T HE
MES SA GE
MAT T ER S
Campaign Effects in Presidential Elections
[Abridged by Author for meeting at Princeton University: 75 pages]
LYNN VAVRECK,
UNIVERSITY OF CALIFORNIA, LOS ANGELES
DRAFT 4.0
April 25, 2008
CHAPTER ONE
PRESIDENTIAL CAMPAIGNS
In 1976, Jimmy Carter ran for President of the United States as a
trustworthy Washington outsider. He was elected by defeating an opponent
who was the consummate Washington insider – a man who pardoned
Richard Nixon and who was appointed to the Vice Presidency and the
Presidency. Gerald Ford was as inside the Washington beltway as one
could get – and he could not get out. Similarly, John F. Kennedy in 1960
recognized his opponent’s culpability in what was called “The Missile Gap”
– the alleged fact that America had fallen behind Russia in weapons
development. Richard Nixon, whose administration had presided over this
“slump” in American productivity, had no evidence it did not exist and
could not counter Kennedy’s claim. Thus, he too, was stuck – although the
economy was doing well and the war in Vietnam had not yet escalated.
What Carter and Kennedy both recognized and exploited was a
dimension of electoral politics on which their opponents faced constraints.
Although Ford and Nixon were members of incumbent administrations and
the nation’s economy was growing over the course of their stewardships,
their challengers defeated them. This book is about how candidates
campaign, what effects these campaigns have on voters, and how the
context of elections conditions all of these things in important ways.
2
What we know about why presidential candidates lose elections
mostly centers around campaign strategies or candidate style. Pundits are
quick to blame electoral losses on a poor campaign strategy or a candidate’s
inability to connect with voters on the stump. Journalists, however, are
rarely heard suggesting that a candidate lost because his or her policies were
unpopular. Why did Michael Dukakis lose the 1988 presidential election
after a 17-point lead coming out of his convention? It must be because
George Bush out-campaigned him, say the experts. The New York Times
Editorial Board wrote:
He [Dukakis] was not destined to lose at all, and did so only
because he ran a dismal campaign. . . ‘Why didn’t he say …’
became a virtual motto of endless exasperation – when the
Dukakis campaign gave leaden answers or no answers at all
to accusations about the Pledge of Allegiance or prison
furloughs or to questions like, ‘How would Governor
Dukakis feel if his wife were raped and murdered?’
(November 9, 1988, p. A34)
Similarly, why did Bill Clinton win the 1992 election at a time when the
incumbent Bush was popular and the economy was recovering from a
recession? Many pundits answered that it was Clinton’s “War Room”
campaign strategists who out-maneuvered the Bush campaign on a daily
basis, or that Bush himself was somehow a “bad” campaigner:
Something odd happens to Mr. Bush when he vaults into
‘campaign mode.’ His good manners fall away and he
stands revealed as Nasty-man. . . this time it went from Redbaiting to juvenile expostulations like, ‘My dog Millie knows
more about foreign affairs than these two bozos.’
(November 5, 1992, p. A34).
What exactly is a good campaign or a good campaigner? These
concepts have certain ephemeral qualities about them. Experts cannot
precisely detail what makes a campaign “good”, except maybe that it
produced a winner; and, they know a good campaigner when they see one.
Such explication is not helpful. Most notably it ignores the fact that one
3
candidate can be a good campaigner in one year (Bush in 1988) and a lousy
one in a later year (Bush in 1992). If campaign success were merely a
function of the candidate’s ability to strategize about how to beat the
opponent and then execute that strategy effectively, we would not expect to
see such differences in the successes of candidates like Bush (1988,1992),
Carter (1976,1980), and Nixon (1960,1968,1972).
There must be more to the story about why campaigns are successful
than strategy and execution. My aim in this book is to explore more
systematically the types of campaigns run by candidates for president of the
United States by paying special attention to the constraints candidate’s face
when running their campaigns.
The most important constraint for all candidates is the condition of
the nation’s economy. Other constraints include previously taken issue
positions or personal characteristics. I focus on illustrating why candidates
like Carter and Kennedy exploit their opponents’ weaknesses, while Ronald
Reagan and Dwight Eisenhower talk mainly about a booming economy,
largely ignoring their opponent’s presence in the race – and why these
campaign strategies are predictable well in advance of the election and
ultimately successful. My hope is that presidential campaigns may come to
be viewed not as exercises in strategy executed by idiosyncratic candidates
whose personal capabilities and whims influence success, but as logical,
rational and often pre-determined means toward an end.
BASIC QUESTIONS
Much attention has been given recently to presidential campaigns
because they are “too negative” or “too long” or cost “too much” money.
Proposed reforms include removing or lessening the role of money, giving
candidates free television advertising time, urging newspapers to report
when candidates are lying or misleading voters in their advertisements, and
asking candidates to sign “codes of conduct” or compacts to promise “good
behavior”. These complaints and reforms presuppose that money, ads,
4
newspaper coverage, and campaign tone all matter to voters on Election
Day. Or more generally, that what goes on during campaigns for the
Presidency matters to voters at all. The extent to which discussion about
these reforms increases in the absence of scholarly understanding about
whether and how campaigns “matter” to voters is striking. While
journalists, pundits, and voters may be confident that presidential
campaigns influence election outcomes, political scientists have not always
been so sure.
Party identification is still the greatest and most powerful predictor of
vote choice across any demographic group (Miller 1994; Bartels 2000).
Most people do not pay much attention to politics or campaigns, even when
faced with making a decision about their president. And, worse yet,
perhaps because of their lack of attention, it seems that many voters are
uninformed about where the candidates for president stand on various and
important issues. Some argue that since party identification is known,
presumably, before the campaign starts, and voters do not attend to
campaigns when they are happening, campaigns must not matter to voting
outcomes.
The study of campaigns, however, is not that simple. What does it
mean, for example, to say that a campaign “matters”? Does a campaign
matter if it changes someone’s choices or vote decision? This is simple –
probably so. However, does the campaign matter if it reinforces a voter’s
decision – if it makes the voter more confident of his or her choice? What
if campaigns change the focus of national discussion; do they matter then –
even if a campaign does not produce a winning candidate or change
people’s minds? If campaigns teach voters about the current state of the
economy or of education policy or of trade policy in America, does it matter
that voters have learned during the campaign even if their voting decisions
were left unaffected? Any sophisticated and systematic analysis of
campaigns and their effects must deal with these questions before moving
on to assess whether the process is in need of change.
5
Many have argued that investigating the effects of campaigns is so
complex as to be nearly impossible. Campaigns, because of their dynamic,
contemporaneous, competitive, and cumulative nature take place in a
research environment that is difficult to control. My own view is slightly
more optimistic. If political scientists can theorize about voting behavior
then we can theorize about campaign behavior and effects. Moreover,
starting from a theory about how and why campaigns can matter, we can
observe patterns of behavior among candidates that add to our
understanding of the dynamic, contemporaneous, competitive, and
cumulative nature of the campaign environment. That is what I attempt to
do in the pages that follow.
WHAT’S COMING
This book is divided into three parts. Part I sets the stage for the
other two by introducing theories of voting behavior and extrapolating them
into a theory of how and why campaigns can matter (Chapter 2). From this
theory, a campaign typology is developed that introduces two types of
campaigns, the clarifying campaign (in which candidates clarify their
position on an already important issue – the economy) and the insurgent
campaign (in which candidates attempt to reset the agenda from the
economy and onto an issue on which their opponent has previously
committed to a position less popular than their own). Each campaign type
consists of unique behavior by candidates and unique effects on voters
(Chapter 3). For example, in the clarifying campaign I expect candidates to
talk mainly about the economy – simply taking credit for the good
economic times or laying blame for the bad times. Voters, then, should be
very certain of the candidate’s positions on economic issues by the end of
the campaign as they learn about them throughout the process. In the
insurgent campaign the candidate is expected to talk mainly about an
insurgent issue (one on which his opponent is less popular then him). If the
6
insurgent candidate is successful, voters, over time, should begin to believe
that this issue is more important to their voting decision.
In Chapter 3 I also explain how the theory suggests candidates sort
themselves into these two categories. Of critical importance here is the
state of the nation’s economy. The candidate who is predicted to win the
election based on a simple economic forecast runs the clarifying campaign.
This could be the incumbent in a good economy or the challenger in a bad
economy. The predicted economic-loser runs the insurgent campaign. I use
a simple economic forecast to predict what types of campaigns the theory
suggests candidates will run over the period from 1952-2000.
Part II begins in Chapter 4 in which I detail the content of modern
presidential campaigns by conducting an analysis of campaign
advertisements, stump speeches, and campaign news coverage for every
election since 1952. Despite traveling with the candidates every day during
the campaign, the data reveal a great divide between what candidates say
and do and what the media report about them. In Chapter 5 I explain
candidate behavior in campaigns using the theory and typology as
foundations. Using content analyses from The New York Times, candidate
advertisements, and candidate stump speeches, I am able to compare what
candidates actually talked about during their campaigns with what the
theory predicts they should have talked about. The analytic power of the
theory is tested as I discover that eventual winners are most often those
candidates who behave as the theory and typology predict they should.
Candidates who violate the typology’s prescriptions lose elections.
In Chapter 6, Part III, I turn the investigative light onto the behavior
of voters in elections, searching for the unique effects associated with the
clarifying and insurgent campaigns. Using public opinion data from 1976
to 2000 provided by The National Election Study, I assess whether
candidates who conform to the typology’s prescriptions were able to
influence voters in specific and meaningful ways. Finally, in Chapter 7 I
engage larger theoretical questions about context and constraints. Can
7
incumbent presidents manipulate national context? Is there an insurgent
issue for every insurgent candidate or are some candidates just lucky? And,
finally, how do these findings fit with common notions about democracy
and elections?
Much of the theorizing and analyses in the beginning and latter part
of the book is based upon the use of statistical and mathematical models.
As much as possible I have tried to present material in the main sections of
the book in terms that citizens interested in politics can understand. The
nuances of mathematics and modeling, along with detailed information on
the data, models, and estimates from which my conclusions are drawn can
be found in the book’s appendices.
8
CHAPTER SIX
CONTENT V. CONTEXT:
MICRO-LEVEL TESTS OF THE THEORY
So far, I have described presidential campaign behavior as
predictable long before the candidates are known or the first general
election campaign dollar is spent. I have described a campaign
environment in which the economy plays a starring role and both candidates
react to it. I have shown that candidates seem to understand this world and
generally behave as the typology predicts they should, or they lose
elections. Presidential candidates, at least some of them, understand that
national context matters and that decisions about where to center their
campaigns have consequences. Some of them are better at it than others,
and some of them are exemplary (Kennedy, Nixon, Carter, and George W.
Bush). Candidate behavior in campaigns is predictable and systematic,
across candidates, across elections, and across varying national contexts.
But what I have yet to demonstrate is that this candidate behavior
matters to voters in systematic and important ways. I have shown that it
matters to outcomes, in general – but are the individual level mechanisms
implied by the spatial model in Chapter 2 real? Does the clarifying
candidate have to talk about the economy more than his opponent for voters
to connect him to the economic prosperity? Or is it enough for this
candidate that he happens to be in this fortunate position? How difficult is
it for insurgent candidates to increase the importance of issues in voters’
minds – hopefully issues on which they benefit among voters more than
their opponent? Is it even possible for insurgent candidates to prime issues
among the electorate? Even those insurgent candidates who chose
unwisely should have some success making those issues more important
9
among voters, but what does the evidence suggest? In order to understand
why the theory explains candidate behavior so well, it is necessary to
explore individual level data that can demonstrate how voters are affected
by the different types of campaigns.
Before I introduce these data, I want to set some expectations. Any
argument is only as good as the evidence on which it is based. Sometimes
we lose site of this. For a long time now, the divide between scholars of
politics and practitioners of politics over whether and how much campaigns
matter has grown. The people who work on campaigns for a living, the
candidates, even the voters know that campaigns are important. Their
evidence is anecdotal and personal, but real. And the people who analyze
campaigns, study them, and look deeply for patterns of effects have decided
campaigns matter only in minimal ways, and only at the margins. Their
evidence is observational, limited in quantity, but also real. Both sets of
evidence suffer from limitations, and I fear it is the limitations, not the
evidence itself, that is motivating the arguments. Too often we hear
anecdotal stories about a campaign strategist’s successes while tending not
to hear about the failures. Similarly, we hear about the lack of change in
voters’ attitudes or survey responses over the course of “the campaign”
without hearing that the campaign is being defined as a few weeks before
the actual election.
