) after being randomly assigned into one of three conditions. Two treatment groups were given information about Senator Nelson’s earmark activities:

(General Positive) Last fiscal year, Senator Bill Nelson of Florida brought $654,441,700 in earmarks (also known as pork projects) to the State. He ranked 4th in the Senate for the most money brought home in that fiscal year. (Particularized Positive) Last fiscal year, Senator Bill Nelson of Florida brought $654,441,700 in earmarks (also known as pork projects) to the State. He ranked 4th in the Senate for the most money brought home in that fiscal year. Over $900,000 of the money he brought home to Florida directly funded programs at Florida State University (FSU) and Tallahassee Community College (TCC).

The control group was given no additional information. After the treatment, all subjects were asked two questions: “How would you describe your views towards Senator Bill Nelson,” and “How strongly do you approve or disapprove of the job Senator Bill Nelson is doing?” These two evaluation items were combined to create an averaged approval measure.9 The questions were combined to lend accuracy to Member evaluation. Cronbach’s ❛ conﬁrms the uni-dimensional

nature of the latent construct: α = 0.887.10 9These two questions were both evaluated on a 7-point Likert scale. 10Replication of the results using uncombined dependent variables reveals the results remain consistent (see appendix).

55 The vignette was designed to give subjects a slightly positive view of earmarks, or at the very least, to avoid negative framing. All of the information displayed was factually correct, and supplied by the data collection efforts of Taxpayers for Common Sense.11

5.4.2 Results of Study 1

The results of study 1 (table 5.1, column 1), which addressed Senator Nelson’s general and higher education earmark efforts, conﬁrm hypotheses one, two, and three. First, general information about

earmarks was found to increase support for the Senator by over half a point (.533, p < 0.01). In other words, merely reminding recipients that their Senator had brought federal funds to their state did indeed bolster support. Second, we see that personally relevant local beneﬁts notably increase support for the responsible member. In fact, as table 5.1 shows, the specialized education frame

bolstered support for Senator Nelson by more than a full point on a seven-point scale (1.115, p < 0.001). Finally, the results confirm the third hypothesis: knowledge of personalized benefits has a stronger effect than simply being aware of general pork allocations. That said, while the coefficient is larger, it must be shown that the two treatments are statistically distinguishable from one another; this can be achieved through a simple F-test. As predicted, the two treatments are statistically different F(1, 246)= 7.51, p <.01.

As a robustness check of independence, a non-parametric Mann-Whitney test was conducted (Mann & Whitney 1947). This stricter alternative to the t-test does not assume an equal- interval scale of measurement, has greater efﬁciency, and does not assume that the underlying sample distributions are normal. Rather, the test assumes that the groups are independent, that the responses are ordinal, and that both groups are equal if the null is true. The test again conﬁrms that

the two treatments are statistically distinguishable: z = −2.91, p = .004.

Two areas of concern regarding the ﬁndings warrant further discussion: the inclusion of

11< http://www.taxpayer.net >

56 Table 5.1: The Effect of Positively Framed Pork Treatments on Support for Sen. Bill Nelson (D-FL), Summer 2011 Variable Coeff. (std err) Coeff. (std err) General (Positive) Pork Treatment 0.533** 0.495 (0.203) (0.522) Educ. (Particularized Positive) Pork Treatment 1.115*** 0.338 (0.205) (0.523) Party ID (Dem) 0.100 0.023 (0.066) (0.104) Party x Generalized Pork Treatment 0.010 (0.155) Party x Education Pork Treatment 0.265 (0.162) Constant 3.959*** 4.203*** (0.249) (0.355) R2 0.111 0.122 N 250 250 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

the party identification control variable, and the potential for party identification to moderate the findings. Randomization inherent in the experimental design makes it extremely unlikely that other explanatory variables are correlated with the treatment. This does not prevent chance assignment of “too many people of a particular type [to] one of the treatment groups” (Ansolabehere & Iyengar 1995, 172). In the experimental setting, if the data generating process includes party identification and it is omitted, we will not have an omitted variable problem, but its inclusion will reduce the estimated error. In other words, the inclusion of relevant variables can lead to nontrivial increases in efficiency (Franklin 1991). Nonetheless, this model and all subsequent models were also estimated with the party identification variable omitted; the results remain substantively unchanged.12 The results from these aforementioned models are featured in the appendix. The second issue of concern is the potential for increases in support being an artifact of congruence, or lack thereof, between the Member and the respondent, rather than the earmark treatments themselves. The treat-

12Additionally, the distribution of party ID across the two treatment and control conditions suggests random assignment. The general, particularized, and control groups party identiﬁcation scores had an average of 3.051, 2.846, and 3.151, respectively; a maximum difference of only .304.

57 ments attempted to avoid this by omitting the party of the member from the vignettes. Nonetheless, the concern still poses a potential threat to the findings. Consequently, to test for such a relationship a series of party-treatment interactions were included. If the treatment is contingent upon the party of the respondent we would expect such an interaction to be statistically significant. As table 5.1 (column 2) reveals, none of the treatment and party identification interactions are statistically significant, suggesting the absence of a moderating effect.

The findings of table 5.1 are visually explored below. Figure 5.2 reports the treatment effects of positive general and positive particularized (education) frames of pork allocations. When informing respondents of the amount of earmarked dollars brought back to the state, we see a significant increase in support for Senator Nelson, and when these allocations are personally relevant (in this case, students being told of pork for higher education), approval notably climbs. More- over, despite the overlapping confidence intervals, statistical comparisons (see above) indicate that the particularized treatment is statistically different from the generalized treatment, confirming hypothesis three.

Figure 5.2: Treatment Effects of Pork on Support for Sen. Nelson (Positively Framed)

58 5.4.3 Study 2

A second experiment was conducted in the Fall of 2011. It attempted to build upon and expand the findings of the previous study. Namely, the second experiment desired to ascertain changes in respondent evaluation of a U.S. House member after being informed of his earmark activities. Ideally respondents would have been asked to evaluate Representative Steve Southerland’s (R, FL-2nd) earmarking efforts, as he was the current congressman representing the district in which Tallahassee resides. However, at the time of the experiment Rep. Southerland was a freshmen Member, which eliminated the possibility of relying on factual earmarked allocations. Conse- quently, rather than relying on fictitious information, Rep. Southerland was replaced with Rep. Jeff Miller (R, FL-1st). The assumption in this replacement is that Miller will function as an adequate proxy. This rests on the fact that there exists a general lack of awareness when it comes to a constituent knowing who is their member (Jacobson 2009, 123-24; Mann 1978, 30-34). Fur- thermore, both congressmen share the same party affiliation, and their districts are geographically proximate.

In addition to adding certainty to the findings of the first study by running a similar experiment on Representatives, the second study tests whether positive evaluations resulting from earmarks are a matter of simply providing subjects with more information, or the result of an appreciative issue public specific to an individual policy arena (e.g. isolated solely to college students rewarding for pork for higher education). As Stein & Bickers (1994, 380-81) point out, awareness of outlays “should vary with voters’ general attentiveness [conditional on] the substantive nature of the grant award; [t]eachers might be expected to be more aware of new grant awards for education,” for example. Since the population in all three studies is college students, it is expected that information about allocations that benefit institutions of higher learning would evoke positive feelings. What remained unanswered was if such an affect would translate to other policies that are considered personally advantageous. To answer this question, an additional treatment was added regarding military allocations. Later in the survey, respondents were asked “[a]re you or a member of your immediate family actively serving in the military?” The treatment vignettes were the

59 following:

(General Frame) Last fiscal year U.S. Representative Jeff Miller for Florida’s 1st Dis- trict secured $21.5 million in earmarks (also known as pork projects) for his district. (Particularized, Education Frame) Last fiscal year U.S. Representative Jeff Miller for Florida’s 1st District secured $21.5 million in earmarks (also known as pork projects) for his district. A sizable amount of the money he brought home to Florida directly funded programs at institutions of higher learning. (Particularized, Military Frame) Last fiscal year U.S. Representative Jeff Miller for Florida’s 1st District secured $21.5 million in earmarks (also known as pork projects) for his district. A sizable amount of the money he brought home to Florida directly funded military defense programs and projects.

Once again, the control group was given no information. All respondents were then asked to describe their views towards Jeff Miller, and to provide a job approval rating. These two responses were combined to form an overall evaluation measure.13

The first two treatments (General and Education) mirror those of the first study; only the dollar amounts of the projects brought home are notably different.14 The third treatment (Mili- tary), when paired with the follow-up question regarding association with the military, attempts to capture the influence of self-interested behavior on the desirably of pork akin to the higher education vignette. Those in the General treatment category are expected to have a more favorable opinion of the member as compared to the control group (hypothesis one), as are those in the Education treatment and those with military affiliation in the Military treatment (hypothesis two). Additionally, information about particularized benefits (i.e. the Military and Education treatments) will positively affect respondent appreciation as compared to the General treatment (hypothesis three). Finally, there are no a priori expectations regarding the size of the effect of the Education treatment as compared to the Military treatment.

13Like the ﬁrst study, these two questions were both measured on 7-point Likert scales. Cronbach’s α again conﬁrms the unidimensional nature of the combined measure: α = .757. Disaggregated results are substantively the same, and can be found in the appendix. 14Readers will also notice that the Education treatment is vague in regards to the destination of federal dollars devoted to higher education.

60 5.4.4 Results of Study 2

Study 2 explored the effect of general, education, and military pork information on respondents’ views of Representative Miller; together they conﬁrm the second and third hypotheses. However, contrary to the ﬁrst hypothesis, table 5.2 (column 1) reveals that simply telling respondents about earmarks did not affect support for the responsible member.

Beginning with the first hypothesis, unlike the previous experiment involving Senator Nel- son, general information about earmark allocations was not a statistically significant predictor of changes in support. It is not clear why the general frame was significant in the previous iteration but not in the the one at hand; however, it may be due to the slight difference in the wording of the frame (i.e. mentioning Senator Nelson was the fourth most successful at securing earmarks).15 The second treatment informed respondents of earmarks going to institutions of higher learning; such information led to an increase in positive evaluations of the responsible Member, confirming the second hypothesis. As table 5.2 details, learning of pork allocations that benefit colleges and universities resulted in almost a full point increase in favorability (0.84, p < 0.001), ceteris paribus.

Readers may question whether the aforementioned treatment affect is generalizable to other issue publics. If earmark dollars devoted to issues of personal interest truly increase support for the member who secured the funds, such an effect should exist for other issues pertinent to the respondent. To solidify the initial ﬁndings, an additional treatment was included (similar to the education frame) in the experiments. This was accomplished through the Military Treatment frame and a follow-up question asking if the respondent, or a member of their immediate family, was actively serving in the armed forces. As table 5.2 (below) reveals, for those not associated with the military, the treatment had no statistically signiﬁcant effect (0.057, p = not signi ficant). How- ever, for those who received the military-pork treatment associated with the military (i.e. Military

15One might suspect that the cause lies in relying on Representative Miller, instead of Representative Southerland, however, if this was the case, we should expect also to ﬁnd insigniﬁcant particularized treatments, but this clearly did not occur.

61 Afﬁliation x Military Frame), there is a sizeable increase in support equal to an almost one point (0.98, p < 0.001), ceteris paribus.

Finally, when taken together, the education and military frame conﬁrm the third hypothesis: local allocations of personal beneﬁt result in larger increases in support for favorability than generalized allocations. The education frame and military frame are comparable in size (0.84 and 0.98, respectively) and both are notably larger and statistically different from the general frame, which

an F-test conﬁrms for both the education frame, F(1, 431)= 30.7, p <.001, and the military frame (for those in the military) F(1, 431)= 7.41, p <.01.16

Table 5.2: The Effect of Positively Framed Pork Treatments on Support for Rep. Miller (R, FL- 1st), Fall 2011 Variable Coeff. (std err) Coeff. (std err) General Frame (Positive) 0.123 0.338 (0.129) (0.319) Education Frame (Positive) 0.835*** 0.889** (0.123) (0.305) Military Frame (Positive) 0.057 0.543 (0.128) (0.311) Party ID (Dem) -0.077* -0.017 (0.033) (0.066) Military Afﬁliation -0.220 -0.201 (0.140) (0.141) Mil. Afﬁliation x Mil. Frame 0.978*** 0.918** (0.283) (0.285) Party x Generalized Treatment -0.068 (0.096) Party x Education Treatment -0.011 (0.092) Party x Military Treatment -0.157 (0.092)

16As with the previous specification, a stricter Mann-Whitney test was conducted. This test also confirms that the Education Treatment (z = −4.932, p = .0000), and the Military Treatment for those affiliated with the military (z = −1.841, p = .0656) are statistically distinguishable from the General Treatment.

62 Table 5.2 - continued Variable Coeff. (std err) Coeff. (std err) Constant 4.352*** 4.156*** (0.140) (0.229) R2 0.150 0.157 N 438 438 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis. Like the previous study, the concern of party-treatment moderating effects warrants the consideration of several interactions. As column 2 of table 5.2 reveals, the interactions are once again statistically insigniﬁcant, suggesting no moderating affect between the party identiﬁcation of the respondent and the treatment outcomes.

Figure 5.3: Treatment Effects of Pork on Support for Rep. Miller (Positively Framed)

The results of study 2 are born out in Figure 5.3. Here the effect of the general treatment is compared to the individual interest frames of military earmarks and education earmarks. As the

63 ﬁgure details, while information about earmarks in general (the general treatment) is not statistically signiﬁcant, for respondents informed about personally relevant projects (military pork for those in the military and education pork for all student respondents), such outlays notably bolster ones’ opinion of Representative Miller. Moreover, this effect does not appear to be limited to a particular policy arena, rather, as predicted by the second hypothesis, the change in support hinges on personal interest.

5.4.5 Study 3

The ﬁnal study was conducted in the Spring of 2012 to explore the consequences of negatively framing earmark allocations. As discussed previously, a large proportion of media coverage of pork projects is negative in nature, yet much academic literature assumes an appreciative public regardless. The ﬁnal experiment puts this assumption to the test.

Much like the ﬁrst experiment (Summer 2011), the third study was interested in respondent evaluations of Senator Bill Nelson (D-FL). However, this third iteration explored two negative vignettes and a control condition; the two treatments were the following:

(General Negative Treatment) Taxpayers for Common Sense, compiles a yearly list of earmarks (also known as pork projects) obtained by Members of Congress. According to them, Sen. Bill Nelson (FL) secured over $650 million in projects. Taxpayers for Common Sense argues for the elimination of earmarks to increase accountability and reduce wasteful spending. (Particularized Negative Treatment) Taxpayers for Common Sense compiles a yearly list of earmarks (also known as pork projects) obtained by Members of Congress. According to them, Sen. Bill Nelson (FL) secured over $650 million in projects. Al- most $1 million of the money he brought home to Florida directly funded programs at Florida State University (FSU) and Tallahassee Community College (TCC). Taxpayers for Common Sense argues for the elimination of earmarks to increase accountability and reduce wasteful spending.

Following treatment or control (no text) conditions, subjects were once again asked to

64 evaluate Senator Nelson on approval and job performance as measured on 7-point Likert scales.17 Much like the first study, treated subjects were given information regarding earmark allocations generally, or targeted to higher education. The key difference in the final study is the association of earmarks with “wasteful spending.” The expectation is that the positive gains secured by alert- ing constituents to general and particularized earmarks is attenuated by negative press coverage (hypotheses four and five).

5.4.6 Results of Study 3

The results of study 3 are revealed in table 5.3 (column 1). The analysis conﬁrms hypotheses four and ﬁve.

First, regarding hypothesis four, even a mildly negative framing of earmarks attenuates positive gains in Member favorability. Looking at the general negative frame, we see that the treatment had no statistically signiﬁcant effect on evaluation of Senator Nelson; this is in stark contrast to the half-point gain in the previous general (positive) treatment. As hypothesis four predicted, negative connotations surrounding the general acquisition of pork does indeed attenuate the positive gains. Additionally, as hypothesis ﬁve predicted, a negative framing of personally relevant pork projects attenuates gains in Member support. However, unlike the general negative frame, we see that despite the negative connotations, personally relevant pork projects (education

dollars) nonetheless increase support by almost half a point (0.47, p < 0.001). Overall, the effect of negative framing attenuates support as compared to a positive frame, but nonetheless bolsters support amongst those seeing the outlays as personally pertinent.

Like the previous models, the concern surrounding party-treatment effects is tested with the inclusion of a number of interactions between the treatment condition and the party identiﬁcation of the respondent. As with the previous two models, column 2 of table 5.3 reveals that neither of

17Cronbach’s α once again conﬁrms the unidimensional nature of the combined measure: α = .886. Disaggregated results are substantively the same, and can be found in the appendix.

65 the interactions is statistically signiﬁcant.

Table 5.3: The Effect of Negatively Framed Pork Treatments on Support for Sen. Bill Nelson (D-FL), Spring 2012 Variable Coeff. (std err) Coeff. (std err) General (Negative Frame) Pork Treatment -0.167 0.153 (0.136) (0.329) Education (Negative Frame) Pork Treatment 0.467*** 0.641* (0.131) (0.315) Party ID (Dem) 0.205*** 0.257*** (0.040) (0.068) Party x Generalized Treatment -0.107 (0.100) Party x Education Treatment -0.058 (0.096) Constant 3.609*** 3.454*** (0.151) (0.222) R2 0.091 0.094 N 507 507 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

The findings of table 5.3 are visually explored below. Figure 5.4 reports the treatment effects of positive frames (study 1, estimates indicated with squares) and negative frames (study 3, estimates indicated with circles) as they pertain to individual interest in pork allocation. Again, when informing respondents of the amount of earmarked dollars brought back to the state, we see a significant increase in support for Senator Nelson. However, this effect is no longer significant when pork was framed as wasteful. Both Senator Nelson and Representative Miller shared the significant boost in approval when respondents were informed of earmarked dollars that are of personal interest (in this case, students being told of pork for higher education). However, consistent with hypothesis five, positive gains are attenuated when subjects were given the same information within a slightly negative context. While personally relevant (education) information still led to a gain in support, when the message turned negative, there was a notable decline in such gains.

66 Figure 5.4: Treatment Effect of Pork on Support for Sen. Bill Nelson

Interestingly, the negative frames are very similar to positive frames in their degree of change in support for Senator Nelson. Specifically, while generalized negatively framed pork projects were not statistically significant, particularized benefits still maintain a stronger substantive effect than a generalized negative frame. And like the first study, these effects statistically

differ from one another (F(1, 503)= 21.91, p <.001).18

5.5 Conclusion

This work explored a long held assumption surrounding federal outlays: that earmarks brought home to a state or district will invariably win the favor of appreciative recipients. A number of

18As with the previous two speciﬁcation, a stricter Mann-Whitney test revealed the negative education treatment (z = −4.334, p = .0000) is statistically distinguishable from the general treatment.

67 scholars have implied or directly stated that pork stands as a robust tool available to incumbents to reduce electoral uncertainty (see Popkin 1991; Stein and Bickers 1995; 1995; Jacobson 2009; Sellers 1997 for prominent examples). This conclusion has long been asserted in spite of the overwhelming aggregate disapproval of earmarks. These suppositions are based on traditional notions of representation: that while voters may despise the institution as whole, they appreciate their Member. While this makes intuitive sense, such proposals have not been subjected to empirical testing. By relying on a series of experiments, this work conﬁrms that earmarks may aid candidates, but that such gains are conditioned on the framing language used to describe the projects.

Ultimately the experiments presented above recast conventional views of earmarks as a universally positive benefit for Members wishing to bolster electoral security. Rather, the ability of a Member to successfully sell earmarks to their constituents, like most political issues, is heavily dependent on the framing of the issue, not just the message. Basic factual information describing the amount of money brought home met with mixed results. While such information was not a statistically significant aid in the House example, when these monies were put in the added context of how that Member ranks in comparison to his peers (namely, the example of Senator Nelson ranking fourth in earmarks procurement), we saw significantly elevated opinions. When this information was personally relevant (e.g. college students enlightened to university earmarks, and those associated with the military being told of pork pertaining to defense) Members gained significantly from these efforts. Finally, when earmarks are framed in a negative light, we see that the gains from general information disappear; yet, if the earmarks are personally relevant, the information continues to aid the Member, regardless of negative connotations.

The results of this work contribute to the current literature by working to conﬁrm assertions that pork can aid Members seeking reelection, while correcting preexisting academic assertions that pork is a universally positive good. Members looking to win appreciation should not assume that more is always better. Rather coverage of projects is what determines popular reaction. Skilled politicians must identify which policy arenas are of particular importance to their constituents, work to secure projects in those policy realms, and ensure prominent media coverage of those

68 efforts. Casual analysis suggests that many Members do indeed pursue this goal. For example, Adam Smith (WA-9th) represents a Washington district with numerous environmental and farming interests. He single-handedly worked to secure $24.1 million in earmarks for projects related to environmental quality, and features these efforts prominently on his website.19 Appeals to issue publics are a Members’ best hope to reap support from pork.

