The Pennsylvania State University
The Graduate School
The Mary Jane and Frank P. Smeal College of Business

ESSAYS IN POLITICAL MARKETING

A Dissertation in
Business Administration by David LeBaron

 2008 David N. LeBaron

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2008

The thesis of David N. LeBaron was reviewed and approved* by the following:

William T. Ross, Jr.
Professor of Marketing
Dissertation Co-Advisor
Co-Chair of Committee

John C. Liechty
Associate Professor of Marketing
Dissertation Co-Advisor
Co-Chair of Committee

Peter Ebbes
Assistant Professor of Marketing

David J. Huff
Clinical Assistant Professor of Supply Chain and Information Systems

Johannes Baumgartner
Professor of Marketing
Head of the Department of Marketing

*Signatures are on file in the Graduate School

iii
ABSTRACT
Political marketing sits at the nexus of two disciplines, political science and marketing, but is not entirely accepted by either. The present research looks at the origin, development, and evolution of political marketing and examines how the adoption of a political marketing orientation is impacting the practice of political campaigns. The role of political marketing in actually changing voters’ preferences is also examined, showing that grassroots marketing efforts seem to have the greatest effect, especially with undecided voters. Finally, voter segments are derived for the last five presidential elections in the United States (1988-2004) using latent class analysis (LCA). The interpretation and implications of these segments are discussed and several avenues for future research are suggested.

iv
TABLE OF CONTENTS
LIST OF TABLES ................................................................................................................... vi
ACKNOWLEDGEMENTS ..................................................................................................... vii
Chapter 1 The Elephant Donkey in the Room: Political Marketing in the Marketing
Paradigm .......................................................................................................................... 1
What is Political Marketing?............................................................................................ 3
The Evolution of Political Marketing .............................................................................. 5
The Marketing of Political Marketing.............................................................................. 10
Other Marketing Considerations ...................................................................................... 13
Why Political Marketing? ................................................................................................ 15
Politics as Big Business ........................................................................................... 16
Politics Marketing’s Influence ................................................................................. 18
In Praise of Academic Political Marketing ...................................................................... 20
Conclusion and Outline of Dissertation ........................................................................... 22
Chapter 2 How Much Does Political Marketing Matter? ....................................................... 24
The Minimal Effects Hypothesis ..................................................................................... 26
The “Campaigns Matter” Perspective .............................................................................. 29
What Does “Do Campaigns Matter?” Mean? .................................................................. 34
Data and Analysis ............................................................................................................ 35
Who are the Switchers? .............................................................................................. 36
From Not Voting to Voting......................................................................................... 42
Discussion ........................................................................................................................ 44
Limitations ....................................................................................................................... 46
Future Research................................................................................................................ 46
Chapter 3 Political Taxonomies: Voter Segments in Presidential Elections, 1988-2004 ....... 48
Why Segment Voters? ..................................................................................................... 50
Voter Segmentation.......................................................................................................... 52
Latent Class Regression ................................................................................................... 55
American National Election Studies ................................................................................ 59
Dependent Variable .................................................................................................... 61
Independent Variables ................................................................................................ 62
Demographic Covariates............................................................................................. 65
Results .............................................................................................................................. 65
1988 Election .............................................................................................................. 66
1992 Election .............................................................................................................. 71
1996 Election .............................................................................................................. 76
2000 Election .............................................................................................................. 80
2004 Election .............................................................................................................. 84
Marketing Application ..................................................................................................... 88

v
Discussion ........................................................................................................................ 90
Limitations ....................................................................................................................... 92
Further Research .............................................................................................................. 93
References ................................................................................................................................ 95
Appendix A Policy Variables Used in Analysis by Year ....................................................... 108
Appendix B Selecting the Optimal Number of Segments ...................................................... 111
Appendix C Segment Descriptions by Year ........................................................................... 114
Appendix D Derived Segments by Year ................................................................................. 115

vi
LIST OF TABLES
Table 1-1: Campaign Receipts and Expenditures, 1980-2004 ................................................. 17
Table 2-1: Incumbent Party’s Share of Major Party Vote, 1948-2004 ................................... 28
Table 2-2: Timing of Vote Decision ....................................................................................... 30
Table 2-3: Interest in Political Campaigns Among Campaign Deciders ................................ 32
Table 2-4: Where Campaign Deciders Get Information ......................................................... 32
Table 2-5: Demographics Only Probit Model......................................................................... 38
Table 2-6: Political Fundamentals Probit Model .................................................................... 39
Table 2-7: Political Marketing Probit Model .......................................................................... 40
Table 2-8: Political Marketing with Interactions Probit Model .............................................. 41
Table 2-9: Campaign Effects on Unlikely Voters ................................................................... 43
Table 2-10: Campaign Effects on Nonvoters.......................................................................... 44
Table 3-1: 1988 Model for Dependent Variable Results ........................................................ 70
Table 3-2: 1988 Segment Profiles........................................................................................... 71
Table 3-3: 1992 Model for Dependent Variable Results ........................................................ 75
Table 3-4: 1992 Segment Profiles........................................................................................... 76
Table 3-5: 1996 Model for Dependent Variable Results ........................................................ 79
Table 3-6: 1996 Segment Profiles........................................................................................... 80
Table 3-7: 2000 Model for Dependent Variable Results ........................................................ 83
Table 3-8: 2000 Segment Profiles........................................................................................... 84
Table 3-9: 2004 Model for Dependent Variable Results. ....................................................... 87
Table 3-10: 2004 Segment Profiles......................................................................................... 88

vii
ACKNOWLEDGEMENTS
There are many, many people whom I wish to thank for their kind support and assistance throughout my years in the doctoral program in Marketing at the Smeal College of Business
Administration. Foremost among these, of course, is my wife, Andrea LeBaron. None of this would have been remotely possible without her support, encouragement, patience, and understanding. She has truly been an angel as we worked our way through the long and arduous process of completing a dissertation. My children, Emily, James, and Ben, also contributed much by being there to cheer me up and by making life so much happier and enjoyable.
Special thanks also belong to the members of my committee: Bill Ross, John Liechty,
Peter Ebbes, and David Huff. Bill served as an advisor and a cheerleader and offered much advice as to how I should proceed. More importantly, however, he listened to my concerns and on numerous occasions gave me the extra impetus I needed to keep going. John gave much helpful advice, not only on technical matters, but also helped me keep things in their proper perspective. Peter was always willing to let me drop by his office unannounced and would gladly answer my questions or steer me in the right direction. His kindness and insight were especially appreciated as we taught Marketing Research courses at the same time. Finally, David was a great sounding board who helped me stay on task. Our weekly meetings were a great help to me and did much to bolster me intellectually and spiritually. I owe a great debt of gratitude to all of you for the service you have rendered.
I was fortunate to have many friends who helped me along the path. I am particularly grateful for my office mates, Jeff Kukitz, Bryan Johnson, and Chris McKeon. It was truly a blessing to share an office with people who could talk about marketing and sports with equal ease. To be honest, I learned far more from discussions of the latter than the former. I was also lucky enough to make friends with students from other departments whose empathy helped me

viii through many rough patches. Craig Crossland and Bret Fund were especially helpful in that regard. Thanks for all you did.
Finally I need to thank my parents, Dale and Laura LeBaron, who lent support and even made the long trek from Utah to Pennsylvania on more than one occasion. Thank you, Mom and
Dad, for all you have done throughout my life. My brother, Curtis LeBaron, was also an invaluable resource. He truly understood what I was going through and selflessly gave his time to listen and lift me up when I was struggling. Again, thank you, for everything.

Chapter 1
The Elephant and Donkey in the Room: Political Marketing in the Marketing
Paradigm
Political marketing has become more prevalent during the last forty years. An example of this trend can be found in the following illustration. In May 2007, a focus group consisting of six
Democrats and six Republicans was asked to imagine the leading presidential contenders in the following predicament: Each candidate is standing in a line at an airport, but, “there is only one seat left on the plane, and there are six people ahead of the candidate. What does each of them do?” The focus group participants' responses are simultaneously amusing and insightful. Some thought Rudy Giuliani would refuse to stand in the line, demand to speak to the pilot, or even yell, “9/11!” John McCain, they said, would raise a ruckus, demand a seat on the flight, or simply refuse to abide by the rules. Mitt Romney and Barack Obama were seen as quietly and politely negotiating their way to the front of the line. John Edwards would attempt to bribe the people in front of him or “lay on the Southern charm.” Hillary Clinton, according to the participants, would, “Tell everyone ahead of her that the flight had been canceled,” or “Send her people to take care of the details” (Seelye 2007). While there is no practical way of knowing exactly how each candidate would respond in this particular scenario, the focus group's responses did reveal what potential voters thought of the candidates' personalities. Interestingly enough, being perceived as nice or charming in 2007 did not necessarily guarantee success in the 2008 presidential primaries.
Although voters probably do not cast their ballots on the basis of how someone might behave while in line at the airport, it is nevertheless instructive to see how methods developed for commercial marketing can shed light on politics and politicians. The use of marketing tools and

2 techniques in a political context, commonly known as political marketing, has become increasingly common in advanced democratic countries, i.e. United States, Canada, Western
Europe, during the last thirty years (Lees-Marshment 2001, Newman 1994). The Margaret
Thatcher and Ronald Reagan campaigns of 1979 and 1980 respectively are generally regarded as the onset of the consistent and conscientious application of marketing methods in the political domain. Since that time, candidates running at all levels of government have become more marketing savvy and attuned to the wants and of voters (Hamburger and Wallsten 2006, Newman
1999). All signs indicate that this trend will continue into the foreseeable future which means marketing will play an increasingly important role in electoral outcomes (Newman 2001). It is imperative, therefore, for marketers to understand how their expertise and skills are impacting political processes.
The rising influence of marketing in politics rests on the assumption that political campaigns actually do matter. In other words, marketing efforts actually persuade voters in meaningful and systematic ways. At first glance, that might seem like an obvious proposition. A close study of the literature, however, reveals that there is some disagreement on that point
(Gelman and King 1993, Holbrook 1996). So-called “minimal effects” proponents contend that the net effect of campaigns is negligible, whereas others assert that political campaigns have important and measurable effects on voting behavior. Surprisingly, academic marketers have largely overlooked political marketing or relegated it to the status of an interesting but not particularly relevant or key subfield within the larger marketing discipline (Luck 1969). That is not to say that no work has been done in political marketing, rather not enough work has been done and much exciting and useful research remains to be done. The aim of this essay is to help fill the void in the marketing literature by defining what political marketing is, studying how it has evolved, and arguing as to why the field of political marketing merits further attention. It is hoped that a thorough examination of these issues will demonstrate that political marketing is

3 more than just the, “vigorous but somewhat neglected stepchild of marketing and political science” (Mauser 1983).

What is Political Marketing?
In the broadest sense, political marketing is simply, “Marketing designed to influence target audiences to vote for a particular person, party, or proposition” (American Marketing
Association 2007). Mauser (1983) adds, “Political marketing is a set of practical procedures for identifying effective campaign strategies based upon modern marketing principles and techniques.” Newman (1999) elaborates on this definition by describing political marketing as,
“the application of marketing principles and procedures in political campaigns by various individuals and organizations. The procedures involved include the analysis, development, execution, and management of strategic campaigns by candidates, political parties, governments, lobbyists, and interest groups that seek to drive public opinion, advance their own ideologies, win elections, and pass legislation in response to the needs and wants of selected people and groups in society.” Along similar lines, Butler and Collins (1994) argue that political marketing is, “The marketing of ideas and opinions which relate to public or political issues or to specific candidates.
In general, political marketing is designed to influence people's votes in elections...Political marketing, however, employs many of the same techniques used in product marketing, such as paid advertising, direct mail and publicity.” Finally, Lees-Marshment (2001) argues that political marketing is about, “political organizations adapting business-marketing concepts and techniques to help them achieve their goals.”
The various definitions contain several commonalities. First, all agree that political marketing is intrinsically linked with political campaigns. From a marketing perspective, political campaigns can be seen as nothing more than attempts to identify consumer/voter wants and

4 needs, which can then be satisfied with the promise of a political service or product. Second, the ultimate goal of political marketing is to persuade people to vote a certain way. This necessarily implies that elections must be free and fair in order for political marketing to be meaningful.
Third, there is a consensus among scholars that political marketing is the process of applying tools developed for the commercial marketplace into the political realm. It is not uncommon, therefore, to hear political candidates talk about different clusters or segments of the voting public
(Baines 1999, Newman 1994). This is clearly a situation where marketing jargon and practices have become entrenched in political lexicon and activities. Fourth, political marketing is much more than political advertising. One common misconception which is prevalent among political science circles is that advertising and marketing are essentially synonymous (Franz, Freedman,
Goldstein and Ridout 2008). Clearly, as any undergraduate marketing student has learned, marketing is much more than just sales and advertising (Perreault and McCarthy 2003). Such is the case with political marketing--it encompasses a wide range of activities. Political advertising may be the most prominent and expensive of those activities, but it is nevertheless only one of many things that fall into the political marketing category.
There is one substantial disagreement in the definitions of political marketing: Does the term political marketing include only campaigns or does it cover governance as well? The
American Marketing Association (2007) flatly asserts that, “Attempts by an individual or organization to only educate or change attitudes are not political marketing.” Others have taken a more expansive view by incorporating the persuasion of elected officials in their definitions
(Butler and Collins 1994, Mauser 1983, Newman 2001). There are several reasons as to why the latter approach is more beneficial. If one were to accept the AMA's delineation of political marketing as applying to only campaigns, it would be difficult to describe what governments are doing when they conduct surveys, convene focus groups, and interview constituents. Clearly, governments use marketing techniques to guide policy and reach decisions (McDaniel and Gates

5
2007). It should be obvious that any definition of political marketing must include campaigns and governing. Political marketing can, therefore, be defined as the application of marketing techniques to any political setting. This definition is in line with Newman's definition and is sufficiently broad to cover the full range of political activities that candidates and incumbents are engaged in.

The Evolution of Political Marketing
In addition to being a set of activities, political marketing can be thought of as a general outlook or orientation (Lees-Marshment 2001, Sparrow and Turner 2001). Just as the business world has changed its orientation over the decades, political organizations have undergone changes in their thinking and outlook. In 1799, for example, Thomas Jefferson wrote a political creed that was widely disseminated among the general population and served as a basis for his platform in the 1800 presidential election (Wikipedia 2007a). In 1824, the Harrisburg
Pennsylvanian conducted the first documented piece of marketing research when it asked readers how they intended to vote in the upcoming presidential election (McDaniel and Gates 2007). In
1828, some of John Quincy Adams' supporters produced a series of pamphlets, called the “Coffin
Handbills,” that attacked the character and conduct of his opponent Andrew Jackson, and purportedly contributed to the early death of Mrs. Jackson just six weeks after the election
(Wikipedia 2007b). Since those early days of the American Republic, nearly every aspirant to elected office at the federal, state, or even local level has engaged to varying degrees in some form of marketing. While there has always been some link between marketing and politics, the relationship between the two has evolved and grown increasingly complex with the passage of time. Political marketing, particularly in the United States, has gone through three distinct phases: product-oriented era, sales-oriented era, and marketing-oriented era.

6
In marketing lingo, the early years of American political marketing were characterized by a product orientation. A product-oriented political organization “argues what it stands for and believes in. It assumes that voters will realize it is right and vote for it” (Lees-Marshment 2001).
Voters are important but no meaningful attempt is made to find out what they want. When a product-oriented party loses an election, the loss is blamed on a failure to convince enough people or it is attributed to the notion that the electorate was not ready for the message the party delivered. One distinguishing feature of the production era was the “smoke-filled backrooms” where party bosses and political elites met to decide who would be nominated and, in some cases, elected. The main focus during this time was the organization and what was good for its interests and sustained competitive advantage. In short, the party was paramount and the public was secondary. The production focus lasted well into the 20th century. Warren G. Harding, for example, conducted his 1920 presidential bid as a “Front Porch Campaign,” with large groups of voters coming to his house in Marion, Ohio to hear him speak. Even Harry Truman's famous
1948 “Whistle Stop" tour was an endeavor to persuade people that his positions were correct and that voters should support his positions. In some regards, Truman's campaign bridged the gap between the product-orientation and sales-orientation eras.
The beginning of the “sales orientation” era of political marketing had its roots in the
1948 election but came to full fruition in the 1952 U.S. presidential campaign (Newman 1994).
During that particular election cycle, Dwight Eisenhower's staff performed a rudimentary segmentation of the voting public and sent out different campaign contribution appeals to the segments they had identified (Baines 1999). In 1960, Kennedy aides “organized piles of three-byfive index cards, each representing a targeted voter to mobilize in the crucial West Virginia and
Wisconsin primaries" (Newman 1999). Although the use of segmentation and targeting suggests a nascent marketing orientation, the candidates' emphasis was aimed more towards identifying segments that would be receptive to their messages, as opposed to viewing voters as consumers.

7
A key component of a sales-oriented political party is the belief that, “consumers will not buy enough of the firm's products unless it undertakes a large-scale selling and promotion effort”
(Kotler and Armstrong 2004). In this framework, political organizations do pay closer attention to the voting populace, but their focus is still on selling and promoting the products they are offering. The current paradigm, marketing orientation, began in earnest during the 1980 presidential campaign. Ronald Reagan hired Richard Wirthlin to conduct polls and generate market intelligence. Wirthlin created a proprietary database called Political Information System or PINS, which combined “quantitative, qualitative, institutional and historical data,” to predict who would vote and how they would vote (Mark 2004). The seminal 1980 campaign marked the beginning of the systematic application of commercial marketing in the political realm. Since
1980, a plethora of marketing methods have found their way into political campaigns. Some of these techniques include blanketing specific areas with automated phone messages from a celebrity or popular politician, asking panels of voters to complete surveys immediately after campaign events, and using hand-held dials to gauge voter reactions to comments made in debates (Ferrick 2007, Harnden 2007). By 2004, the Republicans initiated a voter database called
“Voter Vault” and implemented a voter turnout project named the “72 Hour Project,” that increased Republican turnout on Election Day. The Democrats have also created a voter database,
“Demzilla” and have developed a voter turnout database based on the Republicans' model
(Dreazen 2006, Balz 2006).
Having a marketing orientation, however, means more than adopting the “latest and greatest” marketing techniques. It involves a shift in thinking that inverts the emphasis in the product orientation. According to Kohli and Jaworksi (1990), a market orientation is, “The organizationwide generation of market intelligence pertaining to current and future customer needs dissemination of the intelligence across departments, and organizationwide responsiveness

8 to it" (italics in original). The primary implication of a market orientation is firms will have a consumer first approach; the whole organization is geared toward satisfying consumers' or voters' wants and needs. While vestiges of a sales-oriented approach still linger, marketing orientation has become the dominant framework in recent years. Two noted political journalists, Tom
Hamburger and Peter Wallsten (2006), observed with respect to the 2004 presidential election,
“the old-model campaign sought to educate voters about a candidate...the new strategy almost completely reversed that approach. Instead of trying to educate voters about Bush, Rove worked at educating himself and his staff about voters.”
There are several consequences related to the adoption of a marketing orientation. First, campaigns have become a much more controlled, professionalized affair. As Scammell (1999) notes, “the modern campaign is capital intensive, relying on a much smaller base of volunteers, much tighter central direction of campaign operations, increased reliance on non-party experts from media and marketing, and far less face-to-face communication with voters.” This stands in stark contrast to the more traditional approach of “boots on the ground,” and a necessarily less centralized campaign structure. There are some potential drawbacks to this approach: voters may feel less involved which, in turn, may engender apathy, and there is some evidence to suggest that voters are less responsive to mass media persuasion attempts. While there are many advantages to increasing the broad appeal of a campaign, it may come at the cost of losing the personal touch which is very effective in persuading certain segments of voters.
Second, one of the central tenets of a marketing orientation is it involves potential customers at all stages of the conceptualization, design, development, and production process. In political marketing terms this means that voters' input is used in deciding what policies to adopt or pursue. While this runs counter to the traditional notion of a politician or party taking a firm stand and sticking to it, there is some evidence that both Thatcher and Reagan (both widely considered to be doctrinaire or dogmatic politicians) were willing to let voter input guide specific

9 policy choices within a framework (Scammell 1999). Some have speculated that this approach will ultimately lead to the “death of political courage” (Lees-Marshment 2001). Obviously this will not be taken to the extreme where everything is decided by marketing research, but it can be a positive development which may counteract some of the potential negativity created by professional party organizations.
Third, an important consequence of a marketing orientation is political parties are now,
“organized on a permanent war footing...coordinated by a permanent staff of communications and marketing professionals. The permanent campaign requires a permanent market research program” (Sparrow and Turner 2001). A marketing orientation is essentially a permanent campaign, which requires parties and candidates to remain constantly in campaign mode. The rise of the permanent campaign, as some observers have noted, has not been greeted with universal enthusiasm (Henneberg 2004, O’Shaughnessy 2002). The very concept of a political marketing orientation is viewed in certain quarters as a pejorative term that sums up everything that is wrong with politics today. For instance, some feel that marketing is a fundamentally manipulative activity, i.e. “spin doctoring” or the machinations of “skilled humbugs” that have no place in the public arena (Key 1966). Others have argued that votes and policies should not be bought and sold in the marketplace as if they were soap or some other commodity (Marland 2003,
Rothschild 1978a). Still others contend that political marketing is yet another symptom of the dumbing down of democracy and a sign that the rich, “rather than the best," are taking control,
“and this offends democratic notions" (O’ Shaughnessy 2002). Regardless of critics' worries and remonstrations that marketing is the death knell of good politics, it appears that the permanent campaign has become a permanent feature of advanced democracies.

