Content uploaded by Mark J Wattier
Author content
All content in this area was uploaded by Mark J Wattier on Dec 03, 2014
Content may be subject to copyright.
The Clinton Factor: The Effects of Clinton’s Personal Image in 2000 Presidential
Primaries and in the General Election
Mark J. Wattier
Department of Government, Law, and International Affairs
Murray State University
Murray, KY 42071
270.762.2695
Mark.Wattier@MurrayState.edu
Paper presented at the annual meeting of the American Political Science Association,
August 2002, Boston.
2
Under normal circumstances factors directly associated with the candidates
running for office determine whom citizens will support. Social scientists have
investigated an impressive array of these candidate-related decision factors (Campbell et
al. 1960; Hinckley 1980; Hinckley 1981; Kelley and Mirer 1974; Nimmo and Savage
1976). The 2000 election cycle provides an opportunity to study a factor that was,
perhaps, unrelated to a candidate. This cycle provides an opportunity to study whether
perceptions of a person not on any ballot have significant effects.1
During the administration’s darkest hours Vice President Gore publicly defended
President Clinton. The steady flow of news and talk of Clinton’s inappropriate
relationship with Monica Lewinsky from the time it became public knowledge through
the impeachment process surely took its toll on Clinton’s personal popularity. But was
Gore’s reputation tarnished, too? Did evaluations of President Clinton as a person affect
Gore’s electoral fortunes? Bob Edwards, the host of National Public Radio’s Morning
Edition, put this question to Gore most succinctly: “Are [voters] blaming you for the
scandals of the last eight years” and are they “getting back at [Clinton] through you”
(Edwards and Gore 2000)?
The purpose of this paper is to examine whether, and to what extent, evaluations
of Clinton as a person (the “Clinton Factor”) influenced how citizens decided for whom
to vote. Gore had to compete in 43 states with presidential primaries to win the
Democratic Party’s nomination (Mayer 2001). Since Gore won each of these contests it
might be assumed that the Clinton Factor was insignificant. The analysis reported here
will demonstrate that these primary election results essentially concealed some politically
significant effects. Since election forecasters had projected Gore as the November
winner through the use of historically accurate methods and techniques, unique, election-
specific factors might have had an effect.2 Various authors have suggested that the
Clinton Factor and the Gore campaign’s response to it contributed to Gore’s defeat
(Abramowitz 2001; Ceaser and Busch 2001; Pomper 2001; Sides 2002; Tseng 2002;
Weisberg and Hill 2001). The analysis reported here will investigate those arguments.
This paper examines several specific questions: Why did Democratic primary
voters prefer Gore or New Jersey Senator Bradley? Did their evaluations of President
Clinton as a person affect their votes? Why did voters in the November general election
3
prefer Gore or Texas Governor Bush? Did the Clinton Factor sway general-election
voters one way or another? Since presidential primaries are state elections and since
“battleground” states are vital to winning the Electoral College vote (Sides 2002), the
paper also examines where (i.e., in what states) the Clinton Factor made a difference.
The results reported here are of interest for several reasons. First, the nomination
process determines “the kind of person who [will] occupy the White House” (Ranney
1977). Since presidential primaries are central to that process, how primary voters decide
warrants greater attention than it has received. Second, the 2000 election cycle posed
several unusual problems, including how Gore would cope with the Clinton Factor and
how Bradley’s health would affect the race. Third, voting specialists who may have
regarded retrospective voting as a universal phenomenon will be surprised by the null
findings of this study (Weisberg and Hill 2001). And, finally, since this study examines
the same decision factors for Spring presidential primaries and for the Fall general
election, it is possible to compare the decision dynamic of primary voters and general
election voters. We know from studies of their social and demographic traits that “spring
voters” are not like “fall voters” (Kritzer 1980; Lengle 1981; Ranney 1968; Ranney 1972;
Ranney and Epstein 1966). We do not know whether their decision dynamic also differs.