To make stronger arguments we need better evidence. In order for
analysts to argue that campaigns are important to voters, we need evidence
showing the initial perceptions and affinities voters have for candidates, and
by initial, I mean in the early stages of the nominating process, before the
nominees are announced. Further, we need these data for lots and lots of
people in order to isolate what might be a small (in size) yet important
effect on people’s decisions. Then, we ideally need many interviews with
these same people over the course of the campaign and over many years of
elections in order to track how opinions change leading up to vote choice.
And, the interviews would be quite long so that we could ask people about
10
all the possible sources of impact on their political attitudes and behavior.
The costs of such a study would be astronomical and as such, it is not
surprising that no study like this has been done in America.
What has been done, and done quite well, is no less impressive in
size, continuity and scope. Every presidential election year since 1952 (and
most midterms, too), the American National Election Study (ANES)
completes at least one thousand interviews with American citizens before
and after each presidential election.1 The project is currently funded mainly
by the National Science Foundation, but had many grantors over the years.2
The content of the study changes year to year, but a surprising number of
items have been retained over time. These data are analysts’ only and best
shot at uncovering a pattern of effects for presidential campaigns that span
the last half-century. It is to these data that I append my data on the content
of campaigns and news coverage in an effort to learn whether the things
candidates say and do in campaigns affect voters in elections. Of course,
these data are not perfectly suited for this task, and the limitations will
significantly impact the conclusions I can draw about campaign effects, but
they will allow me to clearly identify patterns in some cases and in others,
to carefully suggest relationships at work. Because the nature of the
argument I am making is temporal – and over a 55-year period of time – the
ANES is the best source of political time-series data available.
CAMPAIGN EFFECTS: WHAT DO VOTERS SAY ABOUT CANDIDATES AND
DOES CAMPAIGN DISCOURSE MATTER?
Among other things, the ANES contains measures that reflect the
way ordinary Americans talk about presidential candidates. In each election
since 1952, the ANES has asked a nationally representative sample of
survey respondents the following set of questions about each of the major
party candidates:
11
Now I'd like to ask you about the good and bad points of the major
candidates for President. Is there anything in particular about [insert
candidate's name] that might make you want to vote for him? What
is that? Anything else?
Is there anything in particular about [insert candidate's name] that
might make you want to vote against him? What is that? Anything
else?
Interviewers code up to five reasons for voting for and against each
candidate for a total of up to 20 reasons for the two major party candidates.3
These open-ended questions have the distinct advantage of allowing
respondents to convey whatever voting considerations they may hold.
Closed ended questions, in contrast, restrict respondents to a set of
predetermined considerations that the survey writers have identified in
advance. The flexibility of the open-ended questions is especially important
in tracing changes in the bases of presidential voting over time, since survey
writers are likely to do a better job some years than others in anticipating
the considerations that will be uppermost in Americans' minds.4
The candidate likes/dislikes questions are coded by the ANES into
hundreds of different substantive categories, which I recoded into the same
groups as those that describe the candidate and news content of campaigns:
the economy, domestic policy, foreign policy and defense, and traits or
character. The combination of data on campaign driven content in
advertisements and speeches, news coverage of campaigns, and voters’
evaluations of candidates makes this a unique and powerful dataset for
testing whether the things that candidates talk about in their campaigns
influence voters in elections. In other words, before I can argue that what
clarifying candidates say about the economy matters to voters, I want to
demonstrate that what candidates say during campaigns makes it into voters
thoughts about the candidates during the campaigns.
In order to test the fidelity of this relationship, I reduce these openended data to the proportions of comments in each issue area for each
12
candidate in any given year. If voters are affected by the campaign such
that the more candidates talk about something, the more voters either
believe that thing is important or learn more about the candidate’s position
on that issue (the two modes of campaign influence derived from the spatial
model of voting) then voters’ evaluations of the candidates ought to reflect
the things candidates are raising in their campaigns. And to some degree,
the comments ought to be responsive to changes in the composition of
candidate’s campaigns – maybe not over the course of a single election, but
historically we would expect that voters’ make fewer comments about
defense spending in years when the candidates do not talk about defense
spending than in years when this issue is discussed with great frequency.
Even if voters are merely sampling off the “top of their heads” (Zaller 1992)
we would expect there to be more considerations about defense spending
available for random sampling when candidates or the media talk about
defense spending a lot compared to a paucity of considerations about
defense spending at the top of one’s head when candidates do not mention
the issue much at all.
To show the relationship between what candidates and media talk
about during campaigns and what voters say about the candidates, I present
the figures below, grouped by the four issue areas under consideration.
Each figure plots the percent of voters’ comments about an issue on the
vertical axis and the percent of candidate or news discourse about that same
issue on the horizontal axis. A 45 degree-line would represent perfect
fidelity – if the points fall on (or near) the 45 degree-line, a one percent
increase in candidate or news discourse on a topic is reflected in a one
percent increase in voters’ comments on that topic. As I mentioned earlier,
we might not expect this relationship to be perfectly reflective, but it serves
as a benchmark from which we can evaluate the fidelity of these
relationships.5
As is evident in Figure 6.1, that kind of high fidelity relationship
rarely exists. In some cases, voters rarely talk about issues (the economy)
13
making it difficult for changes in candidate or news discourse to register
effects. In other cases, voters talk a lot about the topic (traits), which also
makes it difficult for candidate or news content to have an effect. Most of
the relationships, however, are positive, with the exception of trait and
character coverage in the news, which seems to negatively affect how much
voters talk about these things. Other relationships seem flat, indicating no
fidelity between candidate or news content and voters’ thoughts (the
economy, domestic, and foreign policy and the news, for example). The
conclusion for campaign-voter fidelity seems to be that voters are subtly
responsive to changes in campaign content, but their comments do not
necessarily reflect the overall composition of the campaign very well.
14
Figure 6.1: Fidelity between Advertisements, Speeches,
News, and Voter Evaluations, 1952-2000
Each row of the figure represents a different issue and each column
is a mode of campaign communication. There are several ways to evaluate
the trends in Figure 6.1. One way is to examine the tightness of the fit of
the data, in other words, how closely are the data points scattered about the
regression line? It is clear that the economy is the issue for which the
relationship between campaign discourse and voters’ thoughts is the tightest
15
– regardless of whether the information comes from ads, speeches, or news.
This suggests that the issue itself drives the relationship, not a specific
method of campaign content delivery. Domestic policy, on the other hand,
trends more parallel to the 45-degree line than the economy, but the data
have a lot of dispersion around the regression line. This suggests that the
general relationship moves in the right direction, but there is a lot variation
in the relationship between domestic policy campaign content and voters’
evaluations of candidates on this topic. The relationship between ad,
speech, and news content and voters’ evaluations is positive and
significantly different from zero for advertising on the economy, foreign
policy, and traits. Speech content is only significantly related to voters’
comments for domestic and foreign policy, and news content is not
significantly related to voters’ thoughts on any of these issues.
Surprisingly, news coverage seems to have very little if any
relationship to the kinds of things people can say about the candidates. As I
described earlier, one explanation for this may be the media’s constant need
to define stories in compelling and sensational ways. It is possible that
news stories, while interesting and worth watching, do not provide voters
with the kind of material they can recite when asked what kinds of things
they like or dislike about the candidates other than turning the generalities
of the news stories into comments about candidates’ traits, which voters
seem more than willing to do regardless of whether the candidates or the
media talk about character.
To further investigate the unique forces driving voters’ evaluations
of candidates, I analyze the open-ended responses and ascertain whether
ads, speeches, or news coverage is mainly affecting people’s thoughts about
candidates.6 The regression coefficients in Table 6.1 describe how voters
are influenced by a one percent increase in candidate discourse about the
issues in each column. In these models, the ads, speeches, and news
coverage compete along with incumbency status and the party of the
candidate to have a unique effect on voters’ thoughts about the candidates.
16
As the shaded cells in the table indicate, of the candidate discourse items,
advertising is the one that most often influences voters. Incumbency status
also has a small effect, presumably because people know more about
candidates who have been in office already and when an incumbent starts
talking about an issue, voters do not need to pay careful attention to fill in
the blanks about what they might be saying.
This may make it easier for people to mimic the composition of
incumbents’ campaigns. As the earlier figures demonstrate, news coverage
of campaigns does not seem to affect voters’ thoughts about the candidates
on any of these issues.
17
Table 6.1: Do Ads, Speeches, or News Coverage
Drive Voters’ Evaluations of Candidates?
Economy
Foreign
Domestic
Policy
Policy
Traits
Advertising
.17 (.05)
.16 (.07)
.04 (.16)
.55 (.13)
Speeches
-.08 (.05)
.10 (.09)
.44 (.20)
-.04 (.21)
News
.03 (.08)
-.04 (.09)
.10 (.28)
-.30 (.17)
Incumbency
.03 (.01)
.05 (.02)
-.02 (.04)
-.06 (.04)
Republican
-.00 (.00)
.04 (.02)
.06 (.05)
-.00 (.02)
Constant
.01 (.01)
.00 (.02)
-.06 (.06)
.42 (.08)
R2
.57
.48
.32
.45
Coverage
There are 25 cases for each column because Goldwater speeches were not available. Cell
entries are OLS regression coefficients with standard errors in parentheses. Standard errors
are clustered for campaign year. Campaign content is measured in percent of appeals in
ads, speeches, or news coverage about each issue in the columns. The content analysis is
done for each candidate in every year from 1952-2000. Voters’ comments are from the
ANES open ended likes-dislikes questions. Shaded cells indicate relationships significantly
different from zero at a .05 level of confidence.
To detail this relationship with a bit more nuance, I am going to
focus solely on campaign advertising, since that is the method of campaign
communication with the strongest relationship to voters’ evaluations. The
theory of campaigns suggests that clarifying and insurgent candidates will
be sending voters’ different kinds of signals during campaigns, thus I
examine the relationship between economic content in advertising and
voters’ comments separately for clarifying and insurgent candidates. If the
individual-level mechanisms derived from the theory of campaigns exist,
18
voters will have more to say about the relationship between clarifying
candidates and the nation’s economy as compared to the same relationship
for insurgent candidates because clarifying candidates should talk about the
economy more than anything else and more than their opponents. The
analyses in chapter 4 confirmed that most clarifying candidates behave this
way. The question now is whether this practice has specific micro-level
effects on voters.
In Figure 6.2, I present graphs showing the fidelity between
campaign rhetoric and voters’ comments, but this time, the data are broken
out by campaign type – clarifying or insurgent. The top row contrasts
clarifying candidates from insurgents and the bottom row separates
clarifying candidates who actually talked about the economy from those
clarifying candidates who did not talk about the economy. This latter
comparison is an attempt to test whether it is enough for candidates merely
to be in the incumbent party during a healthy period of economic growth, or
whether these economically advantaged candidates actually also have to
talk about the state of the nation’s economy in their campaigns. As with the
analysis of the errors in basic forecasting models, these data show that
talking about an economic advantage is helpful to candidates.
19
Figure 6.2: Fidelity of Economic Content in Advertising and
Voter Evaluations for Different Types of Candidates, 1952-2000
The simple linear relationship between economic content in
campaign advertising content and in voters’ evaluations shows that the
relationship is tightest for clarifying candidates. For every one percent
increase in economic advertising, the composition of voters’ evaluations
contains nearly two-tenths of a percent more economic content. Changes in
clarifying candidates’ level of advertising on the economy explain nearly
half the variation in economic content of voters’ comments. In contrast, the
changes in voters’ evaluations of insurgent candidates is closer to a onetenth of a percent increase for every one percent increase in advertising
content about the economy. More telling, however, is the fact that only 22
percent of the variation in voters’ comments is explained by the changes in
advertising content for insurgent candidates. People’s thoughts about the
20
nation’s economy and the candidates better reflect the composition and
dynamics of the clarifying candidates’ campaigns, as the theory predicts.