10
The Marketing of Political Marketing
One of the most striking aspects of the scholarly literature in political marketing is how difficult it has been, and continues to be, for political marketers to achieve legitimate status in the two disciplines where their research would most naturally fit: political science and marketing.
One scholar notes that, “Political marketing is a vigorous but somewhat neglected stepchild of political science and marketing” (Mauser 1983). There have been some exceptions to that pattern, most notably Kotler and Levy, who, more than 35 years ago, advocated for the widening of marketing's disciplinary boundaries (Kotler and Levy 1969). Despite pleas to broaden the boundaries of marketing to include political phenomena, the majority of research done in political marketing has appeared in specialized marketing journals, or journals that are not considered “A" level publications by most American marketing departments. For example, there have been a handful of politics-oriented articles in Journal of Consumer Research and Journal of Marketing, but the preponderance of political marketing work has appeared in other publications such as {\it
European Journal of Marketing, Psychology & Marketing, and Journal of Consumer Marketing
(Ahmed and Jackson 1979, Klein and Ahluwahlia 2005, Newman and Sheth 1985, Niffenegger
1989, Sparrow and Turner 2001). The relative paucity of articles in leading journals suggests that marketing scholars either do not take political marketing seriously, or they view it as a subset of more important frameworks like social marketing. It may also imply that political marketing scholars have not marketed their specialty particularly well.
Political scientists, for their part, have been even more reticent in their acceptance of political marketing (Lees-Marshment 2001, Scammell 1999). In the political science paradigm, political marketing is merely descriptive work that is a subset of a broader electioneering framework. For example, political scientists are interested in campaigns but the actual events are deemed to be of limited value. Similarly, they study the fact that governments shape and mold public opinion but are more interested in the consequences than in the how and why. This

11 attitude has begun to change somewhat in recent years. Hillygus and Shields (2008) examine how marketing techniques are used to prime “wedge issues” which can be used to peel away slices of voters from their stated party affiliation. Others have looked at “microtargeting” or segmentation and targeting at the individual level, and how it is employed to drive electoral outcomes (Hamburger and Wallsten 2006). This recent attention to marketing in politics has helped to bridge the gap between the two disciplines, but the gulf still remains wide.
Some have endeavored, with limited success, to expand the marketing framework to include political marketing (Niffenegger 1989). The question still remains, however, is political marketing really marketing in the traditional sense? From a narrow marketing standing, the answer is clearly “no.” Luck (1969) argues that there are too many differences to bridge between standard marketing and political marketing and any attempts to broaden marketing stretch the discipline "too far.” Luck is one of only a handful of authors who have publicly argued against political marketing, but there does seem to be an entrenched bias against or lack of interest in the subject. If, on the other had, one adopts a more inclusive view of marketing, such as the frameworks delineated by Kotler and Levy (1969) and Bagozzi (1975), the answer is a resounding “yes”--political marketing does belong in the marketing discipline. Kotler and Levy
(1969) state, "[marketing] is a pervasive societal activity that goes considerably beyond the selling of toothpaste, soap, and steel. Political contests remind us that candidates are marketed as well as soap.” Bagozzi sees marketing as exchange between parties. He argues, “Marketing exchanges often are indirect, they may involve intangible and symbolic aspects, and more than two parties may participate” (Bagozzi 1975). Based on their interpretations of marketing, political marketing is a part of the paradigm even though it is not focused on goods or services.
While there is an exchange relationship in politics--a candidate promises certain benefits or policies in exchange for a person's vote--the nature of that exchange differs significantly from the business-to-consumer or business-to-business orientation of traditional marketing. As a

12 further example of the challenges that political analysis pose, consider the four P's of marketing.
It is fairly easy to apply the concepts of promotion (campaign advertisements) and product
(candidates, policies, promises of future action) to the political realm. There are, however, no easily interpreted counterparts for the price and place elements. A vote might be an easy proxy for price or it might be a campaign contribution or it could be something entirely different
(Niffenegger 1989). Place is also difficult to evaluate because there is no clear distribution channel from the producer to the final consumer. It is not surprising therefore to suggest that,
“The `Four Ps"...need considerable stretching to make much sense in politics” (Scammell 1999).
One scholar has even suggested that the four P's framework be changed to include pushing, pulling, polling, and product (Newman 19994). Clearly the inclusion of political marketing into mainstream marketing will require modifications and extensions of existing frameworks which may be the main reason why some marketing academics still remain opposed to expanding the discipline's boundaries (Newman 2002).
Fortunately, for those interested in political marketing as a field of study, there is a growing cadre of scholars who do not view political marketing as the unwanted stepchild of two respectable social sciences (O’Shaughnessy 2002, Scammell 1999). There has been a gradual but perceptible and positive shift in attitudes, at least in the marketing world, in favor of political marketing. In 1996 and 2001, the European Journal of Marketing dedicated special issues to political marketing and Psychology & Marketing did likewise in December 2002 (Harris 1996,
Newman 2002). An edited volume, The Handbook of Political Marketing, was published by
Sage Publications in 1999 and the Journal of Political Marketing has been published bimonthly since 2001. There are annual conferences that focus on political marketing and several European universities have established centers devoted to the study of the subject. In short, the field is growing and becoming more germane. While these recent developments are encouraging for students of political marketing, there is still some convincing yet to be done.

13
Other Marketing Considerations
There are many areas within mainstream marketing that can yield insights when applied to the political marketing realm. Foremost among these are understanding negative evaluations, specifically negative advertising, new product development, and service failures. Negative advertising, for example, has become one of, if not the most, controversial issues in political marketing (Wattenberg and Brians 1999). It has been blamed for decreasing voter turnout, a general increase in cynicism toward the political process, and has been disparaged as, “a corrupted form of democratic discourse--something we would be better off without, and something we need to be on guard against” (Ansolabehere and Iyengar 1994, Franz et al. 2008).
Several candidates have attributed their losses in political campaigns to the fact that their opponents “went negative." In November 1996, for example, Bob Dole said the relatively low turnout, which presumably hurt his candidacy was a result of, “a lot of negative ads. People do get turned off with negative ads” (Wattenberg and Brians 1999).
Although there is general agreement among scholars and voters alike that negative advertising is somewhat distasteful and unpalatable, there is no consensus as to what its actual effects are. Ansolabehere et. al (1994) demonstrate, in a controlled experimental setting, that likely voters who watch negative ads are less likely to vote. This effect is even more pronounced among political independents. Other election scholars, however, dispute this claim and argue that negative advertising not only does not suppress turnout, it might actually increase turnout
(Freedman and Goldstein 1999). A third approach suggests that negative advertising is a complicated issue and there is little support for either the mobilization or demobilization schools of thought (Wattenberg and Brians 1999). Negative advertising is an area that merits further attention from marketing scholars. Some literature exists on negativity in evaluating political candidates or on the potential effects of political advertising (Klein and Ahulwahlia 2005,

14
Rothschild 1978a, Rothschild 1978b). There is, however, much work that can be done in the political world which can subsequently be applied to the commercial domain.
There is a strong marketing literature in the area of new product development. As of yet, political organizations have not adopted the insights from the new product development into political campaigns. The current paradigm in politics is to let candidates go through a sifting process of working their way through the party ranks or gaining support through primary elections or both. At the end of the process, a candidate emerges who is best able to navigate through the system. This process does not guarantee that the nominee is the ideal candidate for a general election. Political organizations might profitably benefit from adopting some marketing research techniques to “design” the best candidate. For instance, multidimensional scaling methods could be used to identify consumer segments and whether or not particular candidates are satisfying the wants and needs of those clusters. Political parties could then seek out the type of people who could meet those desires and thus increase the probability of electoral success.
Another area within marketing that can be used in politics is the notion of service failure.
Service failures occur when there is a disparity between what a service provider offers and what the customers expect. As Hess, Ganesan, and Klein (2003) point out, “Even the best service providers produce errors in service delivery.” In politics, it is simply impossible, and maybe even undesirable, for elected representatives to please all of their constituents. Given that some people will be displeased, it is worthwhile to investigate how politicians can best deal with these voters or consumers. The service failure literature suggests that customer relationships provide a buffer when things go wrong (Berry 1995). Political organizations and figures can learn how to best mitigate the voter displeasure that inevitably follows when difficult policy choices are made.
The service failure literature can also provide guidelines for would be competitors who can then exploit the dissatisfaction in future electoral contests.

15
The brief discussion of negative advertising, new product development, and service failure, highlights that fact that there are many subfields within marketing that can be valuable for political marketers. Marketing is a broad field and the lessons learned from commercial practice can be beneficial for political practice and vice versa. No attempt is made in this essay, however, to examine the breadth of marketing applications. There are many exciting avenues for further research. The success of future research hinges, in part, in how eagerly scholars embrace the concept of political marketing. The next section examines the reception political marketing has received thus far and how it can fit within the broader academic marketing paradigm.

Why Political Marketing?
Earlier it was noted that political marketing scholars have not fully articulated why political marketing should be taken as seriously as other issues within marketing. In other words, there is no clearly stated argument as to why political marketing should be viewed on a par with the marketing of laundry detergent or automobiles or any other product class. Most political marketing scholars seem to operate from the premise that it is self-evident as to why academics and practitioners should heed developments in the field. Unfortunately, this is insufficient. More work needs to be done to explicitly argue why political marketing warrants special attention.
There are several reasons why political marketing should be integrated more fully into the marketing framework. These reasons are: 1) Political campaigns actually do matter; 2) Politics has grown into a relatively large industry that affects almost every other aspect of economic life in the United States; and 3) The adoption of marketing in politics has produced profound changes within politics itself. Each of these issues will be investigated in further detail below.

16
Politics as Big Business
Mark Hanna, who was the campaign manager for President William McKinley in 1896 and 1900, reportedly said, “There are two things that are important in politics. The first is money and I can't remember what the second one is” (http://www.opensecrets.org). Since that time, the influence of money has only grown larger and more important. Hanna's statement draws attention to the fact that politics is becoming a large industry or at least a larger industry than it has been in the past. The media give broad and persistent coverage to the issue of money in politics (Greeley
2007). Campaigns are becoming increasingly costly and drawn out and many feel that the constant need for money is corrupting politics (Doyle 2007). That being said, one of the ongoing dilemmas in the elections and campaigns literature is the issue of why there is so relatively little money in American politics (Ansolabehere, de Figueiredo, and Snyder 2003). Politics and campaigning is far more expensive than it used to be, but in comparison to the other industries, it is not overwhelmingly large. Even though politics is ideally a not-for-profit activity, and is smaller than many other endeavors, it is nonetheless a rapidly growing business sector. Table 1 shows the increase in total spending by presidential candidates over the last 24 years of presidential elections.

17
TABLE 1-1
Campaign Receipts and Expenditures, 1980-2004, (millions $, 2004)
Year
1980
1984
1988
1992
1996
2000
2004
Source: www.opensecrets.org

Total Receipts
405.9
369.1
519.6
442.6
510.1
579.8
880.5

Total Spending
231.4
189.3
337.5
256.9
287.4
376.1
717.9

The data reveal an interesting pattern. Expenditures show the greatest increase from one election cycle to the next when there is no incumbent president running (1988 and 2000). The
2004 election is something of an anomaly and represents a dramatic surge in campaign spending.
Part of the explanation lies in the fact that both parties were aware that the election would be closely contested given the 2000 results so they poured more effort and financial resources into the campaign. The amount of money spent of course differs from the amount of money raised.
The major party candidates raised an estimated $880.5 million from private and governmental sources during the 2004 U.S. presidential election cycle (http://www.opensecrets.org). Of the money raised, $498 million came from individual contributions and $149 million came from the federal government. It is estimated that the total spending in the 2008 campaign will exceed one billion dollars (www.opensecrets.org). The increasing price tag of political campaigns naturally begs the question of why this pattern has been observed. The obvious answer is found in marketing. “This reliance on expensive advertising is a leading factor behind the rise in the cost of running for office in the United States. This rising cost is considered by some to discourage those without well-monied connections, or money themselves, from running for office" (Francia
2001).
Most of the campaign money is being spent on television advertisements, promotional materials, personal appearances, and the services of political consultants (Franz et al. 2008). In

18
2004, for example, $655 million of the total $717 million in expenditures went toward the above named categories. That represents an increase of 330 million over the 1992 election media spending, which in turn was a doubling of the amount spent in the 1980 Reagan-Carter campaign
(Kotler and Kotler 1999). Undoubtedly, some of the increased expense can be attributed to increased advertising costs, higher postage rates, and more expensive consultants. Most of the escalating campaign costs, however, come from greater usage of and reliance on traditional media, and the use of more sophisticated marketing techniques like dial testing and tracking polls.
The growth of politics as a business is not limited to the presidential race. Senate, congressional, and even local political contests have become more expensive and are a lucrative industry for those who specialize in guiding candidates to success (Niffenegger 1989). Of course many have decried the influence of marketing in politics, but unless Congress and the President enact major changes to the way elections are funded--which they have thus far been loathe to do--the political business will continue to grow in size and importance. The political business is fundamentally different from other industries. Those who acquire political office have the ability to set policy and enact legislation that could potentially affect every other domain that is study by traditional marketing scholars. If marketing scholars are indifferent to politics and political marketing, they will miss out on the evolution of an important industry.

Political Marketing's Influence
Given the increased need for money in politics that is directly related to increased marketing costs, it is worthwhile to look at how political marketing is impacting the practice of politics. The most obvious impact is found in the length of the election campaign season. As recently as 1968, would-be nominees did not declare their intention to run for the presidency until January or
February of the election year. Only 40 years later, candidates have begun announcing their plans

19 more than 18 months before the general presidential election. The difference can be attributed to, in large measure, the increased need for money which is a direct result of increased reliance upon modern marketing techniques. In 1968, Hubert Humphrey described fundraising as a,
“disgusting, degrading, demeaning experience.” Most candidates today would probably agree with that assessment, yet, “More and more candidates are being forced to devote significant portions of their personal schedules on fundraising activities, including meeting with and making telephone calls to PAC directors and individuals who make large contributions” (Francia 2001).
Once they are in office, politicians are required to spend significant amounts of time on fundraising activities for their own reelection bids and for members of their parties who are seeking other elected positions. The amount of time that fundraising requires, again this can be tied back to marketing, has a direct influence on the type of person who runs for president.
People who run for president in the foreseeable future will need to meet one of three criteria.
First, candidates may need to be independently wealthy, like Ross Perot in 1992 or Mitt Romney in 2008, if they desire to eschew the rigors of constantly traveling and requesting funds. Second, candidates who are not rich will need to be the type of people who enjoy the “glad-handing” of politics and who do not mind doing that type of work for an extended period of time. Third, candidates will need to become even more well-connected to political elites in order to tap into the political donor network. This may mean that no person will be able to successfully run for president without first establishing a reputation within political party circles prior to seeking office. The main implication of that is the days of the dark horse candidate or the last minute surprise candidate may have come to an end. Whether these developments are necessarily unpalatable or bad somehow for democracy is an issue that merits further attention (Wleizen and
Erickson 2002).
A second legacy of the rise of political marketing is an increased distrust of politicians and the political process. The growing importance of money in politics has led some to believe

20 that the money is engendering corruption and an elite-driven political system (Glantz,
Abramowitz, and Burkart 1976). The logic follows thusly: political candidates need money to get elected, potential donors (usually wealthy individuals or corporations) have money, the politicians ask for that money, and the donors decide to give that money if they can get something in return. In other words, in this framework, political contributions are functionally purchases or exchanges. The politician sells the promise of an advantageous policy or benefit in exchange for the requisite cash to get elected. This arrangement runs counter to the honest-broker “Mr. Smith
Goes to Washington” ideal type that permeates American political thought. There are two counterarguments for to the notion that political marketing is ultimately leading to the corruption of politics. First, a brief examination of the sources of political contributions shows that most donations come from individuals who give on average less than 2000
(http://www.opensecrets.org). The idea of powerful donors giving large chunks of money to buy influence and preferential treatment is not supported by the data. Second, the intense media scrutiny that accompanies politicians, particularly on the federal level acts as at least a partial check on malfeasance. If anyone is caught trying to manipulate the rules of the game to favor a specific contributor, that will most likely attract media attention and questioning. Even though the increased need for money due to the increasing reliance on political marketing is perceived as corrupting the system, the evidence shows that this is not likely the case. At most, it can be shown that money buys access instead of preferential treatment.

In Praise of Academic Political Marketing
Political marketing has become something of a bogeyman in American politics. Much of the discontent is blamed on marketing related concepts like “mudslinging” or “attack advertising” which conjures up images of rabid politicians with bared fangs waiting to tear apart their

21 opponents (Franz et al. 2008). Despite these negative images, there are real benefits that come from political marketing. Foremost among these benefits is political marketing provides information. Although there is much criticism of “thirty-second TV spots” they are valuable informational cues that educate voters as to what candidates believe and how they would act once elected. Some have even likened political advertising to an “informational multivitamin" that is no substitute for real political food but it can compensate for the fact that most people do not digest large amounts of campaign material (Franz et al. 2008). The simple fact is most people are encumbered with many cares and have neither the time nor the inclination to search for and delve into political news. Political marketing, primarily in the form of television and radio advertisements, provides shortcuts for people to find relevant political information.
A second benefit of political marketing comes directly from political organizations developing a marketing orientation. The marketing concept places the consumer at the center of the exchange relationship. In the political realm this means that voters' needs and wants must be kept foremost in the minds of candidates and leaders. Failure to consider the voter as a consumer can result in losing touch with their wishes which can lead to losing elections and offices. This stands in sharp contrast to the product-oriented approach which focused on what was best for the organization's growth and development. This has necessitated a shift in terms of who the real power brokers are. The power now resides with professional campaign staffs who are in some senses the real “king makers” in modern American politics (Newman 2004). It also means that influence has to some extent shifted back to the voters. Instead of being taken for granted, politicians solicit their advice and feel a need to be responsive to them. Ultimately the adoption of a marketing-orientation approach can be the catalyst for reinvigorating interest in and enthusiasm for elections and government.

22
Conclusion and Outline of Dissertation
The purpose of this essay has been to investigate some issues relating to political marketing and why it should be taken seriously by marketing scholars. The reluctance of mainstream marketing academics to accept political marketing fully has been documented and addressed. Three key reasons--campaigns matter, politics has become a big business, and marketing is impacting the practice of politics-were offered as arguments as to why marketing scholars should look at political marketing with less skepticism and doubt. The lack of acceptance in mainstream marketing, it was noted, is due, in part, to the failure of political marketing devotees to make the subdiscipline more germane to the field of marketing in general. With further research, perhaps political marketing will cease to be an elephant in the room that everyone is aware of, but no one pays particular attention to. It is hoped that the arguments presented in this essay take an important first step toward bridging the gap.
The remaining parts of this dissertation will proceed as follows. The next chapter is an examination of how and why political campaigns actually matter. This is a subject that merits further attention because it is still a mater of some debate. On the one hand, there are those who dismiss campaigns entirely as interesting exercises that are, in the final analysis, not particularly meaningful. On the other extreme are those who argue that campaigns are extremely important and play a major role in deciding who the next leaders will be. The questions of how and why campaigns matter are evaluated using a panel data set that was collected during the 2004
American presidential. It will be shown that campaigns do matter but primarily with people who are undecided about whether they will vote.
The third chapter looks at voter segments over the course of several U.S. presidential elections. The segments are derived via latent class regression using data from the premier political science data set, American National Election Studies (ANES). The main idea is to

23 examine how the voter segments have changed over time and how policy attitudes have shifted and changed too. The research presented in chapter three shows that many voter segments are fairly consistent over time but there are smaller segments which arise in any given election year.
This suggests a two-pronged political marketing approach: a strong get out the vote (GOTV) effort with a party's base, and more persuasive “Why you should vote for me” appeals to the smaller segments. Several new research directions are also discussed.

Chapter 2

How Much Does Political Marketing Matter?
Several weeks after the 2004 American presidential election, Democratic National
Committee Chairman Terry McAuliffe lamented, “They were smart. They came into our neighborhoods. They came into Democratic areas with very specific targeted messages to take
Democratic voters away from us. They were much more sophisticated in their message delivery”
(Edsall and Grimaldi 2004). McAuliffe's “they,” of course, was the Republican Party which had succeeded in its efforts to re-elect George W. Bush as president, and had maintained, for the time being, its majority in Congress. The “smart” thing the Republicans did was to employ a combination of customer relationship management (CRM) databases, traditional political marketing, and sophisticated data analyses to identify and court small groups of persuadable voters-a process known in political circles as microtargeting or narrowcasting (Allen and Carney
2006, Lundry 2004, Seelye 2004). The difference between winning and losing, at least according to Terry McAuliffe, hinged more on the way the Republican message was delivered than it did on the message itself. In essence, he was arguing that it is just as important to deliver the message to the right people as it is to deliver the right message to the people.
The implicit assumption in Mr. McAuliffe's statement is campaigns really do make a difference. In other words, the events of the campaign actually influence the final electoral outcome. Surprisingly there is a great deal of disagreement amongst political scholars and academics as to whether and how much campaigns, and hence political marketing, matter
(Gelman and King 1993). One school of thought, although now largely discredited, suggests that campaigns do not matter at all. This is best exemplified by V.O. Key's (1966) famous assertion that campaigning amounts to nothing more than “manipulation by skilled humbugs.” Some

25 scholars have opted for a more moderate approach, often called the “minimal effects hypothesis," which argues that campaigns might matter but the outcome of campaigns is largely predictable several months in advance. According to the minimal effects framework, simple models, based on the status of the economy and the incumbent president's popularity, can predict the election outcome before the campaign season begins in earnest. If the ultimate goal is to understand who wins elections and why, there is little to be gained from studying the campaigns themselves
(Campbell 2000, Holbrook 1996, Abramowitz 1988).
On the other end of the spectrum sit the political consultants and professional campaigners who see campaigns as determinative. This is hardly surprising given the fact that their livelihoods depend on convincing people that campaigns do matter. This approach, which
Gelman and King (1993) termed the “Journalists' Model" views the campaign as a “horse race" which interprets “each short-term change in the public opinion polls as a serious change in the likely fortunes of the candidates.” Clearly there is a divide in thinking with many political scientists arguing for minimal effects and many professional campaigners arguing that the right campaign can propel any candidate to victory regardless of incumbent popularity or national economic conditions. As two political scientists recently noted, “There is perhaps no wider gulf in thinking than that between academics and political practitioners on the question of “do campaigns matter?” Campaign professionals tend to view election outcomes as singularly determined by the campaign itself, while political scholars often treat election outcomes as a foregone conclusion
(Hillygus and Shields 2008).
It is not entirely accurate to say that all political scientists subscribe to the minimal effects hypothesis (Hillygus and Jackman 2003, Finkel 1993, Iyengar and Simon 2000). There is a growing literature that looks at how campaigns affect voters' decisions. As of yet no consensus has been reached in this important area. There is, however, a sense that campaigns really do matter even if there are competing ideas as to how and why they matter. The purpose of this

26 essay is to delve deeper into this timely and interesting topic. The remainder of the essay will proceed as follows. First, there is an examination of the minimal effects hypothesis and the
“campaigns matter" point of view. Second, there is a discussion of what it means to argue that campaigns matter. Third, a data set from the 2004 presidential election is analyzed to see if there is empirical evidence suggesting that campaigns matter. Finally, there is a discussion of what has been learned thus far and what future research directions can be pursued in this field.

