Christopher Wlezien, author of The Timeline of Presidential Elections: How Campaigns Do (and Do Not) Matter joins the conversation during today’s guest spot. Christopher Wlezien analyzes data from 2,000 national polls between 1952 and 2008 and concluded that a persons preferences become clear in the last six months of the campaign. So what does it mean for this year’s election. Be sure to tune in for the full conversation at 3:30pm and check out an excerpt from his book below.
Election Campaigns and Voter Preferences
Imagine the timeline of a presidential election campaign. We begin the timeline at some early point before the election, perhaps as soon as polls ask voters whom they will support. The timeline ends on Election Day. At the beginning, the polls reveal the electorate’s preliminary vote intentions. On Election Day, at the end of the campaign, the electorate reaches a fi nal verdict. In this book we trace the national vote division as it evolves over the campaign timeline.
We ask: How much does the vote change over the timeline? Is the shift a smooth trajectory, or does the aggregate vote lurch over the timeline in a series of bumps and wiggles? What are the forces that infl uence the vote and when do they occur? When new events affect the vote decision, how long do the effects last? To what extent are their effects temporary and to what extent do they persist to affect the outcome on Election Day? These are some of the questions we address in this book. Their answers inform us about the importance of the election campaigns—often beginning before the national party conventions—on the outcome of the presidential election.
How much do campaigns matter? Here, some division can be seen between the views of political practitioners and journalists on the one hand, and academic scholars on the other. Especially in the heat of the campaign, practitioners and journalists emphasize elections as a battle of rival campaigns, with the winning team determined by campaign quality plus the random shocks from unexpected campaign events. In the extreme, elections are decided by which side is better at the public relations art of persuading voters.
Of course, all observers recognize that campaign outcomes involve more than a combination of salesmanship and luck. When political scien- tists study elections, they tend to emphasize the political environment— often referred to as the “fundamentals” of the campaign. Many concede that campaigns may matter, but that they do so mostly to channel the vote toward a verdict that can largely be seen in advance from the fundamentals.
Always prominent in discussions of campaign fundamentals is the economy’s performance. But the fundamentals can also include the electorate’s net evaluation of the competence and perceived ideological positioning of the major political parties and candidates (Lewis-Beck and Rice 1992; Gelman and King 1993; Holbrook 1996; Campbell 2008a; also see Popkin 1991). In the extreme, the fundamentals of the election are in place before the campaign begins, and the campaign is a mere conduit to drive the voters’ decision to its deterministic decision. Election outcomes are not simply the residue of campaign quality plus a dose of chance. Neither are they the automatic result of deterministic forces that can be foreseen in advance of the campaign. Voters are influenced by a variety of factors, some stemming from the candidates’ campaigns and some beyond the candidates’ control.
The general puzzles that motivate many discussions of elections remain: how much do campaigns affect elections, and how much do the fundamentals shape the campaign and its effect on the voters? (See, e.g., Holbrook 1996, 2010; Campbell 2008a; Stimson 2004; Bartels 2006; Ansolabehere 2006; and Vavreck 2009.) In this book, we translate general arguments about the effects of campaign events into a set of formal expectations. We then analyze all available national polls from the fifteen presidential elections from 1952 through 2008. We have three main goals. The fi rst is to identify the dynamics of the electorate’s vote intentions over the campaign timeline. The second is to assess the extent to which changes in voter preferences over the campaign timeline persist to impact the Election Day vote. The third is to model the sources of electoral change over the campaign timeline. To complement the analysis of aggregate poll results, we also examine individual-level poll responses. This allows us to observe the crystallization of voter preferences over the campaign timeline.
At the beginning of an election year, trial-heat polls reveal little about the eventual Election Day verdict By April, however, the electorate forms impressions of the candidates that bear some resemblance to the final verdict.
As the campaign progresses, the electorate’s vote division typically resembles the outcome that analysts predict from the fundamentals, though not perfectly and sometimes not much at all. The vote division rarely ends where it starts early in the election year, but, except (occasionally) in the af- You are reading copyrighted material published by University of Chicago Press. Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal and injures the author and publisher. Election Campaigns and Voter Preferences / 3 termath of the party conventions, change is usually gradual. As we will see, the relative stability of the electorate’s preferences is often masked by sampling error in polls. And the real movement in the electorate’s preferences often is nothing more than short-term change that fades quickly. Elections are decided by the slow evolution of campaign events that leave an impact that lasts until Election Day.
This book analyzes all available national vote intention polls for the fi fteen presidential elections from 1952 through 2008. The sheer volume of polls—nearly 2,000 of them—allows us to assess the dynamics of aggregate electoral preferences in considerable detail. We can determine whether and how preferences change over the course of the election year; indeed, we can quantify and date the change. We also can determine whether and the extent to which the change in preferences we observe actually lasts to affect the outcome. Finally, we can assess the causes of aggregate preference change. To complete the story, we analyze individual vote preferences at different points of the campaign. This makes more understandable those patterns we see at the aggregate level. We are thus able to offer a comprehensive accounting of preference evolution. What we glean from our analysis provides answers to important questions about electoral preferences: how do they change over time, why do they change, and with what effect on the final outcome?