To establish whether actually talking about the economy more than
anything else in the campaign and more than your opponent pays off for
candidates in terms of the kinds of things voters say about them, in the
bottom row of Figure 6.2, I compare the clarifying candidates who follow
the theory’s recommendations to those who do not. Talking about the
economy in campaign advertising (and by definition being advantaged by
economic conditions) results in twice as big an increase in voters’
responsiveness to candidate discourse about the economy when evaluating
candidates. This is a critical point – the clarifying candidate’s best strategy
according to the theory is to reduce voters’ uncertainty about the connection
between economic conditions in the country and his candidacy. The theory
posits that if the clarifying candidate does this, he has the best chance of
winning the election, since the economy is always important to voters and
most people prefer prosperity to decline. The fact that people’s comments
about clarifying candidates who talk about the economy tightly reflect the
economic discourse of the campaign is strong evidence that talking about
the economy when it advantages your campaign has payoffs beyond merely
being an incumbent in a good economy.
To put this simple bivariate relationship to a stronger test, in Table
6.2 I present the results of the same model I discussed in Table 6.1 – does
advertising drive voters’ evaluations with greater weight for clarifying
candidates who talk about the economy when incumbency, party, and other
measures of campaign discourse are allowed to have an influence on
comments?
21
Table 6.2: The Effects of Campaign Content on Voters
Evaluations of Candidates for Economic Mentions
Clarifying
Insurgent
Advertising
.19 (.07)
.06 (.05)
Speeches
-.09 (.08)
.08 (.11)
News Coverage
.03 (.07)
-.02 (.20)
Incumbency
.03 (.02)
.07 (.01)
Republican
-.00 (.01)
-.00 (.02)
Constant
.00 (.02)
.02 (.03)
R2
.70
.73
N
13
12
Cell entries are OLS regression coefficients with standard errors in parentheses. Campaign
content is measured in percent of appeals in ads, speeches, or news coverage about the
economy. The content analysis is done for each candidate in every year from 1952-2000.
Voters’ comments are from the ANES open-ended likes-dislikes questions. Shaded cells
indicate relationships significantly different from zero at a .05 level of confidence.
Once more, economic mentions in advertisements affect the
composition of voters’ thoughts about clarifying candidates, but not about
insurgent candidates. In fact, candidates have to increase economic content
in their advertisements by five percentage points to get a one percentage
point increase in economic mentions from voters, but insurgent candidates
have to increase economic content by 16 percentage points to get the same
increase from voters if the relationship exists at all.7
22
Table 6.3: Advertising Effectiveness on Economic
Evaluations for Clarifying Candidates
Clarifying
Candidates
Advertising
.26 (.04)
Advertising & No Focus on Economy
-.25 (.08)
Focus on Economy
.08 (.02)
Incumbency
.05 (.01)
Constant
-.07 (.02)
R2
.85
N
13
Further support for the notion that candidates need to talk about the
economy as well as be advantaged by it comes from evidence in Table 6.3.
This table presents the results of advertising effectiveness on clarifying
candidates who focused on the economy more than anything else, compared
to those who did not behave this way. Since the number of cases gets quite
small when we break the clarifying candidates up into these two sets, I
estimate the effects using a single interactive model and reduce the model to
its most parsimonious form.8 The results indicate that there is a significant
difference between the effectiveness of advertising on the economy for
these two groups of candidates. Clarifying candidates who focus on the
economy increase voters’ economic considerations of them at a ratio of 4:1.
That is, the percent of economic mentions among voters commentary about
the candidate goes up 1 percent for every 4 percent increase in the
candidates’ economic discourse. For the clarifying candidates who did not
focus on the economy more than anything else in their campaigns, the
relationship between ads and evaluations is nearly non-existent. The ratio
23
of advertising mentions to voters’ evaluations for clarifying candidates who
do not focus on the economy is 100:1.
This is an important difference because it illustrates a critical feature
of the American presidential election landscape: it is not enough to be the
candidate who is advantaged by good economic times, you have to talk
about those good times and remind voters of your role in fostering the
period of economic growth. Even on national conditions as fundamental as
the state of the nation’s economy, the behavior of candidates can matter to
voter evaluations. These individual level findings corroborate the candidate
and aggregate-level analyses in illustrating for a third time the importance
of campaign rhetoric about the economy for those candidates who are
advantaged by current conditions.
Interpreting coefficients across models often comes with furtive
complications. To illustrate the importance of being advantaged by the
economy and also talking about the economy more than anything else, I
calculated the predicted levels of economic considerations given by voters
broken out by candidate type. The same pattern emerges – the data show
that voters will make more economic comments about the clarifying
candidates, compared to insurgent candidates as the proportion of economic
content in ads and speeches increases. Further, people are responsive to
campaign discourse about the economy for clarifying candidates who make
this the focus of their campaigns, compared to those clarifying candidates
who do not. These results are presented in Figure 6.3. The last graph in
Figure 6.3 shows the relationship between advertising about the economy
and voters’ evaluations of insurgent candidates – separately for those
insurgents who did not talk mainly about the economy and those who did.
24
Figure 6.3: Economic Content in Advertising and
Predictions on Mentioning the Economy
Once more, the theory’s predictions are confirmed. The data show
large differences in predicted levels of economic evaluations between
clarifying candidates who make the economy a priority and those who do
not. But the same difference is not as large for insurgent candidates. This
is yet another test of whether merely talking about the economy (and not
being advantaged by it) is enough to influence voters’ evaluations or
whether the combination of good economic times and focusing voters’
minds on the connection to the clarifying party are both needed. It appears
again, both are needed. It is not enough to be the incumbent in a period of
growth or a challenger in a period of decline – candidates have to make that
connection between their party and the economic times in order for voters’
evaluations to be sensitive on this topic.
25
CAMPAIGN EFFECTS: HOW MUCH DO VOTERS KNOW ABOUT
CANDIDATES’ ISSUE POSITIONS AND DOES THE CAMPAIGN MATTER?
Voters’ comments about the clarifying candidates and the economy
are driven by campaign content, specifically, advertising content. Even for
some other issues, the data demonstrate a relationship between what
candidates say and do in their campaigns and what voters’ think of them
when asked to talk freely about the candidates. Having only 26 candidateelection pairs, however, limits the persuasiveness of this set of findings. In
an attempt to show that the theory’s predictions are robust, I turn now to a
completely different method of assessing whether candidate behavior
matters to voters in the ways the theory of campaigns suggests. The next
several pieces of evidence will use the ANES data, but not the open-ended
questions. Instead, I turn to closed-ended questions about specific policies.
The ANES introduced a regular battery of issue questions in 1972 to
tap in to citizens’ perceptions of candidates’ positions on issues.
Respondents are asked to place themselves and if they can do that, they are
asked to place each of the presidential candidates on the same issues.9
Several issues have been asked repeatedly over the last 34 years and some
of them coincide with the kinds of things candidates were talking about or
getting news coverage of in their campaigns, they are: whether the
government ought to be responsible for making sure every American who
wants a job can find one; whether we should cooperate with Russia;
whether we are spending the right amount of money of national defense;
whether the government ought to increase spending and services or cut
both; and whether women should enter the workforce or stay home. The
general question text, which changes for each issue and for whether the
respondent is placing themselves or the candidates, reads:
Some people feel that the government in Washington
should see to it that every person has a job and a good
standard of living. Suppose these people are at one end of
a scale, at point 1. Others think the government should
26
just let each person get ahead on his/their own. Suppose
these people are at the other end, at point 7. And, of
course, some other people have opinions somewhere in
between, at pints 2,3,4,5 or 6. Where would you place
yourself on this scale, or haven't you thought much about
this?10
Additionally, the ANES asks respondents to place themselves and
candidates on an ideology scale that ranges from very liberal to very
conservative.11
These issue questions give us great leverage on whether voters know
enough to place the candidates on the issues, and if so, whether those issues
are important in evaluating candidates. The first thing that comes to mind,
however, is a simple question: How good are voters at placing the
candidates on these issues and where do they place the candidates relative to
themselves?
Below, I present respondents’ placements of themselves and the
candidates on these scales over the course presidential campaigns from
1972 to 2000. The square marks represent the average of respondents’
placements of themselves on the relevant issue, the circles denote the
respondents’ average placements of Republican presidential candidates and
the triangles are respondents’ placements, on average, of Democrats. The
lines extending away from the mean estimates represent a single standard
deviation in each direction.
27
Figure 6.4: Average Placements of Respondents and
Presidential Candidates on Issue Scales, 1972-2000
28
A few things are at once obvious from Figure 6.4. Primarily, on
average, people do a pretty good job of knowing that Republicans are more
conservative than Democrats, even on complicated issues like equal rights
for women or cooperation with Russia. Only in one case, equal rights for
women, do respondents consider themselves more liberal on average than
both political parties, and this estimate moves considerably over the course
of campaigns, such that by Election Day, people place Democrats farther to
the left and themselves closer to the middle. The second pattern that
emerges from this figure, then, is that, on average, people place themselves
between the two political parties on most issues, and this pattern does not
change as the campaign progresses. The third interesting point to learn
from Figure 6.4 is that there is a wide range of placements for people on all
issues. In no case are people’s placements clustered tightly around the
mean, instead, the lines extending out from the points in the figure suggest
that people vary significantly in their placements of themselves and the
candidates on these issues. The fourth pattern that is evident from this
picture is that there is very little movement on these placements between
Labor Day and Election Day, but this does not mean that the movement is
independent of what candidates are doing in their campaigns.
Another interesting characteristic of these issue-placement scales is
the fact that a lot of people cannot make a placement about themselves or
the candidates on these issues. I assume that means they know so little
about the candidates or the issue that they cannot even guess at what
position on the scale might be reasonable. If the campaigns are reducing
uncertainty of candidates’ issue positions, especially clarifying candidates’
positions on economic issues, then over the course of the campaign more
people should be able to offer placements of the candidates on these scales.
By dividing the campaign into 2 week intervals, I can measure the changes
in how many people can place candidates on the issue during the campaign.
29
Figure 6.5: Changes in Mean Levels of Uncertainty
from September to Election Day
Each bar in the figure represents the average change in respondents’ refusals to place the
clarifying candidate (indicating level of uncertainty) on the related issue from the first two
weeks of September to the two weeks before Election Day compared to their refusals to do
the same for the insurgent candidate. Decreases in uncertainty indicate learning over the
period of the campaign.
In Figure 6.5 I show changes in the average level of non-placement
on six issues from the first two weeks of September of the election year to
the last two weeks before Election day.
The data show that clarifying candidates are in fact able to reduce
the overall level of uncertainty about their positions on economic issues,
like whether the government should guarantee jobs to everyone who wants
to work. Insurgent candidates are not able to reduce the general level of
uncertainty on this issue over the course of the campaign. Clarifying
candidates, however, are also able to reduce people’s levels of uncertainty
about their positions on the left-right ideological placement scale and also
on the level of cooperation they desire with Russia. Insurgent candidates,
30
however, are not able to reduce people’s levels of uncertainty about any of
the issues over the course of the campaign.
Are these differences meaningful? Using all the issue-candidateyear data, I test whether people’s abilities to place the clarifying candidates
on the jobs question is systematically different from their ability to place the
insurgent candidate – and different from either candidate on any other issue.
In general, people are better able to place the clarifying candidates on issues
at the beginning of the campaign by nearly 2 percentage points, and there
does not seem to be anything systematically special about the jobs issue
compared to others. Conditioning on the jobs issue, however, people are
better able to place the clarifying candidate than the insurgent candidate by
one percentage point. The data in the first column of Table 6.4 suggest that
the benefit of being a clarifying candidate is a reduction in voter uncertainty
about your position on guaranteed jobs of about 3.5 points.
Over the course of the campaign, as the clarifying candidates talk
about the economy and help voters connect their party to prosperity, the
data show that people learn more about clarifying candidates on the
economy than they do about either candidate on other issues. The second
column of Table 6.4 presents the changes in average non-placement levels
from September to November. In terms of campaign learning, it is not
enough to merely be the clarifying candidate, the direct effects on this
indicator are indistinguishable from zero. Again, however, conditional on
being the clarifying candidate, people are better able to make placements on
the jobs question. The average reduction in uncertainty for clarifying
candidates on the economy over the course of the campaign is almost 2
percentage points. These findings provide strong support for the theory of
campaigns. Clarifying candidates experience greater reductions in
uncertainty than insurgent candidates on economic issues as they are talking
about these issues over the course of the campaign.