The Minimal Effects Hypothesis
The minimal effects hypothesis has its roots in the pioneering work of Lazarsfeld,
Berelson, and Gaudet (1944) and Berelson, Lazarsfeld, and McPhee (1954). According to their perspective, campaigns show little ability to induce voters to switch their choice for president. In fact, campaigns seem to have the opposite effect. During the course of the campaign, voters appear to become even more committed to their initial choice. These findings spawned a stream of research that became the “minimal effects” school. According to the minimal effects model, presidential elections are largely predictable based on “national conditions” variables which include the state of the economy, considerations as to whether the country is headed in the right direction, and retrospective evaluations of the president's performance in office (Abramowitz
1988). Numerous authors have concluded, with a reasonable degree of success, that models incorporating these factors can accurately predict the eventual winner more than six months before a presidential election (Campbell 2000, Flanigan and Zingale 2006, Holbrook 1996). That being the case, they argue, there is little need for examination of campaigns because, in the final analysis, campaigns do not sway voters in any meaningful or systematic manner and can thus be watched with bemused detachment (Bartels 1992). It is not entirely correct to suggest that

27 minimal effects advocates view campaigns as nothing more than a “tempest in a teapot.” Instead, they contend that net effect of campaigns is at most negligible (Shaw 2006).
Although political marketing scholars would naturally disagree with this conclusion, the
“minimal effects” school of thought has some merit. First, not every election is closely contested and some elections may be decided with or without the candidate running a solid campaign. For example, in 1932 national conditions were such that the electorate was willing and ready to vote for anyone but the incumbent. Similarly in 1964, there was still a great deal of sympathy towards
Lyndon Johnson due to the Kennedy assassination. It might be successfully argued that no amount of campaigning or political marketing was going to have a significant impact on the outcome in that particular election. Holbrook (1996) and Campbell (2000) argue that the 1980 election was similarly dominated by the state of the economy and dissatisfaction with the incumbent. Second, minimal effects proponents point out, correctly, that it is hard to state unequivocally how campaigns affect voting decisions. In other words, it is extremely difficult to state with any degree of certainty that political advertisements influence a certain segment while another group's preferences shift due to relative performance in the presidential debates. Because it is too hard to assess the relative impact of campaign events, minimal effects scholars focus their attention on the variables they deem more salient (Miller and Shanks 1996).
To support their assertions, minimal effects scholars have posited a relatively simple regression model (Abramowitz 1988). This model is denoted as:
V = β 0 + β 1 * Popularity + β 2 * GDP + β 3 * Incumbent + ε
The terms have the following meaning:
V = Incumbent party's percentage share of the vote--only the Democrat and Republican vote share is counted.
Popularity = the incumbent president's approval rating (in percentage points) as measured by the Gallup Poll in November of the election year. In 1960 and 1976, no Gallup Poll

28 was conducted until December. For those years, the approval rating was calculated by averaging the closest pre-election poll with the December results.
G.D.P. (Gross Domestic Product) = the percentage change in real G.D.P. from the beginning of the election year until the beginning of the fourth quarter of that year.
Incumbent = Measures whether the incumbent president's party has been in power for four or eight years. Four years is coded as 0, and eight years as 1. The idea behind this term is there may be an anti-incumbent feeling or a desire for fresh leadership after eight years.
The estimated equation is:
V = 35.234 + 0.294 * Popularity + 1.015 * GDP -3.17 * Incumbent
The original work using this model was published in 1988, thus encompassing 19481984. The results presented in Table 2-1 are an updated version which includes subsequent elections through 2004.
TABLE 2-1
Incumbent Party's Share of Major Party Vote, 1948-2004
Year

Actual Vote
Share

F 3,11

1948
52.37
1952
44.55
1956
57.75
1960
49.92
1964
61.34
1968
49.59
1972
61.79
1976
48.95
1980
44.69
1984
59.17
1988
53.90
1992
46.55
1996
54.74
2000
50.27
2004
51.24
= 24.29, p-value = 3.743e-05, R2=0.8688

Predicted
Vote Share
52.13
43.27
57.50
51.32
60.74
49.28
58.73
50.04
42.46
57.97
51.18
50.94
55.56
52.33
53.37

Difference
0.24
1.28
0.25
-1.40
0.60
0.31
3.06
-1.09
2.24
1.20
2.73
-4.40
-0.83
-2.07
-2.12

29
There are several important findings that come from this basic regression model. First, a simple model using presidential popularity, national economic conditions, and incumbency, fits the data quite well with an R2 = 0.8688. This would suggest that the minimal effects hypothesis may actually be useful. Second, of the 15 elections included in the sample, 11 of them are predicted correctly. Only the 1960, 1976, 1992, and 2000 elections were off target. Three of those elections (1960, 1976, 2000) were some of the most closely contested elections in U.S. presidential history. The fourth election, 1992, was characterized by a very strong showing for a third-party candidate, H. Ross Perot. Clearly there is a case to be made that most elections can be predicted several months in advance. The results, however, do not show conclusively that campaigns do not matter at all.

The “Campaigns Matter” Perspective
For political marketing scholars, and those political scientists who disagree with the minimal effects hypothesis, there is ample evidence to suggest that campaigns really do make a difference ((Finkel 1993, Hillygus and Jackman 2003, Shaw 1999, Wleizen and Erickson 2001).
There are several compelling reasons undergirding this argument. First, the results in the preceding section show there is some room for campaigning, although it may be the case that campaigns do not matter equally in every election. For instance the results for 1964, 1972, and
1984 show that the incumbent won by a large margin in each of those years. It may be that conditions were such that no amount of political marketing efforts by the other party were going to make a difference. Second, there is some evidence of declining partisanship in the United
States, although reports of its demise may be greatly exaggerated. Longitudinal analysis of
American National Election Studies (ANES) data shows that partisanship has dipped a few percentage points over the last thirty years, but it is currently holding quite steady. This would

30 imply that there are more “Independents” or “Moderates” in the voting public. Furthermore, although partisanship is a strong determinant of voting choice, it is not deterministic. There are ample occasions when people who identify themselves as belonging to one party vote for the other party's candidates. One famous example is the “Reagan Democrats” who switched party loyalties to support the candidate who more closely matched their personal and policy preferences in 1980. For the most part, however, partisan preferences do drive vote choice, but in the right situation, those effects can be overcome and voters who are considered brand loyal can become brand switchers (Hillygus and Shields 2008).
A second argument as to why political campaigns matter comes from the fact that at least some voters do not make up their minds as quickly as the minimal effects model would suggest
(Fournier, Nadeau, Blais, Gidengil, and Nevitte 2004). If national economic conditions and the incumbent's popularity explain to a large extent how people vote, then most voters should have already made their vote choice long before the typical campaign season, which usually runs from the end of the first round of presidential primaries in February until Election Day in November.
In the last several rounds of the ANES survey, respondents have been asked to indicate when they made their decision as to who they would vote for. The results, presented in Table 2-2, suggest there is still a great deal of room left for political persuasion:
TABLE 2-2
Timing of Vote Decision
When Decided
N
Knew All Along
273
During Primaries
93
Before Conventions
85
During Conventions
121
Between Conventions and
226
Election Day
On Election Day
18
Source: American National Election Studies 2004

Percentage
33.4
11.4
10.4
14.8
27.6
2.2

31
The numbers in Table 2-2 indicate that nearly two-thirds of voters had not made a final decision before the campaign season begins. Although this does not provide evidence that these voters are not committed partisans or that they will give equal weight to both candidates, it does show that there is an opening for some persuasion to take place. Cross tabulation results show that there is a statistically significant difference between the two groups in terms of interest in political campaigns (χ2= 19.56, p < 0.00). Not surprisingly, the early deciders are also more partisan or committed to their party or ideology. Of the 818 voters polled as part of the ANES
2004, 545 might reasonably be labeled "Persuadable." It should be noted that nearly 30 % of voters wait until after the conventions to make their decision. This presents solid evidence that there is a significant number of people who are watching with interest. That should give the nominees hope that they can reach those voters and make some appeal to win their support.
Again, no claim is made as to how truly receptive they are to political marketing, but at least there is a glimmer of hope for those candidates who might want to approach this segment of the voting population. If voters who fall into the “persuadable" category do not pay much attention to political campaigns, then political marketing truly would be relegated to the margins. ANES data demonstrate, however, that people who do not make their minds up before the campaign has started tend to be involved with politics. Table 2-3 presents the self-reported level of interest in political campaigns among the later deciders. The results in Table 2-3 show that the potentially persuadable segment of the population is not politically disengaged. Over half exhibit a strong interest in politics and a further 40.4 % state at least a passing interest in campaigns. This provides further evidence that there is room for political marketing and campaigns.

32
TABLE 2-3
Interest in Political Campaigns among Campaign Deciders
How Interested
N
Very Much Interested
301
Somewhat Interested
220
Not Much Interested
24
Total
545
Source: American National Election Studies 2004

Percentage
55.2
40.4
4.4
100.0

A third piece of evidence for the campaigns matter hypothesis is how involved the
“persuadable" segment is with various media. Table 4 lists the different media that people use to gather information about the election and candidates. For instance, nearly 90 watched at least one television program that was dedicated to the election. Almost half of respondents listened to political discussion on talk radio and over two-thirds of those who had internet connections in
2004 went online to look up campaign information.
TABLE 2-4
Where Campaign Deciders Get Information
Source
Television
Magazines
Rallies, Speeches
Talk Radio
Internet
*124 respondents had no internet access.

N
Percentage
483
88.6
188
34.5
49
9.0
257
47.2
287
68.2*
Source: American National Election Studies 2004

Table 2-4 suggests some interesting things about the American electorate. First, most information is gleaned in a passive manner. People are seeking for information but they are not inclined to show up at campaign rallies and speeches. Voters may be covertly engaged rather than engaging in overt political behavior. That may, of course, just be a function of fewer opportunities, but undoubtedly more people are choosing to stay home or listen in the car than go attend campaign activities. For politicians this implies that the media message delivered over television, the internet, and radio is where one's efforts should be focused. Most people cannot show up in person or even if they could, they may be disinclined to do so.

33
Although evidence has been presented to suggest that campaigns matter, the issue of how campaigns matter remains a cogent yet more difficult question to answer. Some scholars have made a foray into this area and have presented ideas that draw heavily on the same literature that informs consumer behavior research (Druckman, Jacobs, and Ostermeier 2004, Iyengar and
Simon 2000, Popkin 1994, Wleizen and Erickson 2001). The basic premise underlying the “how campaigns matter" literature is campaigns prime the voting public by emphasizing issues that matter to the candidate or by highlighting core competencies the candidates possesses (Druckman et al. 2004). A key example of this is found in the 1992 presidential election. During that campaign, the Clinton staff adopted the motto “It's the economy, stupid!" as a reminder of the recent recession. Although the economy had technically recovered by November 1992, voters were sufficiently motivated to vote for Clinton, because he had successfully positioned himself as someone who could competently guide the economy back to a more robust status. Other examples include Reagan's famous question, “Are you better off than you were four years ago?” and Theodore Roosevelt's foreign policy slogan, “Speak softly and carry a big stick.”
These examples highlight some key issues regarding priming. First, priming works when the public perceives the candidate as someone who can advantageously handle the issue. Clinton and Reagan were seen as people who could fix the economic malaise. On the other hand, Herbert
Hoover unwisely chose to state that there would be a “chicken in every pot.” The public did not find his claim credible and he was soundly defeated in his reelection bid (Druckman et al. 2004,
Miller and Krosnick 2000, Petrocik 1996). Second, priming works when the public agrees with a candidate's position. Ronald Reagan's slogan that emphasized how people were worse off as a result of the Jimmy Carter presidency resonated with a large enough percentage of the public and helped him win the 1980 election (Riker 1996). Finally, priming matters when the public views the issue that is being primed as important. Examples abound, e.g. Alan Keyes strong stance on abortion in 1996, for instance, of candidates who feel strongly about an issue, who are capable of

34 addressing the issue, but whose campaigns falter because they choose the wrong issue to prime.
In order for the campaign to matter to the electorate, the policies and promises brought to the forefront must excite or interest people on a certain level (Quelch and Jocz 2007).
The extant literature on campaigns has made limited inroads into explaining how campaigns make a difference. Scholars have successfully argued that campaigns do matter in that they persuade voters by bringing certain issues to the foreground of national consciousness, but more work needs to be done in this area. This is a field where marketing academics could apply their work and thus demonstrate how the insights gained from marketing can have broader applicability outside the traditional disciplinary confines.

What Does “Do Campaigns Matter?" Mean?
Thus far, attention has been focused on debates of whether campaigns actually matter.
No attempt has been made, however, to define exactly what it means to say that campaigns do matter (Gelman and King 1993). From a political science perspective, campaigns matter when they have somehow affected the way people vote. Obviously this can mean many things which explains, in part, why political scientists have grappled with this issue for so long. Marketing as a discipline, on the other hand, tends to be more focused on results and less so on process. From a marketing perspective, the question of “Do campaigns matter?" can be boiled down to three inquiries. First, has the campaign in someway activated voters' pre-existing dispositions? In other words, did the political marketing efforts remind Democrats as to why they are Democrats? In this sense, campaigns matter when they reaffirm voters' beliefs. Second, did the campaign convince anyone to switch sides? Research shows that relatively few people ever do switch from
Democrat to Republican or vice versa, but, nevertheless, it still happens. Closely contested elections are won or lost on the margins which implies even small numbers of switchers can make

35 a large difference, particularly in a “winner takes al” electoral system. Third, campaigns can be considered meaningful if they persuade people to vote who might not have otherwise done so.
There has been a steady decline in voter turnout in the United States over the last forty years.
This means there is a huge untapped market that, if properly approached, can lead to large electoral gains. If a candidate can persuade enough non-voters to “purchase” a particular political brand, then that candidate is a much stronger position to gain elected office.
The first question regarding the relevance of campaigns has been analyzed in great depth within political science circles (Campbell 2000, Gelman and King 1993, Holbrook 1996, Iyengar and Simon 2000, Key 1966). No attempt is made presently to take up this issue. The second question, “Who switches sides?”, and the third question, “How can nonvoters be persuaded to vote?" will be examined in some depth. The political marketing implications of the answers to these questions will then be analyzed and discussed in further detail.

Data and Analysis
The data are taken from a panel data set that was produced as a joint effort between
Brigham Young University and the University of Wisconsin during the 2004 presidential election campaign. A nationally representative sample, with oversampling in the key battleground state of
Ohio, was interviewed at three stages during the election cycle: late June into early July, the beginning of September, and immediately after the November election. As with any panel data there was attrition. The initial wave included 2782 respondents, the second wave consisted of
1523 people, and the final wave has a total sample size of 1438. In order to ensure representativeness, the final sample was poststratified along geographic and demographic dimensions. The interviews were conducted over the telephone and lasted approximately thirty minutes. 36
Respondents in each stage were asked a wide variety of questions ranging from religious practices and feelings about the candidates to voting intentions and media usage. In each wave, participants were asked specifically about media usage, i.e. how many hours of television they watch in a typical week, how many hours are spent listening to the radio, how many days in a week does the respondent read the newspaper, etc. Additionally, people were asked about contact with the presidential campaigns. For example “Did you receive an email message from one of the campaigns?” and “Did you receive any letters from the campaigns?” and, if so, how many letters? Other areas that were covered include face-to-face conversations, requests for donations, phone calls, television advertisements, and radio advertisements. Respondents were asked about their partisanship from liberal to conservative, and their partisan affiliation “Strong Democrat” to
“Strong Republican.” Finally, the survey asked people whether they intended to vote and for whom. In the final wave, participants were asked to identify their vote choice.
Panel data are ideally suited for the study of campaign effects. Because the same people are interviewed repeatedly, it can help pinpoint whether campaign communications influenced the way people vote. It also allows researchers to identify which voters switched their voting intentions. This can be accomplished by comparing stated candidate preference in the first wave of the panel with final vote choice in the last wave of the panel. It also permits one to investigate whether intention to vote changed over the course of the campaign. This is done by again comparing a person's response in the first wave with their self-reported actions in the final wave.

Who are the Switchers?
The question of who switches sides is easily observed with panel data. Participants were asked in wave 1 which candidate they intended to support. This was then compared with their actual vote choice in the final wave. For instance a person who states that she will vote for John

37
Kerry in wave 1 and then reports voting for John Kerry in wave 3 is considered a non-switcher.
Likewise a person whose stated intention in wave 1 differs from wave 3 is thus considered to be a switcher. The dependent variable is, therefore, a binary variable with non-switchers coded as 0, and switchers coded as 1. In the final sample, just under % actually switched--meaning there were approximately 100 switchers in total.
Several models were then estimated using a Bayesian binomial probit approach. The choice of a probit model over a logistic regression specification is simply a choice of convenience. As Lynch (2007) notes, “the probit model, from a probability standpoint is easier to work with, because it involves use of the normal distribution." The framework of this model is given by:

The models were estimated with the Zelig package in the R statistical package (Imai,
King, and Lau 2006). The program supplies default priors which were used in all analyses. The independent variables differed from model to model. The first model, a demographics only approach, included age, gender (Female=1), years of education, whether the respondent lived in a
“battleground” state, and marital status. The variables were chosen for the following reasons.
Age is expected to lessen the likelihood of switching because people tend to get set in their ways as they grow older. Education would presumably increase the chances of switching because educated people are assumed to be open-minded and thus more likely to consider alternatives before making a final choice. Battleground state, which is defined as a state that both parties identify as important to winning the presidential election, is included because those states receive

38 more attention and political marketing efforts during the course of a campaign. Finally, gender and marital status are included as typical demographic variables. Because the models are estimated in a Bayesian framework, the results do not show significance as in the typical framework. Instead the 95 % highest posterior density region is reported with 2.5 % as the lower bound and 97.5 % as the upper bound.
TABLE 2-5
Demographics Only Probit Model
Variable
Intercept
Age
Female
Education
Battleground State
Married

Mean
-1.322
-0.085
0.056
-0.032
-0.027
0.157

Std. Dev.
0.316
0.045
0.131
0.036
0.136
0.147

2.5%
-1.939
-0.175
-0.199
-0.104
-0.291
-0.123

97.5%
-0.706
0.004
0.315
0.038
0.238
0.449

The results of the “Demographics Only” model indicate that none of the variables selected is predictive of switching behavior. This means that switchers are not significantly different from the population as a whole. Given the lack of predictive power, this model is enhanced by adding “political fundamentals” variables. These variables are strength of partisanship and general interest in government. It is assumed that the stronger one's partisan identification is, the less likely that person is to switch. Also, the more attention one pays to political matters, the less inclined that person is to switch sides during the course of a political campaign. It should be noted here that strength of partisanship in this context is measured on a four point scale. A “hardcore" Democrat or Republican is coded as 1, a “weak” partisan receives a 2, a “independent but leaning” Democrat or Republican is a 3, and self-identified “True independents” are assigned a value of 4. General interest in politics and government is measured on a four-point scale, with higher numbers indicating less interest or attention.

39
TABLE 2-6
Political Fundamentals Probit Model
Variable
Intercept
Age
Female
Education
Battleground State
Married
Partisanship

Mean
-2.704
-0.043
0.010
-0.010
0.042
0.146
0.280

Std. Dev.
0.473
0.048
0.131
0.036
0.140
0.150
0.073

2.5%
-3.630
-0.139
-0.207
-0.101
-0.235
-0.145
0.137

97.5%
-1.771
0.050
0.305
0.038
0.313
0.447
0.424

Interest

0.227

0.085

0.061

0.393

As one would expect, declining levels of partisanship and interest in government increase the probability of switching sides during a campaign. This comports with the extant literature on the effect of political campaigns (Bai99). From a marketing perspective, it is essential to identify who the less partisan people are and then make appeals to that particular segment of the voting population. The final step is to evaluate whether campaign communications led to any switching behavior. The final model, the political marketing model, includes survey items about media usage which indicates levels of exposure to the campaigns and items about direct contact with the campaigns themselves.