Other Factors That Could Affect Primary and General Election Voters
Determining whether Clinton’s personal image had an effect requires not only
evidence on this hypothesized relationship but also control of rival hypotheses. What
other factors could have affected voters? Clinton’s job performance, candidate images,
electability, ideology, and party identification could have affected the choices of primary
and general-election voters.3
Clinton’s Job Performance
Incumbent office holders are typically voted out of office only when there is
widespread dissatisfaction with their performance. If the incumbent is popular, the party
will usually re-nominate that office holder with little, if any, opposition (e.g., Reagan in
1984 or Clinton in 1996). However, if the incumbent is unpopular and seriously
4
jeopardizes the party’s chance to keep the office, a viable challenger may emerge to
compete for the nomination (e.g., Reagan v. President Ford in 1976 and Kennedy v.
President Carter in 1980). If Vice President Gore were viewed as a surrogate for
President Clinton, primary voters could have based their choice of Gore or Bradley on
their evaluation of Clinton’s job performance (Wattier 1990).4 Prior research also
suggests that general election voters could have based their choice of Gore or Bush on
Clinton’s job performance (Fiorina 1981; Key 1966; Miller and Wattenberg 1985).
In the 2000 election cycle Vice President Gore had a difficult choice—whether
and to what extent to run as a “surrogate” incumbent.5 If Gore ran as a surrogate for
President Clinton, he would have tried to link his campaign to the positive
accomplishments of the administration. That strategy, though, might have invited his
rivals to associate him with Clinton’s personal scandals (i.e., the Clinton Factor).
Running as a non-incumbent would have provided Gore an opportunity to escape the
Clinton Factor. However, that approach might have broken the linkage to Clinton’s
performance.6
It is assumed, albeit tentatively, that Gore campaigned, and perhaps was viewed
by many voters, as a surrogate incumbent.7 Ceaser and Busch (2001, 57) substantially
agree with this assumption:
Potential candidates on the Democrat side did … face what was probably
the closest approximation of incumbency ever witnessed in the person of
Al Gore. As vice president, Gore enjoyed the support of Bill Clinton and
served as an active and trusted member of the administration. In the style
of modern vice presidents, he was assigned many important
responsibilities and took part in most of the crucial decisions. Clinton
chose Gore in 1992 not least because their political views were similar.
The two were joined at the hip politically, to the detriment of intraparty
challengers.
Candidate Images
David Broder and Malcolm Jewell remind us that popular images of candidates
are fundamentally important in every election (Butler and Stokes 1976; Campbell et al.
1960; Conway and Wyckoff 1980; Kelley and Mirer 1974; Marshall 1984; Nimmo and
Savage 1976; Wattier 1983b). To quote our discipline’s favorite political reporter:
Elections are contests between individuals, not between philosophies.
Voters choose between Candidate A and Candidate B, not between
5
liberalism and conservatism, high or low taxes, permissive or restrictive
abortion policies. (Broder 1987, 259)
In 1974, as states began to adopt presidential primaries, Malcolm Jewell opined that candidate
images would emerge as important determinants of choice:
The structure of the presidential primary system makes it possible for a
candidate to win primaries if he has a strong organization, plenty of funds,
shrewd advisers, an appealing campaign style, and a good image on
television. (Jewell 1974, 282)
Subsequent research has shown that images of candidates on primary ballots have had a
stronger effect than electability (Abramowitz 1987; Abramowitz 1989; Norrander 1986),
ideology (Marshall 1984; Norrander 1986; Wattier 1983b), or issues (Marshall 1984;
Norrander 1986; Williams et al. 1976). It is essential that images of candidates be
controlled. Perhaps the most interesting image factors in the 2000 election cycle were
the image liabilities: the Clinton Factor, which has been discussed; and Bradley’s heart
problem, which probably contributed to his defeat in New Hampshire.
During 1999 Bradley emerged as the only viable alternative to Vice President
Gore.8 Bradley’s fundraising efforts kept pace with Gore’s. Bradley actually raised more
money in the third quarter--$6.7 million to Gore’s $6.5 million. Bradley’s poll standings
improved both nationally and in key states. Gore’s lead in Iowa narrowed, and some
polls had Bradley leading in New Hampshire and in New York. Endorsements from
party leaders (Senators Kerrey and Moynihan), former Clinton administrators (Robert
Reich, former Labor Secretary), sports celebrities (20 NBA Hall of Famers and one
future Hall of Famer, Michael Jordon), and interest groups (“Friends of the Earth,” an
environmental group critical of Gore’s record) were added to Bradley’s column. Gore’s
campaign lacked momentum just as Bradley’s train picked up steam (Germer 2001).