What remains unanswered is how, if at all, the efforts of Members make their way into the consciousness of their constituents? Most voters remain unaware of pork, but occasionally prominent earmarks make their way onto the national stage. For example, the Big Dig (the Central Artery Tunnel Project) in Boston, and the Bridge to Nowhere (the Gravina Island Bridge) in Alaska both received extensive national coverage. But, as readers may have noticed, both of these projects had extremely negative connotations. What needs further exploration is how the ﬁndings of this study map onto the national media landscape. Speciﬁcally, scholars have yet to explore the impact of real world media coverage of pork and its effect on recipients.

Future work should seek to use experiments and media studies to ascertain the ability of Members to utilize pork to garner support. In the experimental setting it would be valuable to replicate these findings with a non-college student population, as well as to manipulate the issue areas presented to subjects. In the aggregate, it would be useful to explore not only the impact of coverage of earmarks, but the utility in foregoing earmarks altogether. While this study has confirmed Members may reap benefits from pork, it remains to be seen if fiscal stewardship is rewarded with the same marked appreciation as particularized benefits.

19< http://adamsmith.house.gov/ﬁscalappropriations/ >

69 CHAPTER 6

MYSTERIOUS PORK: THE LACK OF CITIZEN AWARENESS OF EARMARKS

6.1 Introduction

Scholars have long assumed that legislators who acquire federal money for their constituency in the form of earmarks reap great electoral rewards. This assumption is hardly surprising. Legislators go to extreme lengths to woo their constituents, and the practice of earmarking federal money for the beneﬁt of local voters stands as one of the prime examples of reelection-seeking behavior. Given the uncertainty surrounding reelection (Fenno 1978), it comes as no surprise that Members would take great pains to bring home the bacon, regardless of the uncertain impact. In contrast to previous work (e.g. Ferejohn 1974; Shepsle and Weingast 1981; Jacobson 2009), this chapter challenges this assumption. In order to reward legislators for bringing home pork, citizens must be aware of the new projects, aware that the projects were funded with federal dollars, and able to attribute credit for federal money to the proper legislators. I argue that this demands a lot from citizens, particularly on an issue that receives relatively little media coverage.

This chapter begins by ﬁrst reviewing how previous work has conceptualized and measured earmarks. Using newly available data, I reassess previous measures of these pork projects, and ﬁnd that they are only moderately correlated with more accurate measures. Second, I test a

70 minimal condition for voters rewarding pork: that voter awareness of earmarks is correlated with the acquisition of earmarks. In contrast to the expectations of much previous literature, no relationship is found. Third, this chapter shows evidence that voter awareness of pork is driven by media coverage. Finally, this chapter demonstrates that the aggregate, direct effect of earmarks on legislator approval is minimal. Additionally, modest levels of media coverage of earmarks do not affect evaluations of a respondents’ Congressperson. Clearly this does not foreclose the potential for earmarks to affect member evaluations (see Chapter 5), but it does mean that, on the whole, MCs have done a poor job of enlightening their constituents.

6.2 Earmarks and Electoral Behavior

For decades scholars have worked under the assumption that earmarks, and other direct efforts, have an electoral impact. Intuitively, the notion that a member can increase voter support by providing tangible beneﬁts seems plausible. Projects involving university funding, new bridges, or waterfronts should appeal to all voters and seemingly require little sophistication to understand. Earmarked projects can be thought of as a valence issue–every constituent should desire more because they beneﬁt their home district at virtually no cost to the constituent.

Incumbents have a variety of tactics for increasing their vote share in upcoming elections that are unavailable to challengers, especially forms of constituent service. As a result, scholars have speculated that electoral resources likely explain some of the incumbency advantage (King and Gelman 1991). For example, scholars have identiﬁed the franking privilege and casework as means for incumbents to increase their vote share (Cover & Brumberg 1982; Johannes 1984; Serra 1994; Serra & Cover 1992; Romero 1996; King 1991; Johannes & McAdams 1981). Earmarks can be thought of as an ideal form of constituency service, because they provide direct beneﬁts to

71 a large number of constituents, and are easily manipulated by legislators.1 Indeed, many congressional scholars have used this assumption in theories of electoral behavior. For example, Shepsle and Weingast (1981) use the occurrence of pork to explain why we observe oversized coalitions in Congress, as opposed to minimal winning coalitions. Fundamental to this theory is the assumption that earmarks have a substantive impact on legislators; or in the words of Shepsle and Weingast (1981, 110): ““pork,” in various forms, will always serve as a part of the legislators’ response to his voters’ retrospective question, “What have you done for me lately?”” Baron (1990) also suggests a connection between pork spent by House Members and electoral consequences in his study of funding for Amtrak. He considers “constituent preferences” as a possible explanation for distributive resources (Baron 1990, 886), suggesting that the recipients of beneﬁts are able to attribute federal dollars to a particular individual. In his groundbreaking work on pork projects in rivers and harbors legislation, Ferejohn (1974) contended that “projects can be symbolically important in a reelection campaign,” and that bringing home earmarks allows a Member “to ensure himself more freedom in voting on other issues” (1974, 49-50). Finally, Jacobson (2009, 235) argues:

Electoral logic inspires members to promote narrowly targeted programs, projects, and tax breaks for constituents and supporting groups without worrying about their impact on spending and revenues. Recipients notice and appreciate such specific and identifiable benefits and show their gratitude to the legislator responsible at election time.

Clearly there is a shared consensus that the public rewards legislators for bringing home pork and, thus, that earmarks are a valuable tool for reelection. Nowhere has this connection been more thoroughly explored than in the work of Stein and Bickers (1994; 1995). In their groundbreaking work, these authors propose three assumptions pertaining to a potential pork-electoral connection: that allocation of earmarks can be altered by legislators, that constituents are aware of receiving earmarks, and that voters reward a Member who has procured federal pork (1994, 382). Ultimately the authors conclude that only vulnerable members seek to procure more earmarks,

1This stands in contrast to federal outlays, which are a part of the normal budgetary deliberation process. Such normal spending is not easily manipulated by a single legislator, and thus does not likely serve as a common form of constituency service.

72 that only politically attentive recipients are likely to be aware of said beneﬁts, and that awareness of projects results in increased favorability.2 In spite of the innovativeness of Stein and Bicker’s work, there are a number of unanswered questions that remain. First, given the limited availability of accurate data, it is unclear if previous ﬁndings are an artifact of time or data availability. Sec- ond, Stein and Bickers (1994, 383) acknowledge that “[r]epeated news coverage should enhance both voter recognition and evaluation of the incumbent,” but do not empirically test such a proposition. Finally, Stein and Bickers hint at possible differences in the way Members and constituents rely upon party cues to make assessments regarding earmarks–however, this path too is not fully explored. This work attempts to lend clarity to past work on the impact of earmarks, while also exploring remaining questions regarding media and ideology.

6.2.1 Self-Interest, Information, and Electoral Rewards

As detailed above, much previous work assumes that voters will act in their own self-interest and reward legislators in the voting booth for bringing home earmarks. However, evidence of citizens voting based on their own self-interest has been scarce. For example, Campbell (2002) finds that senior citizens are motivated to turnout by concerns over Social Security and describes this as a “rare example of self-interest exerting a significant influence on individual behavior.” Indeed, a number of articles reviewing the effects of self-interest on political attitudes find only weak or null evidence (e.g. Lau & Heldman 2009; Sears et al. 1980).

Psychologically speakinmg, people often fail to act in their self-interest because of infor- mational or cognitive limitations (Simon 1983). As Althaus et al. (2011, 1065) argue, “[f]ew individuals have a direct view of events, and so most people respond instead to the mediated realities constructed for them by new outlets.” Indeed, a substantial body of research demonstrates

2This last ﬁnding is questionable given the inclusion of both pork and awareness measures in predicted incumbent support. As Cain et al. (1984, 123) point out, there is a strong possibility of endogeneity between self-reported awareness of localized beneﬁts, and support for one’s own member. This will be given further consideration below.

73 that media coverage is one of the strongest drivers of political knowledge (e.g. Jerit et al. 2006; Barabas & Jerit 2009). And as Grimmer et al. (2012, 6) have recently pointed out, half of all press releases claim credit for allocations outside of the appropriations process. This literature leads to questions surrounding prior claims asserting a direct casual link between localized benefits and the awareness of said benefits. This chapter contends that increased earmark outlays have no direct, measurable effect on citizen awareness of pork. Rather, citizen awareness will depend heavily on media exposure, which leads to the first hypothesis: As media coverage of a congressperson’s earmarking activity increases, constituents will be more likely to identify that Member as bringing home earmarks. Thus, while citizens may not have direct experience with pork projects (let alone knowledge of where the money comes from), media coverage may enhance voter awareness of these projects.

Assuming that media coverage increases knowledge of earmarks, perhaps media coverage can save the purported electoral beneﬁts of earmarking as well. However, such a conclusion ﬁrst requires consideration of the information required to reward legislators for procuring pork. This asks a lot from citizens, particularly on an issue that, as we will see, receives relatively little media attention.3 Supporting this contention, survey data reveal that the public knows relatively little about earmarks. For example, a majority of citizens readily admit to not knowing the proportion of the national budget devoted to pork projects.4 Moreover, according to the 2008 CCES, only 17% of respondents report being able to recall a project that their legislator brought to their district.5 Given that nearly all legislators brought home federal money for a pork project, this low level of knowledge serves as a reminder that “routine transmission of politically important information should never be assumed” (Althaus et al. 2011, 1076).

3Moreover, as will be demonstrated below, national media coverage of a Senator’s earmarking behavior is only moderately correlated with their actual procurement of earmarked dollars (r = .41, p <.001); local media coverage of these efforts is even weaker (r = .32, p <.001) 4A 2011 CBS poll revealed that 41% of respondents admitted not knowing how much of the national budget was devoted to earmarks; only 16% were able to answer correctly from the six possible choices. 5Note, however, that this likely severely overstates public knowledge, because it does not require them to actually hold knowledge, but simply to report that they hold it. A follow-up question asked respondents claiming to recall projects give speciﬁcs about their recollection. Analysis of this open-ended data revealed that only 13.8% of respondents were able to both claim to recall a project, and give a speciﬁc example.

74 Several scholars have implicitly or explicitly assumed that, despite the cognitive hurdles, greater amounts of pork increase awareness of these projects, as well as bolster support for the responsible Member. The most direct attempt to empirically establish a connection between constituent awareness and response to pork is the cumulative work of Stein and Bickers (1994, 1995). Using the 1988 NES, the authors demonstrate that changes in federal money ﬂowing to a respondent’s locality from the previous to the current year are associated with an increased probability of the respondent saying that their Representative has done something “special” for their district recently. Stein and Bickers (1994, 1995) also ﬁnd that believing a Representative has done something “special” recently is associated with an increased likelihood of voting for the incumbent.

While these studies find that the acquisition of pork yields electoral benefits among highly informed citizens, I argue this finding warrants reconsideration. Until recently scholars have not had access to an accurate measure of earmark acquisitions, forcing reliance upon Federal Assis- tance Award Data System (FAADS) data, which were not designed to measure earmarks. Indeed, all of the previous attempts to establish a connection between earmarks and voter behavior have relied on FAADS data (Stein and Bickers 1994, 1995; Bickers and Stein 1996; Sellers 1997). Armed with new, more accurate data, this chapter reevaluates previous findings. In addition to replicat- ing previous work, two additional contributions are made. First, this work follows up on previous speculation that media coverage may be crucial to Members reaping the electoral benefits of pork (Stein and Bickers 1994). Second, given the changes in institutional procedures, and a number of high profile controversies involving earmarks, I seek to determine the temporal robustness of previous findings.

The next section begins by deﬁning earmarks and describing a newly available measure of the projects. Next, I assess the quality of the FAADS data relied upon by previous work, and provide evidence that claims made on the basis of these data should be regarded with skepticism. Finally, I reexamine previous work on the electoral beneﬁts of earmarks, and build upon it by exploring the effect of media coverage.

75 6.3 Data Sources

Those attempting to study earmarks are presented with three primary sources of data: Citizens Against Government Waste’s Pig Book (CAGW), Taxpayers for Common Sense (TCS) data, and the Federal Assistance Awards Data System (FAADS). FAADS is temporally the most expansive data source. It began in 1981 via Congressional statue to function as a repository for federal assistance and transactions; it is prepared quarterly by the Census Bureau in conjunction with the Ofﬁce of Management and Budget (OMB). The data are sorted according to location of initial receipt (this money could then be reallocated) on an action-by-action basis, or as an aggregate county total. This means that in many instances the data may be attributed down to the state, district, or city location, depending on how the funding allocation is targeted.

While such a system sounds promising for measuring earmarks, these data have substantial limitations. First, and most importantly, FAADS data do not attempt to catalog earmark projects, rather they are a record of all federal direct spending.6 This poses a problem for scholars wishing to differentiate pork from the normal competitive allocation process. Moreover, it requires either relying on the data in its current form, which would introduce a great degree of error, or attempting to ﬁlter out earmarked projects from normal spending. While Bickers and Stein (1996; Stein and Bickers 1994, 1995) attempted to avoid this problem by focusing on the change in the number of new awards, rather than the dollar amount, this raises concerns regarding accuracy, which will be explored further below.7 Finally, FAADS data do not allow scholars to attribute the federal assistance to a speciﬁc Member of congress, nor do they identify the legislation responsible for the project. This ambiguity has limited the study of earmarks by imposing the assumption that voters, electoral challengers, and recipient groups, attribute credit to all Members in both the House and the Senate for projects that occur within a given district or state.

6For example, FAADS data also includes detailed receipts of spending by the Social Security Administration, Department of Defense, and the Department of Homeland Security. 7To their credit, Stein and Bickers (1994, 1995) took great pains to remove non-discretionary spending programs, as well as to ensure the proper attribution of federal outlays to the correct geographic beneﬁciaries.

76 Alternatively, CAGW has been collecting information on pork spending since 1991, and has archived its records in a searchable format on its website. CAGW relies on their own seven- point scheme for determining which projects are considered pork (see Chapter 3); while this process does have its limitations, it is nonetheless consistent. Further, while CAGW has data on pork spending in the aggregate beginning in 1991, it is not until 2008 that the site allows users to attribute pork projects to speciﬁc Members. Consequently, CAGW data only allows researchers to observe national year-to-year trends, amongst states.

The ﬁnal source in tracking earmarks is TCS’s data collection efforts. As previously discussed, Taxpayers for Common Sense rely on changes to institutional rules that require Members to disclose their earmarks. Its data are highly accurate, but temporally limited.

Given FAADS data has long been the only source for information on pork projects, it is not surprising that research on earmarks has been limited (Ferejohn 1974; Bickers & Stein 1996; Sellers 1997; Balla et al. 2002) for several decades. However, as discussed, recent policy changes have streamlined this process. For the ﬁrst time in Congressional history, scholars are able to attribute all earmarks to their sponsoring Member, not just those from speciﬁc policy areas. As a result, recent articles (Lazarus 2009, 2010; Lazarus and Steigerwalt 2009; Engstrom & Vanberg 2010) utilize these new data to explain which members of Congress receive earmarks. However, none have reassessed the potential impact of earmarks on electoral behavior.

6.3.1 Data Accuracy

Scholars have taken to measuring earmarks in a variety of ways. For example, many authors rely on the dollar value or logged value of the projects (Balla et al. 2002; Heitshusen 2001; Berry et al. 2010; Lazarus 2010; Engstrom & Vanberg 2010; Crespin & Finocchiaro 2008), while others rely on the number of projects (Lazarus 2009; 2010; Lazarus and Steigerwalt 2009; Engstrom and Vanberg 2010). Still others have opted to look at changing outlays, such as the change in

77 the ratio of new awards (Stein and Bickers 1994; 1995; Bickers and Stein 1995), the change in the percentage (Lee 2003), the number and dollar value of new awards (Alvarez & Saving 1997; Porter & Walsh 2006), the average number (Shepsle et al. 2009), or the change over a limited time frame (Sellers 1997).

The variation in measurement schemes seems to hinge on how scholars perceive that constituents process information about pork projects, assuming they are aware of the benefits in the first place. Authors have adopted two approaches: evaluation via retrospective comparison and evaluation based on signaling strength. Retrospective comparison proposes voters contrast the amount or significance of projects brought to the state or district today, versus those brought in the past. On the other hand, assessment through signaling strength implies the size (worth or frequency) of local projects affects the ability of recipients to take notice. Neither claim is fully convincing.

This work proposes that a modiﬁed version of the signal strength scheme will best capture constituent awareness, or lack thereof. Speciﬁcally, while the strength of the signal (as measured by the value or number of projects) is an accurate way to assess the potential for recipient awareness; however, this overlooks other information sources besides direct exposure.

It seems highly probable that if constituents are ever to become aware of earmark projects, then the source of that information is likely to be media coverage rather than direct exposure, especially of projects that are diffused across a large geographic area, such as a state. Media coverage of pork places the smallest information burden on recipients, and facilitates awareness regardless of direct exposure to the projects themselves.

This analysis focuses on the monetary value of earmarks to a particular district or state. Additionally, reliance on dollar value is consistent with recent work on earmarks (Porter & Walsh 2006; Lazarus 2010; Engstrom & Vanberg 2010; Berry et al. 2010; Crespin & Finocchiaro 2008; Rocca & Gordon 2013). Nonetheless, the number of projects was considered as well. The results remain substantively unchanged. Alternative speciﬁcations are shown in the appendix.

78 6.4 Methodology

The analysis reevaluates past investigations of earmark awareness by turning to the 2008 Coop- erative Congressional Election Study (CCES), relying upon newly acquired, highly accurate TCS earmark data. The chapter concludes by reexamining models that found a connection between earmarks and Member support, as well as earmarks and voter support.

6.4.1 Model 1: Project Recall: 2008 CCES

The 2008 CCES is part of an ongoing collaborative effort involving 30 teams from varying universities.8 The 2008 study was conducted in two waves: pre-election (October 2008) and post-election (November 2008). It asked a random national sample of 36,500 respondents approximately 120 questions, amongst which was the following: “[c]an you recall any specific projects that your Members of Congress brought back to your area?” Respondents were permitted to respond yes or no. If they responded yes, they were asked a follow-up question asking if they could recall any specific details about the project(s). Over 5,000 respondents answered in the affirmative (16.6%).

CCES survey data present an excellent opportunity to utilize a well-worded question, a large sample, and highly accurate earmark data (namely, TCS). The use of these data also means that district and state level analysis may be conducted; however, given the wording of the question, it is unclear which Member of Congress the respondents are recalling. For that reason, the

model considers the impact of both state and district outlays: Pr (y(Pro ject Recalli,t)= 1|x)= Φ (α + θi,(t−1) + δi,(t−1) + γ + ε). Where θi,(t−1) and δi,(t−1) represent earmarks from a respondents’ Senators and Representative from the previous ﬁscal year, respectively.9 These measures

8Ansolabehere, Stephen, Cooperative Congressional Election Study, 2008: Common Content. [Computer File] Release 3: April 12, 2011. Cambridge, MA: Harvard University [producer] http://projects.iq.harvard.edu/cces 9As previously mentioned, also considered were models measuring the number of pork projects where data are available (see appendix).

79 are calculated using TCS data in the form of logged dollars secured by the responsible member. A

vector of control variables (γ) is also included.10

6.4.2 Measuring Media Coverage

To capture media exposure I created variables measuring newspaper coverage at the national and state level for each Member of Congress. To get a handle on national exposure, several student coders were directed to search Lexis-Nexis Academic Universe for all articles appearing one year prior to each poll. The articles collected from the searches were then hand-coded to identify mentions of speciﬁc Senators and Representatives.11 Articles were considered a “hit” if it suggested that the relevant legislator was directly responsible for, and received, an earmark or pork project in his or her state (see the appendix for an example of a “hit” and a “miss”). The result of this content analysis is a raw count of the number of articles in which a Member was associated with pork spending (the averaged total amongst the two Senators and a raw count for the individual

Representative). The intercoder reliability was quite high (Krippendorf’s α = .96), suggesting a sound coding scheme.

In order to accurately capture media exposure at the local level additional content analysis was performed. Borrowing from Fridkin and Kenney’s (2011) approach, we compiled a list of the most highly circulated newspapers in each state that were also searchable on Lexis-Nexis or NewsBank: Access World News (see appendix for this list of state newspapers and search term wording).12

Unfortunately, Representatives were rarely mentioned in association with pork projects, even in their home state. Consequently we were only able to generate a variable capturing state media exposure for Senators; however, a national media exposure variable was generated for both

10In order to conserve space, the full models are moved to the appendix. Additionally, the CCES features a sample weight scheme. An alternative analysis was conducted using this weighing. The results appear in the appendix; they remain largely unchanged. 11See appendix for the language used in these searches. 12In some cases, the top circulation newspaper was unavailable. In these instances, following previous research (Fridkin & Kenney 2011), we searched the next available newspaper.