40

TABLE 2-7
Political Marketing Probit Model
Variable
Intercept
Age
Female
Education
Battleground
Married
Partisanship
Interest
Letter
Phone Call
Face-to-Face
Email
Radio Ad
TV Ad

Mean
-3.050
-0.024
0.015
-0.009
0.125
0.097
0.284
0.250
-0.150
-0.233
-0.010
0.095
0.152
0.345

Std. Dev.
0.654
0.057
0.149
0.041
0.156
0.158
0.079
0.092
0.162
0.160
0.200
0.205
0.176
0.381

2.5%
-4.348
-0.136
-0.279
-0.091
-0.180
-0.208
0.131
0.072
-0.470
-0.547
-0.410
-0.329
-0.192
-0.348

97.5%
-1.827
0.084
0.314
0.071
0.435
0.410
0.442
0.431
0.163
0.086
0.374
0.478
0.513
1.131

As can be seen, the campaign contact variables fail to add any explanatory power to the question of “Who switches?" In fact the only variables that appear to make a difference are strength of partisanship and general interest in politics and government. As partisanship decreases and attention to politics decreases the probability of being a campaign switcher increases. This suggests that there is limited support for assertions that campaigns matter because they get voters to change their voting intentions. This does not necessarily mean that political marketing really is a mere “tempest in a teapot” but rather that there are other more enduring factors that drive voting decisions. A final step is to include the interaction of strength of partisanship with the political marketing variables and general interest with the political marketing variables. The reasoning behind this model is strongly partisan individuals are less likely to be influenced by campaign contacts and people who are more interested in politics will have sought out election related information and are therefore less likely to switch due to marketing efforts from the campaigns.

41

TABLE 2-8
Political Marketing with Interactions Probit Model
Variable
Partisanship
Interest
Letter
Phone Call
Face-to-Face
Email
Radio Ad
TV Ad
Partisan: Letter
Partisan: Call
Partisan: Face
Partisan: Email
Partisan: Radio
Partisan: TV
Interest: Letter
Interest: Call
Interest: Face
Interest: Email
Interest: Radio
Interest: TV

Mean
1.995
0.147
1.603
-1.203
-0.297
0.344
0.196
4.951
-0.241
0.273
-0.508
-0.306
0.131
-1.731
-0.487
0.160
0.548
0.168
-0.126
0.383

Std. Dev.
0.929
0.659
0.700
0.598
0.756
0.726
0.617
3.947
0.181
0.180
0.257
0.233
0.188
0.959
0.224
0.196
0.281
0.298
0.210
0.662

2.5%
0.609
-1.179
0.290
-2.367
-1.792
-1.089
-0.946
-0.851
-0.597
-0.081
-1.030
-0.776
-0.227
-4.157
-0.926
-0.232
0.028
-0.414
-0.555
-0.953

97.5%
4.311
1.500
3.009
-0.033
1.132
1.744
1.466
14.672
0.111
0.624
-0.006
0.146
0.504
-0.298
-0.047
0.543
1.120
0.746
0.272
1.659

Given space constraints, the first few rows of the table have been omitted because the variables were not significant. Attention is focused on the political marketing variables and the interactions. As can be seen from table 2-8, when the interaction terms are added in, “general interest" is no longer a strong predictor of voting behavior, but “strength of partisanship" remains strong. As a general rule, however, the interactions are of greater interest. In this case, the interaction of “face-to-face" with “strength of partisanship" has a negative effect. This would suggest that direct contact with someone from a campaign seems to dissuade weak partisans. On the other hand, the interaction of “general interest" with “face-to-face" has a positive effect on switching behavior. For those who are interested in politics, a face-to-face conversation is more

42 likely to lead to a change in candidate preference, which runs counter to the expectations stated above. The results of the regression model with interactions suggest two things. First, impersonal campaign contacts (e.g. letters, email, radio and television advertisements) do not have much of an impact on preference switching. Second, personal campaign contacts, particularly face-to-face contact with someone from the campaign, has a mixed effect. For people who are not highly partisan, personal contact seems to encourage their existing choice. For those who are highly interested in campaigns and politics, the direct, personal approach seems to work best (RosRob90). From a practical perspective this means campaigns need to use their face-toface efforts judiciously. Instead of sending out hordes of volunteers to canvass an entire neighborhood, they need to find out beforehand who the interested people are and then limit their efforts to speaking with those people. Of course, in reality, that necessitates a great deal of research beforehand (i.e. finding out who is interested, who is partisan etc.) which may be cost prohibitive. Nevertheless, the results do show that learning about voters can be helpful in understanding how best to contact or approach them.

From Not Voting to Voting
The next question that needs to be answered is do campaigns really matter in terms of persuading people to vote who might otherwise not vote. This is determined in the following manner. In the first wave of the BYU-Wisconsin study, respondents were asked if they planned to vote. Most (n=1328) indicated that they would definitely or probably vote in the upcoming election. A small number (n=110) told interviewers that they would probably not or definitely not vote that November. There were 53 people in the “probably will not vote" category and 57 in the “definitely will not vote" group. The final voting behavior of these groups was compared

43 with the intention they stated in the first wave. Of the “probably will not vote" group a total of 31 actually voted in the general election. Of the “definitely not group" only 12 cast a ballot in
November 2004. Because these groups are different in terms of their stated intentions, binomial probit models were run for each group. In both cases a voter was coded as 1. The independent variables were simply the political marketing variables from the previous section. No attempt was made to analyze who does not vote and why not. There is already a large literature investigating the phenomenon of non-voting behavior.
TABLE 2-9
Campaign Effects on Unlikely Voters
Variable
Intercept
Letter
Phone Call
Face-to-Face
Email
Radio Ad
TV Ad

Mean
-1.141
0.885
0.529
40.617
0.216
0.321
0.511

Std. Dev.
0.586
0.437
0.406
20.598
0.943
0.441
0.611

2.5%
-2.353
0.024
-0.249
4.359
-1.673
-0.547
-0.645

97.5%
-0.043
1.738
1.335
71.644
2.083
1.207
1.765

Interestingly enough, it appears that direct campaign communications do influence people who are debating whether or not to vote. Direct mailings, in the form of letters, have an effect but the largest effect is due to face-to-face communication with someone from the campaign. This implies that old fashioned strategies such as canvassing a neighborhood can lead to increased voting. Unfortunately for presidential campaigns the cost of going door-to-door requires a substantial grassroots campaign or a lot of money. The last step is to see how these same variables impacted the people who are determined not to vote. The results of the final probit regression model are presented in Table 2-10.

44

TABLE 2-10
Campaign Effects on Nonvoters
Variable
Intercept
Letter
Phone Call
Face-to-Face
Email
Radio Ad
TV Ad

Mean
-2.551
1.810
0.383
-7.074
-0.167
0.039
0.506

Std. Dev.
0.838
0.629
0.575
5.843
1.035
0.650
0.868

2.5%
-4.387
0.660
-0.745
-23.330
-2.211
-1.235
-1.079

97.5%
-1.067
3.127
1.499
0.054
1.860
1.329
2.293

Once again direct mail, e.g. a letter, seems to persuade non-voting individuals. There is no significant effect with other marketing variables including face-to-face conversations with campaign representativeness. Further work needs to be done to determine why this is the case.

Discussion
The evidence as to whether campaigns matter is somewhat mixed. There is no support for a direct effect in terms of campaigns persuading people to switch voting preferences. When the interactions are added, personal, face-to-face contact affects voters although differently depending on strength of partisanship and general interest in campaigns. This is an area that warrants further investigation. Overall, the results would suggest to political marketers that there is little to be gained from endeavoring to entice people to switch party affiliations. This recommendation depends, however, on how close the election is going to be. If the candidates are in a tight race, it may be worthwhile to pursue the switching segment. Unfortunately this task is complicated by the fact that there are no easy demographic predictors of who will switch.

45
Perhaps a more fruitful approach is to rally one's “natural” consistency in an effort to get them to vote in greater numbers (Hamburger and Wallsten 2006).
It has been shown that campaigns can indeed influence people to vote who otherwise might not have. People who are unsure of their decision to vote can be persuaded by direct contact from the campaign. The case is not so clear for that segment of the population that is committed to not voting. A small number of them eventually end up voting but the vast majority remain committed to staying away from the ballot box. The evidence suggests that the oldfashioned approach seems to work best. It is vital for political marketers then to look at grassroots campaigns and determine how they can best build up a network that will enhance “Get out the vote” (GOTV) efforts (Bergan, Gerber, Green, and Panagopolous 2005). As has been pointed out elsewhere, however, the rise of the modern campaign has led to a decline in the number of “boots on the ground.” As campaigns have become more professionalized they have become less reliant on volunteers and more dependent on consultants who actively control the image and message of a campaign.
Finally, the title of this essay asks the question, “How much does political marketing matter?” It has been shown that there are situations when political marketing does not seem to have much of an effect. This is perhaps why the “minimal effects” school of thought arose in the first place. On the other hand, there are situations which tend to support the hypothesis that campaigns really do matter. This research has demonstrated that there is merit to both the
“minimal effects” argument and the “campaigns matter argument.” In other words, as with many issues, the truth lies somewhere in the middle. The answer to the initial question, “how much does political marketing matter?” is it can matter a great deal but it depends on the context. In terms of actual campaign practice it would make sense to continue campaigning as though the election results do depend on how well the campaign is run.

46
Limitations
There are several limitations to this research, mostly as a direct result of the quality of the data. First, although the data set is a panel data set, many of the measures are collected only once. For example, there are no repeated measures of many of the political marketing variables.
For many of the political marketing variables there is simply a binary variable indicating whether a person heard an advertisement on the radio or saw a political commercial on television.
Second, there is nothing in the data that indicates the partisanship of the political communications one was exposed to. For instance, all that is known is whether a person received a letter from one of the candidates. There is no indication as to how many were sent by Democrats and how many were sent by Republicans. In future research, it would be critical to have that information.
A final limitation of this research is there is no way of knowing when a voter who actually did switch decided to do so. Nothing is known about the timing of the actual decision to switch. Ultimately the only way to model that information would be to track a large sample of voters on a frequent basis over the course of an entire campaign. A daily tracking poll of the same people would then allow researchers to see at what point a person decided to change their vote or at what point a person decided to become a voter. This type of information could be modeled using Hidden Markov Models (HMM). Some initial forays have been made in that direction, but more work remains to be done (Hillygus and Jackman 2003).

Future Research
The future of “campaign effects" and electoral research is online. As pointed out in the limitations section, there is a pressing need for daily tracking of the same panel. In the past this has been done with traditional polling methods that would typically involve a daily phone call.

47
This method grows burdensome very quickly. A better approach is to recruit a committed panel of voters who are willing, for some type of remuneration, to record their campaign contacts and feelings on a frequent basis. A few firms, primarily Palo Alto based Knowledge Networks, have made steps in that direction. More work is needed, however, to truly study the phenomenon of how much political marketing matters.
Political marketing is an important area within the marketing discipline. As has been demonstrated in this essay, political marketing matters. The degree to which it matters depends on several contextual factors. Future research can hopefully demonstrate more clearly how it matters. That knowledge can then be used to improve the practice of modern political campaigns.

48

Chapter 3
Political Taxonomies: Voter Segments in Presidential Elections, 1988-2004
In recent years, sobriquets like “soccer moms,” “NASCAR dads," “angry white men,” and “mortgage moms,” have become an important part of the American political lexicon
(Flanigan and Zingale 2006, Elder 2007, Seelye 2004). Each of these terms represents a segment of voters who vote more or less en bloc and thus merit serious attention from political candidates.
Soccer moms, for example, are middle class housewives who hurriedly shuttle their children from activity to activity in minivans. They tend to vote Democratic and were key to Bill Clinton's victories in 1992 and 1996. NASCAR dads, on the other hand, are blue collar workers, devotees of stock car racing, and fans of country music (Russell 2006). Like soccer moms, NASCAR dads live in suburbia or the South. Unlike soccer moms, they vote Republican and were a crucial component of George W. Bush's re-election strategy in 2004. With each election cycle, new segments of voters are identified and courted in hopes of gaining victory. The challenge for political candidates in future elections, therefore, is to pinpoint and appeal to the clusters of voters that will ensure electoral success.
The very existence of labels for various clusters within the electorate is an overt acknowledgment that voters are a heterogeneous group. Perhaps motivated by a desire to be liked by everyone or to be seen as uniters, politicians have been reluctant to acknowledge this fact publicly (Baines 1999). Barack Obama, for example, famously said at the 2004 Democratic
National Convention, “The pundits like to slice-and-dice our country into red states and blue states: red states for Republicans and blue states for Democrats. But I've got news for them, too.
We are one people, all of us pledging allegiance to the stars and stripes, all of us defending the
United States of America (Baldassarri and Gelman 2008). Over the last forty years and since
1980 in particular, some presidential candidates have overtly recognized the inherent difficulty

49 and even the impossibility of trying to appeal to everyone. Instead, campaigns have adopted a strategy of finding and targeting the right combination of voter segments. The right combination, of course, is defined as the combination of voters that will win the election for the candidate.
Ronald Reagan's chief pollster Richard Wirthlin, for instance, examined numerous psychographic and demographic characteristics of the voting populace which helped him find a large bloc of voters which was subsequently named “Reagan Democrats.” In 2004, Karl Rove, the chief strategist for George W. Bush, “used computer models and demographic files to locate probable
GOP voters. They looked at what they read, what they watch, what they spend money on” (Balz and Allen 2004). This approach was widely credited for crafting a small majority, but nonetheless a majority, which secured Bush's re-election.
Although voter segmentation has been a critical element of presidential campaigns over the last four decades, relatively little attention has been paid to it in the academic literature. Some work, which will be summarized below, has been done on the different bases of voter segmentation, but no one in the marketing discipline has studied how voter segments have changed over time. The purpose of this essay is to fill that gap in the academic marketing literature by providing an overview of segmentation in the political marketplace. Special emphasis will be placed on examining voter segments over the course of multiple presidential elections in the United States. The segments will be identified via latent class regression analysis.
It is hoped that this research will shed some light on the changing nature of American presidential politics and provide some insight as to how marketing methods can be used to understand political phenomena.

50
Why Segment Voters?
In the simplest sense, voter segmentation is the political analog of market segmentation.
Market segmentation entails, “viewing a heterogeneous market as a number of smaller homogeneous markets, in response to differing preferences, attributable to the desires of consumers, for more precise satisfaction of their varying wants" (Wedel and Kamakura 2000).
Although there are numerous methods and bases for segmenting consumers, the ultimate goal is to distinguish clusters that are homogenous within, heterogeneous without, substantial, and actionable (Perreault and McCarthy 2003). The key is to identify the underlying sources of heterogeneity and then use a targeted response to satisfy the wants and needs of each segment.
Voter segmentation is thus the process of recognizing political heterogeneity and identifying smaller, homogeneous groups in the voting population. Segmentation is a crucial step in understanding what voters are looking for, which is, according to one political marketing scholar, the first guideline to electoral success (Newman 2001).
In addition to helping candidates understand what voters want, voter segmentation studies can help identify which voters are most susceptible to a campaign's marketing efforts. Given that resources are limited, it is important that they be used where they can achieve the greatest return.
For example, a liberal leaning candidate should not spend time and money advertising in known liberal enclaves. Likewise a conservative candidate is unlikely to expend funds and energies in those same areas. Instead, candidates want to narrow their efforts to those areas and voters who are “floating" or persuadable (Baines 1999). One political marketing firm noted in the 2004 that the Democrats segmentation efforts failed because, “Critical voters often didn't receive the level or type of attention necessary, while many of the wrong voters (including Republicans and consistent-voting Democrats) received a disproportionate share of resources and attention"
(YouGovPolimetrix 2008). A second purpose for segmenting voters is to uncover any potential

51
“markets” that previously did not exist or were ignored. In the 2004 election, for example, the
Bush campaign used segmentation and “found" a small pocket of Russian Jews in Mayfield
Heights, Ohio. This area was typically considered a Democratic stronghold but Bush's advisors felt this particular group would be amenable to the Republican message. A staff member was sent there to address a gathering at a local cultural center and in the election, which was held shortly thereafter, this segment voted heavily in favor of George W. Bush (Hamburger and
Wallsten 2006). These votes were crucial in providing Bush with a victory in a closely contested state. From a marketing perspective voter segmentation is a no-lose proposition and makes complete sense. Politically, however, there can be some drawbacks. First, it can be viewed, in the words of Barack Obama, as a divisive strategy. Instead of bringing everyone together, a segmenter is trying to find and exploit differences for political gain. This criticism has some merit, but, given the heterogeneity of the American population, it is simply impractical to appeal to everyone. Second, a segmentation strategy can create situations where candidates come across as trying to be “everything to everyone." Hypothetically speaking the message that appeals to
Democrat Segment A is likely going to be different than the message that appeals to Democrat
Segment B. Given enough segments, it will be difficult for a candidate to stay focused on a simple message and can lead to circumstances of saying things or adopting positions that are contradictory. This is perhaps best illustrated by Mitt Romney's unsuccessful presidential bid in
2008. Romney, whose campaign made extensive use of marketing research methods, adopted certain positions while serving as governor of Massachusetts. In order to win the Republican nomination, however, Romney felt he needed to appear more conservative so he reversed course on numerous social issues including abortion. Ultimately this effort appeared to many as mere pandering or political expediency. Finally, segmentation strategies can possibly lead to problems with governing. If an elected official is facing re-election, he or she may skew policies and

52 government funds towards those segments that will ensure another term in office. Taken to its extreme the use of government power and largesse could create privileged groups that would lock in a particular party's hold on office (Hamburger and Wallsten 2006). Such efforts and outcomes might be inimical to democracy and elections.

Voter Segmentation
With all its virtues and vices, voter segmentation is likely to remain an important part of political campaigns in Western democracies. It is important, therefore, from an academic perspective to study the phenomenon and its consequences. As of this writing, some attention has been paid to the subject in marketing journals. This work has appeared primarily in lesser or non
“A-level" marketing journals. A brief overview of the more important papers highlights some key considerations that need to be addressed in voter segmentation and will suggest avenues for further research.
The existing voter segmentation research covers several topics ranging from why nonvoters do not participate to the effects of trust on voting behavior (Bannon 2003, Dermody and Hanmer-Lloyd 2004, Schiffman, Sherman, and Kirpalani 2002). Other authors examine the use of psychographic and demographic variables as segmentation bases and some use political attitudes to distinguish between classes of voters (Ahmed and Jackson 1979, Newman and Sheth
85). Schiffman et al. (2002) use two well-established political trust measures (incumbent-based and regime-based trust) to see how trust impacts political involvement and actions. They find that people who trust the incumbent and the regime are more likely to engage in thoughtful consideration before voting and are more likely to watch televised debates and engage in political conversations with friends and colleagues at work. The authors used a convenience sample that consisted primarily of the friends and acquaintances of the survey administrators. They do show,

53 however, that there are significant demographic differences between segments. The trust construct is also used in an effort to segment younger voters in Dermody and Hanmer-Lloyd
(2004). They find that the causes of young people's political disaffection are complex, but trust and distrust explain a great deal of youth electoral participation. The authors identify four segments without explaining how they did so. They subsequently propose marketing strategies that can be used to entice young people to be more active in the political process. The strategies are different for each segment.
Smith and Hirst (2001) follow a slightly more elaborate procedure in their segmentation scheme. They use the British Social Attitudes survey that was administered shortly before the
1997 UK elections. The data set contains numerous items on lifestyle, demographics, and political feelings. Given the overwhelming scope of their data set, the authors initially performed a principal components analysis (PCA) to reduce the dimensionality of the data. They identified sixteen factors that were subsequently used to cluster respondents. The authors chose a sevencluster solution. Cluster membership was estimated using attitude data and other demographic variables. The clusters were then examined with regards to feelings about policy issues such as environmental protection, transportation, and public assistance to the poor. The primary focus of the piece by Smith and Hirst is substantive. They endeavor to show that policy variables and demographic variables can be a useful basis for segmenting voters
In their article on voter behavior, Newman and Sheth (1985) attempt to demonstrate that policy preferences are an important determinant of voting choices. They argue that economic policy, foreign policy, social policy, and leadership characteristics, along with several other factors, drive voting behavior. The authors factor analyze survey data and find seven factors, or domains, which are then summed up and divided by the number of items in each domain. In effect they create seven averages that are then used to perform a discriminant analysis that classifies respondents into one of two groups. In the last step, Newman and Sheth conducted a

54 second discriminant analysis using demographic variables. The second model predicted 70.4 % of respondents correctly, which indicates that demographic variables explain some, but not all, of the forces that affect a voter's behavior.
There are several themes running through the voter segmentation literature. First, there are significant demographic differences between voting blocs. On average women vote differently than men, older people vote differently than younger voters, and people of different ethnic and racial backgrounds vote differently. In some ways this makes the political marketer's job much easier in that it makes it easier to identify segments. Second, the underlying dimensions of voting decisions are complex and multi-faceted. Some people might vote because they trust the government and others might vote because they distrust the government. On the other hand some might not vote because they trust the government and others might abstain for the reason that they distrust any government. In other words, voters are a complicated group. Third, the literature demonstrates that segmenting voters actually does shed some light on voting behavior and can be useful tool for political campaigns. Smith and Hirst (2001) and Newman and Sheth
(1985) both demonstrate that dividing the voting populace into groups can help politicians understand those groups and what motivates them. This creates an opening for political marketing efforts which can, from a politician's standpoint, be used to persuade voters to vote for them. There are two glaring omissions in the literature thus far: methods and time. The existing research is based on simple methods and, as of yet, nothing has been done to remedy this deficiency. This essay employs latent class regression analysis to understand voting patterns.
Although this is not a novel technique, having been used in multiple marketing papers, it is, nonetheless, one that is ideally suited to issues of voting (Wedel and DeSarbo 1993, Wedel and
Kamakura 2000). The second omission is time. All of the existing studies treat segmentation as a one-shot process that is conducted solely for an upcoming election. No effort has been made to

55 understand how the voter segments shift and change over time. The second goal of this essay is to correct this oversight by examining several presidential elections in the United States. Thus the ultimate goal is two-fold. First, does the application of newer marketing research techniques cast a clearer light on voting behavior? Second, what does a longitudinal view of voter segments contribute to the understanding of politics in general and political marketing in particular? For instance are conservatives and liberals concerned about the same issues over time? Are conservative and liberal segments stable over time? Is there evidence that people move from segment to segment based on issues or do the segments themselves remain constant? The rest of this essay is devoted to examining and understanding these research questions.