However, late in the fourth quarter, on December 10, 1999, Bradley left the campaign
trail for treatment of a recurring health problem, an irregular heart beat (Dao 1999).
News reports of this condition appeared and January 2000 tracking polls9 showed Gore
leading in New Hampshire (Altman 2000; Dao 2000; Harwood 2000). What effect, if
any, did news of Bradley’s heart problem have upon primary voters in New Hampshire?10
Ceaser and Busch (2001, 83) have speculated: “Bradley’s heart problem (and seeming
6
lack of candor surrounding it) may have swayed enough voters to cost him New
Hampshire.”
Electability
Voters may support the candidate they perceive as having the best chance of
winning an election (Abramowitz 1989; Abramson et al. 1992; Aldrich 1980; Bartels
1988; Norrander 1986). A successful campaign raises a great deal of money, gains
endorsements of party and group leaders, discourages rivals from running, leads in the
public opinion polls, attracts more news coverage, and receives the Fourth Estate’s
imprimatur, “front runner” (Crotty and Jackson 1985; Gurian 1986; Gurian and Haynes
1993; Keech and Matthews 1977). The front-runner label provides a powerful cue:
“Vote for me because I can win!” Voters might respond to that winning message for
instrumental and expressive reasons.
Electability might function as an instrumental factor (Abramowitz 1989;
Abramson et al. 1992; Stone, Rapoport, and Atkeson 1995). Voters not only evaluate
candidates in terms of expected benefits, but also know that a candidate must have a
realistic chance of winning before “fairer income taxes” can be delivered. Supporting a
candidate with high “utility” but low electability seems irrational, especially when
another candidate offers moderate utility and high electability.11 In short, electability
may be a sophisticated, rational-choice consideration (Downs 1957).
Electability might also function as a form of expressive behavior—citizens might
support a candidate because they wish to be associated with a winner.12 They might
observe that one candidate has the “Big Mo,” causing them to act as political lemmings,
jumping on the bandwagon along the way to victory (Abramowitz 1987; Gallup 1940;
Schmitt-Beck 1996). Electability offers voters a simple, relatively easy, and convenient
basis for choosing a candidate. Information is readily available as to which candidate
seems to have the “inside track” to the winner’s circle (Keeter and Zukin 1983).
Ideology
Election campaigns are usually contests among candidates representing various
ideological factions of a party (Polsby 1981). Those candidates frequently make
7
ideological appeals, and voters seem to respond to those appeals (Conover and Feldman
1981; Hedlund 1977-78; Holm and Robinson 1978; Lengle 1981; Levitin and Miller
1979; Norrander 1988; Wattier 1983a). It seems prudent to control for voter ideology.13
But why does ideology affect candidate choice?
Downs was probably the first to suggest that voters use political ideologies to
make decisions. A voter simply “compares [candidate] ideologies and supports the one
most like his own” (1957, 99). Downs assumes that an ideological identification
represents a general political philosophy or belief system that organizes perceptions of
political issues. An ideological vote is a form of instrumental political behavior; citizens
use their vote (“means”) to affect the philosophical direction (“end”) of public policy.14
According to Downs, voting on an ideological basis is an alternative mode of
action when voters are uncertain about the issue positions of candidates. Several aspects
of primary elections may cause this uncertainty and therefore adversely affect the
opportunity of voters to cast issue votes (Abramowitz 1989; Geer 1989; Gopoian 1982;
Keeter and Zukin 1983; Marshall 1984; Norrander 1986; Wattier 1983b; Williams et al.
1976). The crowded primary schedule does not afford candidates an opportunity to
articulate their stands on issues. The early primaries usually have so many candidates
running that voters are not likely to have a great deal of knowledge about each
candidate’s stand on issues. Candidates may emphasize general campaign themes,
thereby avoiding specific statements about sensitive political issues. When candidates do
emphasize their issue positions, the news media tend to ignore them and focus instead on
the “horse race” (Jewell 1974; Patterson 1980; Patterson 1993).