80 Senators and Representatives. Among the national newspapers, each Senator received an average of 1.4 and 2.5 stories covering his/her earmarking activities within the one year prior to the 2006 and 2008 surveys, respectively. Among the state newspapers, each Senator received an average of 3.5 stories.13 The correlation between the national and state newspaper measures is .32 (p <.001), suggesting that coverage of topics of local interest (e.g., behavior of a state’s Senator) may vary considerably between the state level and the national level.

6.5 Results

6.5.1 2008 CCES Survey

In addition to the earmarks and media count measures, the 2008 CCES analysis also included controls for respondent and member characteristics that may affect awareness. This includes respondent media use, news interest, political sophistication, ideology, income, party afﬁliation, age, view of the state of the economy, approval of Congress, and education. Representative controls include ideology, tenure, electoral marginality, membership on the appropriations committee, if the Representative is a chair or ranking minority member, and political party. Finally, Senator controls consisted of average Senator ideology, average tenure, average marginality, number of Senators on the appropriations committee, number of chairs or ranking minority members, and political party. Summary statistics and descriptions of the variables can be found in the Appendix.

Finally, also considered were interactions between respondent and Member ideology and between respondent and Member party.14 These interaction terms were included under the sus- picion that project recall may be affected by projection. In other words, if a constituent shares

13Note that these two measures are not directly comparable, because the national measure combines multiple newspapers, while the state measure uses only a single newspaper. 14Summary statistics and descriptions of the variables can be found in the appendix.

81 their partisanship with their Congressman, that constituent may be more likely to report that MC brought home the bacon. If a signiﬁcant interaction emerges, it suggests that constituent recall is endogenous to the favorability of that member with the constituent.

Table 6.1: Project Recall, CCES 2008 Ability to Recall Speciﬁc Member Projects Coeff (std. err.) Ln(Representative Pork), TCS 0.021** (0.008) Ln(Senators Pork), TCS -0.037 (0.025) National Rep. Media Coverage 0.006 (0.009) National Sen. Media Coverage 0.004 (0.005) State-Level Sen. Media Coverage 0.022*** (0.006) Respondent News Interest 0.272*** (0.018) Respondent Political Sophistication 0.279*** (0.029) Respondent Media Use 0.520*** (0.041) Respondent Ideology 0.007 (0.012) Respondent Democrat -0.073** (0.022) Ideology, Representative (DW-NOMINATE) -0.210 (0.157) Representative Democrat 0.027 (0.111) Average Ideolo., Sen. (DW-NOMINATE) 0.227 (0.189) # of Senate Democrats 0.120 (0.071) Ideolo. x Ideolo., Representative 0.061* (0.027) Democrat x Democrat, Rep. 0.183*** (0.031) Ideolo. x Ideolo., Senators 0.031 (0.031) Democrat x Democrat, Senators 0.003 (0.021)

82 Table 6.1 - continued

Ability to Recall Speciﬁc Member Projects Coeff (std. err.) Additional Controls Moved to Appendix Constant -2.516*** (0.534) Log-Likelihood -11,501 N 27,422 Probit with robust standard errors clustered by Congressional district in parenthesis. ***p < .001; **p < .01; *p < .05; two-tailed.

The findings (Table 6.1) are loosely similar to those of Stein and Bickers (1994; 1995), but differ in important ways. Specifically, greater amounts of pork from a Representative to a district do indeed raise respondent awareness. The same does not apply to pork from Senators, however. This is hardly surprising given that the greater geographic distribution of the benefits for Senators makes attribution difficult. As Figure 6.1 reveals, moving from the minimum to maximum amount of logged earmarked dollars brought home by a Representative increases a respondents’ ability to recall projects by approximately 6 percentage points (from 6.7% to 12.7%). While this is a relatively modest change, it is notably greater than the previous prediction of a 3.4% increase (Stein and Bickers 1995; 132). What is noteworthy here is that while this lends support to previous findings, the effect is limited only to Representatives. Turning to the media variables, while only state media coverage of Senator earmarks is statistically significant, it has a substantial effect. As Figure 6.2 reveals, moving from the minimum number of state newspaper articles (0) to the maximum (28) increases recall 15 percentage points (11% to approximately 26%), ceteris paribus. This is notably larger than the effect earmarks play on recall, suggesting that media may indeed play a strong role.

Given the large sample, many of the controls in the models associated with CCES data proved significant (see appendix for full model specification). Both of the respondent media controls (media use and news interest) were positive and significant. So too were the respondent’s political sophistication, perceived state of the economy,15 education, and support for Congress.

15This may be endogenously caused. Earmark projects, or coverage of those projects, may be the reason a respondent suspects the economy is getter better, but such a connection seems improbable given the low level of self-reported awareness.

83 Not surprisingly, age was a negative predictor of respondent recall ability. In regards to Member controls, only those accounting for Representative tenure and being on the Appropriations Com- mittee were significant (as expected, both were positive). Similarly, the respondent-Member party and ideology interactions were significant only for those involving the respondent’s Representa- tive. Both the ideology and party interactions were positive and significant, suggesting that party congruence is a positive predictor of a respondent claiming to recall earmarks. This suggests that project recall is at least partially projection, and is thus endogenous to candidate favorability. As a result, previous models that include project recall as a predictor of candidate favorability (e.g., Stein and Bickers 1994; 1995) may overstate the relationship between recall and favorability.

Figure 6.1: Project Recall and Representative Pork Dollars (logged)

Overall, the results reafﬁrm past ﬁndings that pork leads to increased citizen awareness of earmarking efforts. However, the results show that increased media coverage also raises self- reported awareness, but only in regards to Senators. This may suggest that the strength of the

84 Figure 6.2: Project Recall and State Newspaper Coverage of Senator Pork pork-awareness connection is attenuated for Representatives, but the measure utilized in this paper is not speciﬁc enough to capture such effects. Finally, the results demonstrate the potential danger in relying on FAADS data, as the results differ dramatically from previous ﬁndings.

The results show strong support for the argument that media coverage of earmarking efforts does indeed increase respondent awareness of their Members’ efforts. As Stein and Bickers (1994, 383) predicted, was that repeated media coverage did enhance voter recognition of earmarks. However, unlike Stein and Bickers (1994, 391), no relationship was found between the party of a Member and an individual’s ability to recall earmarks. However, congruence between respondent and member ideology and party, at least in regards to Representatives, was a positive predictor of awareness. This helps explain the previous “curious ﬁnding” that “Representatives’ constituents in districts represented by Republicans are less likely to be aware of new awards” (Stein and Bickers 1994, 391)—namely, such a relationship may be an artifact of omitted variable bias. What is left open is the question of the effect of earmarks on incumbent favorability. The chapter turns to this question in the next section.

85 6.6 Approval and Support

So far the chapter has examined whether the acquisition of earmarks and media coverage of earmarks affect citizen knowledge, as well as the role of party and ideological congruence. While this is perhaps the strictest test, as awareness of earmarks seems to be a necessary condition for earmarks affecting voting behavior, the work now directly assess whether earmarks increase the probability of voting for the incumbent or translate into greater favorability towards the incumbent.16 Previous results have found a positive connection between earmarks and approval (Stein and Bickers 1994; 1995); Chapter 4 found that previous earmark allocations lead to increased vote share in the next election. I reexamine these ﬁndings with recent, highly accurate data; furthermore, as with previous models, I add controls for media coverage, as well as party and ideological congruence.

6.6.1 Pork and Approval

While pork may not push voters to alter behavior at the ballot box, that does not mean pork is not capable of generating good feelings; indeed, Chapter 5 found strong evidence that knowledge of particularized beneﬁts can increase favorability. What was unanswered by the previous chapter is how successful MCs are at ensuring those good feelings are actualized. Here the analysis again relies on the 2008 CCES, which asked respondents for their approval of each Senator on a ﬁve-point scale. Given there are two Senators per state, CCES asked respondents how they approved of each independently; the models refer to these as approval for Senator A and Senator B. Table 6.2 (below) displays the models predicting incumbent favorability as measured on a 5-point Likert scale.17 As

16One could argue that knowledge of earmarks is not a necessary condition, as voters may update through an online, rather than memory-based, process (Lodge et al. 1995). If this were the case, earmarks could still affect attitudes and behavior, while knowledge would be only weakly correlated with this effect. Additionally, it should be noted that it is only possible to examine voting for incumbents, since only they have had the opportunity to acquire earmarks. The downside of this analysis is that it limits the sample size to only races where an incumbent was up for reelection. 17CCES asked how strongly the respondent agreed with the job their Representative was doing ranging from 1-4, where 4 = strongly approved. Don’t knows were recoded as a median category.

86 the primary independent variable, I rely on the logged TCS measure of earmarks, the most accurate account of pork spending. Additionally, the media coverage variables are included, as well as a vector of controls, which apply to the individual Representative or Senator.18 The ﬁrst column of table 6.2 presents the results of an ordered probit model predicting candidate favorability. The table reveals that neither pork nor media coverage of pork alters respondent evaluations of their Representative. Rather the strongest predictors are party and ideology.19 Finally, it is worth noting that Representative media coverage was negatively signed and approached statistical signiﬁcance.

Table 6.2: Senator and Representative Approval, CCES 2008 Representative Senator A Senator B Variable Coeff. (std. err.) Coeff. (std. err.) Coeff. (std. err.) Ln(Representative Pork) 0.006 (0.004) Ln(Senators Pork) 0.006 -0.003 (0.004) (0.005) National Rep. Media Coverage -0.009 (0.006) National Sen. Media Coverage -0.001 -0.002 (0.003) (0.002) State-Level Sen. Media Coverage -0.007 0.004 (0.006) (0.012) Respondent News Interest 0.047*** 0.008 0.029* (0.008) (0.011) (0.013) Respondent Political Sophistication -0.053** -0.048* -0.059** (0.018) (0.024) (0.018) Respondent Media Use 0.011 -0.043 -0.010 (0.029) (0.027) (0.027) Respondent Ideology 0.004 -0.010 -0.053* (0.010) (0.023) (0.025) Respondent Democrat -0.402*** 0.020 -0.014 (0.019) (0.051) (0.049)

18An alternative approach would be to estimate the effect of self-reported knowledge on favorability. However, this approach is less desirable for two reasons. First, it doesn’t allow a test of whether increased earmarks increases favorability, only a test of whether perceptions of earmarks affect favorability. Second, self-reported knowledge is likely endogenous to favorability. If respondents like their Senator, they may be more likely to report that their Senator has brought home federal money. 19For example, for liberal respondents, the probability of strongly approving of an incumbent rises by 43% when moving from a incongruent to congruent Representative. Similarly, party congruence meant respondents were 15% more likely to strongly approve of their Representative, while incongruent districts had only a 4% chance of strongly approving, certeris paribus.

87 Table 6.2 - continued

Representative Senator A Senator B Variable Coeff. (std. err.) Coeff. (std. err.) Coeff. (std. err.) Ideology, Representative (DW-NOMINATE) -2.040*** (0.121) Representative Democrat 0.073 (0.079) Senator Ideology, A/B -1.841*** -1.816*** (0.387) (0.249) Senator Democrat, A/B -0.138 0.489** (0.097) (0.159) Ideolo. x Ideolo., Representative 0.695*** (0.026) Democrat x Democrat, Rep. 0.795*** (0.028) Ideolo. x Ideolo., Senators A/B 0.603*** 0.589*** (0.105) (0.076) Democrat x Democrat, Senators A/B 0.471*** 0.534*** (0.065) (0.041) Additional Respondent and Member Controls Moved to Appendix Log-Likelihood -39,600 -38,900 -37,800 N 27,735 27,785 27,497 Dependent variable is Member approval on a 5-point Likert scale with 4 being “strongly approve.” Ordered probit with robust standard errors clustered by Congressional district for the Representative model, and by State for the Senator models (in parenthesis). ***p < .001; **p < .01; *p < .05; two-tailed.

Columns two and three show the results of models predicting Senator approval. Again, neither the pork variable nor media controls is a statistically significant predictor of approval. Yet ideology and party are again strong determinants.20 Contrary to previous findings (Stein and Bickers 1994; 1995), the analyses reveal no statistically significant relationship between earmarks and candidate support.21 Nor do the media exposure measures suggest a statistically meaningful connection between exposure to information about pork projects and approval. In other words, MCs face an uphill battle if they wish to garner support from press releases about pork (Gimmer

20Liberal respondents are 17% (Senator A) and 33% (Senator B) more likely to strongly support their Senator when they share ideological disposition. Party had a similar effect, bolstering support by 8% (Senator A) and 20% (Senator B), ceteris paribus. 21Stein and Bickers actually claim that a change in the proportion of new awards is not related to change in incumbent evaluation, however the coefficient on their measure is statistically significant at the 0.1 level (one-tailed). Rather, the authors cite awareness of new projects as the meaningful predictor; as previously discussed, relying on awareness to predict incumbent support generates significant endogeneity concerns.

88 et al. 2012), or media coverage.

The results, however, do ﬁnd that ideological and party congruence are factors strongly inﬂuencing candidate support. In short, it appears that pork has no notable effect on a respondent’s feelings towards their Member in regards to evaluation. While media coverage did affect project recall, it seems likely that the effect was too weak to then translate into greater favorability. What remains to be explored is if this translates into an equally weak connection when it comes to voting.

6.6.2 Earmarks and Support for the Procuring Incumbent

The final analysis asks whether securing particularized benefits results in measurable electoral gains for the responsible incumbent. Additionally, this formulation allows for the expansion of existing literature in three ways. First, the findings here can build upon the work on Stein and Bickers (1994; 1995) by considering the Senate in addition to the House. Second, the analysis will once again inquire into the impact of media coverage of earmarks on incumbent electoral support. Finally, I consider the roles of party and ideology in conjunction with earmarks.

The analysis once again relies upon the 2008 CCES, with the respondent as the unit of analysis.22 Additionally, as in previous models, I include the pork-media exposure variables, ideology and party congruence measures, and a host of controls. Table 6.3 presents the results. Overall, the results reveal no evidence that increasing the amount of pork to a district or state comes close to reaching the conventional levels of statistical significance. Interestingly, the amount of media coverage of pork projects brought home by Representatives was negatively related to likelihood of supporting the incumbent. This has interesting implications for the findings of Chapters 4 and 5. Namely, while Chapter 4 found a positive, significant relationship between earmarks and overall support in the next election, in the incumbent setting, media seems to attenuate this process. This is in keeping with the finding from Chapter 5 that negative framing has a strong attenuating effect

22The CCES over-samples potential voters, which explains the high number of voting respondents. This might bias ﬁndings towards ﬁnding a relationship between pork and vote; that will prove not to be the case, however.

89 on support, especially for projects that target issue areas that are not personally relevant to the recipient.

Table 6.3: Pork and Incumbent Support, CCES 2008 Vote for Incumbent in 2008 Representative Senators Coeff. (std. err.) Coeff. (std. err.) Ln(Representative Pork) 0.002 (0.007) Ln(Senators Pork) -0.163 (0.090) National Rep. Media Coverage -0.026** (0.009) National Sen. Media Coverage -0.008 (0.006) State-Level Sen. Media Coverage 0.011 (0.014) Respondent News Interest -0.059** 0.061*** (0.020) (0.018) Respondent Political Sophistication -0.061 0.126* (0.038) (0.057) Respondent Media Use -0.151* -0.074 (0.059) (0.055) Respondent Ideology -0.114*** -0.011 (0.023) (0.032) Respondent Democrat -0.954*** -0.617*** (0.031) (0.035) Rep. Ideo. (DW-NOMINATE) -5.107*** (0.242) Rep. Democrat -0.023 (0.107) Incumbent Sen. Ideology -2.824*** (0.611) Incumbent Democrat Sen. 0.598 (0.468) Ideolo. x Ideolo., Rep. 1.609*** (0.058) Democrat x Democrat, Rep. 1.905*** (0.043) Ideolo. x Ideolo., Senator 1.090*** (0.045)

90 Table 6.3 - continued

Vote for Incumbent in 2008 Representative Senators Coeff. (std. err.) Coeff. (std. err.) Democrat x Democrat, Senator 1.327*** (0.060) Additional Respondent and Member Controls Moved to Appendix Constant 1.225*** 2.439 (0.205) (1.798) Log-Likelihood -5,509 -5,817 N 16,076 12,401 Probit with robust standard errors clustered by Congressional district for the Representative model, and by State for the Senator models. ***p < .001; **p < .01; *p < .05; two-tailed.

Figure 6.3 displays this finding visually. Holding all other values at their mean or mode where appropriate, moving from no national coverage of an individual Representative securing pork to the maximum (23 articles), the likelihood of voting for that incumbent declines from 35% to 17%. This has two implications: first, contrary to conventional assumptions, pork may actually hurt incumbents. Representatives, unlike Senators, need to satisfy a much smaller population, but this also implies a narrower group of benefit recipients. Moreover, the initial investigations found a high degree of negativity in media coverage of pork.23 Awards particularized to a given district, when viewed through a national lens, are unlikely to be covered favorably. Second, scholars should not assume that voters view earmarks securing by Representatives and Senators as identical in effect. Chapter 5 attests to this conclusion. Past research has been forced to rely on measures that consolidate State and district projects into a single measure. Such specifications may prove inaccurate, especially in regards to pork’s impact on voters.

23The initial investigations into the tone (positive or negative) of media coverage found that approximately 43% of national senator media articles were negative. I expect this effect to be more pronounced for national coverage of Representatives as being consistent with traditional notions of favoring ones’ own Member, while detesting the institution.

91 Figure 6.3: Representative Media Coverage (of Earmarks) on Incumbent Support

6.7 Conclusion

The findings differ significantly from previous attempts to ascertain the effect of earmarks on public opinion and electoral support. First, the results reaffirm previous findings that pork has a direct effect, albeit weakly, on the (self-reported) ability of a recipient to recall projects. Second, as predicted, I find evidence confirming that media coverage of earmarks does indeed affect the recall ability of recipients. Third, I was unable to confirm preexisting research that discovered a connection between earmarks and Member favorability. Finally, findings reveal that pork is not universally positively viewed by the public; increasing national media coverage of a Representative securing pork actually hurts that incumbent’s reelection prospects.

Scholars have long assumed that members of Congress pursue earmarks because citizens reward their Members for bringing home the bacon. This conclusion rests on the assumption that voters are aware of, and will reward, earmarked dollars. However, Members reaping electoral gains from pork face a pair of inherent obstacles: citizens are often unaware of pork, and are unlikely to know whom to credit or blame for earmarked projects. I reassessed the previous scholarly

92 assumptions that citizens are aware of localized beneﬁts by proposing that media coverage of pork, and not the projects themselves, are primarily responsible for citizen awareness.

As this study revealed, earlier work contending that greater allocations of pork lead to a corresponding escalation in awareness of those projects is well founded. Additionally, the unexplored suppositions advanced by Stein and Bickers (1994; 1995) that media, ideology, and party are likely to play a role also proved correct.

Unlike Stein and Bickers (1994; 1995), however, acquisition of earmarks did not mean- ingfully affect member evaluation. I suspect that previous findings to the contrary were due to a combination of model specification problems, questionable measurement (i.e. the reliance on FAADS data), the omission of media and ideological congruence measures, and potential temporal restrictions. Given the low level of self-reported awareness of local pork projects, the absence of a relationship can be seen as consistent with notions that voters possess minimal amounts of specific information about candidates. That is not to say that pork may never affect MC evaluation, merely that aggregate data suggest it is not resulting in a measurable change.

Finally, contrary to previous work, no relationship was found between the amount of pork to a given district and incumbent support. Therefore, I suspect this difference in ﬁndings to be attributable to the same reasons mentioned above. Furthermore, the results suggest that media coverage of these earmarks has a negative effect. In other words, national media coverage of local projects actually hurts House incumbents.

So what does this mean for Members of Congress, and more importantly, for elections? For incumbent Members, there is little reason to think that current earmark efforts will directly reap electoral rewards. What explains these results, and how can they be reconciled with the ﬁndings of Chapter 4? The simplest explanation is that incumbent support and general voter approval are inherently different concepts. Incumbents may reach a saturation point where additional earmarks have diminishing returns. An alternative explanation is that earmarks work to increase reelection chances through a different causal pathway than voter approval. Indeed, many have speculated that earmarks are merely one side of a quid pro quo between legislators and special interests (a topic

93 that is given further treatment in Chapter 7). With the recent availability of accurate earmark data, this remains an interesting and important question to investigate.