Latent Class Regression
Latent class regression analysis is ideally suited for situations where there is heterogeneity within the population of interest. As Gelman et al. (2004) explain, latent class procedures can be used when, “the population of sampling units consists of a number of subpopulations within each of which a relatively simple model applies." (GelRub04) This stands in contrast to to the typical regression model which assumes that the population of interest is essentially homogeneous. In other words, there is only one class or group within the population and a single model can adequately describe the entire population. While this model is useful in many political marketing studies, it is inadequate for the purpose of understanding the many groups and complexities in the voting public. For example in this research, a basic multiple regression would assume that the effect of the policy variables on voting is the same for all voters. Clearly this is not the case. Some people will care more about economic issues and some people will care more about foreign policy or vice versa.

56
The basic idea behind latent class regression is the parameters of a regression model differ across unobserved, or latent, classes. A latent class model “specifies a finite mixture of normal regression models where the dependent variable is specified as a linear function of a set of explanatory variables" (Wedel and Kamakura 2000). Instead of using one regression to model the whole population, a separate regression model is run for each subgroup. There are two types of variables within latent class models: predictors and covariates. Predictors function the same as they do in regular regression models--they explain the dependent variable. Covariates, on the other hand, are variables that explain or predict membership within a latent class or segment. In terms of this research, policy preferences (feelings about economic policy, foreign affairs, moral issues, and civil rights) are the explanatory variables which are used to predict voting behavior.
Cases with similar policy preferences are divided into groups or latent classes. Covariates such as demographics (e.g. age, race, marital status, etc.) are then used to profile or describe the segments. In the past, this procedure was accomplished in multiple stages. The advantage of latent class regression is that it, “simultaneously estimates the number of market segments, their size and composition, and the segment-specific regression coefficients (DeSarbo and Wedel
2006). In other words, everything can be estimated in a single step.
The final goal of latent class regression is to uncover or reveal the “correct" number of segments in the population. Of course the results are meaningful only if the segments meet the criteria for good or well-behaved segments: identifiability, substantiality, accessibility, stability, responsiveness, and actionability (Wedel and Kamakura 2000). Another important feature is profitability. In a political marketing sense this means that the identified segments must be substantial enough that it is worth spending time and money to court the voters in that particular group. While there are no hard and fast rules for what constitutes a “useful" segment, in this research only those segments that are approximately 3% of the population or larger will be

57 examined. The 3% rule is chosen arbitrarily based on the notion that anything substantially smaller than 3% is simply too small to merit a political marketer's attention.
The basic latent class regression framework is modeled as follows (DeSarbo and Wedel
2006):
k = 1,..., K unknown market segments

b jk = the value of the jth regression coefficient for the kth segment

= the variance term for the kth segment

= the size or proportion of segment k

In this framework, it is assumed that y i is distributed as a finite sum or mixture of conditional univariate normal densities. This can be denoted as:

The objective is to estimate the parameters

, and b jk for each of the segments.

These parameters can be estimated by maximizing a likelihood function, which is, “based on the assumption of a conditional normal distribution of the dependent variable within each segment"(DeSarbo and Wedel 2006). In addition to estimating segment-specific parameters, it is valuable to assign individuals to one of the derived segments. Segment membership probabilities, $p_{ik)$ are obtained using a Bayesian procedure. Although the formula for

58 segment membership probability is not reproduced here, the key point is segment membership is assigned in a probabilistic manner. Individuals are assigned to a segment based on the highest probability. As an example, a person who belongs to segment 1 with probability 0.60 and segment 2 with probability 0.40 will be assigned to segment 1.
Perhaps the greatest challenge in latent class regression is selecting the optimal number of segments. As Wedel and Kamakura (2000) state, “the problem of identifying the number of segments is still without a satisfactory statistical solution." Although there is no objective or scientific criterion for choosing the right number of segments, there are several guidelines or heuristics. The most frequently used heuristics are the various information criteria. The more popular of these are Akaike Information Criterion (AIC), a slightly modified version of AIC, called AIC3, and Bayesian Information Criterion (BIC) (DeSarbo and Wedel 2006, Wedel and
Kamakura 2000). The formulas for these information criteria are:
AIC = -2 ln L + 2P
AIC3 = -2 ln L + 3P
BIC = -2 ln L + P (ln(n))
The optimal number of segments, K*, is chosen when these information statistics reach a minimum value. Unfortunately as DeSarbo et al. note, in many instances these heuristics “do not all render consistent selection of the optimal K*” (DesWed06). There are very few guidelines as to how to proceed when the information criteria suggest different numbers of latent classes.
Andrews and Currim (2003) demonstrate in simulation studies that AIC3 performs the best in terms of recovering the correct number of segments. In this research when there is disagreement, the optimal number of segments will be selected using AIC3. Furthermore, as previously mentioned, only segments that are around 3% or more of the population will be considered. It may be that a 15-segment solution outperforms a 7-segment solution, but many of the groups do

59 not meet the 3% rule. Tables with the various information criteria for the latent class regression from 1988 to 2004 are included in the appendix.

American National Election Studies
The American National Elections Study (ANES), which is funded by the National
Science Foundation, is the premier political science survey. The data set is commonly used in political science research which makes it attractive to use in a political marketing study. It is conducted biennially to coincide with presidential and off-year congressional elections.
According to the ANES website, the mission of the American National Election Studies (ANES) is to “inform explanations of election outcomes by providing data that support rich hypothesis testing, maximize methodological excellence, measure many variables, and promote comparisons across people, contexts, and time. The ANES serves this mission by providing researchers with a view of the political world through the eyes of ordinary citizens. Such data are critical, because these citizens' actions determine election outcomes."
The first ANES survey was completed in 1948, the second survey was done in 1952, and
ANES data have subsequently been collected and compiled every two years beginning with the
1956 presidential election. The survey provides a snapshot of American attitudes toward policy issues, political candidates, social issues, government, and demographic variables (Sap99). The survey is conducted for the most part using face-to-face interviews and is completed in two stages: pre-election and post-election. In some years respondents have been paid for their participation although this is not always the case. The survey is given to a sample of eligible voters (18 or older by election day November 2, 2004) drawn from all over the United States that is post-stratified by region, age, and education (American National Election Studies 2006). The

60 post-stratification necessitates the use of weights when running the latent class regression analysis. Even though the ANES is conducted every two years, the respondents are not necessarily the same from survey to survey. The data are cross-sections of American political attitudes and actions for a given year, not a longitudinal tracking of the same voters over an extended period of time (Bartels 1998). There are, however, some items that form a "core" of American political beliefs that are given from year to year. These core issues include questions about economic policy, foreign policy, “moral issues" which focus mostly on gay rights and abortion, and attitudes toward civil rights and racial equality. As is to be expected, the composition of ANES questions changes as circumstances warrant and conditions vary. For example, the initial 1948
ANES asked questions about feelings toward the Soviet Union. The 1992 survey focused on the
Persian Gulf War and issues related to the economy. Naturally, the 2004 survey asked questions pertaining to the wars in Afghanistan and Iraq. Every four years, the ANES questionnaire includes several psychographic items that cover political activities, political interests, and political opinions. For example there are questions about attendance at campaign rallies, making contributions to a candidate or party, and smaller acts such as wearing a campaign button or having a political bumper sticker. There are also questions about how much time the respondent spent watching campaign events, how many newspaper articles he or she read, and whether the respondent read campaign information online. Other questions include "thermometer" ratings of politicians, federal officials, religious and ethnic groups, and various segments of society such as gays, women, and southerners.
The size of the sample varies from year to year, ranging from n=654 in 1948 to n=2488 in 1992. In the most recent survey, which was conducted in 2004 a total of 1212 pre-election surveys were completed. (The 2006 ANES time series data were not collected in the same manner as had been done in previous years.) 146 participants failed to complete the post-election

61 survey thus reducing the effective sample size to 1066. Of that group, 828 reported that they had voted in the 2004 presidential election. Of those 828, only 17 voted for a third party candidate like Ralph Nader. For the purposes of this analysis, only those who voted for a major party candidate, either Republican or Democrat, are included in the sample. Further research may delve into who chose not to vote or who chose to vote for a third party candidate. The present analysis makes no attempt to answer these questions.
Although the ANES goes back to 1948, only the presidential elections from 1988 to 2004 are included in this analysis. There are several reasons for this choice. First, the idea behind this essay is to evaluate voter segments over the course of several elections. The 1988-2004 time frame includes five presidential elections which is sufficient to answer the stated research questions. Second, the year 1988 represents something of a shift in American politics. It marked the end of the Reagan years and the beginning of the Bush-Clinton-Bush era. Third, there is a certain degree of “stability" in both the numbers and types of questions asked in ANES surveys from 1988 forward. For example in 1968 and 1972 there are questions about school busing to achieve racial balance in schools, school prayer, etc. In other words, the 1988 survey is fairly similar to the surveys conducted in subsequent years whereas some of the previous years' questionnaires are quite different.

Dependent Variable
The dependent variable is constructed from feeling thermometer questions in ANES. In each survey, respondents are asked to rate several groups (e.g. Democratic and Republican parties) and several people, including the major parties' candidates, on a scale of 0 to 100.
Survey participants are told that a score between 0 and 50 indicates the respondent “doesn't care much for that person," and scores over 50 show that the survey taker, “feels favorable and warm

62 toward that person." A score of exactly 50 is indicative of neutral feelings. The dependent variable is calculated by subtracting the feeling thermometer score for the Democratic candidate from the feeling thermometer score for the Republican candidate. For example in the ANES
2004 data, a person who gave John Kerry a 90 and George W. Bush a 10, would have a value of 80. Similarly a person who rated Bush a 100 and gave Kerry a 0 would have a value of 100.
Given the way the variable is constructed, scores range from -100 to 100 and can be treated as continuous. There are two issues that need to be addressed with this choice of dependent variable.
First, a person who gives both candidates a 100 receives the same value as a person whose feeling about the candidates are measured as 0. This appears problematic at first, but a person who has identical feeling thermometer evaluations for both candidates, be they “cold" or “hot", is indifferent between the two, and it will be difficult to predict that individual's vote. In any case, the number of people who are identical in their feeling thermometer scores is very small and do not affect the final results in a meaningful way. Second, it may appear more logical to use a binary dependent variable in this type of analysis with, for instance, a Democrat coded as 1 and
Republican as 0. Finite mixture or latent class regression models with a binomial dependent variable, however, cannot be identified unless there are replications. Unfortunately the data used for this analysis is cross-sectional so there are no repeat observations for individuals, thus necessitating the creation of a continuous dependent variable.

Independent Variables
One of the long running debates in political science is whether voters make their decisions based on issues and and policies or whether something more basic and fundamental, such as partisan identification, is driving the vote decision (CarSti80, Abr95, Pet96, Kro88). On

63 the one hand, some have argued that voters weigh the evidence and vote for the person who they best think will represent their views. This view, known as policy voting, presumes that an individual's vote, “is the final result of a sophisticated decision calculus; that it represents a reasoned and thoughtful attempt by voters to use policy preferences to guide their electoral decision"(CarSti80). This view of voters is often held up as the ideal for democracy. On the other hand, some political scientists have argued that voters do not like to engage in the mental effort associated with comparing candidates' positions with their own and then reaching a conclusion.
As proof of this hypothesis they run regression models with vote choice as the dependent variable and one or two policy-related questions from ANES as the predictors. The coefficients from these models are rarely, if ever, statistically significant. Furthermore, ANES data reveal that many people have made up their minds even before a campaign begins. For example, there are people known as “yellow dog Democrats" who would rather vote for a “yellow dog" than any
Republican. Clearly these are not the type of people who carefully weigh the options before making a choice. All of this suggests that policy voting is a quaint theoretical abstraction but not a practical reality.
Although there are “yellow dog Democrats" and their equivalent on the Republican side, there are still many voters who to varying degrees conform to the idealized notion of sophisticated, reasoning voters. Ansolabehere, Rodden, and Snyder (2008) state that many of the models that purport to show that policy issues are not significant predictors of voting behavior are in fact characterized by high degrees of measurement error. In other words, models that fail to find significant policy effects may simply be mismeasured. There are many causes of measurement error in survey data including ambiguous wording, “vague response categories or categories that do not reflect the individual's actual attitudes, inattentiveness on the part of the respondent, and even typographical errors"(Ansolabehere et al. 2008). Whatever the cause of the measurement error, the net effect is to bias the estimate towards zero and in many cases the signs

64 on the coefficients will be incorrect or nonsensical. The easiest solution to this problem is to use multiple items to measure the same policy domain. As an example, ANES contains multiple questions about race relations. If one were trying to understand the impact of racial attitudes on vote, the most basic approach would be to include a single question as a predictor. Unfortunately this method yields non-significant results, thereby causing the researcher to erroneously conclude that racial attitudes do not matter. Instead the better way is to include all race related items either through factor analysis or by creating scales. This approach yields policy-related regression coefficients that are statistically significant.
Given that policy issues can matter when people make voting decisions and that the use of scales mitigates measurement error when examining policy issues, the independent variables were constructed in the following manner. First, a schema, taken from Baldassarri and Gelman
(2008) was used to categorize the policy-related variables in the various ANES surveys. This framework contains four general policy areas: economic, foreign policy, moral issues, and civil rights or racial attitudes. If a variable could not be clearly classified as belonging to one of these areas it was excluded from the analysis. For instance questions about isolation were not included because there are people of all political stripes who advocate an isolationist foreign policy.
Second, all of the variables were coded so that they would be consistent in terms of ideology.
Each variable was checked, and recoded if necessary, so that the responses would be liberal on one end of the scale and conservative on the other end. The third step was to sum up the variables into a single index. By way of demonstration, the 2004 ANES includes four questions that measure economic policy attitudes on seven-point scale, with an extremely liberal position coded as 1 and an extremely conservative position coded as 7. On economic issues a person on the far left would have a composite score of 4 and a person on the far right would have a composite score of 28. The final step was to standardize the dependent and independent variables. A complete list of all the variables is included in the appendix.

65
Demographic Covariates
A key element of any successful segmentation study is to identify the segments. This is most often accomplished by looking at the demographics of the segments (Wedel and Kamakura
2000). The political science literature suggests that the following demographic variables are important when it comes to understanding votes: age, gender, race, education, income, marital status, and region, more specifically southern or non-southern. In addition to these variables, the following covariates were added: self-reported interest in politics and campaigns, and religious observance (as measured by frequency of church attendance). Because this research is being conducted after the elections, vote choice is included as a covariate as a way of validating or checking segment composition. Interest in political campaigns is important from marketing perspective because it may make less sense to appeal to people who do not care about politics or campaigns all that much. Religion is included because there are numerous popular media accounts suggesting that Republicans have a lock on the devoutly religious in the United States.
Including religiosity as a covariate is a simple way of seeing whether this assumption is correct or whether there is indeed a “religious left."

Results
The results for each presidential election year are presented in the tables below. Tables with information criteria statistics are included in the appendix. Another table is included in the appendix that lists the segments derived for each election. Some explanation is in order as to what the tables presented in the main body of the text mean. The first table for each election year beginning in 1988 presents the regression results with the predictors. The second table is the demographic and psychographic variables that profile each segment. An explanation and brief

66 discussion is given following each election year's tables. It should be noted that the optimal number of segments was 7 for every year.

1988 Election
The 1988 election was transitional in the sense that it marked the first time in 20 years that no incumbent was standing for re-election. The presidential contest was between George
H.W. Bush and Democrat Party nominee Michael Dukakis. Bush won with 53.37% of the popular vote and won 426 electoral votes which represents victory in 40 states. Although the
1988 election was the end of the Reagan years many saw the George H.W. Bush administration as a continuation of Reagan's era. In fact the subsequent election was cast as an opportunity to make a clean break from the Reagan-Bush years.
Seven segments are identified in the 1988 ANES data set. Segment 1, which comprises about 22% of the sample is middle-aged, strongly male, is overwhelmingly white, is married, comes from regions other than the South, is middle class given the median income for 1988 was $
27,225, is religious, and displays a moderate degree of interest in politics. Based on the regression coefficients for the dependent variable, this segment is fairly moderate on economic matters, foreign policy, and civil rights issues, but tends to be slightly liberal on moral issues. It would be safe to call this group “Moderate Republicans." Segments 2, is fairly similar to segment 1 in many regards. A typical member of segment 2 is middle-aged, white, married, nonsouthern, and attends church regularly. Unlike segment 1 members, however, segment 2 members are more likely to be female, they are poorer (with 50.06% below the median income), but they display a much stronger interest in politics. In terms of policy issues, segment 2 is virtually identical to segment 1 except there is a greater importance attached to conservative foreign policy positions. Segment 2 is assigned the label of “Working Class Republicans."

67

A representative segment 3 member is middle-aged, female, white although much more likely to be a minority than members of Republican segments, married, non-southern, poorer than average, attends church either every week or a few times a year, and is interested in politics. On the policy side, segment 3 members are strongly conservative on economic issues, liberal on foreign policy and moral issues, and marginally conservative on racial or civil rights issues.
Segment 3 is given the tile of “Working Class Democrats." In contrast to segment 3, segment 4 is completely Republican. A normal segment 4 person is most likely to be female, white, married, non-southern, wealthy, very religious, but not very concerned with politics. This segment displays a moderate bent in foreign policy but is conservative in economic policy, moral issues, and civil rights. Given the profile, segment 4 could reasonably be called the “Religious Right."
The remaining three segments are all Democratic groups although it appears their policy beliefs and profiles are quite different. Segment 5, for instance, is male, white, married, nonsouthern, middle-class or even upper middle-class, religious, and somewhat nonchalant about politics. Policy wise, segment 5 members are fairly neutral displaying slightly liberal tendencies on economic and foreign policy matters, and mildly conservative tendencies on moral issues and civil rights. Segment 5 members show a greater inclination for regular church attendance than any other segment except for the “Religious Right." For that reason, this group is labeled as the
“Religious Left." Segment 6 is something of an anomaly. The coefficients on all policy related items are nearly zero, which may indicate a “middle of the road" mindset. A typical segment member is middle-aged, female, white, married, non-southern, lower middle class or middle class, equally religious as segment 1 and 2 members, and ambivalent about politics. Given this background, it seems reasonable to assign the name “Moderate Democrats" to this group.

68
Segment 7 is quite small, comprising only 4.51 % of the sample, and is different from the other segments in terms of gender and race. A standard segment 7 member is definitely female, much more likely to be a minority, married, non-southern, middle-class, not particularly religious, but quite interested in politics. Surprisingly, segment 7 people are strongly conservative in all policy domains, but voted overwhelmingly for the Democratic Party candidate. This may suggest, at least in 1988, that minorities were more attuned to Republican principles but for whatever reason did not vote for the Republican party. Given this group's policy proclivities, it makes sense to view this set of people as “Conservative Democrats." Interestingly enough, the existence of this group suggests from a marketing perspective that Republicans may have had an opportunity to make inroads among minority voters.
The 1988 results show that no single policy issue dominated the election. The political science literature argues that if no issue is dominating and the outgoing president is popular the election is essentially going to go in favor of the person who represents a continuation of the same policies and conditions (Abramowitz 1988). Clearly in 1988 voters saw no need to change the direction of the country and thus opted to keep Bush in power as a way of carrying on the generally good economic climate of the nation.

TABLE 3-1
1988 Model for Dependent Variable Results
Segment
1
0.2679

Segment
2
0.2034

Segment
3
0.2027

Segment
4
0.2804

Segment
5
0.0241

Segment
6
0.0035

Segment
7
0.829

Intercept

0.4042

1.0609

-1.2038

0.3756

-0.6599

-0.2059

-0.4236

2690.849

Economic

0.0535

0.0637

0.3087

0.1477

-0.0144

-0.0001

0.15

42.0263

Foreign
Policy
Moral Issues

0.1301

0.2251

-0.2205

-0.0114

-0.0085

-0.0003

0.952

-0.0871

-0.0238

-0.1679

0.2163

0.0104

0.0006

Civil Rights

0.1892

0.1297

0.0927

0.1553

0.0385

-0.0003

R²

Wald

p-value

9.8e-578

Wald(=)

p-value

663.0604

5.80E-140

5.10E-07

40.7255

3.30E-07

146.5406

2.20E-28

145.5921

6.60E-29

0.2277

25.0576

0.00074

24.9992

0.00034

0.2418

20.8199

0.004

20.0403

0.0027

TABLE 3-2
1988 Segment Profiles

Class Size
Vote
Democrat
Republican
Mean Age
Gender
Male
Female
Race
White
Non-White
Marital Status
Not Married
Married
Region
Non-South
South
Church Attendance
Never
A few times a year
Once or twice a month
Almost every week
Every week
Income
0-24,999
25,000-49,999
50,000-74,999
75,000-89,999
90,000+
Interest in Politics
Very Much
Somewhat
Not Much

Segment Segment Segment Segment Segment Segment Segment
1
2
3
4
5
6
7
0.2185
0.2022
0.1884
0.1398
0.1074
0.0987
0.0451
0.0951
0.9049
48.3922

0.0079
0.9921
47.2197

0.9885
0.0115
47.4962

0.000
1.000
45.0286

0.9734
0.0266
43.6928

0.9136
0.0864
48.8366

0.9703
0.0297
40.9568

0.6044
0.3956

0.4586
0.5414

0.3749
0.6251

0.3844
0.6156

0.625
0.375

0.4362
0.5638

0.0038
0.9962

0.9724
0.0276

1.000
0

0.726
0.274

0.965
0.035

0.7991
0.2009

0.8169
0.1831

0.5828
0.4172

0.4307
0.5693

0.4409
0.5591

0.4547
0.5453

0.0288
0.9712

0.3304
0.6696

0.4004
0.5996

0.3283
0.6717

0.7814
0.2186

0.7048
0.2952

0.6457
0.3543

0.7164
0.2836

0.7763
0.2237

0.803
0.197

0.8061
0.1939

0.1025
0.3175
0.134
0.1413
0.3047

0.1291
0.2606
0.1479
0.1178
0.3445

0.0925
0.2543
0.2258
0.1013
0.3261

0.0706
0.2784
0.1222
0.1134
0.4154

0.077
0.303
0.1458
0.1195
0.3547

0.147
0.2485
0.1541
0.1713
0.2792

0.2395
0.3986
0.1791
0.1514
0.0314

0.2417
0.5176
0.1826
0.0581
0

0.5006
0.3443
0.1115
0.031
0.0126

0.6925
0.2435
0.0406
0.0072
0.0162

0.3207
0.1711
0.2855
0.0076
0.215

0.2897
0.5462
0.1522
0
0.0119

0.4018
0.4686
0.047
0.0519
0.0307

0.0804
0.6107
0.2669
0.0184
0.0237

0.3824
0.5984
0.0192

0.5609
0.3572
0.0818

0.4673
0.4403
0.0924

0.1377
0.5127
0.3496

0.2944
0.5728
0.1328

0.3199
0.4524
0.2276

0.4326
0.4426
0.1248

1992 Election
The 1992 election was unique in that there was a strong, if not totally viable, third party candidate, H. Ross Perot. Perot entered the race as an independent who touted his business experience as a way of re-energizing the economic malaise that seemingly afflicted the country at the time. Ultimately, the Democratic nominee, Bill Clinton, defeated the incumbent George
H.W. Bush by a margin of 5 million votes, capturing 43% of the popular vote compared to Bush's
38%. In terms of the electoral college, Clinton won 368 votes and 32 states and Bush won 168 votes and 18 states. Perot received 19% of the vote but carried no states thus earning no electoral college votes.
As in the previous election, the seven-segment solution was chosen for the 1992 data.
Segment 1, which is 30.75% of the population, is middle-aged, female, white, married, nonsouthern, mostly middle-class (1992 median income = $30,636), and not particularly religious.
Interest in politics was measured differently in the 1992 ANES and is thus not included in the analysis. Segment 1 is conservative in most policy areas, especially foreign policy, which may reflect strong support for the 1991 Gulf War. This group does not seem to be overly concerned with economic policy or moral civil rights issues. Segment 1 is probably the “Moderate
Republican" crowd. Segment 2 is strongly Democratic, younger than segment 1, predominantly white male, evenly split between not married and married, is mostly non-southern, relatively wealthy, and highly religious. This segment is conservative on racial matters, and slightly conservative on economic policy, foreign policy, and moral issues. Given its tendency to vote
Democrat but espouse conservative policy positions, this group is labeled as, “Conservative
Democrats."