31
Table 6.4: Reductions in Mean Levels of Issue
Uncertainty for Eight Issues
Mean Nonplacement
Change in Mean
Non-placement
Clarifying
-.018 (.005)
-.002 (.003)
Jobs
-.024 (.015)
-.007 (.003)
Clarifying*Jobs
.010 (.005)
-.008 (.003)
Constant
.351 (.041)
.015 (.004)
N
94
94
R2
.66
.42
Difference in Mean
-.031
-.017
Placements for
Candidate Types
Data are respondents’ abilities to place each candidate on eight different issues: guaranteed
jobs, defense spending, cooperation with Russia, services v. spending, the role of women,
aid to minorities or black, government supplied health insurance, and the ideological leftright scale in every year from 1972-2000. Not all issues are asked in each year. The
dependent variable is coded 1 if the respondent cannot place the candidate on the issue,
thus negative coefficients indicate reductions in uncertainty levels. Cell entries are
regression coefficients generated by regressing whether the candidate in question was the
clarifying candidate, whether the issue is jobs, the interaction of these two variables, and
indicators for each year (coefficients suppressed), except 1972. Changes over the course of
the campaign are defined as the difference between the first two weeks of September and
the last two weeks of November. Standard errors are clustered on issue.
To be clearer, it is not the case that clarifying candidates exhibit greater
reductions in uncertainty over the course of the campaign in general, these
effects are particular to the jobs issue.
Average rates of non-placement are indicative of people’s levels of
uncertainty about candidates over the course of the campaign, but one of the
advantages of individual level data is the ability to allow different people to
have different reactions to conditions. For example, people who pay more
attention to the campaign in the media may show greater reductions in
uncertainty about issues than those who pay little attention. Education may
32
play a role in how much people learn about the candidates. To control for
these various factors, I model issue non-placement separately for the issues
mentioned above by including controls for attention to the campaign in the
media, time of interview, the interaction of these two things, which is
essentially the test of campaign learning, and various demographic variables
(age, race, gender, and education). The average predicted levels of nonplacement, broken out by particular issues confirm the story told by the
mean levels of non-placement above. There does not appear to be anything
special about just being the clarifying candidate, one of the largest reduction
in uncertainty over the course of the campaign about any issue comes from
placing clarifying candidates on the jobs issue (similar learning is
demonstrated for cooperation with the Russians). This effect, however, is
seen clearly among those people paying high levels of attention to the
campaign in the media.
The comparison between the people who pay no attention to the
campaign in the media and those who pay a lot of attention is another test of
the theory. For those who pay no attention to the campaign in the media,
the data show average reductions in the change in clarifying candidate
uncertainty from September to November for jobs, services, and
cooperation with Russia. Insurgent candidates show reductions in the
average change in uncertainty on defense spending and cooperation with
Russia. The other issues show increased average levels of non-placement
over the course of the campaign for both candidates. This is essentially the
effect of the passage of time absent any media campaign coverage.
The second panel of the figure, for people paying high levels of
attention to the campaign in the media, shows reductions in uncertainty, on
average, for all issues and all candidates. This is evidence of campaign
learning. From September to November, people who pay attention to the
campaign in the news are better able, on average, to place candidates on all
issues at the end of the campaign. This is not true for people who do not
encounter the campaign in the news.
33
Figure 6.6: Changes in Average Predicted Probability of Nonplacement over the Course of the Campaign by variations in Level of
Attention to Campaign in Media
Each bar in the figure represents the average change in respondents’ predicted probabilities
of non-placement for the clarifying candidate on the related issue from the first two weeks
of September to the two weeks before Election Day compared to the average predicted
probabilities of doing the same for the insurgent candidate. Decreases in uncertainty
indicate learning over the period of the campaign. Predicted probabilities of nonplacement are generated from 14 probit analyses with willingness to place each candidate
on each issue as the dependent variables. Explanatory variables include level of attention
to the campaign in the media, date of interview, demographic controls, and indicators for
campaign year. Standard errors are robust and clustered on campaign year.
The largest reduction in uncertainty over the course of the campaign
is for the placement of clarifying candidates on the economy and on
cooperation with Russia. The reduction in uncertainty for clarifying
candidates is more than twice as large as the reduction for insurgent
candidates. People are, on average, approximately five percent more likely
to place the clarifying candidate on the issue of guaranteed jobs at the end
of the campaign than they are at the beginning of the ca,paign. For
34
insurgent candidates, the average increase in likelihood of placement on this
issue is only about one and half percent, on average. This is another piece
of evidence supporting the theory. Even controlling for demographics,
year-specific effects, and media attention, the effects for clarifying
candidates on the jobs issue are distinguishable from the others.12
It is not enough to be an incumbent party nominee in a good
economy or a challenger in a time of economic decline. Candidates who
make this connection the most important message of their campaign
discourse affect what voters think and know about them and ultimately,
they win elections. I demonstrated this pattern at the candidate level in
Chapter 5; and, as a test for mechanisms driving the candidate level pattern,
I show effects of candidate discourse at the individual level in this Chapter.
The economy matters, but how candidates react to it matters, too. The
Clinton campaign in 1992 seemed to understand this. In Bill Clinton’s 1992
campaign headquarters a sign hung that read, “It’s the economy, stupid”.
This adage turns out to be more right than maybe even Clinton and his
strategist James Carville imagined. The economy matters because it
determines which candidates can talk about which issues in campaigns; it is
always an important consideration to voters; and, it is critical that
candidates advantaged by the economic situation talk about it during the
campaign more than they talk about anything else and more than their
opponents, even if they are incumbents. Clinton and Carville probably
knew this, as the sign was inward facing, to remind people who were
talking to the press, writing ad copy, or drafting speeches, and to remind the
candidate himself that the campaign’s message was simple: remind voters
that times were not as good as they seemed at the hands of George H.W.
Bush.
INSURGENT CANDIDATES:
CHANGING THE DEBATE BY INCREASING THE IMPORTANCE OF ISSUES
35
The previous section detailed the connections between what
candidates talk about on the campaign trail and what voters say about
candidates during the same time period. Citizens make more comments
about the economy when asked to describe the candidates, and they are
better able to place clarifying candidates on economic issue scales if the
clarifying candidate talks mainly about the economy in ads or speeches.
But, sometimes clarifying candidates lose elections. These losses are not
merely caused by errors on the part of clarifying candidates (a few win even
though they do not talk mainly about the economy). The other important
element of electioneering is, of course, what the insurgent candidate is
doing along side the clarifying candidate.
Recall that the insurgent candidate needs to focus the election off of
the economy and onto some other issue. Specifically, an issue on which he
is closer to the average American than his opponent, and an issue on which
his opponent is committed to an unpopular position. It also helps if public
opinion on the issue is lopsided, that is, that opinion on the issue is not split
between the two candidate’s positions. It does little good to increase the
importance of an issue on which half the people will favor the opponent’s
position. The kinds of things insurgent candidates choose to talk about as
their insurgent issues vary across year and party, but typically they fall in to
two broad categories: foreign and domestic policy.
In the early fifties, Stevenson tried to refocus the elections on to
Eisenhower’s mishandling of the Korean War. In 1960, Kennedy painted a
picture of both foreign and domestic decline as he compared the US to the
Soviets. Mondale wanted to portray Reagan in 1984 as reckless and
aggressive in his possible reactions to the Soviets. And in 2000, George
Bush offered leadership for solving domestic problems that was bi-partisan,
kinder, compassionate – yet conservative and market driven. A few
insurgent candidates centered their campaigns around economic issues
(McGovern, Dole) and thereby, according to the theory, helped their
opponents win the election.
36
How successful are the insurgent candidates at increasing the
importance of their chosen insurgent issues? Using the same NES data,
collected over the period from 1972 to 2000, I create a measure for
“insurgent issue” that varies by election year. Thus, the insurgent issue
questions I use in the remainder of the book are constructed such that each
year relates to that specific candidate’s insurgent issue. So, the observations
on the insurgent issue for 1980, for example, always refers to foreign policy
– specifically, our relations with the Soviet Union and the willingness of the
U.S. to go to war. For 2000, they always deal with domestic policy –
education, social security, and healthcare. To organize the analyses that
follow in a tractable manner, I define three different classes of insurgent
candidates, which escalate in accordance with how many of the insurgent
candidate prescriptions the candidate follows. At the high end of the
adherence scale are those insurgent candidates who meet all the insurgent
candidate criteria laid out by the theory. That is, they talk about something
other than the economy; they are closer to most voters than their opponent
on this issue; and, their opponent is committed to his unpopular position on
this issue. I defined these candidates as Kennedy, Nixon (1960), Carter
(1976), and George W. Bush (2000). For the remainder of the book, I
classify these candidates as fully meeting the theory’s criteria. The
remainder of the candidates I classify as those who violate at least one of
the theory’s prescriptions. Because decisions about whether the candidates
satisfy the theory’s corollaries may seem subjective to some (it is a matter
of some judgment to determine whether Eisenhower actually was “soft on
communism” or whether Reagan truly was “a hawk” who preferred military
solutions to diplomatic ones), I relax the requirements and create a second
class of candidates who satisfy the first and arguably most important criteria
– they talk about something other than the economy. These insurgent
candidates include Stevenson, Kennedy, Goldwater, Carter in 1976, George
W. Bush in 2000, George H.W. Bush in 1992, Dukakis, Mondale, and
37
Carter in 1980. I will refer to this class of candidates as the “not economy”
group in subsequent tests.
If the theory is a good explanation of how insurgent candidates win
elections, the candidates who meet the theory’s criteria will show the
greatest effects of priming their insurgent issues during the campaign. Each
year, the NES asks respondents to name the nation’s most important
problem. I use this measure as a first test for whether the various classes of
insurgent candidates are successful at increasing the importance of their
insurgent issues over the course of the campaign.
Figure 6.7 plots the mean percent of respondents who named the
insurgent issue as the nation’s most important problem by date of interview
(broken into roughly two week segments) for each of the classes of
insurgent candidates. Care should be taken not to make too much of the
time trend in these figures, since the NES does not employ a sampling
design that generates representativeness at the daily, weekly, or monthly
level. There is, however, little reason to believe that the people who
participate in the face-to-face interviews in the first weeks of the period are
systematically different from those who participate in the last few weeks.
Thus, any interesting trends based on time of interview should be taken as
suggestive of a larger pattern with this caveat in mind.
38
Figure 6.7: Insurgent Issue as Nation’s
Most Important Problem during Campaign
The first thing to notice about the data in Figure 6.7 is that early in
the campaign period, the mean levels of incidence are clustered together,
around 23 to 29 percent of respondents name the insurgent issue as the
nation’s most important problem. Over time, however, a compelling
blueprint emerges. Insurgent candidate who violate the theory’s
prescriptions are not able to increase the importance of their issue in voters’
minds. The line tracing their proportion over the campaign is nearly flat –
and remains the lowest of the three lines. In direct contrast, the insurgent
candidates who fully meet the theory’s criteria show a marked jump in the
proportion of respondents who think their issue is the most important in the
country. Nearly half of those asked (47 percent) in the last two weeks of
September think the insurgent candidate’s issue is the most important issue
in the nation. This is an 18 point increase over the percentage of people
who think this in the first two weeks of September; and, a 28 point increase
over the issues of candidates who violate the theory’s criteria (for the same
39
period of interviews). As evidence that this pattern is not being driven by
sampling error within the time periods, the same differences exist between
the first two weeks of September and the first two weeks of October.13
Further support for the robustness of these trends comes from the data for
insurgent candidates who talk about something other than the economy.
The percentage of respondents who think their insurgent issue is the most
important reaches a high of 34 in the last half of September and drops
slightly to 31 during the first weeks of October. These numbers are
between the numbers for insurgent candidates who get it wrong and those
who get it right. In other words, getting it half-right (talking about
something other than the economy, but not something on which you are
uniquely advantaged vis-à-vis your opponent) lands a candidate half way
between the individual level effects of those who violate and those who
fully meet the theory’s criteria. Finally, all the classes of insurgent
candidates show declines in the percentage of people who name the
insurgent issue as the nation’s most important in the last weeks of the
campaign. Only the insurgent candidates who get it all right end up with an
overall increase (6 points) in the importance of the insurgent issue. Like the
volatility and reversion Gelman and King (1994) illustrated with public
opinion polls over the course of the campaign, these individual level data
show movement during the campaign period with reversion to the starting
position.