As Chapter 5 detailed, and this study suggested, the effect of earmarks may be more nuanced than that of a magic wand to be waived in exchange for electoral support. While recipients may be appreciative of earmarks, negative media coverage may undermine such gains.

94 CHAPTER 7

PORK AND CAMPAIGNING: FINANCIAL BENEFITS OF EARMARKING

Rep. Jeff Flake (R-Ariz.) said the budget system almost demands that lawmakers use earmarks to win support from voters. He said his colleagues also award earmarks with the expectation that they will generate campaign contributions from happy recipients and their lobbyists. 1

Congressional earmarks frequently make their way onto the national stage, typically as an example of government waste and excess. Not surprisingly, earmarks are, on the whole, abhorred by the American public. The notion of individualized projects beneﬁting only a small segment of the nation evokes negative sentiments, especially when the projects are perceived as narrow, localized, or frivolous. For example, Representative Don Young (AK) and late Senator Ted Stevens (AK) both devoted signiﬁcant effort to secure funds to connect Ketchikan, Alaska to its airport via a bridge (the so-called “Bridge to Nowhere”). More recently, Hillary Clinton’s attempts to win the Democratic presidential nomination were hindered by popular criticism of her efforts to build a Woodstock museum in her home state of New York with money secured via earmarks. Therefore, it is not surprising that numerous citizen groups, such as Citizens Against Government Waste (CAGW), Taxpayers for Common Sense (TCS), and OpenSecrets.org (OpenSecrets), have rallied against such projects. Often joining the cause of citizen groups, media outlets tend to lambast pork- barrel spending, as illustrated by the quote above. This raises an obvious question: if earmarks are

1(Robert O’Harrow Jr., The Washington Post, “Earmark Spending Makes a Comeback; Congress Pledged Curbs in 2007” June 13, 2008).

95 eschewed so often in the press, why do members so actively seek them?2 This chapter builds on existing research (Gimmer et al. 2012; Stein & Bickers 1995; Rocca & Gordon 2013) by exploring the reciprocal connection between earmarks and campaign ﬁnancial support.

The common critique of earmarks by activist groups is the contention that earmarks create a quid pro quo relationship between members and special interest groups that fund political campaigns. Until recently, political scientists have done very little to address whether this concern is well founded. Thankfully the literature on earmarks has undergone a renaissance spurred by the availability of new data. Much of this initial work (Stein & Bickers 1994, 1995; Lazarus 2009, 2010; Lazarus & Steigerwalt 2009) has been devoted to the ability of Members to secure projects, and the impact of those projects on electoral security. Bucking this trend, Rocca and Gordon (2013) recently provided the first attempt to disentangle the relationship between interest groups and campaign contributions. In their analysis Rocca and Gordon (2013) looked to a subset of total earmarks (military spending), and investigated one causal possibility in the endogenous relationship: that pork-barrel projects lead to campaign contributions. While extremely promising, their work left unanswered whether the reverse relationship (campaign contributions lead to earmarks) might be be true. This is not surprising given the difficulty in finding a solution to the endogeneity problem. In short, attempting to disentangle the reciprocal relationship between pork and campaign contributions remains a formidable task.

The aim of this chapter is to shed light on the relationship between earmark acquisition by Members of Congress, and contributions to those Members by special interest organizations. Speciﬁcally, this work is concerned with the validity of numerous media claims that suggest a pervasive connection between campaigns and pork. For example, the New York Times attested to a link between military earmarks and a $200,000 donation for Rock Santorum,3 the New Orleans Times- Picayune brought attention to Senator Mary Landrieu for her apparent trade-off of a $2 million

2For example, in ﬁscal year 2008 all but 20 House members sought earmarks, with the average Representative securing over $6 million in earmarked dollars. 3Michael Luo and Mike McIntire. Jan. 15, 2012 “Donors Gave as Santorum Won Earmarks.” New York Times.< http://www.nytimes.com/2012/01/16/us/politics/as-rick-santorum-secured-earmarks-2006-donations-ﬂowed- in.html?pagewanted=all&_r=0 >

96 earmark for a reading program in exchange for $30,000 in campaign contributions,4 and Senator Arlen Specter was forced to defend his earmark acquisitions for the labor and service industries in exchange for $226,000 from labor political action committees in 1994 to the The Philadelphia Inquirer.5 This chapter builds upon the work of Rocca & Gordon (2013) by exploring the inverse causal relationship; namely, that contributions breed earmarks. This is achieved via a unique approach to solving the endogenous question of pork acquisition and campaign contributions: by relying on naturally occurring conditions to disentangle the relationship. Ultimately this chapter finds that, in the aggregate, contributions do indeed prompt MCs to secure more earmarks. These findings have several very important implications. First, this work sheds light on many aggregate questions left open by previous research. Namely, it attempts to continue where recent research has left-off by exploring the inverse effect of earmarks (i.e. contributions breeding earmarks) over several years. Second, it confirms many of the media’s fears: contributions do indeed prompt MCs to seek more pork. Finally, the process of this investigation necessitated the creation of a unique dataset on contributions to representatives over several years, which will be made publicly available.

7.1 The Entangled Campaign Process

In 2007, a plurality of the Supreme Court held that the Bipartisan Campaign Reform Act’s (BCRA) restrictions on electioneering communications drawn from general treasury funds violated the First Amendment, but the Court failed to deliver a clear rationale as to why.6 A split decision three years later helped resolve this ambiguity. In the now renowned Citizens United decision, the

4Bruce Alpert and Bill Walsh. Jan. 9, 2008. “Ethics group targets Landrieu earmark - Reading program donated to campaign” The Times-Picayune. 5Steve Goldstein. Oct. 24, 2007. “Specter, GOP colleague spar over legislative ’earmarks’; South Carolina’s Jim DeMint said Specter steered a no-bid grant to a supportive union.” The Philadelphia Inquirer. 6See Federal Election Commission v. Wisconsin Right to Life 551 U.S. ___ (2007).

97 Court backpedaled in its jurisprudence, opening the doors to communications from independent groups, while sustaining BCRA’s ban on corporations and unions from using their treasury funds for direct political advocacy.7 In the years following the Citizens United decision, commentators and scholars have speculated about the the potential consequences from the recent ruling. The concern, of course, is the pervasive role of money in the political process. While the effects of this recent court decision are still unfolding, the larger concern that “large contributions [...] can corrupt or, at the very least, create the appearance of corruption of federal candidates and ofﬁceholders” has been problem lingering since the the founding, as well as a catalyst for major legislation since the beginning of the 20th century.8

Following infamous examples of political corruption and the rise of political “bosses,” states and the federal government attempted to respond with legislation curtailing the amount of money going to candidates; these restrictions were mostly intensely directed towards the source of such donations. However, these efforts have met with mixed success. At its core is the concern that large special interest donors would be able to purchase candidates favorable to their industries in exchange for enormous sums of money; while limits on the size and type of donations have stymied such direct efforts, there exists a glaring avenue of opportunity for such brinkmanship: earmarks—namely, the potential for Members to serve their donors via particularized projects, many of which beneﬁt the private sector in exchange for contributions, and for donors to respond in kind.

The questions presented in this chapter exist at the intersection of two literatures: interest group support for MCs, and the ability of and incentive for individual MCs to secure pork projects.

From the interest group perspective, the literature has found contrasting evidence: both that interest groups only seek access to MCs who have a direct geographic connection (Wright 1985, 1989), as well as support for MCs with no direct geographic linkage (Grenzke 1988). Nonethe- less, the motivation remains the same: building a coalition of like-minded MCs who are willing

7Citizens United v. Federal Election Commission 558 U.S. 310 (2010). 8Quoting McConnell v. FEC, 540 U.S. 93, 34 (2003).

98 to grant groups access in order to provide information to Members (Wright 1996; Schlozman & Tierney 1983; Tripathi et al. 2002). This, of course, allows interest groups to directly affect policy outcomes. For their cooperation members receive both policy input, and reelection support. This behavior is directed at congressional allies, foes, and undecided Members alike. Allies are targeted to bolster support and provide policy input (Bauer et al. 1963; Hall & Deardorff 2006), while foes and the undecided are targeted to persuade, or at the very least, to mitigate the lobbying of opposition groups (Austen-Smith & Wright 1992; Hojnacki & Kimball 1998). Research regarding whether such efforts actually affect ﬂoor votes is mixed. Numerous scholars have found no connection between contributions and vote behavior (Grenzke 1989; Wright 1985; Wawro 2001), while others have found a relationship with varying degrees of strength (Chappell 1982; Langbein 1993; Saltzman 1987; Frendreis & Waterman 1985). Whether lobbying efforts substantially affect the policy process is outside the scope of this work; it sufﬁces to say that lobbying does exist in various forms, and that one such form is the exchange of money (campaign contributions) in exchange for access to the political apparatus.

Unlike the extensive literature on special interest lobbying, the pork literature is considerably less developed, due largely to a lack of transparency. Substantial work on an individual MC’s ability to secure localized beneﬁts began with Ferejohn’s (1974) seminal work on rivers and harbors legislation. That said, it wasn’t until the 1990s that researchers began to apply quantitative methods to attempt to connect electoral consequences with earmarks. For example, Stein & Bick- ers (1994) provided an initial investigation of earmarks and electoral support, which was followed by their comprehensive book on the topic (1995) and follow-up piece on challenger deterrence (Bickers & Stein 1996). Following the change in House and Senate rules regarding the disclosure of earmark requests, the literature saw new life breathed into it. Jeffery Lazarus (2009; 2010) and Lazarus & Steigerwalt (2009) began a number of investigations into which MCs are more adept at acquiring pork. They found that a Member’s institutional positions (ex. Appropriation Committee members, party leaders, ranking minority members, etc.) were the strongest predictors of individual ability to secure pork projects. Armed with new data, many of the questions thought to have

99 been answered are being revisited. For example, while a number of authors have proposed that voters are directly aware of beneficial pork projects (Levitt & Snyder 1997; Stein & Bickers 1994, 1995; Sellers 1997), this work (Chapter 6), as well as that of a number of other scholars (Gimmer et al. 2012; Samuels 2002; Rocca & Gordon 2013) has expressed doubts, or proposed alternative causal explanations. Given that earmarks may serve a purpose other than direct electoral benefit, what might that purpose be? This work attempts to answer that question by exploring the relationship between earmarks and rewards from non-voting beneficiaries: special interest contributors.

7.2 Pork and Campaigning: Financial Beneﬁts of Earmarking

In recent years scholars have made great advances in the study of legislative earmarks. Greater availability of data, furnished by changes to disclosure requirements for the House and Senate have allowed scholars to peer more deeply into the causal process of how and why MCs are able to bring home federal dollars. One prominent explanation often found in the media, but little explored by researchers, is the potential for MCs to bring home dollars in exchange for campaign support, or, conversely, for campaign contributions to be given in exchange for future projects. Stein & Bickers (1995) were the first to attempt to empirically determine the existence of such a relationship. Stein & Bickers (1995, 99-103) selected a semi-random sample of nine public laws, identified distributive domestic assistance programs contained in those laws, and identified which interest groups testified during committee hearings related to the nine laws. This authors then used factor analysis to cluster the Political Action Committee (PAC) groups. Ultimately, the authors found that PACs that contributed to members, and who also testified at the committee hearings for the law, had a modest to weak effect on altering a Member’s voting behavior.

Stein & Bickers (1995) work is, as the authors admit, a best effort to connect PAC involve- ment to earmark acquisition given the limited data available. Stein & Bickers were unable to rely

100 upon accurate earmark data attributable to each member; moreover, PAC contributions, while collected by the FEC, were not easily searchable until recently. As a result, the authors were unable to attempt to answer the fundamental question of campaign contributions and pork acquisition.

This early attempt was revisited in earnest by the recent work of Rocca & Gordon (2013). These authors gathered data from citizen groups on earmarks for projects related to national defense, as well as earmarks to individual MCs from PACs related to defense. The authors attempt to solve the endogeneity problem by relying on a two-stage least squares (2SLS) approach with the square root of total earmarks received by individual MCs acting as their instrumental variable. Ultimately, the authors ﬁnd that the amount of defense earmarks received by individual members does indeed leader to greater contributions; for example, the authors found that MCs receiving the mean amount of defense earmarks received 35% more in defense contributions versus those securing no defense earmarks (2013, 250).

Rocca & Gordon (2013) possessed much richer data than Stein and Bickers (1995), but faced the same endogeneity problem; ultimately the authors used a statistically robust instrument to remove the endogeneity. That said, there were several imposed assumptions that suggest the need for further analysis. For example, the authors imposed a tacit temporal assumption: that earmark acquisition and campaign contributions occur simultaneously since the authors rely on data from the same time period (i.e. without lags). Additionally, the authors opted to include a number of controls that may also be endogenous (such as military personnel as a percent of the district). That said, Rocca and Gordon have done the discipline a great service in laying the groundwork towards answering a question plagued with reverse causation.

Given the limited exploration of the topic, a number of questions remain; for example, it is unclear whether the reverse causal process exists: that contributions lead to future earmark acquisition. Also, given the richness of contribution data and several years of earmark data, it is possible to move beyond looking to one election cycle or one policy area and explore the relationship of pork and contributions across different congressional sessions.

This chapter attempts to ﬁll the gaps left by scholars regarding earmarks and campaign con-

101 tributions in a unique and innovative way. Speciﬁcally, I wish to consider the possibility of a quid pro quo relationship between campaign contributions and pork-barrel projects. This chapter will explore the role contributions play in altering MC behavior to secure pork. Much like attempting to determine the relationship between earmarks and electoral outcomes, this conceptualization is also plagued with endogeneity. For example, it is possible that earmarks obtained at time t may be an effort to garner campaign contributions from groups to aid in the election at time t +1. However, it is also possible that contributions given to a candidate at time t −1, are what prompted said candid to pursue earmarks (at time t) in the ﬁrst place. This chapter proposes a unique solution to testing the inverse relationship examined by Rocca & Gordon (2013): that contributions foster pork.

There are two rival, equally likely theories regarding earmarks and campaign support. The first proposes that an unknown proportion of earmarks brought home to a district or state is the result of a Member attempting to appease a particular interest, or set of interests, with the understanding that such behavior will result in future rewards in the form of campaign contributions (Projects for Contributions Theory). Alternatively, it is also possible that interest groups may finan- cially assist an electoral bid with the understanding that the Member will use his or her position to secure projects that will be of a benefit to the campaign donor (Contributions for Projects Theory). This reciprocal relationship is portrayed in Figure 7.1. As mentioned, Rocca & Gordon (2013) found strong evidence affirming the projects for contributions theory, but left future researchers to answer the reciprocal theory (contributions for projects). This work answers that call by deriving a testable hypothesis and empirically testing for the presence of such a relationship.

Figure 7.1: The Earmark-Contribution Endogeneity Problem

102 7.2.1 Contributions for Projects Theory: Are Earmarks the Result of Cam-

paign Contributions?

The inverse causal relationship proposes that MCs secure earmarks with the understanding that grateful donors will reward their efforts with increased contributions. This test relies on a naturally occurring difference between Members: incumbency. The proposed solution to the problem of endogeneity between contributions and earmarks relies on looking at MCs who are newly elected to ofﬁce at time t, and their inherent incapability of securing earmarks at time t-1. By relying solely on this sub-sample of the total elected population, it is possible to ascertain whether the amount of donations at time t results in a corresponding increase in the amount of pork at time t+1. For example, Representative McNerney (CA, 11th) was ﬁrst elected in 2006, consequently he was unable to have previously secured pork projects in exchange for contributions. To re- state this as a hypothesis: H2: When a freshman MC is awarded with campaign contributions at time t − 1, we should see a corresponding rise in earmarks at time t. To state this empirically:

Y (Λ(Porki,t)) = α + β1 (η(Contributionsi,t−1)) + controls, where η(Contributionsi,t−1) is equal to the contributions given at time t −1, and where Λ(porki,t) is equal to the amount of pork secured by a Member of congress iff he or she is a freshmen.

There does exist the possibility that donors contribute to campaigns working under the assumption that should the candidate prove victorious he or she will repay the donations with earmarks, but this would be fraught with uncertainty. Because of the fact that the Member is a freshman, the ability of a newly elected MC to secure pork is unknowable.9 It is impossible for contributors to know how the freshman member will behave when victorious; this includes committee assignments, which issues areas the Member will deem important, and whether the Member will be a workhorse or show horse.

There is one possible source of information on potential future pork efforts for special

9Given this uncertainty donors would need to give to all challengers in the hopes of securing earmarks; this would be both exceedingly expensive, and, even if true, increases the likelihood of a Type II error (i.e. a null ﬁnding), but not a Type I error.

103 interest donors: previous legislative service in a state assembly. It may be possible that freshmen MCs who have served in state legislatures have secured earmarks in the past that beneﬁted special interest groups. If true, donor groups would have to assume that the newly elected MC would continue such activities if victorious; however, such a possibility would be fraught with uncertainty, since, as mentioned above, it is impossible to know the future behavior or resources of a newly elected Member. Nonetheless, a separate model will be estimated looking solely at Freshmen MCs who have not previously served as a state representative or senator.10

7.3 Data and Methods

Given the proposed causal process, the unit of analysis must be the individual freshmen MC; therefore, the two variables of interest are contributions from special interest groups (the dependent variable), and earmarks secured by individual MCs (the key independent variable). As this section will explain, the former of the two necessitated the construction of a unique dataset.

As discussed at length in Chapter 3, measuring earmarks has been made markedly easier thanks to changes in House and Senate disclosure rules, as well as the efforts of several citizen groups (namely, Taxpayers for Common Sense, OpenSecrets.org, and Citizens Against Govern- ment Waste). In fact, thanks to the efforts of these groups, earmark data are readily available for free over the internet. That said, accurate earmark data are temporally limited by the changes in disclosure. Earmark data exists for fiscal years 2008, 2009, and 2010.11 Fiscal years span from October 1st of the previous calendar year to September 30th of the year for which the fiscal year is named (ex. fiscal year 2008 spans from October 1, 2007 to September 30, 2008). This translates to earmark data spanning the 110th Congress (2007-2009), and the 111th Congress (2009-2011).

10Special thanks to Carol Weissert and Daniel Milton for their efforts in collecting information on previous legislative service. 11Fiscal year 2011 data do exist, but recently all but a handful of House Republicans have taken a pledge not to rely on earmarks. While this will make for several interesting future research projects related to MCs foregoing earmarks, it means the 2011 data are not representative of a normal legislative session.

104 Thanks to federal disclosure laws, campaign contribution data are also readily available via the FEC’s website. The FEC divides contributions into five categories: itemized individual contributions, unitemized individual contributions, party committees contributions, other committees contributions, and candidate contributions. For the vast majority of candidates, their primary sources of contributions are individual (itemized and unitemized) and other committee contributions (i.e. interest group contributions). As discussed, this work is concerned solely with the latter (interest group contributions). These data are searchable beginning in 1996, but detailed contribution records (including itemized reports of which special interest group was responsible) are not available until 2004. Early records often omit the unique committee identification code assigned by the FEC to each interest group, or attribute the donation to the correct group only sporadi- cally. Moreover, while the data are readily available on the FEC website, there is no easy way to access large quantities of data, rather it was necessary to conduct unique searches by member by year. Thankfully, Florida State University’s Reach Intensive Baccalaureate Certificate (RIBC) program provided a student data collector to assist in this process.12 The result was over 2,400 unique searches with a combined database of over 800,000 contributions. This database includes all interest group contributions to MCs from 2004 to 2012. Following publication of this work, the database will be made available publicly.

Given individual freshmen MCs are the unit of analysis, a limited number controls accounting for institutional position are needed.13 Controls are added for an MC’s ideology (relying on Poole and Rosenthal’s (2001) DW-NOMINATE score),14 a dummy indicating whether the Member serves on the appropriation committee (1 if yes, 0 otherwise), and a dummy indicating whether the member was elected from a marginal district (relying on Jacobson’s (2009) work, a member was considered from a marginal district if their last electoral victory was 60% or less). Additionally, given the times series cross-sectional (TSCS) nature of the data, a year dummy was included.

12Thanks to Joshua Gendal for his efforts in collecting the data. 13This is, of course, due to the fact that freshmen MCs are not party leaders, committee chairs, and the like. 14Given the small sample, ideology was selected over a party control variable because of model ﬁt. Party and ideology were correlated at .94, so only ideology was used.

105 7.4 Results

The results of the empirical test assessing the contributions for earmarks theory are featured below. Once again, individual MCs are the unit of analysis; however, only freshmen from the 110th and 111th Congresses are considered. The dependent variable is the sum of earmarks obtained by individual members (logged) for the first year of the 110th and 111th Congresses (fiscal years 2008 and 2010). The key independent variable is sum of contributions from the previous fiscal year (i.e. 2005-2006 contributions and 2007-2008 contributions); this ensures a 1-year lag between when contributions ended and pork acquisition began.15 Additionally, a number of previously discussed controls were included.