72
Segment 3 is completely Democratic and is liberal on economic policy, foreign affairs, and moral issues. Surprisingly, segment 3 has a marginally conservative position on civil rights.
A typical member of this cluster is middle-aged, female, white, married, non-southern, poorer than average, and not particularly religious. Given their liberal views and lower socioeconomic status, members of this group are considered “Working Class Democrats." Segment 4, like segment 3, is totally Democratic. The mean age is 52.75, the group is predominantly female, white, married, and not from the south, although like segment 3 there is a strong southern aspect.
Unlike segment 3, this group is highly religious. This segment is economically conservative and conservative on foreign policy and civil rights issues also. Segment 4 is liberal on moral issues which indicates support for abortion and gay rights. The group might be termed the “Religious
Left."
Segment 5 is something of a mixed bag. An average member of this cluster is younger, male, white, married, non-southern, and below or near median income. People from this group are more religious than their peers and split their votes 59% Democrat and 41 % Republican.
This group is strongly conservative on social issues like civil rights and moral issues and is conservative in economic and foreign policy. This segment is somewhat difficult to pinpoint. It appears, once again, to be conservative in terms of policy but the majority voted for the
Democrat. Perhaps the best way to describe this group is to assign the label “Independents."
Segment 6 is a conservative group that cares deeply about economic issues and adheres to conservative viewpoints on foreign policy, moral issues, and to a lesser extent civil rights. This segment is young, male, white, married, non-southern, much wealthier than their peers, and unreligious. Given this group's conservative nature and wealth, perhaps the best title for segment
6 is “Blue Blood Republicans."
Segment 7 is somewhat problematic. The archetypal person in segment 7 is young, female, non-white, married, non-southern, solidly middle class, and almost never attends church

73 meetings. With this profile, one would expect this segment to have voted almost exclusively
Democrat and to espouse more liberal views. As it happens, segment 7 is strongly conservative on economic matters and civil rights and only slightly liberal on foreign policy. This group represents only 2.38% of the population and its very existence seems anomalous. Further investigation is warranted to see whether segment 7 actually does exist or is simply an artifact of the data.
As in 1988 it appears that the voting public is a complex group. There are clearly identifiable segments but they tend to hold mixed policy views. Most groups seem to be conservative in some regards but liberal in others. Another pattern that is emerging is the fact that religious observance is not the strict province of the right. There is a religious left in 1992 and it is even more pronounced than in the previous presidential election. Some of the anomalous results may be explained by the fact that 1992 was a transitional election. The incumbent, George
H.W. Bush represented a continuation of the Reagan era and had grown unpopular even among some Republican segments. Another factor was Ross Perot's entry into the race. Perot's presence is not accounted for in this model.

TABLE 3-3
1992 Model for Dependent Variable

R²
Intercept
Economic
Foreign
Policy
Moral
Issues
Civil
Rights

Segment Segment Segment Segment Segment Segment Segment Wald p-value Wald(=) p-value
1
2
3
4
5
6
7
0.1515
0.3496
0.1747
0.1493
0.7458
0.6462
0.8831
0.8582 -0.6804 -0.3394 -1.5075
0.1502
0.5823
-0.371 637.441 2.10E-133 622.6357 3.10E-131
-0.0078
0.1079 -0.1296
0.1171
0.1306
0.3357
0.4446 144.351
6.20E-28 117.4688
5.50E-23
0.2396
0.0727 -0.0101
0.0483
0.0663
0.1888 -0.0161 59.3137
2.10E-10
32.7895
1.20E-05
0.1555

0.0913

-0.108

-0.154

0.1954

0.1991

-0.1542

44.9091

1.40E-07

39.4964

5.70E-07

0.0416

0.2316

0.0683

0.1507

0.4757

0.0317

0.4136

85.1215

1.20E-15

39.4857

5.70E-07

TABLE 3-4
1992 Segment Profiles

Class Size
Vote
Democrat
Republican
Mean Age
Gender
Male
Female
Race
White
Non-White
Marital Status
Not Married
Married
Region
Non-South
South
Church Attendance
Never
A few times a year
Once or twice a month Almost every week
Every week
Income
0-24,999
25,000-49,999
50,000-74,999
75,00-104,999
105,000+

Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 Segment 6 Segment 7
0.3075
0.2195
0.1559
0.1497
0.0766
0.067
0.0238
0.0002
0.9998
49.9925

0.9999
0.0001
40.5557

1
0
49.2039

1
0
52.7469

0.5876
0.4124
37.4671

0
1
33.8251

0.6814
0.3186
35.0995

0.4101
0.5899

0.6125
0.3875

0.2628
0.7372

0.3762
0.6238

0.6697
0.3303

0.6724
0.3276

0.1091
0.8909

0.9799
0.0201

0.8615
0.1385

0.7147
0.2853

0.7674
0.2326

0.7611
0.2389

0.9015
0.0985

0.3455
0.6545

0.2608
0.7392

0.5027
0.4973

0.2804
0.7196

0.3564
0.6436

0.4297
0.5703

0.2267
0.7733

0.1018
0.8982

0.6841
0.3159

0.8164
0.1836

0.6081
0.3919

0.6744
0.3256

0.7082
0.2918

0.9307
0.0693

0.764
0.236

0.3946
0.1581
0.1227

0.0971
0.0866
0.1114

0.368
0.1475
0.1762

0.2379
0.0958
0.1193

0.0993
0.1001
0.1822

0.621
0.1409
0.1347

0.934
0.0657
0

0.1411
0.1835

0.179
0.5259

0.1075
0.2008

0.1422
0.4048

0.2139
0.4045

0.0759
0.0275

0.0002
0

0.28
0.393
0.1982
0.0921
0.0366

0.28
0.3626
0.2103
0.0706
0.0765

0.4357
0.3313
0.1671
0.0381
0.0278

0.5237
0.2839
0.1297
0.0453
0.0173

0.3217
0.3471
0.1407
0.0993
0.0912

0.0403
0.246
0.363
0.1866
0.1641

0.1869
0.5371
0.2757
0.0002
0.0001

76
1996 Election
The 1996 election was Bill Clinton's re-election campaign. Once again Ross Perot entered the race but was not nearly the factor he was in 1992 and he garnered only 8% of the popular vote. Economic conditions were favorable for Clinton's re-election and the Republicans nominated a candidate, Bob Dole, who failed to inspire the electorate. Clinton captured 49% of the popular vote and 379 electoral college votes compared to Dole's 41% of the popular vote and
159 electoral college votes. Although the 1996 was a relatively easy victory for the Democrats, the events of Clinton's second term made for a hotly contested election in 2000.
As in all other years, the optimal number of segments in the 1996 ANES is seven.
Segment 1 is middle-aged, evenly divided between males and females, predominantly white, married, non-southern albeit with a strong southern component, slightly below average in income
(1996 median income = $35, 492), and mildly interested in politics. This segment is mildly conservative on economics, foreign policy, and moral issues, and holds an even more conservative position on civil rights. This group might be termed “Conservative Democrats."
Segment 2, on the other hand, is strongly Republican. The typical segment 2 member is middleaged, male, white, married, not from the south, wealthier than average, strongly religious, and quite interested in politics and government. This group has a strong conservative streak in economic matters is slightly liberal in foreign policy and just barely conservative in moral issues and civil rights. For those reasons, this group is called “Moderate Republicans."
Segment 3 is another group that defies easy description. A representative member of this segment is middle-aged, male, white, married, non-southern, relatively wealthy, less religious than their peers, and somewhat interested in politics. This group is conservative in all policy domains. Segment 3 is most likely the “Moderates" group of the voting public. Unlike segment
3, segment 4 is strongly Democratic. This group is young, female, white, married, non-southern

77 although there is a strong southern presence, lower than average on the income scale, as likely to be non-church attending as church attending, and somewhat interested in politics. In terms of issues, this segment is conservative in economic policy and foreign policy, but liberal in moral issues and civil rights. Segment 4 is made up of fiscally conservative but socially liberal people and is thus given the name “Moderate Democrats."
Segment 5 is heavily Republican, slightly older, female, white, married, evenly divided between south and other regions, strongly religious, and marginally interested in politics and campaigns. This group is in favor of some government intervention in economic matters, but supports a conservative foreign policy and conservative civil rights agenda. Moral issues do not appear to be a major area of interest for this segment. This cluster is given the title “Social
Conservatives." Segment 6 is exclusively Democratic, older, female, white, married, nonsouthern, slightly below average in income, religious, and keenly interested in politics. This group which will be called the “Religious Left," favors free market economics, but is liberal on moral issues, civil rights, foreign policy.
Lastly, segment 7 is a strongly Democratic group. Group members are young, predominantly female, white, married, southern, not particularly religious, not as well as other segments, and not overly interested in politics. Segment 7 is conservative on economics and civil rights but is liberal on foreign policy and moral issues. This group might arguably be called
“Southern Democrats." As with the previous elections, the main takeaway message is voters' policy preferences are complex. There is only one group that is completely conservative in all policy domains and there are no groups that are completely liberal in all policy areas. This may suggest two things. First, perhaps the United States as a whole is moderately conservative.
Second, it may be that the traditional liberal and conservative labels which are being applied in this research are simply inadequate. It may be that these concepts are far more complicated than political scientists have thought.

TABLE 3-5
1996 Model for Dependent Variable Results

R²
Intercept
Economic
Foreign
Policy
Moral
Civil Rights

Segment Segment Segment Segment Segment Segment Segment Wald p-value Wald(=) p-value 1
2
3
4
5
6
7
0.141
0.4164
0.461
0.1845
0.2493
0.1735
0.9525
-0.7164
0.9945
0.1147
-0.3296
0.8581
-1.6038
-0.8515 1247.115
4.60E- 651.1379
2.20E265
137
0.0579
0.3889
0.2034
0.1778
-0.2351
0.0711
0.0615
84.3926
1.70E54.1523
6.90E15
10
0.0535
-0.0093
0.0363
0.066
0.2799
-0.089
-0.5323 532.6421
7.70E- 376.1847
3.70E111
78
0.0323
0.0733
0.1028
-0.0125
0.008
-0.0658
-0.0983
16.5364
0.021
16.532
0.011
0.1284
0.0338
0.1128
-0.2072
0.256
-0.0454
0.2062
44.0394
2.10E31.1784
2.30E07
05

TABLE 3-6
1996 Segment Profiles

Class Size
Mean
Vote
Democrat
Republican
Mean Age
Gender
Male
Female
Race
White
Non-White
Marital Status
Not Married
Married
Region
Non-South
South
Church Attendance
Never
A few times a year
Once or twice a month Almost every week
Every week
Interest in Politics
Very Much
Somewhat
Not Much
Income
0-24,999
25,000-49,999
50,000-74,999
75,00-104,999
105,000+

Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 Segment 6 Segment 7
0.3243
0.2481
0.1241
0.1073
0.0983
0.0505
0.0474
-0.738
1.3868
0.247
-0.42
1.0133
-1.6133
-0.5543
0.9999
0.0001
51.0271

0.0002
0.9998
46.4627

0.4184
0.5816
48.9915

1
0
34.5473

0.0175
0.9825
51.7183

1
0
51.5453

1
0
32.7093

0.4992
0.5008

0.6568
0.3432

0.674
0.326

0.1899
0.8101

0.2223
0.7777

0.1414
0.8586

0.2416
0.7584

0.8319
0.1681

0.9611
0.0389

0.9272
0.0728

0.7315
0.2685

0.9735
0.0265

0.8254
0.1746

0.6853
0.3147

0.4462
0.5538

0.2429
0.7571

0.1988
0.8012

0.4396
0.5604

0.3573
0.6427

0.4232
0.5768

0.2675
0.7325

0.7479
0.2521

0.7306
0.2694

0.6469
0.3531

0.5225
0.4775

0.6606
0.3394

0.8145
0.1855

0.2879
0.7121

0.3087
0.2087
0.1233

0.1572
0.1475
0.0991

0.2469
0.2081
0.1099

0.2264
0.1432
0.2105

0.1384
0.1028
0.213

0.2425
0.1005
0.1908

0.2881
0.1682
0.2164

0.1327
0.2267

0.1874
0.4087

0.1557
0.2795

0.191
0.2289

0.153
0.3928

0.1221
0.3441

0.0629
0.2644

0.451
0.5263
0.0228

0.5053
0.428
0.0667

0.182
0.6917
0.1263

0.0111
0.7153
0.2735

0.3075
0.5971
0.0954

0.868
0.1316
0.0004

0.2656
0.7134
0.021

0.3635
0.3091
0.1677
0.0917
0.0679

0.059
0.3495
0.3078
0.1589
0.1248

0.0697
0.3227
0.3329
0.1559
0.1187

0.4033
0.3919
0.1641
0.0282
0.0125

0.5508
0.4236
0.0256
0
0

0.4281
0.3263
0.1294
0.0682
0.048

0.4307
0.26
0.3091
0
0.0002

80
2000 Election
The 2000 election was the most closely contested presidential contest in 40 years and was the first time since 1888 that the popular vote winner lost the electoral college vote. The
Democratic nominee, Al Gore, won 48.38% of the popular vote but failed to gain an electoral college victory winning 266 votes out of 538 possible. The eventual winner, George W. Bush, garnered 47.87% of the popular vote and was able to gain 271 electoral college votes thereby gaining the presidency. Unlike the previous two elections, there were no significant third party challengers. Once again, latent class regression identifies seven segments in the voting public.
Segment 1 is Democratic, middle-aged, female, white, married, non-southern, lower on the income scale, not particularly religious, and very interested in politics and campaigns. This group is conservative on economic policy, moral issues and foreign policy (although just barely so), and slightly liberal on civil rights issues. Segment 1 is made up primarily of “Moderate Democrats."
Segment 2, is heavily Republican, younger than segment 1, male, almost exclusively white, married, split between southern and non-southern, not extremely religious, and modestly interested in politics. This segment is strongly conservative with regards to economic matters and civil rights, mildly conservative on foreign policy and slightly liberal on moral issues. This group is most likely “Moderate Republicans."
Segment 3 is Republican, middle-aged, male, white, married, non-southern, middle-class, very religious, and interested in politics. This segment supports government involvement in the economy and some form of government help for minorities. Segment 3 is strongly conservative in foreign policy and moral issues. Given this profile, the term “Social Conservatives" is an appropriate label for this particular cluster. Segment 4 is a younger group, mostly female, white, not married, split between southern and non-southern, has a lower household income than most

81 especially when compared to the 2000 median income of $ 41,990, is fairly religious but only somewhat interested in politics. Segment 4, which can be called “Traditional Democrats," is conservative in all areas except civil rights.
Segments 5 and 6 are similar in many regards. Both groups are young, white, female, not married, well below median income, not especially religious, and not that interested or involved in political matters. The main differences between the two groups is how they voted and where they live. Segment 5, which is strongly conservative on civil rights and decidedly non-southern is most likely a “Conservative Democrat" segment. Segment 6, on the other hand, is almost equally likely to be southern or non-southern is actually strongly liberal on civil rights matters. This group voted 60-40 in favor of the Republican candidate and is thus given the name of “Fiscal
Conservatives" which implies socially liberal positions.
Segment 7, which is a relatively small group, is completely Democratic leaning, middleaged, female, white, married, southern, poor, not religious, and interested in politics. This group is liberal in civil rights but strongly conservative, particularly on economics and moral issues.
This group might appropriately be termed “Southern Democrats." There is a strong conservative angle with this segment but at the same time there is a preference for more government involvement in solving racial problems. This group is relatively small and is somewhat puzzling.

TABLE 3-7
2000 Model for Dependent Variable Results

R²
Intercept
Economic
Foreign
Policy
Moral
Issues
Civil
Rights

Segment Segment Segment Segment Segment Segment Segment Wald p-value Wald(=) p-value 1
2
3
4
5
6
7
0.0347
0.4427
0.2611
0.2647
0.8871
0.9342
0.9613
0.7918
-1.0003
0.398
0.868 -0.3835 -0.3708
0.5379 -0.9882 1110.964 1.30E-235 820.7862 5.00E-174
0.1035
0.3417 -0.2589
0.1223 -0.0052
0.198
0.7091
157.414
1.10E-30
93.5142
5.60E-18
0.018
0.0474
0.2737
0.1235
0.0984
0.1841
0.1587
42.0504
5.10E-07
7.8881
0.25
0.131

-0.0031

0.4064

0.1504

0.1494

0.1591

0.3575

67.1668

5.50E-12

19.5836

0.0033

-0.0613

0.2417

-0.0576

-0.1044

0.6302

-1.2132

-0.8618

519.8384

4.30E-108

436.4891

4.00E-91

TABLE 3-8
2000 Segment Profiles

Class Size
Mean
Vote
Democrat
Republican
Mean Age
Gender
Male
Female
Race
White
Non-White
Marital Status
Not Married
Married
Region
Non-South
South
Church Attendance
Never
A few times a year
Once or twice a month Almost every week
Every week
Interest in Politics
Very Much
Somewhat
Not Much
Income
0-24,999
25,000-49,999
50,000-74,999
75,00-104,999
105,000+

Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 Segment 6 Segment 7
0.303
0.2188
0.1842
0.1438
0.0554
0.051
0.0438
-0.9857
0.7157
1.1219
-0.4115
-0.2309
0.1973
-1.3309
0.9973
0.0027
54.7488

0.0399
0.9601
47.2551

0.0641
0.9359
51.6521

0.9943
0.0057
38.2245

0.6517
0.3483
32.2606

0.3762
0.6238
26.5915

0.9407
0.0593
55.9293

0.3954
0.6046

0.7914
0.2086

0.3427
0.6573

0.3184
0.6816

0.2666
0.7334

0.0002
0.9998

0.417
0.583

0.8103
0.1897

0.9531
0.0469

0.9831
0.0169

0.6835
0.3165

0.7463
0.2537

0.9733
0.0267

0.9993
0.0007

0.3973
0.6027

0.2523
0.7477

0.2565
0.7435

0.6476
0.3524

0.5282
0.4718

0.566
0.434

0.2547
0.7453

0.8837
0.1163

0.5831
0.4169

0.7925
0.2075

0.5381
0.4619

0.9258
0.0742

0.5252
0.4748

0.4364
0.5636

0.2926
0.1702
0.1918

0.2753
0.1971
0.1351

0.1325
0.1387
0.1403

0.1212
0.0689
0.3161

0.3772
0.1262
0.1962

0.3033
0.3048
0.1475

0.3171
0.3302
0.0006

0.1223
0.2231

0.1263
0.2662

0.1275
0.4609

0.1739
0.3199

0.2969
0.0035

0.2435
0.0009

0.0119
0.3401

0.6082
0.3725
0.0193

0.4621
0.4498
0.0881

0.635
0.3208
0.0442

0.2552
0.6257
0.1191

0.1061
0.5906
0.3034

0.1933
0.6063
0.2004

0.4997
0.4496
0.0507

0.3905
0.3915
0.1169
0.0603
0.0408

0.4563
0.3831
0.1388
0.0218
0

0.3029
0.2834
0.2014
0.1603
0.052

0.7411
0.2457
0.0132
0
0

0.4116
0.3785
0.0972
0.0464
0.0663

0.9085
0.0915
0
0
0

1
0
0
0
0

84
2004 Election
The most recent presidential election, as of this writing, was another closely contested race. The incumbent, George W. Bush, eked out a narrow victory gaining 50.73% of the popular vote and 286 electoral college votes. His challenger, John F. Kerry won 48.27% of the popular vote and 251 electoral college votes. As with previous elections, latent class regression identified a seven-segment solution. Segment 1 is comprised of people who voted Democrat, are middleaged, female, white, married, non-southern, middle-class, not particularly religious, but strongly interested in politics. This group is mildly conservative on all issues, except foreign policy. In that area, they prefer diplomacy over military force as a general rule and non-intervention abroad.
This group is “Conservative Democrats." Segment 2, is the opposite of segment 1 in many ways.
Nearly everyone in this segment voted Republican, is middle-age, evenly split along gender lines, white, married, non-southern, wealthy, not particularly religious, and not that interested in politics. For this segment, conservative moral stances and conservative foreign policy positions are important. The other policy domains have relatively small, but positive--indicating a conservative bent, coefficients. This segment is made up of “Blue Blood Republicans."
Segment 3 is Democratic, younger, female, split between non-white and white, split between married and non-married, non-southern, lower middle class to middle class, not religious, and somewhat interested in politics. This segment is conservative in all policy areas and is thus called “Traditional Democrats." Segment 4 is strongly Republican, middle-aged, female, white, married, non-southern, wealthy, religious, and very interested in politics. This cluster is slightly liberal on economic issues and slightly conservative in other areas. This segment is marginally liberal on economic policy and slightly conservative in the other policy domains. For that reason, they are given the name “Moderate Republicans."