The evidence from Figure 6.7 suggests that insurgent candidates
who behave as the theory suggests are successful at priming their issue in
the minds of voters. But, the question remains: Does the importance of this
issue matter to a vote choice on Election Day? In order to learn whether
increasing the importance of the insurgent issue helps these insurgent
candidates at the polls, I present the results in Figure 6.8. The vertical bars
represent the predicted probability of vote for the clarifying candidate (the
predicted winner) as a function of whether respondents named the insurgent
issue as the nation’s most important problem.14 I present these comparisons
40
for each of the three classes of insurgent candidates and, across different
levels of respondents’ political information.
41
Figure 6.8: Vote for Clarifying Candidate by Insurgent Issue as
Most Important Problem during the Campaign
Low Information Voters
High Information Voters
The first two bars of Figure 6.8 show that insurgent candidates who
meet the criteria for a good insurgent issue are able to keep support for the
clarifying candidate below 50 percent if they are successful at priming their
issue. If they are not successful at priming, support for the clarifying
candidate jumps by nearly 10 percentage points, and crosses the 50 percent
mark. This relationship is more pronounced for low-information voters,
who seem especially susceptible to the campaign. If the insurgent
candidates who fully meet the criteria can prime their issue, lowinformation voters are more likely to vote for them by 14 percentage points.
Taken in concert with the results presented in the previous figure, a story of
insurgent campaign effects takes shape. Insurgent candidates who meet the
theory’s criteria are successful at priming the insurgent issue’s importance
over the course of the campaign – and, voters who believe the insurgent
issue is the nation’s most important problem are more likely to vote for the
42
insurgent candidate, especially if they are low political information voters.
This means that as the importance of the insurgent issue increases over the
course of the campaign, support for the insurgent candidate goes up. This
campaign effects story is particular to insurgent candidates who choose
wisely as they pick an insurgent issue and who are successful at priming its
importance in the minds of voters. Insurgent candidates who talk about
something other than the economy, but do not pay heed to their advantage
over their opponent on this issue, are again between the level of effects for
those candidates who totally get the issue wrong and those who fully get it
right. Support for the clarifying candidate among these elections remains
above 50 percent in the whole sample, even if the insurgent candidate
primes his issue. Among low political information voters, however, support
for the clarifying candidate drops to 46 percent if the insurgent candidate is
successful at priming his issue. Once again, low-information voters show
themselves to be most susceptible to campaign efforts. The campaign
effects for these candidates who almost get it right are also much smaller, in
the range of 2 to 5 points, in the whole sample and among low-information
voters, respectively.
The use of respondent’s self-report of the nation’s most important
problem to explain changes in vote share is not ideal as it is possible that the
campaign is influencing both of these things in real time, such that it may
not be possible to isolate the causal direction in this relationship. Maybe
ideas about the nation’s most important problem drive vote, but maybe vote
preference drives ideas about the nation’s most important problem. This
problem is somewhat mitigated by the fact that the effects of insurgent
campaigns seem to stack neatly in accordance with the level of the
insurgent candidates adherence to the theory’s prescriptions. The worstcase interpretation suggests that insurgent candidates who behave as
predicted increase votes and also prime their issue; instead of prime their
issue and then increase votes. The models in this section all include
indicator variables for election year, which is critical in tracing the
43
possibility that there is just something particular about a certain year that
leads people to vote for the insurgent candidate and think his insurgent
issue is the most important in the nation. In the presence of these controls,
the results in Figure 6.8 suggest that candidate behavior in campaigns is
important to electoral outcomes. Again, it is not enough to talk about an
issue other than the economy if you are the insurgent candidate. Insurgent
candidates have to choose insurgent issues on which they benefit from onesided public opinion and on which their opponents are committed to
unpopular positions.
In addition to making the insurgent issue more important to voters
over the course of the campaign, insurgent candidates can also gain
advantages by persuading people that the relative distance between the
voter’s position on the insurgent issue and the candidate’s position is closer
for insurgent candidates than it is for clarifying candidates. In other words,
they can attempt to demonstrate or even change their advantage on the
issue. This is a test of the insurgent candidate’s ability to persuade voters.
To elucidate whether insurgent candidates can do this over an 8-week
campaign period, I return to the issue position scales I used in the last
section. For each issue, as the theory illustrates, I take the difference
between the candidate’s position and the voter’s position for each issue and
square the value such that all distance measures are expressed in positive
numbers.15 Greater values indicate there is a larger distance between the
voter and the candidate. I do this for both the clarifying and insurgent
candidates on each issue. To answer the question, “to which candidate is
the voter closest?” I simply subtract the clarifying candidate’s squared
distance from the insurgent candidate’s squared distance. This gives me a
relative measure of proximity for this issue. Negative values of this
difference in distances indicate the voter is closer to the insurgent candidate.
Positive numbers indicate the opposite. For example, if the voter placed
both of the candidates and him or herself at exactly the same point on the
issue scale (regardless of where that was on the scale), the squared
44
differences for each candidate would be zero and the difference in distances
would be zero. If a voter placed both candidates at positive 6 and him or
herself at 1, each squared distance would be 25 and again, the difference in
distances would be zero. On the other hand, if a voter placed the clarifying
candidate at 6, the insurgent at 1, and him or herself at 4, the squared
distance to the clarifying candidate would be 4 (6-4=2 and 22 = 4) and the
squared distance to the insurgent candidate would be 9. Thus the difference
in distance would be nine minus four – five. Since this is a positive
number, we know that the voter is closer to the clarifying candidate. The
difference in distance measure can range from -36 to +36. In truth, most
values fall near the zero mark, but are not actually zero. Certainly, none are
greater than +/-10.
If insurgent candidates are successful, they not only prime the
importance of their insurgent issue, but they choose an issue on which they
are closer to voters than the clarifying candidate – or, they try to convince
voters of this over the course of the campaign. Figure 6.9 presents the
changes in average values of the difference in distance measure for the
insurgent issues over the course of the campaign. Again, I present these
data for the three different classes of insurgent candidates and broken out in
the bottom two panels by voters levels of political information. Of primary
importance is the fact that the classes of insurgent candidate again stack in
the correct order. The insurgent candidates who violate the typology
consistently show small advantages for the clarifying candidate (positive
values on the Y axis) and very little change over the course of the
campaign. Insurgents who talk about something other than the economy,
but pay no heed to their advantage on the issue fall between the other two
types of insurgents. Not surprisingly, the trend for these candidates hovers
around the zero point as the campaign goes on. This illustrates the
possibility that the clarifying candidate is demonstrating to voters that he
holds exactly the same position as the insurgent candidate – which is why
the theory’s second corollary is so important: the clarifying candidate
45
should be committed to his unpopular position. Remember Eisenhower
pointing out that he in fact was not soft on communism by referencing the
Berlin Air Lift and declaring, “I shall go to Korea, I shall end the War!”
Similarly, Reagan argued against the claim that he was reckless and prone
to start a nuclear war by holding up as evidence the SALT II Nuclear NonProliferation Treaty that he co-authored and signed with the Soviets only
months before the election started. Even the starting point for this middle
class of insurgent candidates suggests these candidates did not pick issues
on which they were naturally and clearly advantaged. Compare their
starting points in early September to the starting values for the insurgents
who fully adhere to all of the theory’s criteria.
46
Figure 6.9: Differences in Squared Issue Distances over Campaign
47
On average, insurgents who get it right finish the campaign 1 unit closer to
voters on the insurgent issue compared to the clarifying candidate. Those
insurgent candidates who do not choose a winning issue but talk about
something other than the economy end the campaign with voters believing
that the clarifying and the insurgent candidate hold the same position on the
insurgent issue (zero difference). Among low-information voters, the trends
once again are magnified. During the height of the campaign, the lowinformation voters report differences as large as -4 points on the difference
in distance measure. This large insurgent-advantage goes away over time,
however, even for the insurgents who meet all the theory’s criteria. Among
high information voters, insurgent candidates end up with only a small
advantage, but an advantage none-the-less.
The differences in the dynamics between low and high information
voters are also informative. Low information voters show considerable
volatility throughout the campaign period, becoming very close to the
insurgent candidate for those candidates who choose a winning issue but
then reverting back to a position of indifference between the two candidates
on this issue by the end of the campaign. The high information voters begin
the campaign closer to the insurgent candidate on his winning issue, but
over the course of the campaign, the high information voters steadily
become indifferent between the two candidates on this issue, although the
insurgent candidate keeps a slight advantage among high information
voters, whereas this advantage is completely lost among low information
voters. It is not hard to imagine an explanation for these trends: clarifying
candidates who try to neutralize the insurgent issue are doing a good job of
convincing voters that the advantages of going with the insurgent candidate
on this issue are not great. Although these data are speculative, the fact that
this same reversion to indifference does not happen for insurgent candidates
who violate all of the theory’s mandates suggests that there is something
particular about the non-economic insurgent issues that candidates choose,
48
which leads voters to erode their support for the insurgent candidates on
these issues.
Low-information voters may not be the only susceptible voters in
the population. Campaigns may also exhibit greater effects among selfdescribed independents, or those with no political party affiliation. Figure
6.10 presents the results for differences in distances broken out by party
type.
49
Figure 6.10: Differences in Squared Issue Distance during Campaign
by Respondent Partisanship
Of primary importance again is the fact that the lines, across all
categories of party identification, stack in the correct order. Among
respondents who identify with the clarifying candidate’s party (this can
change in a given election year, sometimes it is the Democrats and
sometimes it is the Republicans) the stacking is less clear, but mainly,
insurgent candidates that behave as expected fare better in voters’ minds
than those who violate the theory’s prescriptions. It is also reassuring that
the collection of lines for each type of party identification sort out in the
expected rank ordering: Clarifying partisans see themselves as always
closer to the clarifying candidate on the insurgent issue. Insurgent partisans
see themselves are always closer to the insurgent candidate on the insurgent
issue. And independents see themselves as somewhere in the middle. In
fact, independents end up in the last few weeks of the campaign rating
themselves as slightly closer to the clarifying candidate if the insurgent
50
candidate talks about something other than the economy, but, they end up
slightly closer to the insurgent candidate if these candidates also talk about
an issue on which they are closer to most voters and on which their
opponent is committed to an unpopular position. Independents also show a
familiar trend over the first few weeks of the campaign – they are most
affected by the candidates’ behaviors in the latter part of September, but
then slowly revert to ending points that are close to where they started. This
is the same trend that exists for low information voters. It seems that low
information voters and independents are the most easily influenced by the
campaign in its initial period, and the advantages they give to candidates
during this time dissipate as the campaign wears on. Again, it is difficult to
isolate the cause of this reversion. Campaigning from the opposition seems
like the most obvious explanation, especially since the probability of voting
for the insurgent candidate remains positive (despite the reversion) for
insurgents who choose winning issues, but becomes negative for those
insurgents who choose poorly. Other more complicated explanations are
imaginable, including the possibility that even low information and
independent voters are predisposed to support one or the other of the
candidates, and despite being swayed by short term information during the
campaign, these movable voters return to their original preferences on
election day. Fortunately, each of these explanations is interesting from a
campaign perspective. Unfortunately, one of them implies that campaigns
matter a great deal to voters (the first explanation), and the other implies
that they do not matter at all.
The observable range of the differences in distances on the insurgent
issue goes from +5 (clarifying partisans) to -7 (insurgent partisans). How
important are these relative distances in predicting the vote? To assess this,
I model vote for the clarifying candidate as a function of the difference in
distances on two issues – the insurgent issue and the clarifying issue (which
is most often an economic question about guaranteed jobs for those who
want to work). Controls in this model include gender, age, and party
51
identification.16 Results show that both issues have substantively and
statistically significant effects on the vote. The closer a voter is to the
clarifying candidate on either issue, the more likely they are to vote for that
candidate, and vice versa. These results are presented graphically, for
different levels of differences in Figure 6.11.
These results are presented for three different levels of differences
on the insurgent issue: no difference between the candidates, a small
difference of 4 units closer to the insurgent candidate than the clarifying
candidate, and larger difference of 25 units closer to the insurgent
candidate. For each level of difference, I calculate the probability of voting
for the clarifying candidate holding all the other variables at their means.