Party Identification
A person’s identification with a political party has been viewed as a fundamental
decision factor in practically every kind of electoral contest in which candidates from
rival parties compete for popular support (Bartels 2000; Butler and Stokes 1976;
Campbell et al. 1960; Converse and Pierce 1986; Hinckley 1980; Miller 1956; Miller
1991). Studies of primary elections have not examined the effects of party identification,
focusing, instead, on other decision factors.15 There are at least two plausible reasons to
exclude party identification from “models” of candidate choice in presidential primaries:
8
(1) since primaries are intramural contests, candidates would not, under normal
circumstances, employ partisan cues to win support; and (2) since primaries are party
affairs, only loyal party members would attend them.
Some anecdotal evidence, recalled from observations of nomination contests for
some thirty years, suggests that candidates frequently make partisan appeals by raising
questions about a rival’s party loyalty.16 An example is Walter Mondale’s use of
rhetorical terms such as “real Democrats” and “anti-Democrats” to suggest that he was a
Democrat and that his principal rival, Gary Hart, was not (Reeves 1984). Candidates
frequently attack their rivals by suggesting that legislative votes for the opposition party’s
programs represent acts of partisan disloyalty (e.g., the finger pointing episode between
Mondale and Glenn over Glenn’s vote for Reagan’s tax cuts). Republican candidates
often claim that they are “Reagan Republicans.” Taking New Hampshire’s “no-tax ”
pledge” also seems to function as a Republican loyalty oath. In the 2000 election cycle
Gore and his surrogates attacked Bradley for leaving the U. S. Senate, suggesting that
Gore had “stayed to fight” while Bradley had quit the partisan fight (Dao, 1999a). In
short, partisan cues are present in nomination contests.
The presumption that partisanship does not vary among primary voters is also
questionable. Most states, as Carr and Scott have noted, “require voters to state a party
affiliation on registration to be eligible to vote in a party primary election” (Carr and
Scott 1984, 470). However, variation in partisanship could emerge for any number of
reasons. Polling precincts are notoriously under staffed, which might give rise to lax
administration of closed primaries. Although the National Democratic Party has tried to
limit participation to Democrats only, exceptions are routinely granted (e.g., Wisconsin
and Montana). Some states enforce partisan affiliation by requiring voters to make a
public request for a primary ballot, a procedure that could transform a de jure closed
primary into a de facto open primary (Jewell 1983). There is the widespread use of open
primaries by Republican state parties because the GOP has not required its state
organizations to close their primary doors to independents (Cook 2000). Even with
partisan registration, if state law permits someone to engage in same-day registration, a
great deal of mischief is possible.17 Finally, Geer’s analysis suggests that “even in
closed primaries … self-identified independents and partisans of the opposition party still
9
constitute a sizable segment of the electorate” (Geer 1986, 1019). Estimating the effects
of party identification seems necessary because variation from partisan to non-partisan is
possible.
Data and Measures
An analysis of exit-poll data originally collected by Voter News Service (VNS) was
performed to study the effect of the Clinton Factor on primary voters. Its effect on
general-election voters was analyzed with survey data collected in the 2000 National
Election Study (NES). VNS collected these data for a consortium of news organizations
(ABC, AP, CBS, CNN, Fox News, and NBC). Primary voters, chosen randomly just
after they had voted, were given self-administered questionnaires (see Appendix A for a
description of sampling procedures). NES data were collected for a consortium of
academic institutions (i.e., ICPSR). Eligible citizens, also chosen randomly, were
interviewed (some in face-to-face interviews; others in telephone interviews) before and
after the presidential election.18
VNS data serve the practical interests of journalists who provide immediate, election-
night forecasts and interpretations. The VNS data collection process is driven by news
judgments and deadlines, and, evidently, these concerns are best served by a simple, self-
administered questionnaire. The NES data collection process is driven by the theoretical
and methodological interests of scholars, and these concerns have created—over some
fifty years--a complex, interviewer-administered questionnaire. Even though these
studies have different populations, designs, and questionnaires, both data sets include
survey items for the decision factors that are hypothesized to affect candidate choice in
primary and general elections.