85
Segment 5 voted mostly Democrat in the 2004 election. The typical segment member is young, male, white (although non-white is strongly present in this group), not married, nonsouthern, not religious, and interested in politics. In terms of policies, segment 5 members are liberal on civil rights but very conservative on foreign policy and moral issues. Given the mixed results, this segment might be called “Independents." Segment 6 members are typically middleaged, male, white, married, southern, less wealthy than other Republican groups, not frequent church goers, and interested in politics. candidate. This segment is slightly conservative on economic policy and foreign affairs, more strongly conservative on moral issues, but liberal on civil rights. Segment 6 could reasonably be called “Southern Republicans."
Finally, segment 7 is comprised of older voters, who are white, female, not married, nonsouthern, strongly religious, wealthy, and somewhat interested in politics. This segment is moderately interested in economic and foreign policy issues, is strongly conservative on economic policy and foreign affairs, conservative on moral issues, but strongly liberal on civil rights matters. Given the profile of a typical segment member, this group is labeled the
“Religious Left." Segment 7 is relatively comprises only 3.34% of the voting population which means they would normally not merit any special attention from the campaign. Given the closeness of the 2004 election, however, even small groups became important.
One of the great controversies of the 2004 election was moral issues. During that year, numerous ballot initiatives, which were primarily “Defense of Marriage” amendments to state constitutions were passed. The results from the latent class analysis shows that moral issues were relatively important to every segment and that every segment supported the conservative position.
It is not surprising therefore that in 2006, the Democrat Party specifically sought candidates who held socially conservative viewpoints.

TABLE 3-9
2004 Model for Dependent Variable Results

R²
Intercept
Economic
Foreign
Policy
Moral
Issues
Civil
Rights

Segment Segment Segment Segment Segment Segment Segment Wald p-value Wald(=) p-value 1
2
3
4
5
6
7
0.1299
0.3512
0.6689
0.4759
0.884
0.4088
0.9768
-1.0562
0.473 -0.1383
1.3646 -0.6872
1.4973 -0.3585 2054.583 7.3e-440
1866.58 2.1e-400
0.1026
0.074
0.1189 -0.0622
0.115
0.062
0.343 113.3948 1.80E-21
71.8446 1.70E-13
-0.0592
0.1143
0.1567
0.1762
0.434
0.017
0.5882 346.6967 6.30E-71 176.1681 2.20E-35
0.1186

0.2417

0.1427

0.0729

0.2568

0.234

0.2113

79.3831

1.80E-14

13.9953

0.03

0.0633

0.0712

0.1868

0.0546

-0.1607

-0.117

-0.3351

98.3016

2.40E-18

93.4655

5.80E-18

TABLE 3-10
2004 Segment Profiles
Class Size
Mean
Vote
Democrat
Republican
Mean Age
Gender
Male
Female
Race
White
Non-White
Marital Status
Not Married
Married
Region
Non-South
South
Church Attendance
Never
A few times a year
Once or twice a month Almost every week
Every week
Mean
Interest in Politics
Very Much
Somewhat
Not Much
Income
0-24,999
25,000-49,999
50,000-79,999
80,000-119,999
120,000+

Segment 1 Segment 2 Segment 3 Segment 4 Segment 5 Segment 6 Segment 7
0.3048
0.2927
0.1286
0.1233
0.0745
0.0426
0.0334
-1.1122
0.7677
-0.2748
1.5171
-0.7091
1.5506
-0.0147
0.9733
0.0267
50.2949

0.0003
0.9997
46.2361

0.8323
0.1677
42.1102

0.0301
0.9699
47.1419

0.8381
0.1619
34.8455

0
1
54.9548

0.6037
0.3963
58.5185

0.3709
0.6291

0.5185
0.4815

0.4082
0.5918

0.4443
0.5557

0.9809
0.0191

0.8897
0.1103

0.2121
0.7879

0.7158
0.2842

0.8984
0.1016

0.5017
0.4983

0.8354
0.1646

0.6332
0.3668

0.9805
0.0195

0.7333
0.2667

0.4217
0.5783

0.2742
0.7258

0.4816
0.5184

0.3246
0.6754

0.5272
0.4728

0.0672
0.9328

0.7702
0.2298

0.7466
0.2534

0.747
0.253

0.7085
0.2915

0.7504
0.2496

0.6061
0.3939

0.4257
0.5743

0.9271
0.0729

0.3203
0.1813
0.159

0.2867
0.1282
0.1722

0.4954
0.1601
0.1077

0.2988
0.0902
0.1088

0.3848
0.1156
0.2634

0.3602
0.2464
0.2254

0.0294
0.1088
0.024

0.1288
0.2106
2.7283

0.1577
0.2552
2.9664

0.1154
0.1215
2.2074

0.1115
0.3908
3.2053

0.1082
0.1279
2.4788

0.0001
0.1679
2.3691

0.2563
0.5816
4.2518

0.6269
0.3348
0.0383

0.3581
0.5435
0.0984

0.2184
0.5626
0.219

0.7779
0.2149
0.0072

0.6136
0.3539
0.0326

0.5873
0.2971
0.1156

0.4919
0.4407
0.0674

0.2044
0.2556
0.2479
0.1874
0.1048

0.1103
0.2019
0.337
0.2088
0.142

0.313
0.3069
0.2317
0.107
0.0414

0.1111
0.1902
0.2374
0.2303
0.2311

0.1798
0.299
0.2833
0.1955
0.0425

0.4904
0.1093
0.2992
0.1011
0.0001

0.1015
0.278
0.1751
0.2031
0.2423

88
Marketing Application
The main point of any segmentation study is to define and describe segments so that a marketer can know how to approach that group most profitably. The segments that were estimated via latent class regression, “provide some clear indications for differentially targeting each segment"(Wedel and Kamakura 2000). The purpose of this section is to show how the results obtained in the previous section can provide a guide for future political marketing strategies. In other words, although it may not extrapolate perfectly, one could use the 2004 results to make some predictions and strategy suggestions for the 2008 presidential election.
In 2004, almost every segment had a conservative economic policy stance. This trend appears to continue over the course of several elections. Most American voters tend to favor a limited role for government in economic matters. In order to appeal to the largest number of voters, a candidates in 2008 should emphasize how they will adopt policies that will enable the free market to operate. Of course there are many interpretations of what those policies could be, but the underlying theme is Americans prefer free markets. Any campaign promises or policy platforms that appear to threaten that fundamental tenet are certain to be met with resistance.
Furthermore, in 2004 all of the segments were conservative in the area of moral issues. Even though a candidate may believe that abortion should always be permitted and gay marriage legalized, it would be politically unwise to make those beliefs the central talking points of a campaign (assuming of course that attitudes toward moral issues have not changed dramatically since 2004).
In terms of differentiated offerings, the Democrat nominee should make a strong appeal to “Southern Republicans" who in 2004 voted for Bush but are liberal in their views toward civil rights. Given that this group espouses views that are not in line with the typical Republican approach toward civil rights, these voters may be “cross-pressured" in the political science

89 parlance. Cross-pressured voters hold attitudes that run against the mainstream of the party which suggests, at least in theory, that a candidate from the other party can exploit that difference
(often called a “wedge issue") to appeal to that segment. Likewise, segment 3, “Traditional
Democrats" are more conservative than the Democratic party leadership and candidate.
Republicans could successfully leverage that difference by highlighting how Republicans are more in tune with that segment's actually attitudes and feelings. If voters actually listen to what candidates say, this may create enough doubt that even a Traditional Democrat might be willing to vote for a Republican.
In addition to “stripping" away voters from the other party, understanding voter segments can be helpful in a party's “get out the vote" (GOTV) effort. It is generally accepted that the key to Republican victory in 2004 was the direct result of campaigning directed at GOTV
(Hamburger and Wallsten 2006). For instance, segment 1 in 2004, Conservative Democrats, tend to be more liberal in their foreign policy beliefs. Given the unpopularity of the Iraq War among many Democrats, it would be prudent to remind this segment in 2008 that the Democrats are opposed to the war and want to end it quickly. This could be accompanied with a reminder to the effect that “unless you vote, the war will continue." This appeal would of course not work with
Republican segments and might not even work with certain groups of Democrats. The key is to understand what policies appeal to which groups and then highlight or downplay the candidate's position as they relate to the predispositions of the segment. The goal is to send the right reminders and campaign materials to the appropriate segments so that the usefulness of campaign resources can be maximized. Without segmentation, campaigns will waste time and money in appealing to the wrong groups or to the right groups in the wrong way.

90
Discussion
There are several key findings that come out of the present study. First, there is some consistency in the types of groups that were found over the course of the elections from 19882004. In every year it was possible to find traditional Democrat and Republican groups as well as moderate Democrat and Republican groups. Furthermore, the profiles of these segments were generally consistently over time. Another interesting finding is their views on the four policy domains, economic, foreign affairs, moral issues, and civil rights, seem to be consistent over time as well. This indicates that there is a core of committed partisans in the United States and they do not seem to be disappearing over time. From a marketing perspective this means that both
Republican and Democrat campaigns are starting with a solid base to work with. Given that elections are won and lost on the margins, it means that it is increasingly important to use a twopronged strategy: mobilize the core constituency while simultaneously identifying smaller groups that might be amenable to a campaign's message.
There is a certain degree of face validity in this research. This research has confirmed several longstanding dogmas in American presidential politics. First, minorities tend to vote overwhelmingly Democratic. Second, younger voters are more liberal in their policy views and hence more likely to vote Democrat. Third, the conventional wisdom of a center-right American populace seems to be validated. In no election year was there any group that was strictly liberal in its policy preferences. Clearly some people are socially liberal and economically conservative and some are economically liberal and socially conservative. There are also groups that socially and economically conservative but no segments were identified that are socially and economically liberal. This also suggests that further research is needed to understand what it means to be liberal and what it means to be conservative. Fourth, non-married people tend to vote
Democrat as well.

91
That being said, there are some findings that go against the conventional wisdom. First, wealth does not correlate very neatly with party preference. One ongoing area of research in political science is why rich people are perceived as Republicans yet wealthier states in recent years have sided with the Democrat(GelSho07). This research supports the idea that wealth does not follow strict party lines. As has been shown, there are working class Republican and working class Democrat segments in more than one election. Similarly, there are well-to-do segments of both political parties. This suggests that there are more complex factors at play such as where one grew up or what one's parents' political views were. Economic status alone cannot predict vote choice in a very precise manner.
Another key finding is the existence of the religious left. In several of the election cycles there is a segment that attends church frequently and voted for the Democratic nominee. This runs counter to the popular media portrayal of active religious people as belonging almost exclusively to the conservative, and hence Republican, side of politics. Of course one limitation of this analysis is it lumped together religious people of all creeds into one group. Further research should look at whether “Born Again" voters that attend church weekly are as likely to belong to the religious left as Roman Catholics who attend church weekly. One suspects that the answer is no. Nonetheless, it has been shown that religious people, as measured by frequency of attendance at religious services, belong to both the right and the left in American politics.
One of the key research questions in the beginning of this paper was to ask whether latent class regression analysis can shed any additional light on political phenomena. It appears that this method can be profitably used but because the analysis relied strictly on latent class regression, it is hard to make any comparisons. For example if one were to use discriminant analysis over the same time frame, it may be possible that different results would be reached. In future research it may be beneficial to employ a variety of methods including traditional clustering techniques. The results can then be compared to the results obtained in this research to see if any further

92 conclusions can be drawn which might otherwise not have been reached. This remains a topic for future examination and research.

Limitations
There are of course several limitations inherent in segmenting voters using self-reported policy survey items. Issue voting remains something of a problematic area. There are no guarantees that the voters will actually make their decisions based on issues alone. Certainly they can be a factor, but there are other “shortcuts" available which also have strong explanatory power in understanding elections. Instead of asking people about their policy predispositions and preferences, it may be just as useful to ask a simple question of “Do you consider yourself liberal or conservative?" with a follow up question of “On scale of 1 to 7, how liberal or conservative?"
Another issue that remains problematic is the construction of the policy domains using multiple items. Unfortunately, the four main policy areas were not measured with the same questions in every election cycle. This means that the researcher has to make numerous subjective decisions as to how to classify policy items. Ideally there would be more consistency over time with the specific questions. Even better, the exact same questions would be used from one election to the next. Fortunately there is some of this in the ANES data, but more would be preferred. A further issue is the actual labeling and description of segments. As with many statistical techniques, defining and describing clusters seems to be as much an art as a science.
One wishes there were more objective methods for deciding on the optimal number of segments.
This also leads into issues concerning the use of latent class regression analysis. An initial run through the data resulted in a different number of segments being selected for each election cycle.
For instance the initial analysis leads one to believe that there were six segments in 1988, seven in

93
1992, 1996, 2000, and eight segments in 2004. Over the course of several analyses, the number of segments remained stable for 1992, 1996, and 2000, but fluctuated for 1988 and 2004. It was decided after numerous attempts that the seven-segment solution seemed appropriate even if there was a degree of uneasiness in that decision. Again, “best practices" were followed in the sense that multiple analyses were run and different starting points were used in the analyses so there is a reasonably strong degree of confidence that the solutions are as correct as possible.
Finally, the ideal situation would be to estimate the effect of policy preferences on binary vote choice (0=Democrat, 1=Republican). As mentioned previously, however, this type of estimation requires replications which are not available. Instead, a feeling thermometer difference score is constructed, which correlates very highly with actual vote choice. Ideally one would have panel data that tracks the same voters over an extended period of time. Some panel data have been collected but usually not for more than four which encompasses generally encompasses two presidential elections and one off-cycle congressional election. In the future it would be preferable to have data that covers a wider time horizon.

Further Research
There are two directions for further research in voter segmentation. First, there is the possibility of a developing a model that allows for the simultaneous estimation of binary vote choice with policy variables as the explanatory variables. This model is currently under development and will take some time to fully complete. A second avenue is “microtargeting."
Microtargeting is sometimes called “narrowcasting" and has become a trendy topic in political consulting circles. The basic premise behind microtargeting is that using publicly available voters files and combining that with market research information, one can build models of individual voting behavior. Instead of having to rely on segments, it may be possible to unravel

94 each voter's “political DNA" to predict how that person will vote. This area is especially promising and opens the door for hierarchical Bayes methods which are ideally suited for making predictions at the individual level. This is an area that will be pursued in the near future.
The introductory paragraph of this essay talked about the different monikers that have become part of the American political lexicon. An argument has been made that understanding voter segments is important and that an evaluation of voter segments over the course of several elections can reveal many interesting things. Hopefully this essay is an important starting point for the application of marketing methods in political marketing research. Further research may be able to clarify voter wants and needs more fully and thus enable campaigns to become truly responsive to voters' wants and needs. Such a development would surely be a boon for democratic government.

95

References
Abramowitz, Alan. (1988), “An Improved Model for Predicting Presidential
Election Outcomes,” PS: Political Science & Politics, 21, 843-848.

---- (1995), “It's Abortion, Stupid: Policy Voting in the 1992 Presidential Election,” The Journal of Politics, 57, 176-186.

Ahmed, Saruddin. A. and Douglas N. Jackson (1979), “Psychographics for Social Policy
Decision: Welfare Assistance,” Journal of Consumer Research, 5(4), 229-39.

Allen, Michael and James Carney (2006), “The G.O.P.'s Secret Weapon,” Time, 168(15), 24-27.

American National Election Studies (2006), “American National Election Studies,” (accessed
May 15, 2006) [http://www.electionstudies.org].

Andrews, Rick. L. and Imran S. Currim, Imran (2003), “Retention of Latent Segments in
Regression-based Marketing Models,” International Journal of Research in Marketing, 20, 315321.

Ansolabehere, Stephen, Shanto Iyengar, Adam F. Simon, and Nicholas Valentino (1994), “Does
Attack Advertising Demobilize the Electorate?” American Political Science Review, 88(4), 829838.

96
----, Jonathan Rodden, and James M. Snyder Jr. (2008), “The Strength of Issues: Using Multiple
Measures to Gauge Preference Stability, Ideological Constraint, and Issue Voting,” American
Political Science Review, 102 (2), 215-232.

----, John M. de Figueiredo, and James M. Snyder Jr. (2003), “Why is there so Little Money in
American Politics?” Journal of Economic Perspectives, 17(1), 105-30.

American Marketing Association (2007). “Political Marketing, (accessed June 15, 2007)
[http://www.mar

Bagozzi, Richard P. (1975) “Marketing as Exchange,” Journal of Marketing,
39(4), 32-39.

Baines, Paul R. (1999), “Voter Segmentation and Candidate Positioning” in Handbook of
Political Marketing, Bruce I. Newman, ed., Thousand Oaks, CA: Sage Publications, Inc., 403421.

Baldassarri, Delia and Andrew Gelman (2008), “Partisans without Constraint: Political
Polarization and Trends in American Public Opinion,” American Journal of Sociology, 100, 1-20.

Balz, Dan (2006) “Democrats Aim to Regain Edge in Getting Voters to the Poll,” 8 October,
Washington Post, A6.

97
---- and Mike Allen (2004), “Four More Years Attributed to Rove's Strategy,” 7 November The
Washington Post, A1.
Bannon, Declan P. (2003), “Voting, Non-voting and Consumer Buying Behaviour:
Non-voter Segmentation (NVS) and the Underlining Causes of Electoral Inactivity,” Journal of Public Affairs, 3(2), 138-151.

Bartels, Larry. M. (1992), “The Impact of Electioneering in the United States” in Electioneering:
A Comparative Study of Continuity and Change, David Butler and Austin Ranney, eds., Oxford,
UK: Clarendon Press, 244-277.

Bergan, Daniel, Alan Gerber, Donald Green, and Costas Panagopolous (2005), “Grassroots
Mobilization and Voter Turnout in 2004,” Public Opinion Quarterly, 69,760-777.

Berry, Leonard L. (1995), “Relationship Marketing of Services-Growing Interest,
Emerging Perspectives, “Journal of the Academy of Marketing Science,” 23, 236-245.

Butler, Patrick and Neil Collins (1994), “Political Marketing: Structure and Process,” European
Journal of Marketing, 28(1), 19-34.

Campbell, James E. (2000). The American Campaign, College Station, TX: Texas A&M Press.

Carmines, Edward. G. and James A. Stimson (1980), “The Two Faces of Issue Voting,”
American Political Science Review, 74, 78-91.

98
Center for the Study of Elections and Democracy (2004), “The 2004 Election Panel Study, BYU
Center for the Study of Elections and Democracy and UW-Madison Wisconsin Advertising
Project,” Electronic resources from the EPS Web site (http://csp.polisci.wisc.edu/BYU_UW/).
Madison, WI: University of Wisconsin, Wisconsin Advertising Project [producer and distributor].

Dermody, Janine and Stuart Hanmer-Lloyd (2004), “Segmenting Youth Voting Behaviour through Trusting-Distrusting Relationships: A Conceptual Approach,” International Journal of
Nonprofit and Voluntary Sector Marketing, 9(3), 202-217.

DeSarbo, Wayne S., Wagner Kamakura, and Michel (2006), “Latent Structure Regression,” in
The Handbook of Marketing Research, Rajiv Grover and Marco Vriens, eds., Thousand Oaks,
CA: Sage Publications, Inc., 394-417.

Doyle, Michael (2007), “Top Court Takes a Look at Campaign Finance Reform,” New
Star-Ledger.

Dreazen, Yochi J. (2006), “Democrats Playing Catch-up, Tap Database to Woo Potential Voters,”
31 October, Wall Street Journal.

Druckman, James N., Lawrence R. Jacobs, and Eric Ostermeier (2004), “Candidate Strategies to
Prime Issues and Images,” Journal of Politics, 66(4), 1180-1202.

Edsall, Thomas B. and James V. Grimaldi (2004), “On Nov. 2, G.O.P. Got More Bang for its
Billion, Analysis Shows,” 30 December, Washington Post, A03.

99
Elder, Janet (2007), “Soccer Moms are so 1996. Try Wal-Mart Women,” 23 September, The New
York Times, 14.
Ferrick, Tom (2006), “You, too, Can Expect a Call from Bill,” 1 October, The Philadelphia
Inquirer, page B01.

Finkel, Steven. E. (1993), “Reexamining the 'Minimal Effects' Model in Recent
Presidential Campaigns, “Journal of Politics, 55(1), 1-21.

Flanigan, W. H. and Zingale, N. H. (2006), Political Behavior of the American Electorate,
Washington, DC: CQ Press, 11th edition.