As table 7.1 (left column) reveals, contributions from the previous fiscal year are a statistically significant, positive predictor of earmark procurement; in short, hypothesis two is confirmed. While the coefficient on contributions is small, the average member secured approximately $460,000.16 Additionally, a number of the controls proved significant. Liberal freshmen were much more likely to to secure earmarks; unsurprisingly members that were appointed a position on an appropriations committee also secured more pork, and finally, those from marginal districts also brought home more earmarks, no doubt in an effort to reduce electoral uncertainty.

As discussed in section 7.2.1, there exists one source of information regarding earmarks for special interest donors to base the contributions: earmarks secured while serving in a state legislature. Such a realization is unlikely since it would force PACs to assume that any past behavior could replicated in a totally different environment (i.e. Congress); a check for validity is nonetheless needed. Table 7.1 (right column) reconsiders the analysis featured in the left column for freshmen MCs that have not previously served as a state senator or representative. As the results

15Recall that fiscal years extend until September 30th of the listed year. For example, fiscal year 2006 ended September 30, 2006, which was paired with earmarks from fiscal year 2008, which began October 1, 2007, thus achieving a 1-year lag. 16Total contributions for freshmen was normally distributed (skewness of .146; kurtosis of 2.53) and did require need transformation, however, the measure of individual earmarks was not. Consequently, those values were logged to achieve greater normality.

106 Table 7.1: The Effect of Contributions on Freshmen Representatives Securing Earmarks All Freshmen Freshmen Without Prev State Leg Exp Variable Coeff. (std. err.) Coeff. (std. err.)

Total Contributions(t−1) (in $100,000) 0.283* 0.328* (0.125) (0.136) Ideology (DW-Score) -4.665** -3.409* (1.470) (1.420) Appropriations Comm. 1.152+ - (0.572) - Marginal District 2.543** 3.038** (0.901) (0.859) 110th Congress Dummy 1.849* 0.869 (0.892) (0.895) Constant 8.771*** 9.534*** (1.270) (1.408) R2 0.293 0.323 N 112 76 Dependent variable: Individual MC earmarks (logged). OLS regression. Robust standard errors clustered by State. ***p < .001; **p < .01; *p < .05; two-tailed. show, despite a loss of 36 Members (32%), the effect of total contributions on pork procurements remains unaffected. In fact, the effect for this sub-subsample of freshmen MCs we see a stronger effect from contributions.17

The results of table 7.1 (left) are graphically displayed in Figure 7.2 below. The top of the two figures displays the marginal effect of moving from minimal contributions for freshmen Members in the 110th Congress ($0), to the maximum ($1.15 million); the points at the bottom of the figure display the distribution of contributions for 110th freshmen (also known as a rug). As the figure reveals, moving from minimum to maximum contributions resulted in a change of 10.3 logged earmark dollars (approximately $30,000) to 13.6 logged dollars (approximately $775,000), ceteris paribus.18 Figure 7.2 (bottom) reveals the results for table 7.1 for the 111th Congress. Mov-

17As mentioned by Lazarus (2009) and discussed in footnote 4, there exists the possibility of censorship at 0. However, in the context of freshmen Members, the number of $0 pork allocations is extremely low. Nonetheless, a tobit analysis of table 7.1 was conducted in the appendix. The results remain signiﬁcant, and the coefﬁcients are comparable. 18Holding all continuous variables at their mean, and dichotomous variables at their mode.

107 ing from minimum contributions ($0) to the maximum ($1 million) results in a change in earmark acquisition of 8.45 logged dollars (approximately $4,700) to 11.7 logged dollars (approximately $122,000), ceteris paribus.19

Figure 7.2: Marginal Effect of Contributions on Earmark Acquisition, 110th and 111th Congresses

19Again, holding all other values at their mean or mode, where appropriate.

108 7.5 Conclusion

The aim of this chapter was to shed light on the relationship between earmark acquisition by Mem- bers of Congress, and contributions to those Members by special interest organizations. Speciﬁ- cally, to investigate claims commonly made by media sources that there exists a relationship between members securing earmarks and contributions from special interest groups. The fear often expressed by commentators is that a secretive quid pro quo relationship exists amongst donors and Members. Until recently there was no way to accurately investigate such a contention, other than anecdotal accounts, because of the limited availability of earmark data that could be attributed to Members. Thanks to changes in institutional disclosure rules, it is now possible to investigate such claims.

While the data are now available, there remained a glaring problem: disentangling the endogenous relationship been campaign contributions and earmark procurements. Recently Rocca & Gordon (2013) took a ﬁrst step at attempting to disentangle this relationship in the defense spending context by relying on a statistical workaround. This work continues from their recently laid groundwork. Unlike Rocca & Gordon (2013), however, this chapter explored the inverse relationship by turning to naturally occurring conditions that solved the problem of reverse causation. Specially, this work explored the question left open by Rocca & Gordon (2013) of whether contributions affect earmarks by looking at a subset of the House: freshmen Members.

Ultimately, this chapter makes two contributions to the study of earmarks and elections. First, a unique dataset has been built from FEC data that catalogs all contributions to Members of the U.S. House from 2004 to 2012. This database is easy to use, searchable, and will be made available publicly to scholars. Second, my investigation was able to ﬁnd a means to successfully test the relationship between contributions and pork. This investigation found past contributions led to greater earmarks in the future. In other words, popular concerns appear to be warranted. Contri- butions to freshmen Members resulted in those Members securing more earmarks as compared to

109 their counterparts receiving less special interest funding.

This is by no means a definitive end to questions surrounding earmarks and contributions. A number of questions remain. First, while the solution was imperfect, scholars have yet to find a means to uncover a relationship between aggregate Member pork and campaign contributions. Perhaps general analysis of this type will prove difficult or impossible; issue area analysis like that of Rocca & Gordon (2013) may be best vehicle for continued research. Second, in fiscal year 2011 all but a handful of House Republicans have opted not to seek earmarks. It remains unclear if this course of action has affected public opinion or contributions. Finally, while highly accurate earmark data (via TCS) only extends over three fiscal years, other citizen groups (such as Citizens Against Government Waste) have state-level data extending to the mid-1990s. This may make possible a similar analysis to be conducted in the context of the Senate.

110 CHAPTER 8

CONCLUSION

The purpose of this work was to explore the role of earmarks in several facets of American politics, including which Members of the House and Senate secure earmarks, whether citizens find such projects desirable, how the public becomes informed of such efforts, and whether earmarks facilitate Member contributions from special interest groups, or vice versa. The findings of this work fall principally into two categories: those concerning the effect of earmarks on recipients (i.e. constituents), and those regarding the ability of MCs to secure pork to facilitative a political need or desire. Consequently, this conclusion will proceed by briefly discussing the theory surrounding why pork is tolerated, and what purpose it serves, followed by a consideration of each aspect of earmarks’ impact on the political process: serving the institutional needs of members Members, and serving as a means to alter electoral outcomes. Finally, this chapter will conclude by discussing the current state of research on earmarks, and where scholars should look to advance our understanding.

Before accurate measurements of earmark outlays were possible, it was assumed that earmarks only beneﬁted select members of Congress. This prompted many scholars to conclude that only particular policy arenas would reap the beneﬁts of Member efforts since the distribution of policy focuses by Members was not uniform. For example, Stein & Bickers (1995, 141) claimed that a oft assumed “myth” about pork was that it was widely distributed; rather, the authors contended that “[a]n overwhelming majority of federal assistance programs are distributed to a relatively small number of congressional districts.” Others stressed the importance of institutional

111 position (Ferejohn 1974) or partisanship (Levitt & Snyder 1997). Analysis of earmark data from TCS reveals this is not the case. Indeed, recent scholarly work, including the evidence presented in Chapter 4, conﬁrms “universalism[:] [e]veryone gets a share of the logroll and an opportunity to credit claim (Engstrom & Vanberg 2010, 981).

Given the ubiquitous nature of earmarks in Congress, the obvious questions that follows become why do Members desire such projects, and, given there is a benefit, why does the majority permit the minority to secure pork? In answer to the first question, as Chapter 2 proposed, Members seek pork to minimize perceived electoral risk by acquiring federal outlays. This has the potential to serve three purposes: deterring would-be challengers by showing institutional strength, serving as a simple means for MCs to demonstrate to their constituents that they are workhorses that diligently serve the voters, and, perhaps most controversially, to cultivate a relationship with special interest groups that may benefit from the allocations. This work attempted to explain many of these proposed causal stories.

Regarding the second question, concerning the majority’s tolerance of minority pork; as many scholars have pointed out (Balla et al. 2002; Engstrom & Vanberg 2010; Levitt & Snyder 1997; Crespin & Finocchiaro 2008) members of the majority have little choice but to accept similar behavior from the minority. To do otherwise would open the majority party to criticism as wasteful. Moreover, majority party MCs with long-term legislative goals will quickly realize they may not always be the majority party in their chamber; building good relations at present establish a norm of reciprocity.

8.1 Earmarks and Institutions

The ﬁrst chapter of this work to empirically analyze earmarks attempted to answer a fundamental question proposed by Ferejohn (1974, 235): “what is it in Congress that leads to the [...] distribu-

112 tions of expenditures” that we observe in current political landscape? Chapter 4 expands on recent works (Lazarus 2009, 2010; Lazarus & Steigerwalt 2009; Engstrom & Vanberg 2010) that have explored earmark distributions in the House and the Senate by looking at the distribution of earmarks over several fiscal years. As the results revealed, variables predicting earmark acquisition appear to be stable over time, even across Congresses. For the House, Members in positions of power (e.g. members of the appropriations committee, party leaders, ranking minority committee members, etc.) proved most successful at securing earmarks. On the other hand, contrary to the findings of both Lazarus & Steigerwalt (2009) and Engstrom & Vanberg (2010), earmark acquisition in the Senate appears much more volatile. In fact, membership to the Appropriations Committee proved to be the only consistent predictor of acquisition in the Senate. Regarding pork procurement, two major conclusions can be drawn: first, not surprisingly, the House and the Senate are distinctly different when it comes to Representatives and Senators securing earmarks; second, while pork is ubiquitous, some MCs are better at securing earmarks than others. The question then becomes, given this variation, do voters actually desire and respond to these outlays, or might earmarks provide a benefit to members other than direct electoral support? These questions were the focus of Chapters 5, 6, and 7.

8.2 Earmarks and Electoral Outcomes

After introducing earmarks, and providing theoretical expectations about who is most adept at securing them, this work explored earmark’s role as a means to affect constituent opinion. This begin with Chapter 5 asking a rather simple question: do constituents appreciate these particularized beneﬁts? Scholars and Members of Congress alike have long operated under the assumption that earmarked projects would garner appreciation from constituents; this despite numerous polls re- vealing that earmarks do not enjoy popular support. This raises a simple, unexplored question: do constituents appreciate pork projects that beneﬁt them? While such a connection seems intuitive, it

113 has never been empirically veriﬁed. As Chapter 5 revealed, the answer to question of appreciation is nuanced. Constituents do indeed reward MCs with increased support if the projects beneﬁt an issue area for which the constituent is a member of the issue public. This holds true even when knowledge of the outlays is framed as wasteful spending. However, when rewards are given to issues that are not especially important to the recipient, this boon to support disappears. In short, the conclusion of the chapter is this: MCs looking to rely upon pork as a means to supplement electoral support must carefully target the projects to the desires of their district or state. It is not enough for MCs to bring home federal dollars and hope the projects rally support around the Member. As Chapter 6 proves, most citizens remain unaware of aggregate earmarks, therefore Members must be strategic with their procurement efforts if they wish to see results.

In lieu of Chapter 5’s affirmation that earmarks can affect MC support, Chapter 6 ventured to serve two related purposes: first, to build upon the work of Gimmer et al.’s (2012) exciting findings pertaining to Member press releases and pork; second, to address the next logical question following the analysis of Chapter 5: given citizen opinions can be affected by pork, how might this process take place? This was achieved by proposing that earmarks and public opinion are not connected via a direct causal pathway, rather, the media functions as in intermediary between Member efforts and constituent knowledge. Ultimately Chapter 6 confirmed media’s role; media coverage of earmarks was indeed a statistically significant predictor of citizen awareness of pork projects. This chapter confirms the need for “legislators [to] regularly announce small expenditures from bureaucratic agencies [as] evidence that they value the opportunity to frequently claim credit,” since such stories are made with media outlets in mind (Gimmer et al. 2012, 15). While the effect of directly enlightening constituents may be small, Members are nonetheless incentivized to seek any aid to electoral support.

Finally, Chapter 7 considered if pork might be serving an alternative purpose other than directly altering constituent opinion. Speciﬁcally, the ﬁnal chapter investigated what many citizen groups call the sinister side of pork: the relationship between earmarks and special interest campaign support. Citizen groups and media sources have long claimed that earmarks may be

114 brought home to benefit special interest groups, or, conversely, the PACs may be giving to campaigns in exchange for future project dollars. While media outlets have provided several anecdotal examples, there has been but one attempt to investigate if either relationship exists. Namely, the recently work of Rocca & Gordon (2013) that investigated one direction of this endogenous relationship (whether earmarks lead to PAC contributions) for a subsample of total earmarks (defense earmarks). The authors found that it did. Building upon these findings, Chapter 7 provided an innovative solution to testing the inverse relationship. By looking solely at freshmen members of Congress (who by their nature are unable to have secured pork in the past), the chapter found that past contributions do indeed lead to greater earmark acquisitions in the future. In short, Chapter 7 prompts two conclusions: first, earmarks have an indirect route, via PAC contributions, that affects the political process; second, the fears of many pundits appear to be well founded. Just as Rocca & Gordon (2013) discovered, earmarks and campaign support are indeed intertwined.

8.3 The Future of Research Regarding Pork Politics

While this work explored many facets of earmark allocations, there remain a number of interesting questions that are yet to be explored. A discussion of avenues of research left open is worthy of brief consideration.

First, while there has been much research on the effect of pork on recipients, as well as modest work on the role of earmarks and special interest, there has been no effort investigate if earmarks serve to scare-off would-be challengers. Much work has been done on strategic nature of incumbents (Black 1972; Jacobson & Kernell 1983; Jacobson 1989; Krasno & Green 1988; Squire 1995; Bianco 1984; Stone & Maestas 2004), but scholars have yet to investigate earmarks as a strategic resource for challenger deterrence. While the average voter is unlikely to be aware of earmarks, challengers are well versed in the actions of their (potential) incumbent opponents. The

115 difﬁculty for future scholarly work lies in disentangling the role of institutional position; namely, that House Members in positions of power secure more earmarks.

Second, Chapter 5 left open a rather obvious but unexplored question: whether MCs actively pursue earmarks that beneﬁt the largest issue publics in their state or district. For example, should we expect that MCs from military districts to secure more military outlays than their counterparts from non-military districts? The answer would presumably be yes, but committee access may greatly inﬂuence such efforts.

Finally, beginning in ﬁscal year 2011 all but a handful of House Republicans took a pledge to forgo earmarks, while their Democrat counterparts continued their earmark efforts as per usual. This provides a unique chance to investigate changes in MC behavior. Namely, this would allow scholars to detect if the disappearance of pork to a district previously receiving earmarks resulted in measurable changes, such as declines in contributions from PACs, altered awareness earmarks for citizens in these districts, or had a measurable impact on electoral success.

Scholars have just begun to scratch the surface when it comes to earmarks. Regardless of one’s opinion, earmarks are likely to continue for quite some time. It must be the goal of future work to continue to investigate how federal outlays are shaping the face of the political landscape.

116 Appendix A

SUPPORTING MATERIAL TO CHAPTER 4

Ability of Freshmen to Secure Pork

The table below is referenced in footnote 10 in section 4.3. Speciﬁcally, it shows the empirical results of table 4.1 with an added variable indicating whether the MC is a freshmen (1 if yes, 0 otherwise). Not surprisingly, being a freshmen MC results in signiﬁcantly less earmark dollars.

Table A.1: Freshmen Representatives Securing Earmarks, Fiscal Years 2008-2010 Tobit Variable Coeff (std err) Freshman Rep. -1.532* (0.649) Democrat 2.880*** (0.477) Appropriations Comm. 2.481*** (0.447) Majority Cardinal 0.714+ (0.364) Minority Cardinal 0.149 (0.791) Party Leader 0.152 (1.295) Committee Chair -0.389 (0.735) Ranking Min. Member 2.261** (0.798) Tenure 0.038 (0.024)

117 Table A.1 - continued

Tobit Variable Coeff (std err) Marginal 1.096* (0.484) FY2009 Dummy -1.513*** (0.262) FY2010 Dummy -2.018*** (0.339) Constant 11.865*** (0.521) σ 4.915 (0.258) pseudo-R2 0.024 N 1,307 Tobit regression’s censor point at 0 with the dependent variable as logged earmark dollars Two-tailed standard errors, clustered by Congressional district. ***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10

Consideration of a Two-Step Heckman Model When Prediction Earmark Al- locations

The table below features the alternative casual process mentioned in Chapter 4 for the House and Senate contexts; namely, the possibility that an alternative process exists for explaining why some MCs have $0 in earmarks. If Members actually have $0 in earmarks because of a data generating process that differs from MCs that secured pork, a modiﬁed version of the tobit model would be more appropriate. Namely, a two-step hurdle model (in this instance a Heckman two-step process) would more accurately explain the acquisition process while accounting for the correlated errors between the two steps of the process. The table below considers alternative estimations of tables 4.1 and 4.2.

First, as the tables above reveal, for both the Representative (bottom table, left column) and

Senate (top table, right column) estimates, the Inverse Mills ratio (λ) is insigniﬁcant in the Senate context, and on the border of conventional levels in the Representative context. This suggests that

118 Table A.2: Heckman Two-Step Estimation of Earmark Acquisition Representatives Senators Representatives Senators Variable Coeff. (std. err.) Coeff. (std. err.) Coeff. (std. err.) Coeff. (std. err.) Dependent Variable: Secured Earmarks, Step One Dependent Variable: Ln(Earmark Dollars), Step Two Democrat 0.992*** -0.597+ 0.327* -0.313 (0.180) (0.338) (0.143) (0.579) Appropriations Comm. 0.554 1.729 1.571*** 1.685+ (0.827) (14.769) (0.160) (0.896) Majority Cardinal 0.542 0.048 0.258 0.719 (1.933) (13.887) (0.178) (0.438) Minority Cardinal 0.070 3.908 0.273 0.595 (0.239) (6.335) (0.195) (0.612) Party Leader -0.155 0.230 0.951** 0.671+ (0.940) (0.423) (0.364) (0.384) Committee Chair -0.429 0.394 0.276 0.860+ (1.114) (0.307) (0.190) (0.500)

119 Ranking Min. Member 0.561 0.202 0.542* 0.319 (1.245) (1.176) (0.273) (0.446) Tenure 0.016 -0.005 0.021** -0.000 (0.010) (0.012) (0.008) (0.015) Marginal 0.198 0.014 0.385** 0.114 (0.172) (0.175) (0.122) (0.238) FY2009 Dummy -0.653*** -0.232+ 0.020 0.035 (0.103) (0.121) (0.083) (0.152) FY2010 Dummy -0.836*** -0.484* -0.365** -0.596* (0.143) (0.222) (0.122) (0.274) Constant 1.092*** 1.154** 14.157*** 15.117*** (0.178) (0.351) (0.187) (0.577) IV Mills (λ) 0.971+ 0.556 (0.556) (1.543) ρ 0.879 0.419 σ 1.104 1.325 N 1,307 296 1,307 296 Heckman Model; 137 (left column) and 55 (right column) censored (at 0) observations. Bootstrapped standard errors clustered by district (left column) and state (right)***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10 for both models, and especially in the Senate context, the errors between the two steps are no different from 0, reducing the concern that selection may be present. Second, as the second step (bottom table) reveals, the same coefficients remain significant (except for Majority Cardinal in the Representative context, which is borderline significant). In short, there is only modest evidence of a selection process at work here, and when a hurdle model is used, we see similar results.

Application of the 2SLS Approach in Regards to Earmarks and Member Sup- port

The table that follows is referenced in section 4.3. It is the result of a reconsideration of table 4.1 with the inclusion of state population as an independent variable predicting pork (both logged value and number of projects) where the units of analysis are individual House MCs. The purpose of the specification is to determine if state population is an exogenous predictor of earmark allocations. As the table below reveals, in both the dollar value and the count specifications, state population is a significant predictor of earmark acquisition.