Along with displaying the expected and now familiar pattern of stacked
lines, the results also show the importance of choosing an insurgent issue on
which the candidate is closer to most voters than his opponent. When
voters see no difference between the two candidates (the difference in
distances is 0), the probability of voting for the clarifying candidate is
greater than 60 percent, regardless of whether the insurgent candidate chose
wisely when picking his insurgent issue. Being equidistant from both
candidates on the insurgent issue does not lead voters to cast ballots for the
insurgent candidate. Relative issue proximity matters to voters in
predictable ways.
52
Figure 6.11: Predicted Vote for Clarifying Candidate by Difference in
Distances and Insurgent Campaign Type
Results are predicted probabilities of voting for the clarifying candidate at different levels
of relative issue proximity. The categories of distance represent 0 units of difference (No
Difference), four units (Slightly Closer), and 25 units (Very Close). Predicted probabilities
are generated from 3 probit analyses using self-reported vote choice as the dependent
variable. Independent variables include demographic controls, respondent partisanship.
Standard Errors are robust and clustered on election year.
The insurgent candidate has to share his position on this issue with voters in
a way that his opponent cannot. Even when voters are just slightly closer to
the insurgent candidate on the insurgent issue (4 units of difference) the
probability of voting for the clarifying candidate drops slightly and nearly
crosses the 50 percent mark for those candidates who heed the theory’s
counsel. But at greater levels of difference (25 units), the benefits become
very clear. For insurgent candidates who choose issues on which they have
a relative advantage and on which their opponent cannot change his
position, a difference in distances of this magnitude results in less than a 10
percent chance of voting for the clarifying candidate. Figure 6.11 illustrates
53
the importance of issue position proximity to election outcomes. To
reiterate, this finding underscores the importance of the insurgent issue and
its characteristics. If there is no difference between the candidates on the
insurgent issue, the clarifying candidate will win the election handily. The
economic advantage is substantial and it has real benefits. If there are small
perceived differences on the insurgent issue that benefit the insurgent
candidate, and the clarifying candidate is committed to his unpopular
position, the insurgent candidate can win the election by a narrow margin.
And, finally, if there are large advantages for the insurgent candidate, he
will win elections.
The effects of the differences in distance on the insurgent issue vary
by levels of political information and by partisanship, just as the average
differences did. Figure 6.12 presents the effects for political information.
In this case, high political information voters show slightly greater rates of
change in the probability of voting for the clarifying candidate as their
proximity to the insurgent candidate changes. Unlike the previous data, this
evidence seems to suggest that the high information people are more
affected by the campaign. In truth, what these data show is that the high
information people know how to use the proximity measures in their
political calculations, while the low political information people are less
affected by these changes in relative distance to the candidates. Low
information voters may be more susceptible to campaign efforts at
persuasion, as demonstrated by their marked shifts on differences in
54
Figure 6.12: Predicted Vote for Clarifying Candidate by Difference in
Distances on Insurgent Issue and Levels of Political Information
Results are predicted probabilities of voting for the clarifying candidate at different levels
of relative issue proximity and two levels of political information or sophistication. The
categories of distance represent 0 units of difference (No Difference), four units (Slightly
Closer), and 25 units (Very Close). Predicted probabilities are generated from 6 probit
analyses using self-reported vote choice as the dependent variable. Independent variables
include demographic controls and respondent partisanship. Standard Errors are robust and
clustered on election year.
distances over the course of the campaign, but once the relative differences
have changed, they affect the low information voters less than they would
those who are more political aware. This is an interesting conundrum for
candidates – precisely those people they are able to persuade are the people
for whom issue proximity matters the least.
Figure 6.13 presents the effectiveness of relative issue proximity for
different partisan identifiers. Once again, insurgent candidates who behave
as the theory suggests do better than those who do not, across all categories
of partisanship. As with the data on average differences over time, these
results show the importance of the self-identified independents to election
55
outcomes and candidate strategies. If independent voters see no differences
on the insurgent issue between the candidates, they are likely to vote for the
clarifying candidate. Again, this is the power of context and the economic
advantage. If independent identifiers see slight differences between the
candidates, they approach a point on indecision, and large differences place
them squarely on the side of the insurgent candidate. In fact, in the face of
large issue differences favoring the insurgent candidate, nearly all voters
have a high probability of voting for the clarifying candidate, with the
exception of clarifying partisan identifiers in elections with insurgent
candidates who violate the theory’s predictions in some way. They remain
loyal to their clarifying candidate even as the advantage on the insurgent
issue turns to favor the insurgent candidate.
56
Figure 6.13: Predicted Vote for Clarifying Candidate by Differences in
Distances on Insurgent Issue and Respondent Partisanship
Results are predicted probabilities of voting for the clarifying candidate at different levels
of relative issue proximity. The categories of distance represent 0 units of difference (No
Difference), four units (Slightly Closer), and 25 units (Very Close). Predicted probabilities
are generated from 9 probit analyses using self-reported vote choice as the dependent
variable. Independent variables include demographic controls. Standard Errors are robust
and clustered on election year.
The behavior of candidates in campaigns does affect voters in
elections. Clarifying candidates who talk about the economy more than
anything else and more than their opponents can reduce voters’ levels of
uncertainty about economic issues like jobs and unemployment. Voters are
better able to associate the economy with clarifying candidates who follow
the theory’s mandates than they are with clarifying candidates who do not.
In further illustration of the importance of candidate behavior, candidates
can prime issues over the course of the campaign if they choose issues on
which they have a solid advantage (closer to most voters and the opponent
is constrained to an unpopular position). Voters increasingly think these
issues are important problems in the country and these assessments affect
57
vote choice in substantively important ways. All of this is particularly true
for voters with low levels of political information or sophistication and
independents. They are the most susceptible to campaign influences,
especially in the early weeks of the campaign. As time wears on, however,
the advantages gained are lost. To make matters worse, these susceptible
voters actually use the new information they receive from campaigns less
than their more informed and partisan counterparts when making a vote
decision.
The individual level data presented here, in some cases only
suggestive of trends in the right direction, nonetheless illustrate nicely the
mechanisms that drive the spatial model of campaign competition. Not
only do aggregate election forecasts and candidate-level tests establish the
theory’s explanatory and predictive capabilities, these individual-level data
corroborate the foundations on which the theory is built. Why is the
economy a strong predictor of election outcomes? Finally, there is an
individual level explanation for this regularity beyond pocketbook voting.
The economy matters because the candidate who benefits from it talks
about it a lot during the campaign and this makes voters more aware of the
condition and this candidate’s relationship to it. Why is it so difficult for a
challenger to defeat an economically advantaged candidate? Because
finding the right issue onto which the electoral agenda can be reset is
difficult and candidates are challenged by it. Those who find such an issue,
however, can make it more important to voters over the course of the
campaign and narrowly edge out the economically advantaged candidate.
Candidates and campaigns matter to outcomes despite the importance of the
nation’s economy.
58
59
Cited Works (Incomplete)
Abramowitz, Alan. 1988. "An Improved Model for Predicting Presidential
Election Outcomes." Political Science 21: 843-847.
Abramson, Paul R., John H. Aldrich, Phil Paolino, and David Rohde. 1992.
“'Sophisticated' Voting in the 1988 Presidential Primaries.” American
Political Science Review. 86 (March):55-69.
Abramson, Paul R., John H. Aldrich, and David W. Rohde. 2002. Change
and Continuity in the 2000 Elections. Washington, D.C.: CQ Press.
Adasiewicz, C., Rivlin, D., and J. Stanger. 1997. "Free Television for
presidential Candidates: the 1996 Experiment." Report No. 11, Annenberg
Public Policy Center, University of Pennsylvania.
Achen, Christopher H. and Larry M. Bartels. 2004. “Musical Chairs:
Pocketbook Voting and the Limits of Democratic Accountability,”
unpublished manuscript.
76
Aiken, Leona S. and Stephen G. West. 1991. Multiple Regression: Testing
and Interpreting Interactions. London: Sage.
Alliance for Better Campaigns. 1998. Internet WWW page available at
URL: http://www.bettercampaigns.org [accessed September 1, 1998].
Alvarez, Michael. 1997. Issues and information in Presidential Elections.
Ann Arbor: University of Michigan Press.
Alvarez, R. Michael and Charles Franklin. 1994. "Uncertainty and
Political Perceptions," Journal of Politics 56: 671-688.
Alvarez, Michael. 1992. Issues and information in Presidential Elections.
Ph.D. Dissertation, Duke University.
Alvarez, R. Michael and Garrett Glasgow. 2000. "Two-Stage Estimation of
Nonrecursive Choice Models," Political Analysis 8:2, 147.
Amemiya, T. 1978. "The Estimation of a Simultaneous Equation
Generalized Probit Model," Econometrica 46:1193-1205.
Andersen, Christopher. 1995. Blaming the Government: Citizens and the
Economy in Five European Democracies. London: M.E. Sharpe.
Ansolabehere, Stephen and Shanto Iyengar. 1995. Going Negative: How
attack Ads Shrink and Polarize the Electorate. New York: Free Press.
60
Ansolabehere, Stephen, Shanto Iyengar, Adam Simon, and Nicholas
Valentino. 1994. "Does Attack Advertising Demobilize the Electorate."
American Political Science Review 88 (4): 829.
Ansolabehere, S. and S. Iyengar. 1996. Negative Campaigning. Ann
Arbor:
The University of Michigan Press.
Ansolabehere, S. and S. Iyengar. 1996. Going Negative. Ann Arbor: The
University of Michigan Press.
Ayres, B. Drummond. 2001. "Bush Looks Early to New Hampshire," in The
New York Times, February 18, A31.
Bartels, Larry M. 1988. Presidential Primaries and the Dynamics of Public
Choice. Princeton, NJ: Princeton University Press.
Bartels, Larry M. 1993. "Messages Received: The Political Impact of
Media Exposure." American Political Science Review 87 (2): 267.
Bartels, Larry M. 1986. "Issue Voting Under Uncertainty: An Empirical
Test," American Journal of Political Science 30: 709-728.
Bartels, Larry M. 1996. "Uninformed Votes: Information Effects in
Presidential Elections." American Journal of Political Science. 40: 194230.
Bartels, L.M. 2000. "Campaign Quality: Standards for Evaluation,
Benchmarks for Reform," in Campaign Reform: Insights and Evidence.
Ann Arbor: University of Michigan Press.
Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. 1954.
Voting: A Study of Opinion Formation in a Presidential Campaign.
Chicago: University of Chicago Press.
Berke, Richard L. 2001. "Gore and Bush Strategists Analyze their
Campaigns," in The New York Times, February 12, A19.
Boller, P.F., Jr. 1996. Presidential Campaigns. Oxford: Oxford University
Press.
Campbell, A. and P.E. Converse, et al. 1960. The American Voter.
Chicago: University of Chicago Press.
61
Bowler, Shaun and David M. Farrell, eds. 1992. Conclusion: The
Contemporary Election Campaign. Electoral Strategies and Political
Marketing. New York: St. Martin's.
Buchanan, Bruce. 1996. Renewing Presidential Politics. New York: Little.
Buchanan. Bruce. 2000. "Regime Support and Campaign Reform," in
Campaign Reform: Insights and Evidence. Ann Arbor: University of
Michigan Press.
Buell, Emmett H., Jr. 1987. "'Locals' and "Cosmopolitans': National,
Regional, and State Newspaper Coverage of the New Hampshire Primary."
In Gary R. Orren and Nelson W. Polsby, eds. Media and Momentum: The
New Hampshire Primary and Nomination Politics (Chatham, NJ: Chatham
House), 60-103.
Buhr, Tami. 2000. "What Voters Know about the Candidates and How they
Learned It: The 1996 New Hampshire Republican Primary as a Case
Study." In William G.
Mayer, ed., In Pursuit of the White House:
How We Choose Our Presidential Nominees (Chatham, NJ: Chatham
House Publishers), 203-253.
Butler, David and Austin Ranney. 1992. Electioneering: A Comparative
Study of Continuity and Change. Oxford: Clarendon.
Cain, B. 1998. Remarks at The University of Virginia's Sorenson Institute
for Political Leadership Conference on Campaign Ethics, January 18-21,
Charlotesville, VA.
Cain, Bruce, John Ferejohn and Morris Fiorina. 1987. The Personal Vote
(Cambridge, MA: Harvard University Press).