Fournier, Patrick, Richard Nadeau, Andre Blais, Elisabeth Gidengil, and Neil Nevitte
(2004), “Timing-of-Voting Decision and Susceptibility to Campaign Effects,” Electoral
Studies, 23, 661-681.

Francia, Peter L. (2001), “Begging for Bucks-Campaign Fundraising,” Campaigns
& Elections, 1-5.

Franz, Michael, Paul B. Freedman, Kenneth M. Goldstein, and Travis N. Ridout (2008),
Campaign Advertising and American Democracy, Philadelphia, PA: Temple University Press.

Freedman, Paul B. and Kenneth M. Goldstein (1999), “Measuring Media Exposure and the
Effects of Negative Campaign Ads,” American Journal of Political Science, 43(4), 1189-1208.

100
Gelman, Andrew and Gary King (1993). Why are American Presidential Campaign Polls so
Variable when Votes are so Predictable?” British Journal of Political Science, 23, 409-451.

----, John B. Carlin, Hal S. Stern, and Donald B. Rubin (2004) Bayesian Data Analysis. New
York: Chapman & Hall, 2nd edition.

----, Boris Shor, Joseph Bafumi, and David Park (2007), “Rich State, Poor State, Red State, Blue
State: What's the Matter with Connecticut?” Quarterly Journal of Political Science, 2(2):345-367.

Glantz, Stanton A., Alan I. Abramowitz, and Michael P. Burkar (1976), “Election
Outcomes: Whose Money Matters?” Journal of Politics, 38(4), 1033-1038.

Greeley, Andrew (2007), “Big Spending Wins out in Elections,” Chicago Sun-Times.

Hamburger, Thomas and Peter Wallsten (2006), One Party Country: The Republican Plan for
Dominance in the 21st Centrury, Hoboken, NJ: Wiley.

Harnden, Toby (2007), “Republicans Attack Obama,” 6 August, The Daily Telegraph, 1.

Harris, Phil (1996), “Editorial,” European Journal of Marketing, 30(10/11):18-21.

Henneberg, Stephan C. (2004), “The Views of Advocatus Dei: Political Marketing and its
Critics,” Journal of Public Affairs, 4(3), 225-243.

Hess, Ronald. L., Shankar Ganesa., and Noreen M. Klein (2003), “Service Failure and

101
Recovery: The Impact of Relationship Factors on Customer Satisfaction,” Journal of the
Academy of Marketing Science, 312, 127-145.

Hillygus, D. Susnhine and Simon Jackman (2003), “Voter Decision Making in Election 2000:
Campaign Effects, Partisan Activation, and the Clinton Legacy,” American Journal of Political
Science, 47(4), 583-596.

---- and Todd Shields (2008). The Persuadable Voter: Wedge Issues in Presidential Campaigns.
Princeton, NJ: Princeton University Press.

Holbrook, Thomas M. (1996). Do Campaigns Matter? Thousand Oaks, CA: Sage Publications,
Inc.

Imai, Kosuke, Gary King, and Olivia Lau (2006), “Zelig: Everyone’s Statistical Software,” http://GKing.Harvard.Edu/zelig. Iyengar, Shanto and Adam F. Simon (2000), “New Perspectives and Evidence on Political
Communications and Campaign Effects,” Annual Review of Psychology, 51:149-169.

Key, Valdimer O. (1966). The Responsible Electorate: Rationality in Presidential Elections,
1936-1960, New York: Vintage Books.

Klein, Jill G. and Rohini Ahluwalia (2005), “Negativity in the Evaluation of Political
Candidates,” Journal of Marketing, 69(1), 131-148.

102

Kohli, Ajay K. and Bernard J. Jaworski (1990), “ Market Orientation: The Construct, Research
Propositions, and Managerial Implications,” Journal of Marketing, 54, 1-18.

Kotler, Phillip. and Armstrong, Gary. (2004). Principles of Marketing. Upper Saddle River, NJ:
Pearson Prentice Hall, 10th edition.

---- and Neil Kotler, (1999), “Political Marketing: Generating Effective Candidates, Campaigns, and Causes, in Handbook of Political Marketing, Bruce I. Newman, ed., Thousand Oaks, CA:
Sage Publications, Inc., 3-18.

---- and Sidney J. Levy (1969), “Broadening the Concept of Marketing,” Journal of Marketing,
33(1), 10-15.

Lees-Marshment, Jennifer (2001), “The Marriage of Politics and Marketing,” Political Studies,
49, 692-713.

Luck, David J. (1969), “Broadening the Concept of Marketing-Too Far,” Journal of
Marketing, 33, 58-63.

Lundry, A.lex (2004), “Microtargeting: Knowing the Voter Intimately,” Winning
Campaigns Magazine, 4(1), 1-3.

Mark, David (2004), “Review of ‘The Greatest Communicator: What Ronald Reagan Taught Me about Politics, Leadership, and Life,’” Campaigns & Elections, 25, 1-3.

103

Marland, Alex (2003),”Marketing Political Soap: A Political Marketing View of Selling
Candidates like Soap, of Electioneering as a Ritual, and of Electoral Military
Analogies,” Journal of Public Affairs, 3(2), 103-15.

Mauser, Gary A. (1983). Political Marketing: An Approach to Campaign Strategy, New York:
Praeger.

McDaniel Jr., Carl and Roger Gates (2007), Marketing Research, Hoboken, NJ: John Wiley &
Sons, Inc., 7th edition.

Miller, Joanne and Jon Krosnick (2000), “New Media Impact on the Ingredients of Presidential
Evaluations,” American Journal of Political Science, 44(2), 295-309.

Miller, Warren E. and Merrill J. Shanks (1996), The New American Voter, Cambridge, MA:
Harvard University Press.

Newman, Bruce I. (1994), The Marketing of the President. Thousand Oaks, CA: Sage
Publications, Inc.

---- (1999), “Preface,” in Handbook of Political Marketing, Bruce I. Newman, ed., Thousand
Oaks, CA: Sage Publications, Inc., xii-xv.

104
---- (2001), “An Assessment of the 2000 U.S. Presidential Election: A Set of Political Marketing
Guidelines,” Journal of Public Affairs, 1(3), 210-216.

---- (2002), “Broadening the Boundaries of Marketing: Political Marketing in the New
Millennium,” Psychology & Marketing, 19(12), 983-986.

---- and Jagdish N. Sheth, (1985), “A Model of Primary Voter Behavior,” Journal of Consumer
Research, 12(2), 178-187.

Niffenegger, Phillip B. (1989), “Commentary: Strategies for Success from the Political
Marketers,” Journal of Consumer Psychology, 6(1), 45-51.

Open Secrets (2007), “Open Secrets,” (accessed December 15, 2007),
[http://www.opensecrets.org]

O'Shaughnessy, Neil. (2002), “Toward an Ethical Framework for Political Marketing,”
Psychology & Marketing, 19(12), 1079-1094.

Perreault, William D. and McCarthy, E. J. (2003). Essentials of Marketing: A Global-Managerial
Approach, New York: McGraw-Hill/Irwin, 9th edition.

Petrocik, Jon R. (1996), “Issue Ownership in Presidential Elections, with a 1980 Case Study,”
American Journal of Political Science, 40(3), 825-850.

105
Popkin, Samuel L. (1994). The Reasoning Voter. Chicago: University of Chicago Press, 2nd edition. Quelch, Jon A. and Katherine E. Jocz (2007), Greater Good: How Good Marketing Makes for
Better Democracy, Cambridge, MA: Harvard Business Press.

Riker, William H. (1996), The Strategy of Rhetoric, New Haven, CT: Yale University Press.

Rothschild, Michael L. (1978a), “Marketing Communications in Nonbusiness
Situations, or Why it's so Hard to Sell Brotherhood like Soap,” Journal of Marketing,
43(Spring), 11-20.

---- (1978b), “Political Advertising: A Neglected Policy Issue in Marketing,” Journal of
Consumer Research, 15, 58-71.

Russell, Alec (2006), “Mortgage Moms May Hold Key to a Democat Win,” 6 September, The
Daily Telegraph, 17.

Scammell, Margaret (1999), “Political Marketing: Lessons for Political Science,” Political
Studies, XLVII, 718-739.

Schiffman, Leon G., Elaine Sherman, and Nicole Kirpalani (2002), “Trusting Souls: A
Segmentation of the Voting Public,” Psychology & Marketing, 19 (12), 993-1007.

Seelye, Katherine Q. (2004), “How to Sell a Candidate to a Porsche-Driving,

106
Leno-loving NASCAR Fan,” 6 December, The New York Times, A16.

---- (2007), “Overheard” 29 May, The New York Times.

Shaw, Daron R. (1999), “A Study of Presidential Campaign Event Effects from 1952 to 1992,” The Journal of Politics, 61(2), 387-422.

---- (2006), The Race to 270, Chicago, IL: The University of Chicago Press.

Sparrow, Nick and John Turner (2001), “The Permanent Campaign: The Integration of Market
Research Techniques in Developing Strategies in a More Uncertain Political Environment,”
European Journal of Marketing, 35(9/10), 984-1002.

Wattenberg, Martin P. and Craig L. Brians (1999), “Negative Campaign Advertising:
Demobilizer or Mobilizer? American Political Science Review, 93(4), 891-899.

Wedel, Michel and Wagner Kamakura (2000), Market Segmentation: Conceptual and
Methodological Foundations. Boston, MA: Kluwer Academic Publishers, 2nd edition.

Wlezien, Christopher. and Erikson, R. S. (2001), “Campaign Effects in theory and practice. American Politics Research, 29(5):415-436.

---- and ---- (2002), “The Timeline of Presidential Election Campaigns,” Journal of Politics,
64(4), 969-993.

107
YouGovPolimetrix (2008), “Understanding People: The Next Generation of Voter Data,”
(accessed July 18, 2008), [www.polimetrix.com].

Wikipedia (2007a), “The United States Presidential Election, 1800,” (accessed June 1, 2008),
[http://en.wikipedia.org/wiki/United_States_presidential_election,_1800].

Wikipedia (2007b), “The United State Presidential Election, 1828,” (accessed June 1, 2008),
[http://en.wikipedia.org/wiki/United_States_presidential_election,_1828].

108
APPENDIX A
POLICY VARIABLES INCLUDED IN THE ANALYSIS BY YEAR
VARIABLES USED IN ANALYSIS, 1988
Demographic Covariates: age, gender, marital status, church attendance, education, race, income, region (South)
Economic Variables: government spending and services, government health insurance, government should guarantee jobs
Foreign Policy: defense spending, cooperation with Russia, pressure on South Africa
Moral Issues: abortion, gay job discrimination laws, adjust morals, more tolerant other lifestyles, emphasis on traditional values, new lifestyles cause problems, women's role
Civil Rights: ensure equal rights, gone too far pushing equal rights, should worry less about equality, big problem if unequal chances, blacks should overcome without any special favors, blacks gotten less than they deserve, blacks should\ try harder to succeed, history has made it harder for blacks to succeed

VARIABLES USED IN ANALYSIS, 1992
Demographic Covariates: age, gender, marital status, church attendance, education, race, income, region (South)
Economic Variables: government spending and services, government should guarantee jobs
Foreign Policy: Defense spending, willingness to use force
Moral Issues: Abortion, gay job discrimination laws, homosexuals allowed in military, gay couples allowed to adopt, adjust morals, more tolerant, emphasis on traditional values, new lifestyles cause problems, women's role

109
Civil Rights: Preferential treatment in job hiring, quotas in college admissions, should ensure equal rights, gone too far pushing equal rights, should worry less about equality, not a big problem if unequal chances, big problem if unequal chances

VARIABLES USED IN ANALYSIS, 1996
Demographic Covariates: age, gender, marital status, church attendance, education, race, income, region (South)
Economic Variables: Government spending and services, government provided health insurance, government should guarantee jobs, environment vs. jobs tradeoff
Foreign Policy: Defense spending, willingness to use force
Moral Issues: Abortion, gay job discrimination laws, homosexuals allowed in military
Civil Rights: Aid to blacks, society should see to equal opportunity, gone too far pushing equal rights, big problem if unequal chances, should worry less about equality, not a big problem if unequal chances

VARIABLES USED IN ANALYSIS, 2000
Demographic Covariates: age, gender, marital status, church attendance, education, race, income, region (South)
Economic Variables: Government spending and services, government provided health insurance, government should guarantee jobs, environment vs. jobs tradeoff
Foreign Policy: Defense spending, level of foreign aid
Moral Issues: Abortion, partial birth abortion, gay job discrimination laws, homosexuals allowed to serve in military, adjust morals, more tolerant, emphasis on traditional values, new lifestyles cause problems, women's role

110
Civil Rights: Aid to blacks, affirmative action, blacks should overcome without special favors, blacks gotten less than they deserve, blacks should try harder to get ahead, past discrimination affects blacks today

VARIABLES USED IN ANALYSIS, 2004
Demographic Covariates: age, gender, marital status, church attendance, education, race, income, region (South)
Economic Variables: Government spending and services, government provided health insurance, government should guarantee jobs, environment vs. jobs tradeoff
Foreign Policy: Defense spending, use of diplomacy vs. military
Moral Issues: Abortion, partial birth abortion, gay job discrimination laws, homosexuals allowed to serve in military, adjust morals, more tolerant, emphasis on traditional values, new lifestyles cause problems, women's role
Civil Rights: Blacks should work their way up like others, history has made it hard for blacks to succeed, blacks gotten less than they deserve, blacks should try harder, should do what is necessary for equal rights, gone too far pushing for equal rights, big problem if unequal chances, should worry less about equality, not a big problem if unequal chances, fewer problems if there were equality

111
APPENDIX B
Selecting the Optimal Number of Segments, 1988-2004
Selecting the optimal solution is part science and part art. As a general rule, the optimal number of segments is chosen when the information criteria (BIC, AIC, AIC3) are at a minimum.
There are many instances, however, when the information criteria do not agree. The researcher, therefore, is left to make a judgment call as to what the “correct" solution is. Another important factor is segment size. A 15-segment solution may be best in terms of information criteria, but the segments are too small. Generally speaking, no segments were included that did not comprise at least 3 % of the population. The following is a brief description as to how the segments were selected for each year.
1988 BIC, AIC, AIC3 were all at low values for the seven-segment solution. Although the AIC3 was lower for the ten-segment solution, several of the segments were less than 2 % of the sample.
1992 AIC3 was at its lowest value for the seven-segment solution. AIC was lower at ten-segments, but segments were too small; BIC was lower at six-segment solution but given the other criteria pointed toward the seven-segment solution, the seven-segment was chosen.
1996 AIC3 reached minimum at ten-segment solution but segments were small, AIC dropped from 1376 at six segments to 1328 at seven segments and then continued to decline until ten-segment solution, BIC jumped after seven-segment solution. This was a difficult call but seven segments is a compromise choice.
2000 AIC3 at minimum with seven segments, BIC climbed after seven segments, AIC's second lowest value was at seven-segment solution.
2004AIC, AIC3 at minimum with seven segments, BIC climbed steadily although there was a large jump from seven segments to eight segments. Given the criteria, the seven-segment solution once again seemed like the best choice. For every year the choices were made using the

112 information criteria as a guide, but because there is no failsafe method, some judgment was used when the criteria did not agree.
1988 Information Criteria Results

Log-likelihood (LL)
Log-prior
Log-posterior
BIC (based on LL)
AIC (based on LL)
AIC3 (based on LL)
CAIC (based on LL)

5-Class
6-Class
7-Class
8-Class
9-Class
10-Class
-789.12 -756.489 -657.707 -637.548
-667.81 -576.594
-8.5842 -11.3096 -57.6537 -55.9172 -21.5122 -61.6974
-797.705 -767.799 -715.361 -693.466 -689.322 -638.291
2195.417 2274.162 2220.607 2324.296 2528.827 2490.403
1758.241 1734.978 1579.415 1581.097
1683.62 1543.187
1848.241 1845.978 1711.415 1734.097
1857.62 1738.187
2285.417 2385.162 2352.607 2477.296 2702.827 2685.403

1992 Information Criteria Results

Log-likelihood (LL)
Log-prior
Log-posterior
BIC (based on LL)
AIC (based on LL)
AIC3 (based on LL)
CAIC (based on LL)

5-Class
-1047.85
-8.7542
-1056.6
2563.954
2227.699
2293.699
2629.954

6-Class
-1022.68
-12.5321
-1035.21
2620.034
2207.358
2288.358
2701.034

7-Class
-982.396
-21.9261
-1004.32
2645.891
2156.793
2252.793
2741.891

8-Class
-979.604
-17.3933
-996.997
2746.728
2181.208
2292.208
2857.728

9-Class
-969.363
-19.448
-988.811
2832.668
2190.726
2316.726
2958.668

10-Class
-925.791
-16.6826
-942.473
2851.944
2133.581
2274.581
2992.944

1996 Information Criteria Results

Log-likelihood (LL)
Log-prior
Log-posterior
BIC (based on LL)
AIC (based on LL)
AIC3 (based on LL)
CAIC (based on LL)

5-Class
6-Class
7-Class
8-Class
9-Class
10-Class
-620.011 -597.336
-556.1 -534.785 -510.272 -451.326
-13.7416 -19.9984 -19.3539 -25.8618 -26.3792 -38.1666
-633.753 -617.335 -575.454 -560.647 -536.651 -489.493
1739.448 1808.831 1841.092 1913.194 1978.903 1975.743
1388.022 1376.672
1328.2 1319.569 1304.545 1220.652
1462.022 1467.672
1436.2 1444.569 1446.545 1379.652
1813.448 1899.831 1949.092 2038.194 2120.903 2134.743

113

2000 Information Criteria Results

Log-likelihood (LL)
Log-prior
Log-posterior
BIC (based on LL)
AIC (based on LL)
AIC3 (based on LL)
CAIC (based on LL)

5-Class
6-Class
7-Class
8-Class
9-Class
10-Class
-397.794 -372.366 -319.134 -310.019 -299.433 -260.849
-12.3193 -14.9245 -25.9805 -23.5379 -26.0907 -25.9484
-410.113
-387.29 -345.115 -333.556 -325.524 -286.797
1252.812 1306.993 1305.568 1392.375 1476.243 1504.112
943.5878 926.7311 854.2679 870.0371 882.8665
839.698
1017.588 1017.731 962.2679 995.0371 1024.867
998.698
1326.812 1397.993 1413.568 1517.375 1618.243 1663.112

2004 Information Criteria Results

Log-likelihood (LL)
Log-prior
Log-posterior
BIC (based on LL)
AIC (based on LL)
AIC3 (based on LL)
CAIC (based on LL)

5-Class
-437.117
-10.0298
-447.146
1356.025
1022.233
1096.233
1430.025

6-Class
-418.015
-10.9925
-429.007
1428.503
1018.029
1109.029
1519.503

7-Class
-381.26
-19.8084
-401.068
1465.675
978.5191
1086.519
1573.675

8-Class
-370.236
-19.6027
-389.839
1554.311
990.4727
1115.473
1679.311

9-Class
-345.504
-19.9397
-365.443
1615.528
975.0075
1117.008
1757.528

10-Class
-309.031
-23.9157
-332.947
1653.264
936.0622
1095.062
1812.264

APPENDIX C
Segment Descriptions by Year
1992
Moderate
Republicans
Conservative
Democrats
Working Class
Democrats
Religious Left

1996
Conservative
Democrats
Moderate
Republicans
Independents

2000
Moderate Democrats

Segment 4

1988
Moderate
Republicans
Working Class
Republicans
Working Class
Democrats
Religious Right

Moderate Democrats

Segment 5

Religious Left

Independents

Social Conservatives

Segment 6

Moderate Democrats

Wealthy Republicans

Religious Left

Traditional
Democrats
Conservative
Democrats
Fiscal Conservatives

Segment 7

Conservative
Democrats

Young Democrats

Southern Democrats

Southern Democrats

Segment 1
Segment 2
Segment 3

Moderate
Republicans
Social Conservatives

2004
Conservative
Democrats
Wealthy Republicans
Traditional
Democrats
Moderate
Republicans
Independents
Southern
Republicans
Religious Left

APPENDIX D
DERIVED SEGMENTS BY YEAR

Conservative
Democrats
Fiscal
Conservatives
Independents
Moderate
Democrats
Moderate
Republicans
Religious
Left
Religious
Right
Social
Conservatives
Southern
Democrats
Southern
Republicans
Traditional
Democrats
Wealthy
Republicans
Working Class
Democrats
Working Class
Republicans
Young
Democrats

1988
X

1992
X

1996
X

2000
X

2004
X

X
X
X

X

X

X

X
X

X

X

X

X

X

X

X
X

X
X

X

X

X
X
X

X
X

X

X
X

X
X

David N. LeBaron
Curriculum Vitae
Department of Marketing
University of Texas San Antonio
One UTSA Circle
San Antonio, TX 78249

Office: 210.458.6333
Fax: 210.458.6335
Email: david.lebaron@utsa.edu

EDUCATION
Ph.D.

Pennsylvania State University, December 2008
Department of Marketing

M.S.

University of Utah, 1996-1999
Department of Economics, completed coursework

B.A.

Brigham Young University, August 1995
Department of Political Science

PUBLICATIONS
DeSarbo, Wayne, Selin Atalay, David LeBaron, and Simon Blanchard (2008),
“Estimating Multiple Consumer Segment Ideal Points from Context Dependent Survey Data,”
Journal of Consumer Research, 35 (June) 142-153.

WORKS IN PROGRESS
“The Elephant (and Donkey) in the Room: Why Academic Marketers Should Study Politics”
“Making the Personal Political: Microtargeting in Presidential Elections”
“Latent Segmentation and Dimension Reduction in Voter Behavior”

HONORS AND AWARDS
Graduate Research Assistantship, Pennsylvania State University, 2002-2008
Top Candidacy Exam Score, Department of Marketing, Pennsylvania State University, January
2005 ($1000 award)
Graduate Teaching Assistantship, University of Utah, 1996-1998