Table A.3: State Population and Representatives Securing Earmarks, Fiscal Years 2008-2010 Tobit Negative Binomial Variable Coeff. (std. err.) Coeff (std.err) State Population (in millions) 0.0262+ 0.008*** (0.000) (0.002) Democrat 2.756*** 0.353*** (0.474) (0.069) Appropriations Comm. 2.519*** 1.151*** (0.441) (0.074) Majority Cardinal 0.822* 0.143+ (0.368) (0.087) Minority Cardinal 0.252 0.141 (0.788) (0.112) Party Leader 0.213 0.629*** (1.291) (0.191) Committee Chair -0.495 0.134 (0.758) (0.123)

120 Table A.3 - continued

Tobit Negative Binomial Variable Coeff. (std. err.) Coeff (std.err) Ranking Min. Member 2.098** 0.366*** (0.780) (0.105) Tenure 0.050* 0.013** (0.023) (0.005) Marginal 0.833+ 0.377*** (0.447) (0.072) FY2009 Dummy -1.312*** -0.043 (0.264) (0.031) FY2010 Dummy -2.025*** -0.287*** (0.341) (0.044) Constant 11.236*** 1.355*** (0.545) (0.084) σ 4.918*** (0.257) α -0.802*** (0.094) pseudo-R2 0.02936 N 1,307 1,307 Tobit regression’s censor point at 0 with the dependent variable as logged earmark dollars Negative Binomial dependent variable is count of individual earmarks. Both two-tailed standard errors, clustered by Congressional district. ***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10

Following the application of a 2SLS model researchers should always verify there was a problem with endogeneity, and that the instrument employed was an adequate solution. Beginning with the endogeneity test, a Durbin-Wu-Hausman (DWH) test was conducted. Under DWH’s null hypothesis, the endogenous regressor is assumed to be exogenous; the test reveals a strong rejection of that null: F(1,434)= 23.73 (p = 0.000). In other words, we can conclude that the relationship between earmarks and vote share in the next election is endogenous (Cameron & Trivedi 2010, 190). Next it is necessary to test whether the instrument is weak. While simple correlations provide some insight, a joint F statistic provides more insight. The rule of thumb is that an F statistic of 10 or greater is preferred. The adjusted F statistic here was 6.295, (p < 0.013), which is a bit on the weak side, suggesting further tests may be needed (Cameron & Trivedi 2010, 198), consequently I turned to a user-written STATA command (condivreg) (Mikusheva & Poi

121 2006). Their program features three tests of the instrument: the Conditional Likelihood Ratio (CLR) test, the Anderson-Rubin test, and the Lagrange Multiplier (LM) test. All three statistics give a conﬁdence set of 0.026, 0.326, with a p value of (0.0001). This is only slightly wider than the 95% conﬁdence interval around the porkt−1estimate (0.002, 0.133), which suggests there is not need to correct for a weak instrument (Cameron & Trivedi 2010, 203).

122 Appendix B

SUPPORTING MATERIAL TO CHAPTER 5

Number of Previous Experiments

One potential concern with using student samples is the possibility that reliance on student pools may draw subjects into repeated exposures of different frames over multiple experiments.1 Con- sequently, a robustness check was conducted to ascertain if the number of previous experiments affected the presented ﬁndings. Below is a reproduced version of table 5.2 (Study 2) with the inclusion of a control for the number of previous experiments taken by a subject. As table B.1 reveals, the number of prior experiments is not statistically signiﬁcant.

Table B.1: The Effect of Pork Treatment on Support for Rep. Miller (R, FL-1st) Variable Coeff. (std. err.) General Frame (Positive) 0.138 (0.131) Education Frame (Positive) 0.856*** (0.126) Military Frame (Positive) 0.063 (0.130) Party ID (Dem) -0.082* (0.034) Military Afﬁliation -0.254+ (0.143) Mil. Afﬁliation x Mil. Frame 1.096*** (0.291) # of Previous Experiments 0.010 (0.068)

1While this presupposes powerful duration effects, the possibility is explored below nonetheless.

123 Table B.1 - continued

Variable Coeff. (std. err.) Constant 4.369*** (0.147) R2 0.162 N 417 ***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10. Two-tailed standard errors in parenthesis.

The table below is a reproduced version of tables 5.1 and 5.3 (studies 1 and 3, respectively) featuring the inclusion of a control variable accounting for the number of previous political science experiments taken by each subject. Again, the number of experiments variable is not statistically signiﬁcant, thus mitigating concerns regarding the sample pool.

Table B.2: Number of Subject Surveys and the Effect of Pork Treatments on Support for Sen. Bill Nelson (D-FL) Study 1 Study 3 Coeff. (std. err.) Coeff. (std. err.) General (Positive Frame) Pork Treatment 0.537** (0.204) Education (Positive Frame) Pork Treatment 1.100*** (0.206) General (Negative Frame) Pork Treatment -0.153 (0.138) Education (Negative Frame) Pork Treatment 0.503*** (0.133) Party ID (Dem) 0.101 0.208*** (0.066) (0.041) # of Previous Experiments 0.022 -0.082 (0.096) (0.074) Constant 3.947*** 3.631*** (0.256) (0.159) R2 0.108 0.096 N 249 495 ***p < 0.001, **p < 0.01, *p < 0.05, +p < 0.10. Two-tailed standard errors in parenthesis.

124 Disaggregated Dependent Variables

The three tables that follow below disaggregate the dependent variable (support for Senator Nelson or Representative Jeff Miller) into its constituent parts: a question presented in the form of a 7- point likert scale measuring Member favorability, and a question presented in the form of a 7-point likert scale measuring Member job approval. As the tables reveal, the results are consistent for both of the constituent parts of the dependent variable, albeit slightly weaker for the job approval rating versus the favorability measure.

Table B.3: Disaggregated Dependent Variables, Study 1 (Summer 2011, Sen. Nelson) Summer 2011 Summer 2011 Summer 2011 Original Favorability Job Approval General Pork Treatment 0.533** 0.511* 0.555** (0.203) (0.223) (0.208) Education Pork Treatment 1.115*** 1.134*** 1.097*** (0.205) (0.225) (0.210) Party ID (Dem) 0.100 0.109 0.091 (0.066) (0.072) (0.067) Constant 3.959*** 3.981*** 3.938*** (0.249) (0.273) (0.255) R2 0.111 0.097 0.103 N 250 250 250 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

Table B.4: Disaggregated Dependent Variables, Study 2 (Fall 2011, Rep. Jeff Miller) Fall 2011 Fall 2011 Fall 2011 Original Favorability Job Approval General Pork Treatment 0.123 0.097 0.150 (0.129) (0.162) (0.124) Education Pork Treatment 0.835*** 1.106*** 0.564*** (0.123) (0.155) (0.119) Military Pork Treatment 0.057 0.111 0.004 (0.128) (0.161) (0.123) Party ID (Dem) -0.077* -0.101* -0.054 (0.033) (0.042) (0.032)

125 Table B.4 - continued

Fall 2011 Fall 2011 Fall 2011 Original Favorability Job Approval Military Afﬁliation -0.220 -0.335 -0.105 (0.140) (0.176) (0.135) Mil. Afﬁliation x Mil. Frame 0.978*** 1.042** 0.914*** (0.283) (0.356) (0.273) Constant 4.352*** 4.441*** 4.263*** (0.140) (0.177) (0.136) R2 0.150 0.158 0.090 N 438 438 438 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

National and State Newspaper Searches and the Tone of Earmarks in the Media

As mentioned in section 5.2.2, a casual analysis revealed that approximately 43% of newspaper articles were negative in tone. This subsection details that analysis further.

Newspaper articles were gathered related to another research project in order to generate national and local earmark media measures. For the national measure I consulted Lexis-Nexus Academic Universe, restricting the analysis only to newspapers in the United States. For local newspapers I consulted the state newspaper with the highest circulation, however, in some cases the top circulating newspaper was unavailable. In these instances, following previous research (Fridkin & Kenney 2011), I searched the next available newspaper. The list of state newspapers is featured in the table below.

126 Table B.5: List of Utilized Local Newspapers State News Paper List State Birmingham News AL Anchorage Daily News AK Arizona Daily Star AZ Arkansas Democrat-Gazette AR Los Angeles Times CA Denver Post CO Hartford Courant CT Delaware News Journal, The Journal DE St. Petersburg Times FL Atlanta Journal-Constitution GA Honolulu Star-Advertiser HI Idaho Statesman ID Chicago Sun Times IL Indianapolis Star IN Cedar Rapids Gazette; Iowa Falls Times Citizen IA Wichita Eagle KS Lexington Herald-Leader KY New Orleans Times-Picayune LA Bangor Daily News ME Baltimore Sun MD Metro Boston MA Detroit News MI Minneapolis Star Tribune MN Jackson Clarion Ledger MS Kansas City Star MO Billings Gazette MT Lincoln Journal Star NE Las Vegas Review Journal NV New Hampshire Union Leader NH Newark Star-Ledger NJ Albuquerque Journal NM New York Times NY Charlotte Observer NC Bismarck Tribune ND The Plain Dealer OH Oklahoman OK Oregonian, The OR Philadelphia Inquirer PA Providence Journal RI Charleston Post and Courier SC

127 Table B.5 - continued

State News Paper List State Aberdeen American News SD Memphis Commercial Appeal TN Houston Chronicle TX The Salt Lake Tribune UT The Manchester Journal VT Richmond Times-Dispatch VA The Seattle Times WA The Charleston Gazette Daily Mail WV Milwaukee Journal Sentinel WI The Cheyenne Tribune-Eagle WY Note, state newspapers with the largest circulation were the ﬁrst choice for analysis, however, when that paper was not available via Lexis Nexus, or World News Bank, we turned the next highest.

Newspaper articles were then coded using a scheme to ensure Members were actually linked to speciﬁc projects. Mentions of a Member and the key words “pork” or “earmark” or “project” were crosschecked. Here is the example “hit” suggesting a connection between a Mem- ber and an earmark project:

In a statement released the day the appropriations bill passed Congress, Rep. Allen Boyd (D-Fla.) heralded his earmark for $1.5 million to continue development work on a non-gasoline-burning outboard engine for the Navy Special Operation Forces’ underwater systems [emphasis added].

Here is an example of a “miss:”

Rep. George Miller, a Democrat who chairs the House Education and Labor Commit- tee and is leading the House effort for a service bill, is known as one tough legislative strategist. But he is positively tender when he describes visiting Habitat for Humanity projects, meeting with Teach for America volunteers or spending time with church groups that have provided relief in natural disasters. "It’s one of the great rewarding things in politics [emphasis added].

The clear distinction being between the use of the term earmark or pork in the proper context, and the proper attribution to the correct Member being searched.

128 Armed with these search results, articles were hand-coded into two groups: articles that mention the pork dollars in a positive light (1), articles that frame the mentioning of pork dollars in a negative context (0).

Political Information Index and Student Sample Concerns

As discussed in section 5.4, a concern with use of student samples, as pointed out by Druckman and Kam (2011), is the finding that college students tend to possess more political knowledge than the general population. Consequently, the Summer 2011 experiment included an index meant to measure respondent political sophistication. This was accomplished by first telling respondents “[n]ow we have a set of questions concerning various public figures. We want to see how much information about them gets out to the public from television, newspapers, and the like.” Next we asked respondents a series of four multiple choice questions: 1) The first name is Joe Biden. What job or political office does he now hold? 2) What about John Roberts? What job or political office is held by John Roberts? 3) What about David Cameron? What job or political office is held by David Cameron? 4) What about John Boehner? What job or political office is held by John Boehner? We then offered respondents 5 possible answers in a multiple-choice format, including a “don’t know” option.

Armed with these four questions, a sophistication index (ranging from 0-4, with a mean of 2.15) of was generated indicating the number of questions the respondent answered correctly. This variable was added to the model featured in subsection 5.4.2. The results of this model are shown below.

As table B.6 reveals, political sophistication was not a statistically signiﬁcant predictor of a change in support for Senator Nelson.

129 Table B.6: Political Sophistication and Positively Framed Pork Treatments on Support for Sen. Bill Nelson (D-FL), Summer 2011 Coeff. (std. err.) General (Positive) Pork Treatment 0.522* (0.203) Education (Particularized Positive) Pork Treatment 1.105*** (0.204) Party ID (Dem) 0.105 (0.066) Sophistication Index 0.107 (0.067) Constant 3.716*** (0.291) R2 0.120 N 250 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

Details on the Lexis-Nexus Search Regarding Media Coverage of Earmarks.

In order to ascertain the degree of media coverage of earmarks, as well as address the concern that there may have been a spike or lull in coverage of earmarks during the any of the experiments, an ancillary search was conducted via Lexis-Nexus. The search was conducted across all major world publications in the United States. It consisted of the following search parameters: “(pork or earmark or project) w/p (rep. or representative or congress* or sen. or senator),” in an attempt to capture articles that include members of Congress being associated with earmarks. Obviously such a measure is not precise, but it gives a generalization of the media landscape. As the ﬁgure below (left) reveals, the discussion of pork projects are relatively stable over time. As mentioned in the main text, the average was 250 hits per month, with a standard deviation of 41.

Similarly, readers may be concerned that alternations in media coverage of earmarks in the weeks leading up to the experiments may better explain the findings. The figure below (right) suggests this is not the case. The difference between weeks averaged 11 articles per week, with the first three experiments seeing a slight rise, and the Spring 2013 experiment seeing a decline.2 The

2Note that the Spring 2013 experiment was used solely in the appendix.

130 standard deviation across the four experiments was 1.8 hits. It is exceedingly unlikely that such minor changes would undermine the robust ﬁndings presented.

Treatment Effects for Florida Voters Versus Nonresidents

The results below distinguish registered Florida voters from non-Floridians, as determined by self- reported voter registration zip codes. Unfortunately, this information was only collected in the Spring 2012 experiment. As table B.7 reveals, while support for Senator Nelson is similar for those in the control group (as indicated by the intercept), the negatively framed education treatment is no longer statistically significant for non-residents. Fortunately, the vast majority of student subjects were residents. The introduction of non-residents works to downplay the significance of the initial findings. In other words, the merging of residents and non-residents works to increase the likelihood of a null finding (i.e. increasing the likelihood of a Type II error, but not Type I). This attests to the robustness of the findings.

Table B.7: Disaggregated Dependent Variables, Spring 2012 (Sen. Bill Nelson) Spring 2012 Spring 2012 Spring 2012 Sen. Nelson Sen. Nelson Sen. Nelson Original FL Residents Nonresidents General Pork Treatment -0.167 -0.172 -0.073 (0.136) (0.149) (0.332) Education Pork Treatment 0.467*** 0.592*** 0.158 (0.131) (0.146) (0.303) Party ID (Dem) 0.205*** 0.228*** 0.079 (0.040) (0.043) (0.106) Constant 3.609*** 3.554*** 3.868*** (0.151) (0.165) (0.367) R2 0.091 0.120 0.014 N 507 412 95 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

131 Figure B.1: Lexis-Nexus Media Search on Earmarks and Congress

132 The Effect of the Surrounding Political Debate on Proper Usage of the Budget

One potential area of concern with any experiment that delves into an investigation of a topical political issue is whether increased media attention has created an inﬂated concern for the issue being studied. In this case, the rise of the Tea Party’s role in politics. Bearing in mind this concern, a question asking subjects their feelings towards the Tea Party (measured on a 7-point Likert scale from “Very unfavorable” to “Very favorable”) was included in the Summer 2011 experiment. The results (see table B.8 below) reveal, tea party support is not a statistically signiﬁcant predictor of support for Nelson (column 1), nor does it appear to moderate the effect of the pork treatment (column 2).

Table B.8: Favorability of the Tea Party on Support for Senator Nelson (FL), Summer 2011 Summer 2011 Summer 2011 Sen. Nelson Sen. Nelson General (Positive) Pork Treatment 0.526* 0.330 (0.203) (0.455) Education (Positive) Pork Treatment 1.111*** 1.250* (0.205) (0.484) Party ID (Dem) 0.069 0.065 (0.072) (0.073) Tea Party Favorability -0.054 -0.065 (0.054) (0.086) Tea Party x General Treatment 0.057 (0.117) Tea Party x Education Treatment -0.039 (0.123) Constant 4.246*** 4.296*** (0.377) (0.463) R2 0.115 0.117 N 250 250 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

133 The Exclusion of the Party Identiﬁcation Control Variable

A final issue of minor concern is the inclusion of party identification as a control variable. As discussed in 5.4.2, the omission of a control variable is mitigated by the random assignment inherent in the experimental design. Nonetheless, inclusion of such variables helps prevent against chance assignment of too many people of a particular group to a given condition. Moreover, it can aid in gaining more precise estimates (Franklin 1991). The results below reconsider the models found in the main text with the party identification controls omitted. As the table below reveals, omission of the the party ID variable does not substantive alter the findings.

Table B.9: Exclusion of the Party Identification Control Variable, All Models Fall 2011 Summer 2011 Spring 2012 Rep. Miller Sen. Nelson Sen. Nelson General Frame (Positive) 0.154 0.523* (0.126) (0.204) Education Frame (Positive) 0.849*** 1.085*** (0.121) (0.204) Military Frame (Positive) 0.082 (0.126) Military Affiliation -0.199 (0.137) Mil. Affiliation x Mil. Frame 0.992*** (0.281) General Frame (Negative) -0.158 (0.140) Education Frame (Negative) 0.476*** (0.134) Constant 4.094*** 4.274*** 4.220*** (0.089) (0.138) (0.095) R2 0.137 0.103 0.044 N 450 250 507 ***p < 0.001, **p < 0.01, *p < 0.05. Regression estimates. Two-tailed standard errors in parenthesis.

134 Appendix C

SUPPORTING MATERIAL TO CHAPTER 6

Examples of a “Hit” and “Miss” in the Search for Media Exposure of Ear- marks

Here is the example hit given to coders:

Here is an example of a “miss” given to guide coders:

The clear distinction being between the use of the term earmark or pork in the proper context, and attributed to the correct Member.

Descriptive Statistics

Below are two tables that contain descriptive information about the dependent and independent variables.

135 Table C.1: Variable Information Utilized in the Analysis of the 2008 CCES Variable Name Obs. Mean Min Max Description Project Recall 32579 0.17 0 1 The dependent variable asking if respondents recall any projects their member has secured. Ln(District Pork, TCS) 32747 16.45 0 18.99 Logged TCS pork dollars by district Ln(Senators Pork), TCS 32747 37.64 18.68 41.80 Sum of total pork dollars to a state from both Senators (logged) Representative 32747 0.63 0 23 Total number of national newspaper articles Media Coverage attributing respondent’s Rep. to a pork project National Senator 32747 3.42 0 28 Average national media count (Senator A&B/2) . Media Coverage one year prior to the survey. State Senator 32747 2.44 0 28 Average state media count (Senator A&B/2) Media Coverage one year prior to the survey. Respondent News 32140 4.80 1.5 5.5 Average of interest in news and politics Interest (1=not very interested, 4= very interested) Respondent Pol. 32458 0.68 0 1 Knowledge of which party controls the House Sophistication and Senate (average of these two questions) Respondent 32747 0.58 0 1 In the last 24-hours did respondent read a blog, Media Use watch the news, read a newspaper, listen to radio (average of the four measures; i.e. .25= 1 of the 4) Respondent Ideolo. 32747 3.20 1 5 Respondent Ideology (1=very liberal, 5=very conservative) Respondent Income 30717 8.20 1 14 Respondent Income (1=(<$10k), 14=(>$150k)) Respondent Democrat 32747 0.02 -1 1 Respondent Party (1=Democracy, 0=Independent, -1=Republican) Respondent Age 32457 51.95 18 81 Respondent age Respondent 32723 0.53 0 4 Respondent view of the changing national View of Economy economy (0=much worse, 4=much better) Respondent Cong. 30502 1.72 1 4 Respondent view of Congress (1=strongly Approval disapprove, 4=strongly approve) Respondent Educ. 32747 3.37 1 6 Respondent Education (1=No High School degree, 6=post-graduate degree) Representative Ideology 32747 0.06 -0.74 0.99 Rep. DW-NOMINATE score, 1st Dimension Representative Tenure 32747 10.71 1 52 Rep. tenure in years Representative Marginal 32747 0.32 0 1 Rep. from marginal district (last victory <60%) Representative Approp. 32747 0.15 0 1 Rep. on Appropriations Committee Representative Chair/RMM 32747 0.09 0 1 Rep. is a Committee Chair or Ranking Min. Mem. Representative Democrat 32747 0.49 0 1 Rep. is a Democrat Senator Ideology 32348 -0.06 -0.60 0.84 Average Senator ideology (DW-NOMINATE score) Senator Tenure 32747 13.15 4 35 Average Senator tenure (SenatorA+B/2) Marginal State, Senators 32747 1.19 0 2 Number of Senators from marginal states (last victory <60%) Appropriations, Senators 32747 0.54 0 1 Number of Senators on the Appropriations Committee 136 Table C.1 - continued

Variable Name Obs. Mean Min Max Description Chair/RMM, Senators 32747 0.58 0 2 Number of Senators that are committee chairs or ranking minority members. Democrat, Sen. 32747 0.09 -1 1 Average Senator Party (Senator A+B/2, where -1=Republican, 0=Independent, 1=Democrat) Ideolo x Ideolo., Rep. 32747 0.26 -3.69 4.99 Interaction of respondent and Rep. ideology Democrat x Democrat, Rep.. 32747 0.06 -1 1 Interaction of respondent and Rep. party Ideolo x Ideolo., Sen. 32348 -0.15 -2.98 4.18 Interaction of respondent and Senator ideology Democrat x Democrat, Sen. 32747 0.061 -1 1 Interaction of respondent and Senator party

Table C.2: Variable Information Utilized in the Analysis of the 2006 CBS/NYT Poll Variable Name Obs. Mean Min Max Description Project Awareness 983 0.32 0 1 The dependent variable asking if respondents if their member sponsored pork legislation Ln(Pork, CAGW) 980 19.17 16.74 20.41 Logged CAGW pork dollars National Media 980 1.46 0 13.5 Average national media count (Senator A&B/2) one year prior to the survey. Respondent Controls Respondent Pol. Sophis. 983 0.47 0 1 Correctly identiﬁed who is in line for the presidency after the Vice President. Respondent Ideolo. 983 2.13 1 3 Respondent Ideology (1=very liberal, 2=moderate, 3=conservative) Respondent Income 931 3.81 1 6 Respondent Income (1=(<$15k), 6=(>$100k)) Respondent Republican 983 1.96 1 3 Respondent Party (1=Democracy, 2=Independent, -3=Republican) Respondent Age 956 52.02 18 93 Respondent age Respondent View of Econ. 983 2.59 1 4 Respondent view of the national economy (1=very bad, 4=very good) Respondent Cong. Approval 896 0.25 0 1 Respondent view of Congress (1=approve of the job Congress is doing, 0=do not approve) Respondent Educ. 976 3.31 1 5 Respondent Education (1=No High School degree, 5=post-graduate degree) Representative-Level Controls Ideology 980 0.17 -0.72 1.18 Rep. DW-NOMINATE score, 1st Dimension Tenure 980 10.42 0 50 Representative tenure Marginal district 980 0.32 0 1 Last election victory margin <60% Appropriations Committee 980 0.16 0 1 Rep. on Appropriations Committee Chair/Ranking Min. Mem. 980 0.09 0 1 Rep. is a Chair or Ranking Min. Member Republican 980 0.59 0 1 Representative is a Republican (1=Repub., 0=Dem.)