Campaign Discourse Mapping Project. 1996. "Assessing the Quality of
Campaign Discourse - 1960, 1980, 1988, and 1992." Report No. 4,
Annenberg Public Policy Center, University of Pennsylvania.
Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald
Stokes. 1960. The American Voter. Chicago: University of Chicago Press.
Campbell, Angus, Philip converse, Donald Stokes, and Warren Miller.
1960. The American Voter. New York: John Wiley.
Campbell, James E. 1990. "Trial-Heat Forecasts of the Presidential Vote."
American Politics Quarterly 18: 251-269.
62
Campbell, J.E. "Ambiguity in the Issue Positions of Presidential
Candidates:
A Causal Analysis" American Journal of Political Science 27: 284-293.
Campbell, J.E. 1992. "Forecasting the Presidential Elections in the States,"
American Journal of Political Science 36: 386-407.
Carter, Jimmy. 1978. The Presidential Campaign 1976. Washington D.C.:
The United States Printing Office.
_____________. 2000. "Partisanship and voting Behavior, 1952-1996."
American
Journal of Political Science. 44:1(January): 35-50.
Center for Global Ethics. 1998. Internet WWW page available at URL:
http://www.gloablethics.org [accessed September 1, 1998].
Christ, C.F. 1966. Econometric Models and Methods. New York: Wiley
and Sons. P.
395-401.
Cohen, D. Undefeated: The Life of Hubert H. Humphrey. Minneapolis.
Downs, A. 1957. An Economic Theory of Democracy. New York: HarperRow.
Cohen, Marty, David Karol, Hans Noel, and John Zaller. 2001. "Beating
Reform: The Resurgence of Parties in Presidential Nominations, 19802000," paper presented
at the Annual Meeting of the American
Political Science Association, San Francisco, CA. August 31-September 1,
2001.
Converse, Phillip E. 1964. "The Nature of Belief Systems in Mass Publics,
"in David E. Apter, Ed., Ideology and Discontent, (New York: Free Press),
206-265.
Cook, T.E. 1998. Governing with the News: The News Media as a
Political Institution. (Chicago: University of Chicago Press).
Crotty, William J. and John S. Jackson III. 1985. Presidential Primaries and
Nominations. Washington, D.C.: Congressional Quarterly Press.
Deming, Clarence. 1889. "Town Rule in Connecticut," Political Science
Quartlerly,
4:3, p. 408-432.
Dennis, Jack. 1992. "Political Independence in America III: In Search of
Closet Partisans." Political Behavior. 14: 261-296.
63
Downs, A. 1957. An Economic Theory of Democracy. New York: HarperRow.
Economist. 2000. "A Declaration of Independents." V.354, no8153
(January 15), 32.
Enelow, J.M. and M.J. Hinich. 1984. The Spatial Theory of Voting. New
York: Cambridge University Press.
Enelow, J.M. and M.J. Hinich. 1981. "A New Approach to Voter
Uncertainty in the Downsian Spatial Model," American Journal of Political
Science 25: 483-493.
Fair, R.C. 1978. "The Effect of Economic Events on Votes for President."
Review of Economics and Statistics 60:322-325.
Farrell, D.M. 1993. "Campaign Strategies." In How Ireland Voted, 1992,
ed. Michael Gallagher and Michael Laver. Dubln: Folens.
Farrell, D.M. and Martin Wortmann. 1987. "Party Strategies in the Electoral
Market: Political Marketing in West Germany, Britain, and Ireland."
European Journal of Political Research 15: 297-318.
Fenno, Richard F., Jr. 1977. Home Style: House Members in Their
Districts (Boston: Little Brown).
Finkel, Steven E. 1993. "Re-examining the 'Minimal Effects' Model in
Recent Presidential Campaigns." Journal of Politics 55 (1): 1-21.
Finkel, Steven E. and John G. Geer. 1998. "Spot Check: Casting Doubt on
the Demobilizing Effect of Attack Advertising," American Journal of
Political Science, 42 (2): 573-595.
Fiorina, M.P. 1981. Retrospective Voting in American National Elections.
New Haven: Yale University Press.
Fowler, Linda L., Constantine Spiliotes, and Lynn Vavreck. 2002
forthcoming.
Franklin, C.H. 1991. "Eschewing Obfuscation? Campaigns and the
Perceptions of U.S. Senate Incumbents" American Political Science Review
85: 1193-1214.
Franklin, Charles H. 1992. "Measurement and the Dynamics of Party
Identification." Political behavior 14 (2): 297-309.
64
Freedman, P. and K.M. Goldstein. 1999. "Measuring Media Exposure and
the
Effects of Negative Campaign Ads," American Journal of Political Science
43:1189-208.
Free Air Time and Campaign Reform. 1997. "The Report of the
Conference held by The Annenberg Public Policy Center of the University
of Pennsylvania and The Free TV for Straight Talk Coalition," Report No.
15, Annenberg Public Policy Center, University of Pennsylvania.
Geer, John G. 1989. Nominating Presidents: An Evaluation of Voters and
Primaries. Westport, CT: Greenwood Press.
Geer, J.G. 1990. Advertising Videotape aired in POS 332, Arizona State
University.
Geer, J.G. 1994. "Campaign Advertisements in Presidential Elections: A
Look at some Evidence," unpublished manuscript.
Geer, J.G. 1998. "Campaigns, Competition, and Political Advertising," In
John G. Geer, ed., New Perspectives on Party Politics. Baltimore: Johns
Hopkins Press.
Geer, J.G. 2000. "What this Country Needs Are Short, Negative
Campaign Ads," in Campaign Reform: Insights and Evidence. Ann Arbor:
University of Michigan Press.
Geer, J.G. 2000. "Assessing Attack Advertising: A Silver Lining," in Larry
M. Bartels and Lynn Vavreck, eds. Campaign Reform: Insights and
Evidence. Ann Arbor: University of Michigan Press (forthcoming).
Gelman, Andrew and Gary King. 1993. "Why are American Presidential
Polls So Variable When Election Outcomes are So Predictable?" British
Journal of Political Science 23: 409-451.
Gerber, Alan S. and Donald P. Green. 2000. "The Effects of Personal
Canvassing, Telephone Calls and Direct Mail on Voter Turnout: A Field
Experiment." American Political Science Review 94:653-664.
Gerber, Elizabeth and Rebecca Morton. 2001."Electoral Institutions and
Party Competition: The Effects of Nomination Procedures on Electoral
Coalition Formation." Working paper.
Germond, Jack W. 1981. Blue Smoke and Mirrors: How Reagan Won and
Carter Lost the Election of 1980. New York: Warner Books.
65
Germond, Jack W., and Jules Witcover. 1985. Wake Us When It's Over:
Presidential Politics of 1984. New York: Warner Books.
Germond, Jack W., and Jules Witcover. 1989. Whose Broad Stripes and
Bright Stars? The Trivial Pursuit of the Presidency, 1988. New York:
Warner Books.
Germond, Jack W., and Jules Witcover. 1993. Mad as Hell: Revolt at the
Ballot Box, 1992. New York: Warner Books.
Graber, D.A. 1988. Processing the News. White Plains, New York:
Longman, Inc.
Greene, Steven. 2000. "The Psychological Sources of Partisan-Leaning
Independence." American Politics Quarterly 28 (October):511-537.
Greene, W. H. 1993. Econometric Analysis. New York: Macmillan,
second edition.
Johnson, R.G. 1998. "Negative Campaigners can Learn from Alabama,"
The Atlanta Journal and Constitution, July 24, D2.
Grether, D.M. and G.S. Maddala. 1982. "A Time Series Model with
Qualitative Variables," in Games, Economic Dynamics, and Time Series
Analysis, M.Diestler, E. Fürst, and G. Schwödiauer, editors. ViennaWurzburg: Physica Verlag, p. 291-305.
Hart, R.P. 1997. "Political Debate: Forms, Styles, and Media," The
American Behavioral Scientist 40: 1095-122.
Hetherington, M.J. 1996. "The Media's Role in Forming Voters' National
Economic Evaluations in 1992," American Journal of Political Science
40:372.
Hibbs, D.A. 1982. "President Reagan's Mandate from the 1980 Elections:
A Shift to the Right?" American Politics Quarterly 10: 387-420.
Heckman, J.J. 1978. "Dummy Endogenous Variables in a Simultaneous
Equation System" Econometrica 46: 931-960.
Hetherington, M.J. 1996. "The Media's Role in Forming Voters' National
Economic Evaluations in 1992," American Journal of Political Science
40:372.
Holbrook, Thomas M. 1994. "Campaigns, National Conditions, and U.S.
Presidential Elections" American Journal of Political Science 38 (4): 25-46.
66
Holbrook, Thomas M. 1996. Do Campaigns Matter? Thousand Oaks: Sage.
Huckfeldt, R. and J. Sprague. 1995. Citizens, Politics, and Social
Communication: Information and Influence in an Election Campaign. New
York: Cambridge University Press.
Iyengar S. and D. Kinder. 1987. News That Matters: Television and
American
Opinion. Chicago: University of Chicago Press.
Jacobson, Gary C. 1992. The Politics of Congressional Elections (New
York: Harper Collins).
Johnston, Richard, Andre Blais, Henry E. Brady, and Jean Crete. 1992.
Letting the People Decide: Dynamics of a Canadian Election. Stanford:
Stanford University Press.
Jones, Jeffrey M. 1998. "Does Bringing out the Candidate Bring out the
Votes? American Politics Quarterly 26: 395-419.
Just, Marion R., Ann N. Crigler, Dean E. Alger, Timothy Cook, and
Montague Kern. 1996. Crosstalk: Citizens, Candidates, and the Media in a
Presidential Campaign. Chicago: University of Chicago Press.
Just, Marion, Ann Crigler and Tami Buhr. 2000. "Shifting the Balance:
Journalist vs. Candidate Communication in the 1996 Presidential
Campaign," in Larry M.
Bartels and Lynn Vavreck, eds. Campaign
Reform: Insights and Evidence (Ann
Arbor: University of Michigan
Press).
Kahn, K.F. and P. Kenney. 1999. The Spectacle of U.S. Senate
Campaigns.
(Princeton: Princeton University Press).
Keith, Bruce E., et al. 1992. The Myth of the Independent Voter. Berkely,
CA: University of California Press.
Kenny, Christopher and Michael McBurnett. 1997. "Up Close and
Personal: Campaign Contact and Candidate Spending in U. S. House
Elections." Political Research Quarterly 50: 75-96.
Key, V.O. , Jr. 1966. The Responsible Electorate. New York: Vintage
Books.
67
Kinder, D., G. Adams, et al. 1989. "Economics and Politics in the 1984
Presidential Election," American Political Science Review 33: 491-515.
Kramer, Gerald. 1971. "Short-term Fluctuations in U.S. Voting Behavior,
1896-1964". American Political Science Review 65 (1): 131-143.
Kranish, Michael. 1999. "Narrowing Field Speeds GOP Race," in Boston
Globe, p. A24, September 28.
Kreps, D.M. 1990. A Course in Microeconomic Theory. Princeton:
Princeton University Press.
Lau, R..R., L. Sigelman, C. Heldman, and P. Babbit. 1999. "The Effects of
Negative
Political Advertisemets: A Meta-Analytic Assessment," American Political
Science Review 93: 851-876.
Lazarsfeld, Paul N., B.R. Berelson, and Helen Gaudet. 1948. The People's
Choice: How the Voter Makes Up his Mind in a Presidential Campaign.
New York: Columbia University Press.
League of Women Voters. 1998. Internet WWW page available at URL:
http://www.lwv.org
[accessed September 1, 1998].
Lee, L.F. 1982. "Simultaneous Equations Models with Discrete and
Censored Variables," in Structural Analysis of Discrete Data: With
Econometric Applications, C. Manski and D. McFadden, editors.
Cambridge, MA: MIT Press.
Lewis-Beck, Michael and Tom Rice. 1992. Forecasting Elections.
Washington DC: CQ Press.
Lorant, S. 1968. The Glorious Burden: The Amerian Presidency. New
York: Knopf.
Kiewiet, Roderick R. and Douglas Rivers. 1984. "A Retrospective on
Retrospective Voting". Political Behavior 6 (4): 369-393.