The VNS questionnaires were reviewed for relevant survey questions. Eleven
exit polls for Democratic primaries included questions on Clinton’s job performance;
party identification; ideology; and images of Gore, Bradley, and Clinton. In these exit
polls respondents were asked, “Which of these qualities best describes why you voted for
your candidate today?” This question was followed by a list of traits. One of the traits
respondents could choose was: “He has the best chance to win in November.” This item
10
is an electability criterion: the format of the question implies that electability was a basis
for choosing between candidates, not a measure of any one candidate’s perceived chance
of winning (Wattier 1990). The exit polls for New Hampshire and Delaware included a
question about Bradley’s heart problem. The Delaware questionnaire did not have
measures of many of the other variables needed, so it was not analyzed.19
Since the NES questionnaire has multiple items for each decision criteria, items
were selected that closely parallel the VNS items. The questions on Clinton’s job
performance (the traditional Gallup item), ideology, and party identification are either
identical or very similar. The VNS “image” items for Bradley, Clinton, and Gore have
only favorable and unfavorable as the possible responses. Therefore, NES feeling
thermometers for Clinton, Bush, and Gore were recoded into positive (51 to 100) and
negative (0 to 49) categories. An NES item on the expected winner was selected as a
measure of electability.
The analytical task was to determine whether the Clinton Factor affects candidate
choice, controlling for other decision criteria. This task was approached with two
multivariate techniques: ordinary least squares (OLS) regression and logistic regression.
Although OLS regression has some well-known limitations (Aldrich and Nelson 1984),
its principal analytical power—its capacity to estimate the relative effects of predictor
variables—was required. Since the measure (R2) of overall model performance in OLS
regression is of limited value, the summary measure produced by logistic regression, the
percentage of cases correctly predicted, was used.20
Results from these multivariate analyses are discussed in the next section of the
paper; some raw statistical results are also presented in Appendix D. The prescribed
coding requirements for binary logistic regression were followed, and Appendices B and
C provide a detailed listing of concepts, question items, and variable codes for the VNS
and NES studies, respectively.
Results
The 11 Democratic primaries were held on two dates, February 1 for the New
Hampshire primary and March 7 (“Titanic Tuesday”) for the other 10 primaries. Table 1
presents the popular vote and the VNS exit poll percentages for Gore and Bradley. The
11
popular vote seems to indicate that Gore’s bid for the nomination was not adversely
affected by the Clinton Factor. Vice President Gore won each contest. The closest
contest was in New Hampshire, where Gore won by 49.7 percent to Bradley’s 45.6
percent. In several March 7 contests—Georgia, Maryland, Massachusetts, Missouri,
New York, and Ohio—Gore won by substantial margins. The Bradley-Gore contest was
more evenly matched in New England (the “Yankee” primaries), but it was not really
competitive after New Hampshire.
Date State Vote Exit Poll Vote Exit Poll
1 February NH 49.7 52.0 45.6 47.0
7 March CT 55.4 55.0 41.5 42.2
7 March GA 83.8 84.0 16.2 16.0
7 March ME 54.0 54.6 41.3 41.3
7 March MD 67.3 66.9 28.5 28.8
7 March MA 59.9 60.4 37.3 36.8
7 March MO 64.6 64.7 33.6 33.0
7 March NY 65.6 65.0 33.5 34.0
7 March OH 73.6 71.8 24.7 24.2
7 March RI 57.2 56.5 40.6 41.2
7 March VT 54.3 55.0 43.9 44.0
The VNS exit polls accurately represent the popular vote in each primary. The exit-
poll percentages track the popular-vote results fairly closely. The only possible exception
is in New Hampshire where the exit poll overestimates Gore’s vote by 2.3 points and
Bradley’s vote by 1.4 points.
The Clinton Factor in Democratic Primaries
How well do personal evaluations of Clinton predict candidate choice in 2000
Democratic primaries? Figure 1 graphs the percent correctly predicted from three
Table 1. 2000 Democratic Primaries
Gore’s Percent Bradley’s Percenta
Popular Vote Source: Mayer (2001, 32).
aBradley ended his campaign on March 9 with Gore having 33% of the delegates needed to win
the nomination (Norrander 2000).
12
logistic-regression analyses: two analyses with only one predictor variable (line with
triangles for Clinton’s image; line with squares for performance) and a multivariate
analysis with seven predictors (black line with circles).
The effectiveness of Clinton’s image as a predictor of candidate choice is quite
erratic. It varies from a high of 83.9 percent in Georgia to a low of 62.0 percent in
Connecticut and Massachusetts. The prediction rate averages 67.3 percent, but this
measure is somewhat misleading (standard deviation, 6.8). The prediction rates in New
England primaries are especially low.