137 Table C.2 - continued

Variable Name Obs. Mean Min Max Description Senate-Level Controls Ideology 980 0.00 -0.55 0.83 Average Senator ideology (DW-NOMINATE score) Tenure 980 11.64 2 34 Average Senator tenure (SenatorA+B/2) Marginal State 980 1.19 0 2 Number of Senators from marginal states (last victory <60%) Appropriations 980 0.62 0 2 Number of Senators on the Appropriations Committee Chair/RMM 980 0.55 0 2 Number of Senators that are committee chairs. or ranking minority members Republican Senators 980 1.03 0 2 Number of Senators that are Republicans Respondent-Senate & Respondent-Representative Interactions Ideolo x Ideolo., Rep. 980 0.42 -1.94 3.06 Interaction of respondent and Rep. ideology Republican x Repub. Rep. 980 1.19 0 3 Interaction of respondent and Representative party Ideolo x Ideolo., Sen. 980 0.04 -1.63 2.48 Interaction of respondent and Senator ideology Republican x Repub, Sen. 980 2.08 0 6 Interaction of respondent and Senator party

Alternative Speciﬁcation: CCES 2008, Number of Projects.

Below is the alternative speciﬁcation of the analysis provided in table 6.1. Here we rely on the number of projects, rather than the logged dollar amount (this alternative speciﬁcation was mentioned in the discussion of data accuracy). The table reveals the results are substantively the same as those presented in the main text.

Table C.3: The Number of Earmark Projects on Predicting Project Recall Ln(Representative Pork) 0.007*** (0.001) Ln(Senators Pork) -0.000 (0.000) National Rep. Media Coverage -0.003 (0.009) National Sen. Media Coverage 0.006* (0.003) State-Level Sen. Media Coverage 0.020*** (0.006)

138 Table C.3 - continued

Respondent News Interest 0.273*** (0.018) Respondent Political Sophistication 0.280*** (0.029) Respondent Media Use 0.524*** (0.041) Respondent Controls Respondent Ideology 0.007 (0.012) Respondent Income 0.004 (0.003) Respondent Democrat -0.074*** (0.022) Respondent Age -0.010*** (0.001) Respondent View of Economy 0.075*** (0.015) Respondent Cong. Approval 0.064*** (0.015) Respondent Education 0.062*** (0.007) Representative Controls Ideology (DW-NOMINATE) -0.222 (0.156) Tenure 0.005 (0.003) Marginal District 0.030 (0.039) Appropriations -0.020 (0.056) Chair or Ranking Min. Mem. -0.000 (0.069) Democrat -0.019 (0.112) Senator Controls Average Sen. Ideology 0.167 (0.186) Average Sen. Tenure 0.003 (0.004) # of Marginal Sen. -0.056 (0.031)

139 Table C.3 - continued

# of Sen. Appropriations 0.036 (0.038) # Sen. Chair or RMM -0.015 (0.030) # of Sen. Democrats 0.097 (0.070) Sen./Rep, & Respondent Interactions Ideolo. x Ideolo., Representative 0.060* (0.027) Democrat x Democrat, Rep. 0.187*** (0.031) Ideolo. x Ideolo., Senators 0.035 (0.031) Democrat x Democrat, Senators 0.001 (0.020) Constant -3.012*** (0.170) Log-Likelihood -11480 N 27,422 Dependent variable is the ability of respondent to recall if their member had done something special on their behalf. Probit with robust standard errors clustered by Congressional district in parenthesis. ***p < .001;**p < .01; *p < .05; two-tailed.

National and State Newspaper Searches

With the help of student coders, I gathered newspaper articles in order to generate the national and local earmark measures. For the national measure we consulted Lexis-Nexus Academic Universe, restricting the analysis only to newspapers in the United States. For local newspapers we consulted the state newspaper with the highest circulation, where available. The list of state newspapers is featured in the table below.

140 Table C.4: Local News Papers State News Paper List State Birmingham News AL Anchorage Daily News AK Arizona Daily Star AZ Arkansas Democrat-Gazette AR Los Angeles Times CA Denver Post CO Hartford Courant CT Delaware News Journal, The Journal DE St. Petersburg Times FL Atlanta Journal-Constitution GA Honolulu Star-Advertiser HI Idaho Statesman ID Chicago Sun Times IL Indianapolis Star IN Cedar Rapids Gazette; Iowa Falls Times Citizen IA Wichita Eagle KS Lexington Herald-Leader KY New Orleans Times-Picayune LA Bangor Daily News ME Baltimore Sun MD Metro Boston MA Detroit News MI Minneapolis Star Tribune MN Jackson Clarion Ledger MS Kansas City Star MO Billings Gazette MT Lincoln Journal Star NE Las Vegas Review Journal NV New Hampshire Union Leader NH Newark Star-Ledger NJ Albuquerque Journal NM New York Times NY Charlotte Observer NC Bismarck Tribune ND The Plain Dealer OH Oklahoman OK Oregonian, The OR Philadelphia Inquirer PA Providence Journal RI Charleston Post and Courier SC

141 Table C.4 - continued

Examples of a “Hit” and “Miss” in the Search for Media Exposure of Ear- marks

Here is the example hit given to coders:

Here is an example of a “miss” given to guide coders:

The clear distinction being between the use of the term earmark or pork in the proper context, and attributed to the correct Member.

142 Table 6.1 Full Model Speciﬁcations

The table below features the full results from table 6.1 in the text. The dependent variable is the ability of the respondent to recall special projects their Member has procured taken from the 2008 CCES.

Table C.5: Pork Dollars and the Ability to Recall Projects, 2008 CCES Full Model Full Model With Speciﬁcation CCES Weights Ln(Representative Pork) 0.021** 0.021*** (0.008) (0.005) Ln(Senators Pork) -0.037 -0.037* (0.025) (0.016) National Rep. Media Coverage 0.006 0.006 (0.009) (0.005) National Sen. Media Coverage 0.004 0.004 (0.005) (0.003) State-Level Sen. Media Coverage 0.022*** 0.022*** (0.006) (0.003) Respondent News Interest 0.272*** 0.272*** (0.018) (0.022) Respondent Political Sophistication 0.279*** 0.279*** (0.029) (0.036) Respondent Media Use 0.520*** 0.520*** (0.041) (0.043) Respondent Controls Respondent Ideology 0.007 0.007 (0.012) (0.012) Respondent Income 0.004 0.004 (0.003) (0.003) Respondent Democrat -0.073** -0.073*** (0.022) (0.018) Respondent Age -0.010*** -0.010*** (0.001) (0.001) Respondent View of Economy 0.074*** 0.074*** (0.015) (0.018) Respondent Cong. Approval 0.065*** 0.065*** (0.016) (0.017)

143 Table C.5 - continued

Full Model Full Model With Speciﬁcation CCES Weights Respondent Education 0.062*** 0.062*** (0.007) (0.009) Representative Controls Ideology (DW-NOMINATE) -0.210 -0.210* (0.157) (0.095) Tenure 0.005* 0.005** (0.003) (0.002) Marginal District 0.036 0.036 (0.040) (0.028) Appropriations 0.121* 0.121*** (0.050) (0.034) Chair or Ranking Min. Mem. -0.021 -0.021 (0.071) (0.034) Democrat 0.027 0.027 (0.111) (0.044) Senator Controls Average Sen. Ideology 0.227 0.227* (0.189) (0.099) Average Sen. Tenure 0.003 0.003 (0.004) (0.002) # of Marginal Sen. -0.064 -0.064*** (0.038) (0.019) # of Sen. Appropriations 0.025 0.025 (0.038) (0.022) # Sen. Chair or RMM -0.015 -0.015 (0.031) (0.016) # of Sen. Democrats 0.120 0.120*** (0.071) (0.029) Sen./Rep, & Respondent Interactions Ideolo. x Ideolo., Representative 0.061* 0.061* (0.027) (0.025) Democrat x Democrat, Rep. 0.183*** 0.183*** (0.031) (0.028) Ideolo. x Ideolo., Senators 0.031 0.031 (0.031) (0.025) Democrat x Democrat, Senators 0.003 0.003 (0.021) (0.018)

144 Table C.5 - continued

Full Model Full Model With Speciﬁcation CCES Weights Constant -2.516*** -2.516*** (0.534) (0.325) Log-Likelihood -11,5001 N 27,422 24,422 Probit with robust standard errors clustered by Congressional district (in parenthesis, full model). Survey weights in weighted model (right column). ***p < .001; **p < .01; *p < .05; two-tailed.

Table 6.2 Full Model Speciﬁcations

The table below is the full speciﬁcation of table 6.2. The dependent variable is a 5-point Likert scale measuring approval of the incumbent Representative and individual Senators.

Table C.6: Incumbent Approval and Earmark Allocations, CCES 2008 Representative Senator A Senator B Ln(Representative Pork) 0.007 (0.004) Ln(Senators A/B Pork) 0.006 -0.003 (0.004) (0.005) National Rep. Media Coverage -0.009 (0.006) National Sen. Media Coverage -0.001 -0.002 (0.003) (0.002) State-Level Sen. Media Coverage -0.007 0.004 (0.006) (0.012) Respondent News Interest 0.047*** 0.008 0.029* (0.008) (0.011) (0.013) Respondent Political Sophistication -0.053** -0.048* -0.059** (0.018) (0.024) (0.018) Respondent Media Use 0.011 -0.043 -0.010 (0.029) (0.027) (0.027) Respondent Controls Respondent Ideology 0.004 -0.010 -0.053* (0.010) (0.023) (0.025) Respondent Income -0.001 0.006* -0.001 (0.002) (0.002) (0.002)

145 Table C.6 - continued

Representative Senator A Senator B Respondent Democrat -0.402*** 0.020 -0.014 (0.019) (0.051) (0.049) Respondent Age -0.007*** -0.006*** -0.005*** (0.001) (0.001) (0.001) Respondent View of Economy 0.055*** 0.003 -0.026 (0.015) (0.039) (0.047) Respondent Cong. Approval 0.218*** 0.363*** 0.362*** (0.016) (0.049) (0.058) Respondent Education -0.013* 0.003 0.004 (0.005) (0.006) (0.005) Representative Controls Ideology (DW-NOMINATE) -2.040*** (0.121) Tenure -0.000 (0.002) Marginal District -0.069* (0.028) Appropriations 0.078* (0.032) Chair or Ranking Min. Mem. 0.064 (0.048) Democrat 0.073 (0.079) Senator A and B Controls Sen. Ideology, A/B -1.841*** -1.816*** (0.387) (0.249) Sen. Tenure, A/B 0.005 -0.003 (0.003) (0.003) Marginal Sen., A/B -0.162* -0.237*** (0.064) (0.057) Appropriations, A/B 0.112 -0.105 (0.060) (0.189) Chair or RMM, A/B 0.021 -0.186*** (0.064) (0.053) Democrat, A/B -0.138 0.489** (0.097) (0.159) Sen./Rep, & Respondent Interactions Ideolo. x Ideolo., Representative 0.695*** (0.026) Democrat x Democrat, Rep. 0.795*** (0.028)

146 Table C.6 - continued

Representative Senator A Senator B Ideolo. x Ideolo., Senators A/B 0.603*** 0.589*** (0.105) (0.076) Democrat x Democrat, Senators A/B 0.471*** 0.534*** (0.065) (0.041) Cut 1 -0.741*** -0.414** -0.917*** (0.121) (0.147) (0.165) Cut 2 -0.185 0.225 -0.331* (0.121) (0.142) (0.164) Cut 3 0.501*** 0.691*** 0.253 (0.121) (0.150) (0.152) Cut 4 1.484*** 1.837*** 1.296*** (0.122) (0.152) (0.154) Log-Likelihood -39600 -38900 -37800 N 27,735 27,785 27,497 Ordered probit with robust standard errors clustered by Congressional district in parenthesis.***p < .001; **p < .01; *p < .05; two-tailed.

Table 6.3 Full Model Speciﬁcations

The table below is the full speciﬁcation of table 6.3 in the text. The dependent variable is whether the respondent voted for the incumbent Representative and Senator.

Table C.7: Incumbent Vote, CCES 2008 Representative Senators Ln(Representative Pork) 0.002 (0.007) Ln(Incumbent Senator Pork) -0.163 (0.090) National Rep. Media Coverage -0.026** (0.009) National Sen. Media Coverage -0.008 (0.006) State-Level Sen. Media Coverage 0.011 (0.014)

147 Table C.7 - continued

Representative Senators Respondent News Interest -0.059** 0.061*** (0.020) (0.018) Respondent Controls Respondent Political Sophistication -0.061 0.126* (0.038) (0.057) Respondent Media Use -0.151* -0.074 (0.059) (0.055) Respondent Ideology -0.114*** -0.011 (0.023) (0.032) Respondent Income -0.001 0.013* (0.005) (0.006) Respondent Democrat -0.954*** -0.617*** (0.031) (0.035) Respondent Age -0.002* -0.007*** (0.001) (0.001) Respondent View of Economy -0.001 0.048 (0.031) (0.065) Respondent Cong. Approval 0.022 -0.040 (0.030) (0.071) Respondent Education -0.007 0.020 (0.010) (0.011) Incumbent Representative Controls Incumbent Ideology (DW-NOMINATE) -5.107*** (0.242) Incumbent Tenure 0.005 (0.003) Incumbent Marginal District -0.150*** (0.042) Incumbent Appropriations -0.052 (0.051) Incumbent Chair or Ranking Min. Mem. -0.044 (0.074) Incumbent Democrat -0.023 (0.107) Incumbent Senator Controls Incumbent Ideology -2.824*** (0.611) Incumbent Tenure -0.022 (0.020)

148 Table C.7 - continued

Representative Senators Incumbent Marginal Sen. -0.044 (0.136) Incumbent Appropriations 0.207 (0.111) Incumbent Chair or RMM 0.337 (0.235) Incumbent Democrat 0.598 (0.468) Incumbent Senator/Representative & Respondent Interactions Ideolo. x Ideolo., Incum. Rep. 1.609*** (0.058) Democrat x Incumbent Democrat, Rep. 1.905*** (0.043) Ideolo. x Ideolo., Incumbent Senator 1.090*** (0.045) Democrat x Democrat, Incum. Senator 1.327*** (0.060) Constant 1.225*** 2.439 (0.205) (1.798) Log-Likelihood -5509.22 -5817.63 N 16,076 12,401 Probit with robust standard errors clustered by Congressional district for the Representative model, and by State for the Senator models (in parenthesis). ***p < .001; **p < .01; *p < .05; two-tailed.

149 Appendix D

SUPPORTING MATERIAL TO CHAPTER 7

Total Contributions for Freshmen: Tobit Analysis

The table below is a reconsideration of the model presented in table 7.1 where the OLS regression has been changed to a Tobit, where the dependent variable (logged earmark dollars) has a censor point at 0 (see 4 for a full explanation as to why a tobit may be needed). As the table reveals, the results undergo no meaningful change. This suggests that while censorship may be present, it is not an issue of major concern given this subsample. For the purposes of simplicity an OLS will sufﬁce.

Table D.1: Tobit Analysis of the Effect of Contributions on Freshmen Representatives Securing Earmark All Freshmen Freshmen Without Prev State Leg Exp Variable Coeff. (std. err.) Coeff. (std err)

Total Contributions(t−1) in $100,000 0.339* 0.361* (0.144) (0.153) Ideology (DW-Score) -5.516** -3.760* (1.803) (1.594) Appropriations Comm. 1.047 - (0.656) - Marginal District 2.974** 3.308** (1.103) (0.975) 110th Congress Dummy 2.092 0.923 (1.058) (0.976) Constant 5.284*** 3.963*** (0.634) (0.654) pseudo-R2 0.058 0.068

150 Table D.1 - continued

All Freshmen Freshmen Without Prev State Leg Exp Variable Coeff. (std. err.) Coeff. (std err) σ 5.284 3.963 (0.634) (0.654) N 112 76 Tobit regression’s censor point at 0 with the dependent variable as logged earmark dollars. Robust standard errors clustered by State. ***p < .001; **p < .01; *p < .05; two-tailed.

151 BIBLIOGRAPHY

Althaus, Scott. 1998. Information Effects in Collective Preferences. American Political Science Review, 92(September), 545–558.

Althaus, Scott, Swigger, Nathaniel, Chernykh, Svitlana, Hendry, David, Wals, Sergio, & Tiwald, Christopher. 2011. Assumed Transmission in Political Science: A Call for Bringing Description Back In. Journal of Politics., 73(4), 1065–1080.

Alvarez, Michael, & Saving, Jason L. 1997. Deﬁcits, Democrats, and Distributive Beneﬁts: Con- gressional Elections and the Pork Barrel in the 1980s. Political Research Quarterly, 50(4), 809–831.

Ansolabehere, S., & Gerber, A. 1994. The Mismeasure of Campaign Spending: Evidence from the 1990 House Elections. Journal of Politics, 56(4), 1106–1118.

Ansolabehere, Stephen, & Iyengar, Shanto. 1995. Going Negative: How Political Advertisements Shrink and Polarize the Electorate. New York: Free Press.

Arceneaux, Kevin. 2005. Does Federalism Weaken Democratic Representation in the United States? Publius: The Journal of Federalism, 35, 297–311.

Arrow, Kenneth J. 1951. Social Choice and Individual Values. John Wiley and Sons, Inc.

Austen-Smith, D., & Wright, John R. 1992. Competitive Lobbying for a Legislator’s Vote. Social Choice and Welfare, 9(3), 229–257.

Balla, Steven J., Lawrence, Eric D., Maltzman, Forrest, & Sigelman, Lee. 2002. Partisanship, Blame Avoidance, and the Distribution of Legislative Pork. American Journal of Political Sci- ence, 46(3), 515–525.

Barabas, Jason, & Jerit, Jennifer. 2009. Estimating the Causal Effects of Media Coverage on Policy-Speciﬁc Knowledge. American Journal of Political Science, 53, 73–89.

Baron, David P. 1990. Distributive Politics and the Persistence of Amtrak. The Journal of Politics, 52, 883–913.

Bartels, Larry. 1991. Instramental and "Quasi-Instramental Variables". American Journal of Polit- ical Science, 35(3), 777–800.

Bartels, Larry. 1996. Uninformed Votes: Information Effects in Presidential Elections. American Journal of Political Science, 40(February), 194–230.