Mannelli, Gilberto Tinacci and Enrico Cheli. 1986. L'immagine del potere:
Comportmaneti, atteggiamenti e strategie d'immagine dei leader politici
italiani. Milano: Franco Angeli Libri.
Marcus, George E., W. Russell Neuman and Michael MacKuen. 2000.
Affective Intelligence and Political Judgment (Chicago: University of
Chicago Press).
68
Marcus, G.E. and M.B. MacKuen. 1993. "Anxiety, Enthusiasm, and the
Vote: The Emotional Underpinnings of Learning and Involvement During
Presidential Campaigns," American Political Science Review 87: 672-685.
Markus, Gregory. 1988. "The Impact of Personal and National Economic
Conditions on the Presidential Vote: A Pooled Cross-Sectional Analysis".
American Journal of Political Science 32 (1): 137-154.
Mayer, W.G. 1996. "In Defense of Negative Campaigning," Political
Science Quarterly 11: 437.
McCallister, B. 1998. "Consultants' Ethics: Politics Survey finds Attitude
of 'Don't Blame Us'", The Washington Post, June 18, C2.
Nie, N.H., S. Verba, and J.R. Petrocik. 1976. The Changing American
Voter. Cambridge, MA: Harvard University Press.
Norrander, Barbara. 1992. Super Tuesday: Regional Politics and
Presidential Primaries. Lexington, KY: University Press of Kentucky.
Ogburn, W.F. and N.S. Talbot. "A Measurement of the Factors in the
Presidential Election of 1928," Social Forces, 8, p. 175-183.
O'Shaughnessy, N.J. 1990. The Phenomenon of Political Marketing.
Houndmills: Macmillan.
Page, B.I. 1976. "The Theory of Political Ambiguity," American Political
Science Review 70:742.
Page, B.I. 1978. Choices and Echoes in Presidential Elections. Chicago:
University of Chicago Press.
Page, B.I. and R.A. Brody. 1972. "Policy Voting and the Electoral Process:
The Vietnam War Issuem" American Political Science Review 66:979.
Page, B.I. 1996. Who Deliberates? Mass Media in Modern Democracy.
Chicago:
University of Chicago Press.
Paldam, Martin. 1991. "How robust is the Vote Function? A Study of 17
Nations over 4 Decades". In Economics of Politics: The Calculus of
Support, ed. Helmut Norpoth, Michael Lewis-Beck, and J.D. Lafay. Ann
Arbor: University of Michigan Press.
Paletz, D.L. 1999. The Media in American Politics: Contents and
Consequences.
69
New York: Longman.
Paolino, Philip and Daron R. Shaw. 2001. "Lifting the Hood on the
Straight-Talk Express: Examining the McCain Phenomenon." American
Politics Research.
29:5 (September): 483-506.
Parmet, H.S. 1972. Eisenhower and the American Crusade. New York:
MacMillan.
Patterson, Thomas E. and Robert D. McClure. 1976. The Unseeing Eye.
New York: Putnam.
Patterson, Thomas E. 1980. The Mass Media Election: How Americans
Choose Their Presidents (New York: Praeger Publishers).
Patterson, Thomas E. 1992. Out of Order (New York: Vintage Books).
Patterson, T.E. 1993. Out of Order. New York: Knopf.
Polsby, Nelson W. and Gary Orren. 1988. "Media and Momentum: The
New Hampshire Primary and Nomination Politics," American Political
Science Review 82:3, 955- 963.
Polsby, Nelson W. 1983. Consequences of Party Reform. New York:
Oxford University Press.
Popkin, Samuel L. 1991. The Reasoning Voter: Communication and
Persuasion in Presidential Campaigns. Chicago, IL: University of Chicago
Press.
Price, D. 1999. "Dole's Star Dims in Presidential Campaign," in The
Detroit News,
pg. A21, October 12.
Project on Campaign Conduct. 1998. Internet WWW page available at
URL: http://www.campaignconduct.org [accessed September 1, 1998].
Putnam, R.D. 1995. "Bowling Alone: America's Declining Social Capital,"
Journal of Democracy 6: 65.
Ragsdale, Lyn. 1998. Vital Statistics on the Presidency: Washington to
Clinton, 2nd ed. Washington, DC: CQ Press.
Ranney, Austin. 1972. "Turnout and Representation in American
Presidential Elections." American Political Science Review. 66
(March):21-37.
70
Rasmussen, E. 1989. Games and Information. Cambridge: Basil
Blackwell.
Rosenstone, S. 1983. Forecasting Presidential Elections. New Haven:
Yale University Press.
Rosenstone, S. 1985. "Why Reagan Won" Brookings Review 3: 25-32.
Rosenstone, S.J. and J.M. Hansen. 1993. Mobilization, Participation, and
Democracy.
New York: Macmillan.
Sanders, David. 1991. "Government Popularity and the Next General
Election,"
Political Quarterly.
Sanders, M.S. 1996. "The Evolution of Individual Decision-Making in
American Presidential Elections," PhD Dissertation, University of
Rochester.
Sanders, M.S. 1996. Beneath Stormy Waters: The Evolution of Individual
Decision
Making the 1984 and 1988 Presidential Elections, PhD Dissertation,
University of Rochester, Rochester, NY.
Schattschnedier, E.E. 1960. The Semisovereign People: A Realists View of
Democracy in America. New York: Holt.
Shaw, Daron R. 1999a. "The Effect of TV Ads and Candidate
Appearances on Statewide Presidential Votes, 1988-1996," American
Political Science Review
93:2, 345.
Shaw, Daron R. 1999b. "A Study of Presidential Camapign Event Effects
from 1952 to 1992," Journal of Politics 61:2 387-422.
Shaw, D.R. 2000. "Is Reform Really Necessary? A Closer look at Media
Coverage, Candidate Events, and Presidential Votes," in Larry M. Bartels
and Lynn Vavreck, eds. Campaign Reform: Insights and Evidence. Ann
Arbor: University of Michigan Press (forthcoming).
Shepsle, K.A. 1972. "The Strategy of Ambiguity: Uncertainty and
Electoral Competition," American Political Science Review 66: 555-68.
Sparrow. B. H. 1999. Uncertain Guardians: The News Media as a
Political Institution. (Baltimore: Johns Hopkins).
71
Stanley, Harold W. and Richard G. Niemi. 1998. Vital Statistics on
American Politics 1997-1998. Washington, D.C.: Congressional Quarterly
Inc.
Stevenson, R. and L. Vavreck. 1999. "Does Campaign Length Matter:
Testing for Cross-national Implications," forthcoming British Journal of
Political Science.
Time Magazine. 1964. "The Fear and The Facts," Time Magazine,
September 25,
p. 16.
Stevenson, Randolph T. 1996. How Parties Compete: Electoral
Performance and Cabinet Composition in Parliamentary Democracies.
Ph.D. Dissertation, University of Rochester.
Taylor, P. 1996. "Free TV Airtime for Candidates is a Partial but
Worthwhile Fix," The New Democrat, September/October.
Tufte, E.R. 1974. Data Analysis for Politics and Policy. Englewood Cliffs,
NJ: Prentice-Hall.
The Washington Post. 1948-1992. Copywrite: Washington D.C.: The
Washington Post
Vavreck, Lynn. 1997. More than Minimal Effects: Differences between
Clarifying and Insurgent Presidential Campaigns in Strategy and Effect,
Ph.D. Dissertation, University of Rochester.
Vavreck, L., C.J.Spiliotes, and L.L. Fowler. 1999. "First in the Nation:
Voter Learning in the New Hampshire Primary," paper delivered at
Dartmouth College Conference on Presidential Primaries, Hanover, NH,
July 1999.
Vavreck, Lynn. 2001. "The Reasoning Voter Meets the Strategic
Candidate: Signals and Specificity in Campaign Advertising, 1998."
American Politics Research. 29 (September): 507-529.
Vavreck, Lynn, Constantine Spiliotes, and Linda Fowler. 2002. The Effects
of Retail Politics in the 1996 New Hampshire Primary. American Journal
of Political Science (July).
Wangan, Edgar. 1983. Polit-Marketing: Das Marketing-Management der
Polischen Parteien. Opladen: Westdeutscher Verlad.
72
Watttenberg, Martin P. 1996. The Decline of American Political Parties:
1992-1994. Cambridge, MA: Cambridge University Press.
West, Darrell. 1993. Air Wares: Television Advertising in Election
Campaigns, 1952-1992. Washington DC: Congressional Quarterly.
White, Theodore. 1961. The Making of the President, 1960. New York:
Atheneum.
Witcover, Jules. 1999. No Way to Pick A President. New York: Farrar,
Straus and Giroux.
Wolfinger, R.E. and S.J. Rosenstone. 1980. Who Votes? New Haven:
Yale University Press.
Woodward, Bob. 1991. The Man Who Would Be President. New York:
Praeger.
Woody, C.H. 1930. The Case of Frank L. Smith, Chicago: Chicago
University Press.
Zaller, John R. 1992. The Nature and Origins of Mass Opinion (New York
or Cambridge: Cambridge University Press)
Zaller, John. 1994. "Politics as Usual: The Rise and Fall of Candidate
Perot," unpublished manuscript.
Zaller, J.R. 1998. "Let's Make Campaigns Fun Again," paper presented at
Princeton University Conference on Campaign Reform, June 18th, 1998.
73
Notes
The ANES has undergone a series of name changes over the years, the most recent from
the National Election Study to its current name in 2005. Prior to that, it was called the
Michigan Election Study (1948-1976)
1
Funding sources included the University of Michigan Survey Research Center and the
Social Science Research Council, Office of the Provost, and Department of Political
Science; the Carnegie Corporation, the Rockefeller Foundation, IBM, the National Science
Foundation, the Ford Foundation, the National Institute of Mental Health, the Markle
Foundation, Russell Sage Foundation, Center for Investigation and Research on Civic
Learning and Education, and Stanford University (IRIS program).
2
3
In 1972, the ANES recorded 3 rather than 5 items for each question.
While concerns are sometimes expressed that these open-ended questions are overly
sensitive to the momentary accessibility of one or anther consideration in respondents'
minds, Geer (1991) shows that respondents' comments do not indiscriminately reflect
recently encountered information.
4
For all of the data presented in the remainder of this chapter, information about numbers
of cases for each election year, for each subset (clarifying v. insurgent, partisanship,
information-level, etc . . .) can be found in the Appendix. For the years under
investigation, the NES data has 25,502 respondents. From 1972 forward, when NES began
asking the 7-point issue scales used at the end of this chapter, the total number of
respondents is 16,870.
5
I do this using OLS regression and controlling for incumbency status and party. I cluster
the standard errors around campaign year.
6
These effects are calculated using the coefficients in the table. The .19 proportional
increase for clarifying candidates results in a 5 to 1 effort (100/19 = 5.26). The similar
effect for insurgent candidates is one-sixteenth (100/6=16.6).
7
If I run a fully interactive model including incumbency and party, the results do not
change substantively, although the coefficient on the direct effect of advertising goes up 6
points and the coefficient on the interaction increases in negativity by 4 points. To allow
the most degrees of freedom as possible, I present the model with only advertising and
incumbency in Table 6.3.
8
After 1996, the ANES sometimes asked respondents to place the candidates even if they
could not place themselves. For consistencies sake, I only use candidate placements for
respondents who could place themselves in all years.
9
This wording was used in some years, and in other years they did not specify “suppose
these people are at one end …” For exact question wording see Appendix.
10
For exact question wording see Appendix.
Results from the issue models show that the interaction between media attention and
time of interview (the test for learning) is significantly different from zero for clarifying
candidates on the economy, but not for any other issues. The results are discernable from
zero for insurgent candidates on no issues.
11
12
74
These differences are robust to varying definitions of time periods. The middle and end
part of the campaign seem systematically different than the beginning.
13
These results are generated from 9 probit analyses using post-election vote report as the
dependent variable. Independent variables include an indicator for whether the respondent
named the insurgent issue as the nation’s most important along with controls for election
year, race, gender, age, and party identification. The probabilities for each class of
insurgent candidate are calculated in separate probit models, as are the results for different
levels of political information.
14
As reported earlier, a fair number of people are unable to place the candidates or
themselves on these issue scales. In the analyses that follow, missing values on issue
scales were imputed using conventional techniques. Results are robust to this process.
15
16
Standard errors are robust and clustered on campaign year.
75