Consider how well Clinton’s performance predicts choice. Its effectiveness mirrors
almost exactly Clinton’s personal image; these two prediction rates are highly correlated
(r = .92). Since performance seems to predict choice as well as personal image, instant
election night analyses may have focused more on performance, surely the more
traditional and perhaps also the more acceptable basis for choosing between Gore and
Bradley. The statistical similarities between performance and personal image may have
concealed the effects of the Clinton Factor.
The addition of other predictors dramatically increases the overall prediction rate
to 81.6 percent. These prediction rates are less erratic (standard deviation, 3.7), and vary
from a low of 77.5 percent in Vermont to a high of 88.8 percent in Georgia. The gaps
between the lines (percentage point differences) visually indicate how much
improvement the seven-variable model makes over the simple one-variable models.
Comparing the image and fully specified models reveals some large percentage-point
improvements in New England primaries: 16.4, 16.3, 19.7, and 14.3 in New Hampshire,
Connecticut, Rhode Island, and Vermont, respectively. The prediction rates observed for
the fully specified model of candidate choice tend to affirm that relevant predictors have
been included.
Figure 2 clearly depicts the effects of the Clinton Factor. Gore’s exit-poll
percents are graphed for primary voters who approved of Clinton’s job performance.
Figure 1 here
Figure 2 here
13
I would have expected to find high percentages, in the high 80s or low 90s, voting for
Gore.21 When primary voters liked Clinton (line with triangles), a high percentage of
them supported Gore in only three primaries—92.0, 85.6, and 82.8 in Georgia, Maryland,
and Ohio, respectively. Consider what happened to Gore’s vote when primary voters
disliked Clinton (line with squares). In two primaries, New Hampshire and Vermont,
Gore’s percent of the vote actually fell below 50 percent. In three primaries—Maine
(51.7), New York (52.9), and Rhode Island (50.8)—Gore’s percent was only slightly
above the 50 percent mark. In Georgia and Ohio Clinton’s image suppressed Gore’s
expected vote, but not to the same extent. Clinton’s image seems to depress Gore’s
expected vote by approximately 19 points (18.9) among voters who approved of
Clinton’s job performance but personally disliked him.
Table 2 displays selected results from logistic-regression analyses that included
all hypothesized decision factors (i.e., the seven-variable model). Only the results for
Clinton’s image and performance are presented in order to examine, in greater detail, the
effects of the Clinton Factor.
State Clinton Image Performance
NH 0.636* 3.263*
CT 0.652 2.085
GA 0.505* 2.004*
ME 0.646 1.369
MD 0.468* 1.559
MA 0.793 2.573*
MO 0.451* 0.980
NY 0.560* 1.896*
OH 0.537* 1.272
RI 0.382 2.627*
VT 0.639 1.787
Table 2. Odds Ratios for Clinton’s Image and Performance
*
p
≤
.05 testing Ho: Odds Ratio
=
1.00.
Note: An Odds Ratio > 1.00 indicates that a factor increases the
odds of voting for Gore; an Odds Ratio < 1.00 indicates that a
factor decreases the odds of voting for Gore. Clinton’s image is
coded 1, unfavorable, and 0, favorable; performance is coded 1,
approve, and 0, disapprove (see Appendix B).
14
The cell entries in Table 2 are the odds ratios for Clinton’s image and
performance.22 These variables were coded with the following expectations: voters who
approved of Clinton’s performance (coded 1) would be more likely to support Gore;
those who personally disliked the President (coded 1) would be less likely to support the
Vice President. I expected odds ratios greater than 1.00 for performance and odds ratios
less than 1.00 for Clinton’s image. As the results in Table 2 indicate, performance is
statistically significant in five of the 11 primaries and Clinton’s image is significant in six
of them. All significant odds ratios are in the hypothesized directions.