152 Bauer, R.A., Pool, I.D., & Dexter, L.A. 1963. American Business and Public Policy: The Politics of Foreign Trade. New York: Atherton Press.

Berry, Christopher, Burden, Barry, & Howell, William. 2010. The President and the Distribution of Federal Spending. American Political Science Review, 104(4), 783–799.

Bianco, William T. 1984. Strategic Decisions on Candidacy in U.S. Congressional Districts. Leg- islative Studies Quarterly, IX:(2), 351–364.

Bickers, Kenneth N., & Stein, Robert M. 1991. Federal Domestic Outlays, 1983-1990 : a Data Books. Armonk, NY: M.E. Sharpe.

Bickers, Kenneth N., & Stein, Robert M. 1996. The Electoral Dynamics of the Federal Pork Barrel. American Journal of Political Science, 40(4), 1300–1326.

Black, Gordon. 1972. A Theory of Political Ambition: Career Choices and the Role of Structural Incentives. American Political Science Review, 66(March), 144–59.

Born, Richard. 1990. The Shared Fortunes of Congress and Congressment: Members May Run from Congress, But They Can’t Hide. Journal of Politics, 52(4), 1223–1241.

Brass, Clinton, Hatch, Garrett, & Peterson, Eric. 2008 (November). Bush Administration Policy Regrarding Congressionally Originated Earmarks: An Overview. Tech. rept. CRS: RL34648. Congressional Research Service: Government Finance Division.

Brewer, Paul R. 2001. Value Words and Lizard Brains: Do Citizens Deliberate About Appeals to Their Core Values? Political Psychology, 22, 45–64.

Cain, Bruce, Ferejohn, John A., & Fiorina, Morris. 1984. The Constituency Service Bias of the Personal Vote for U.S. Representatives and British Members of Parliament. American Political Science Review, 78(1), 110–125.

Cameron, C.A., & Trivedi, P. K. 2010. Microeconometrics Using Stata: Revised Edition. College Station, TX: Stata Press.

Cappella, J.N., & Jamieson, K.H. 1997. Spiral of cynicism: The press and the public good. Oxford University Press, USA.

Chappell, Henry. 1982. Campaign Contributions and Congressional Voting: A Simultaneous Probit-Tobit Model. Review of Economics and Statistics, 52, 77–83.

Chong, Dennis, & Druckman, James. 2007. Framing Public Opinion in Competitive Democracies. American Political Science Review, 101(4), 637–655.

Converse:1964 Converse, Philip E. 2006 [1964]. The Nature of Belief Systems in Mass Publics. Pages 206–261 of: Apter, David E. (ed), Ideology and Discontent. New York: The Free Press.

Cover, Albert C., & Brumberg, Bruce S. 1982. Baby Books and Ballots: The Impact of Congres- sional Mail on Constituent Opinion. American Political Science Review, 76, 347–359.

153 Crespin, Michael H., & Finocchiaro, Charles J. 2008. In Why Not Parties? Party Effects in the United States Senate. Chicago: University of Chicago Press. Chap. Distributive and Partisan Politics in the U.S. Senate: An Exploration of Earmarks, 1996-2005.

Delli Carpini, Michael X., & Keeter, Scott. 1996. What Americans Know About Politics and Why It Matters. Cambridge: Cambridge University Press.

Downs, Anthony. 1957. An Economic Theory of Democracy. Addison Wesley.

Druckman, James. 2001a. On the Limits of Framing Effects. Journal of Politics, 63, 1041–1066.

Druckman, James N. 2001b. The Implications of Framing Effects for Citizen Competence. Politi- cal Behavior, 23(3), 225–256.

Druckman, James N., Hennessy, Cari Lynn, St. Charles, Kristi, & Webber, Jonathan. 2010. Com- peting Rhetoic Over Time: Frames Versus Cues. The Journal of Politics, 72(1), 136–148.

Eagly, Alace, & Chaiken, Shelly. 1993. The Psychology of Attitudes. Harcourt Brace.

Eliasoph, N. 1998. Avioding Politics: How Americans Produce Apathy in Everyday Life. Cam- bridge: Cambridge University Press.

Engstrom, Erik J., & Vanberg, Georg. 2010. Assessing the Allocation of Pork: Evidence from Congressional Earmarks. American Politics Research, 38(6), 959–985.

Erikson, Robert S., & Wright, Gerald C. 2009. Congress Reconsidered, 9th Edition. Congressional Quarterly (CQ) Press. Chap. Voters, Candidates, and Issues in Congressional Elections.

Feldman, Paul, & Jondrow, James. 1984. Congressional Elections and Local Federal Spending. American Journal of Political Science, 28, 147–64.

Fenno, Richard F. 1978. Home Style: House Members in Their Districts. Longman Classics Series.

Ferejohn, John A. 1974. Pork Barrel Politics: Rivers and Harbors Legislation, 1947-1968. Stand- ford University Press.

Finnigan, Tom. 2007 (March). All About Pork: The Abuse of Earmarks and the Needed Reforms.

Fiorina, Morris. 1977. The Case of the Vanishing Marginals: The Bureaucracy did it. American Political Science Review, 71, 177–81.

Fishkin, James. 1997. The Voice of the People: Public Opinion and Democracy. New Haven, CT: Yale University Press. for Common Sense, Taxpayers. 2008. TCS (Taxpayers for Common Sense) Methodology in Com- piling FY08 Earmarks Database. for Common Sense, Taxpayers. 2011 (October). TCS FY2010 Earmark Analysis: Apples-to-Apples Increase in Earmarks. Online Database.

154 Franklin, Charles. 1991. Efﬁcient Estimation in Experiments. The Political Methodologist, 4(1), 13–15.

Frendreis, John, & Waterman, Richard. 1985. PAC Contributions and Legislative Behavior: Senate Voting on Trucking Deregulation. Social Science Quarterly, 66 (June), 401–412.

Fridkin, Kim L., & Kenney, Patrick. 2011. Variability in Citizens’ Reactions to Different Types of Negative Compaigns. American Journal of Political Science, 55(2), 307–325.

Frisch, Scott A. 1998. The Politics of Pork: A Study of Congressional Appropriation Earmarks. Taylor & Francis.

Friske, Susan T., & Taylor, Shelley E. 1991. Social Cognition. New York: McGraw-Hill.

Gerber, Alan. 1998. Estimating the Effect of Campaign Spending on Senate Election Outcomes Using Instrumental Variables. The American Political Science Review, 92(2), 401–411.

Gilens, Martin. 2001. Political Ignorance and Collective Policy Preferences. American Political Science Review, 95, 379–396.

Gimmer, Justin, Messing, Solomon, & Westwood, Sean J. 2012. How WOrds and Money Cultivate a Personal Vote: The Effect of Legislator Credit Claiming on Constituent Credit Allocation. American Political Science Review, 106(4), 703–719.

Green, D., & Krasno, J.S. 1988. Salvation for the Spendthrift Incumbent. American Journal of Political Science, 32(4), 884–907.

Green, Donald, & Krasno, Jonathan S. 1990. Rebuttal to Jacobson’s ’New Evidence for Old Arguments’. American Journal of Political Science, 34, 363–372.

Grenzke, Janet M. 1988. Comparing Contributions to U.S. House Members from Outside Their Districts. Legislative Studies Quarterly, 13, 83–103.

Grenzke, Janet M. 1989. Shopping in the Congressional Supermarket: The Currency is Complex. American Journal of Political Science, 33, 1–24.

Hall, Richard, & Deardorff, A.V. 2006. Lobbying as Legislative Subsidy. American Political Science Review., 100(1), 69–84.

Harbridge, Laurel, & Malhotra, Neil. 2011. Electoral Incentives and Partisan Conﬂict in Congress: Evidence from Survey Experiments. American Journal of Political Science, 55(3), 494–510.

Heitshusen, Valerie. 2001. The Allocation of Federal Money To House Committee Members. American Politics Research, 29(1), 79–97.

Herrnson, Paul S. 1992. Campaign Professionalism and Fundraising in Congressional Elections. The Journal of Politics, 54(3), 859–870.

Herron, Michael, & Shotts, Kenneth. 2006. Term Limits and Pork. Legislative Studies Quarterly, XXXI(3), 383–403.

155 Hibbing, John R., & Theiss-Morse, Elizabeth. 2002. Stealth Democracy: Americans’ Beliefs about How Government Should Work. New York: Cambridge University Press. Hojnacki, M., & Kimball, D.C. 1998. Organized Interest and the Decision of Whom to Lobby in Congress. American Political Science Review, 92(4), 775–790. Iyengar, Shanto. 1991. Is Anyone Responsible? The University of Chicago Press. Jacobs, L.R., Cook, F.L., & Carpini, M.X.D. 2009. Talking Together: Public Deliberation and Political Participation in America. University of Chicago Press. Jacobson, Gary. 1989. Strategic Politicians and the Dynamics of U.S. House Elections, 1946-1986. Amerkan Political Science Review, 83(September), 773–793. Jacobson, Gary. 2006. Campaign Spending Effects in U.S. Senate Elections: Evidence from the National Annenberg Election Survey. Electoral Studies, 25, 195–226. Jacobson, Gary, & Kernell, Samuel. 1983. Strategy and Choke in Congressional Elections. New Haven: Yale University Press. Jacobson, Gary C. 2009. The Politics of Congressional Elections, 7th Edition. Person Education, Inc. Jacobson, G.C. 1980. Money in Congressional Elections. New Haven: Yale University Press. Jerit, Jennifer, Barabas, Jason, & Bolsen, Toby. 2006. Citizens, Knowledge and the Information Environment. American Journal of Political Science, 50, 266–282. Johannes, John, & McAdams, John. 1981. The Congressional Incumbency Effect: Is it Casework Policy Compatible, or Something Else? American Journal of Political Science, 25(3), 512–542. Johannes, John R. 1984. The Serve The People: Congress And Constituency Service. University of Nebraska Press. Kane, Paul. 2008. Candidates’ Earmarks Worth Millions; Of Front-Runners, Mccain Abstained. The Washington Post, February 14, A–SECTION; Pg. A01. King, Gary. 1991. Constituency Service and the Incumbency Advantage. British Journal of Polit- ical, 21, 119–28. King, Gary, & Gelman, Andrew. 1991. Systemic Consequences of Incumbency Advantage in U.S. House Elections. American Journal of Political Science, 35(1), 110–38. Krasno, Jonathan, & Green, Donald P. 1988. Preempting Quality Challengers in House Elections. Journal of Politics, 50(November), 920–936. Krasno, Jonathan, Green, Donald Philip, & Cowden, Jonathan. 1994. The Dynamics of Campaign Fundraising in House Elections. The Journal of Politics, 56(2), 459–74. Krosnick, Jon A. 1990. Government Policy and Citizen Passion: A Study of Issue Publics in Contemporary America. Political Behavior, 12, 59–92.

156 Langbein, L.I. 1993. PACs, Lobbies, and Political Conﬂict: The Case of Gun Control. Public Choice., 77(3), 551–572.

Lau, Richard, & Heldman, Caroline. 2009. Self-Interest, Symbolic Attitudes, and Support for Public Policy: A Multilevel Analysis. Political Psychology, 30(4), 513–537.

Lazarus, Jeffery. 2010. Giving the People What They Want? The Distribution Of Earmarks in the U.S. House of Representatives. American Journal of Political Science, 54(2), 338–353.

Lazarus, Jeffrey. 2009. Party, Electoral Vulnerability, and Earmarks in the U.S. House of Repre- sentatives. Journal of Politics, 71(3), 1050–1061.

Lazarus, Jeffrey, & Steigerwalt, Amy. 2009. Different Houses: The Distribution of Earmarks in the U.S. House and Senate. Legislative Studies Quarterly, 34, 347–373.

Lee, Francis. 2003. Geographic Politics in the U.S. House of Representatives: Coalition Building and Distribution of Beneﬁts. American Journal of Political Science, 47(4), 714–728.

Levitt, Steven, & Snyder, James Jr. 1997. The Impact of Federal Spending on House Election Outcomes. Journal of Political Economy, 105(1), 30–53.

Lodge, Milton, Steenbergen, Marco R., & Brau, Shawn. 1995. The Responsive Voter. American Political Science Review, 89(2), 309–326.

Lupia, Arthur, & McCubbins, Matthew D. 1998. The Democratic Dilemma: Can Citizens Learn What They Need to Know? New York:: Cambridge University Press.

Luskin, R.C. 1990. Explaining political sophistication. Political Behavior, 12(4), 331–361.

Lynch, Megan Suzanne. 2008 (April). Earmark Disclosure Rules in the House: Member and Committee Requirements. Tech. rept. Order Code RS22866. Congressional Research Service (CRS), Government & Finance Division.

Management, & Budget (OMB), Ofﬁce of; U.S. Census Bureau (U.S. Department of Commerce). 2009 (March). Federal Assistance Award Data System (FAADS): Users’ Guide for Federal Fiscal Year 2009, Revised March 2009. Rev. march 2009 edn. Federal Programs Branch, Governments Division, U.S. Census Bureau, Washington, DC 20233-6800.

Mann, H. B., & Whitney, D.R. 1947. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics, 18(1), 50–60.

Mann, Thomas E. 1978. Unsafe at Any Margin: Interpreting Congressional Elections. American Enterprise Institute for Public Policy Research.

Mayhew, David. 1974a. Congressional Elections: The Case of the Vanishing Marginals. Polity, 6, 295–317.

Mayhew, David R. 1974b. Congress: The Electoral Connection. Yale University Press.

157 McLeod, Douglas M., & Shah, Dhavan V. 2008. Framing the Patriot Act. University of Wisconson. Unpublished Manuscript.

Mikusheva, A., & Poi, B.P. 2006. Tests and Conﬁdence Sets with Correct Size When Instruments are Potentially Weak. Stata Journal, 6, 335–347.

Mondak, Jeffrey. 1994. New Directions In Political Psychology. Greenwich, CT: JAI Press. Chap. Cognitive Heuristics, Heuristic Processing, and Efﬁciency in Political Decision Making.

Nabi, R.L. 2003. Exploring the framing effects of emotion. Communication Research, 30(2), 224.

Nelson, Thomas E., Clawson, Rosalee A., & Oxley, Zoe M. 1997. Media Framing of a Civil Liberties Conﬂict and Its Effect on Tolerance. American Political Science Review, 91, 567–583.

O’Keefe, D.J. 2002. Persuasion: Theory & Research. Vol. 2. Sage Publications, Inc.

Parker, Glenn, & Davidson, Roger. 1979. Why Do Americans Love Their Congressmen So Much More Than Their Congress? Legislative Studies Quarterly, 4(1), 53–61.

Popkin, Samuel L. 1991. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. Chicago: University of Chicago Press.

Porter, Rob, & Walsh, Sam. 2006 (May). Brieﬁng Paper 16: Earmarks in the Federal Budget Process. Harvard Law School: Federal Budget Policy Seminar.

Price, Vincent, & Tewksbury, David. 1997. New Values and Public Opinion. Ablex Publishing Corporation. Chap. 13, pages 173–212.

Price, Vincent, David, Calarissa, Goldthorpe, Brian, McCoy, Marci, & Cappella, Joseph N. 2006. Locating the Issue Public: The Multi-Dimensional Nature of Engagement with Health Care FREform. Political Behavior, 28(1), 33–61.

Riker, William H. 1962. The Theory of Political Coalitions. New Haven: Yale University Press.

Riker, William H, & Ordeshook, Peter C. 1968. A Theory of the Calculus of Voting. American Political Science Review, 62, 25–42.

Rocca, Michael S., & Gordon, Stacy B. 2013. Earmarks as a Means and an End: The Link Between Earmarks and Campaign Contributions in the U.S. House of Representatives. The Journal of Politics, 75(1), 241–253.

Romero, David W. 1996. The Case of the Missing Reciprocal Inﬂuence: Incumbent Reputation and the Vote. The Journal of Politics, 58(November), 1198–1207.

Saltzman, G.M. 1987. Congressional Voting on Labor Issues: The Role of PACs. Industrial and Labor Relations Review, 40(2), 163–179.

Samuels, David J. 2002. Pork Barreling is Not Credit Claiming or Advertising: Campaign Finance and the Sources of the Personal Vote in Brazil. Journal of Politics, 64, 845–863.

158 Schlozman, K. L., & Tierney, J. T. 1983. More of the Same: Washington Pressure Group Activity in a Decade of Change. The Journal of Politics, 45(2), 351–377.

Schneider, Saundra, & Jacoby, William. 2003. Public Attitudes Towards the Policy Responsibilities of the National and State Governments: Evidence from South Carolina. State Politics and Policy Quarterly, 3, 246–269.

Sears, David O., Lau, Richard R., Tyler, Tom R., & Harris M. Allen, Jr. 1980. Self-Interest vs. Symbolic Politics in Policy Attitudes and Presidential Voting. American Political Science Review, 74, 670–684.

Sellers, Patrick. 1997. Fiscal Consistency and Federal District Spending in Congressional Elec- tions. American Journal of Political Science, 41(3), 1024–1041.

Serra, George. 1994. What’s In It For Me? The Impact Of Congressional Casework On Incumbent Evaluation. American Politics Quarterly, 22, 403–420.

Serra, George, & Cover, Albert D. 1992. The Electoral Consequences Of Perquisite Use: The Casework Case. Legislative Studies Quarterly, 17, 233–246.

Shepsle, Kenneth A., & Weingast, Barry R. 1981. Political Preferences for the Pork Barrel: A Generalization. American Journal of Political Science, 25(1), 96–111.

Shepsle, Kenneth A., Van Houweling, Robert P., Abrams, Samuel J., & Hanson, Peter C. 2009. The Senate Electoral Cycle and Bicameral Appropriations Politics. American Journal of Political Science, 53(2), 343–359.

Sigelman, Lee, & Zeng, Langche. 1999. Analyzing Censored and Sample-Selected Data with Tobit and Heckit Models. Political Analysis, 8(2), 167–182.

Simon, Herbert A. 1983. Reason in Human Affairs. Stanford, CA:: Stanford University Press.

Sniderman, Paul, Brody, Richard, & Tetlock, Philip. 1991. Reasoning and Choice: Explorations in Political Psychology. Cambridge: Cambridge University Press.

Squire, Peverill. 1995. Candidates, Money, and Voters–Assessing the State of Congressional Elec- tions Research. Political Research Quarterly, 48(4), 891–917.

Stein, Robert M., & Bickers, Kenneth N. 1994. Congressional Elections and the Pork Barrel. The Journal of Politics, 56(2), 377–399.

Stein, Robert M., & Bickers, Kenneth N. 1995. Perpetuating the Pork Barrel: Policy Subsystems in American Democracy. Cambridge University Press.

Stone, Walter, Maisel L. Sandy, & Maestas, Cherie. 2004. Quality Counts: Extending the Strategic Politician Model in Incumbent Deterrence. American Journal of Political Science, 48(3), 479– 495.

159 Thompson, Lyke, & Elling, Richard. 1999. Let Them Eat Marblecake: The Preferences of Michi- gan Citizens for Devolution and Intergovernmental Service-Provision. Publius: The Journal of Federalism, 29, 139–153.

Tripathi, M., Ansolabehere, S., & Snyder, J. M. 2002. Are PAC Contributions and Lobbying Linked? New Evidence from the 1995 Lobby Disclosure Act. Business and Politics, 4(2), Article 2.

Wawro, Gregory G. 2001. A Panel Probit Analysis of Campaign Contributions and Roll-Call Votes. American Journal of Political Science, 45(3), 563–579.

Wood, Wendy. 2000. Attitude Change: Persuasion and Social Inﬂuence. Annual Review of Psy- chology, 50(1), 539–570.

Wright, John R. 1985. PACs, Contributions, and Roll Calls: An Organizational Perspective. Amer- ican Political Science Review, 79, 400–414.

Wright, John R. 1989. PAC Contributions, Lobbying and Representation. The Journal of Politics, 51, 813–829.

Wright, John R. 1996. Interest Groups and Congress: Lobbying, Contributions, and Inﬂuence. Boston: Allyn and Bacon.

Zaller, John. 1992. The Nature and Origins of Mass Opinion. New York: cambridge University Press.

160 BIOGRAPHICAL SKETCH

The son of two military parents, Travis Braidwood was born in Portsmouth, Virginia, but was raised in several cities domestically and abroad. He received a Bachelor of Arts in Political Science and in International Relations from the University of West Florida (UWF) in Pensacola, Florida in 2007. He studied in UWF’s Master of Arts program for one year before transferring into Florida State University’s PhD program in 2008.

His research interests include Congress, elections, voter behavior, political psychology, and research methodology. He is particularly interested in the effect of earmarks (also know as pork projects) on Congress and public opinion, as well as the language of plebiscites and ballot initiatives.

161