In some ways, the timeline of a presidential campaign is like a season of major league baseball. In the spring, each team and its fans believe they have a chance at getting to the playoffs and winning the World Series. It may seem that, with a few good breaks, any team can go all the way. In the end, however, the teams with the highest caliber of talent at the start of the season usually get to the playoffs. The list of postseason entries still can include occasional surprises, which makes the long season interesting.
The electoral parallel is obvious. In the spring of election year, parties and candidates, plus their supporters, see a pathway to victory. Political journalists and pundits speculate that the outcome will depend on who runs the smartest campaign and how the outcome may turn on chance events. But the outcome can typically be foreseen from the fundamentals of the campaign; the candidate favored by the economy and presidential performance usually wins. Surprises are possible, however. Just as in baseball, the season must be played out to determine who wins.
We show how, over the timeline of presidential campaigns, the electorate’s collective vote choice undergoes a slow evolution. Most of our analysis concentrates on what polls show within 200 days of the election. The You are reading copyrighted material published by University of Chicago Press. Unauthorized posting, copying, or distributing of this work except as permitted under U.S. copyright law is illegal and injures the author and publisher.
200th day before the election (ED-200) occurs in mid-April. At that time, the likely major-party candidates are identifi ed and matched up in the pollsters’ trial-heat questions in each of our 15 election years. Those polls, however, are an uncertain guide to the final outcome. After the fact, we know that polls at ED-200 can explain slightly less than half the variance in the fi nal vote division. (We also show that polls from the beginning of the election year have virtually no redictive power, which means that preferences start to come into focus as the nomination process unfolds.)
The polls from April also provide a useful guide to the Election Day winner, as the polls are “right” more often than not. Polls as of ED-200 “erred” only in 1980, 1988, and 1992, while showing a plurality for the final winner in eleven other instances.1 (In 2004, the ED-200 polls showed a virtual tie.) In short, the early polls are fairly useful for identifying the winner. Where they “err,” we do not blame the early polls but rather attribute it to the flow of subsequent events. In other words, the campaign seems to matter.
As everyone who closely follows election polls knows, the numbers bounce around a lot from day to day, and can vary from poll to poll within the same reporting period. Much of this is noise from sampling error. Our book attempts to extract the signal of the ever-moving division of voter preferences over the campaign. This electoral movement is slow—far less than one might think from comparing two polls from nearby dates. With a series of graphs and statistics, we track this slow evolution of voter preferences.
We also identify some of the sources of this slow evolution. Early polls typically start with one candidate ahead by a more one-sided margin than the fi nal vote. Seemingly, if surprise snap elections were called in April, voters would give lopsided verdicts often quite different from their Election Day verdict. We see three periods in the campaign timeline during which aggregate preferences get reshuffl ed more than usual. The fi rst is during the early stages of the primary season, when voters often are fi rst learning about some of the nominees. The second is during and after the period when the political parties hold their national conventions. The third is the fi nal run-up to the election during the fi nal campaign week, a time when many voters decide. In each instance, the electoral verdict tightens, moving closer to 50–50. Of course it is important to know what drives electoral change over the campaign. We show that even by April, trial-heat polls incorporate considerationsthat no longer matter by Election Day. Yet April polls also contain information that persists to become part of the fi nal electoral verdict. Between April and November, something happens. In some fashion, the campaign delivers the economic and political fundamentals to the voters.
It also delivers less tangible information that analysts cannot readily identify. It is well known that presidential election outcomes are predictable, at least up to a point. Despite all the media attention paid to the many events and drama during campaigns, there are certain things that powerfully structure the vote on Election Day. At the individual level, party identification is of great importance. Other factors also matter at the individual level, including class, other social cleavages, and policy preferences, to name but a few
of the many things that structure individuals’ votes. The point is that voters tend to line up in fairly expected ways on Election Day (see Gelman and King 1993; Campbell 2008a; Andersen, Tilley, and Heath 2005). To point out the obvious, the electoral verdict changes from election to election. It is not that everyone changes or even that most do, as the bulk of partisans vote for their parties or candidates of their parties year in and year out. The ones that change tend to be those who are least attached to particular parties. These “fl oating voters” are more likely to reflect shortter considerations, such as the state of the economy or the more general performance of the incumbent president (Zaller 2004).2 There is more to election outcomes than the recent degree of peace and prosperity, but incumbent presidential performance in these domains tells much of the story (see, e.g., Fair 1978; Hibbs 1987; Erikson 1989; Lewis-Beck 2005; Holbrook
2010). Political factors, including the candidates’ policy positions, are also important (Erikson, MacKuen, and Stimson 2002; Vavreck 2009).
1.1.1 The Fundamentals
A common view is that the campaign delivers the fundamentals (e.g.,
Gelman and King 1993; Campbell 2008a). The fundamentals are typically described as a set of economic and political circumstances known long before the election, so that the results are knowable in advance, perhaps before the eventual outcome is evident in the polls. The campaign effectively brings home the fundamentals to voters. If the fi nal result departs from what the fundamentals predict, then the campaign must have failed to fully enlighten voters by Election Day.3 Our view is different. We conceive of the fundamentals not by their content but by their persistence. The fundamentals are those things that cause a long-term shift in voter preferences—long-term, that is, for the length of
You are reading copyrighted material published by University of Chicago Press.
Unauthorized posting, copying, or distributing of this work except as permitted
under U.S. copyright law is illegal and injures the author and publisher.