These results indicate how the Clinton Factor produced mixed political
consequences for Vice President Gore. In three primaries—New Hampshire, Georgia,
and New York—both Clinton variables are significant, which suggests that Clinton’s
performance helped Gore (odds ratios > 1.00) about as much as Clinton’s image (odds
ratios < 1.00) hurt him, politically. In three primaries—Maryland, Missouri, and Ohio—
Clinton’s image is significant, but performance is not. In these primaries Clinton’s
negative personal image decreases the odds of voting for Gore, as expected, but this
negative effect is not offset by a positive effect from performance. In only two
primaries, Massachusetts and Rhode Island, does it seem that Clinton’s performance
helps the Vice President without a negative effect from Clinton’s image. Finally, in
three primaries—Connecticut, Maine, and Vermont—both Clinton variables are not
statistically significant.
Figure 3 displays selected results from the OLS regression analyses. It displays
the beta weights (standardized regression coefficients) for Gore’s image, Bradley’s
image, Clinton’s image, and Clinton’s job performance.23 Clinton’s personal image has
a statistically significant effect in eight of the 11 primaries; it affects whether primary
voters selected Gore or Bradley in all but Georgia, Maine, and Massachusetts. Its overall
effect is stronger than performance, party identification, and ideology (see pooled results
in Appendix D). In Ohio and Missouri its effect is also stronger than electability. These
results also indicate that performance is the least important decision factor. In only three
primaries—New Hampshire, Connecticut, and Massachusetts—is Clinton’s job
performance statistically significant. In every primary except Massachusetts, Clinton’s
image has a stronger effect than Clinton’s performance.
15
Images of the candidates, especially Gore’s image, are considerably more
important than Clinton’s image and performance. Gore’s image is the most important
factor in eight of the 11 primaries. In Maine and Rhode Island the two candidate image
variables have similar effects. Bradley’s image has its strongest effect in his home state
of Missouri.
Bradley’s health is a significant factor in New Hampshire; voters who viewed
Bradley’s health as a consideration tended to support Gore, a beta weight of .16 (see
Appendix D). This factor is not as important as the two candidate image factors, but it
has a somewhat stronger effect than either Clinton’s image or performance.
This finding raises an intriguing question: What if Bradley’s heart problem had
not become an issue (Ceaser and Busch 2001)? This hypothetical question was addressed
with CLARIFY, a software program that estimates what a model would predict under
different, user-specified conditions.24 After estimating coefficients with only statistically
significant factors (i.e., excluding ideology and party identification), I set Bradley’s
health to “0.” This arbitrary action statistically removes health as a consideration for
New Hampshire voters. The predicted values for Gore’s percent are 45.5 percent (OLS
regression) and 39.8 percent (logistic regression). If the statistical assumptions and
techniques of CLARIFY are reasonable, we would have to conclude that Bradley might
have won New Hampshire.
The Clinton Factor in the 2000 General Election
Why did voters in the November general election prefer Gore or Texas Governor
Bush? Did the Clinton Factor sway general-election voters one way or another? Where
was the Clinton Factor important?
A recent paper by Sides (2002) has demonstrated that voters experienced the 2000
presidential campaign differently depending on whether they lived in one of the
“battleground” states (see note to Table 3). These states were where the Clinton and
Figure 3 here
16
Bush campaigns vigorously competed because the outcome was in question. As Sides
(2002, 17) has stated:
A higher percentage of voters in battleground states report reading about
the campaign in the newspaper as well as seeing campaign advertisements
and programming, though here the [difference] in exposure to advertising
is most substantively meaningful: in battleground states 84.6 percent of
respondents report seeing campaign advertisements, compared to 69.2
percent in non-battleground states. Similarly, voters in battleground states
are more likely to be contacted or receive mail from a party. For example,
44.8 percent of respondents in battleground states were contacted, as
opposed to 34.8 percent of those in non-battleground states. Finally, there
is evidence that respondents in battleground states were themselves more
active in the campaign, in that more tried to influence others’ vote and
participated in a campaign activity of some sort.
Such dramatic differences in campaign experiences and behaviors clearly suggest that the
battleground states are where to test for Clinton-Factor effects. While some groups (e.g.,
Democratic Leadership Council) were quick to criticize Gore for not using President
Clinton more effectively, Thomas Mann opined: “Clinton would have done more harm
than good with swing voters in battleground states” (Ceaser and Busch 2001, 31).
Was Clinton’s personal image a significant factor in the battleground states?
Results of separate logistic regression analyses for voters on “safe” ground and for those
on “battle” ground, displayed in Table 3, indicate that Clinton’s image had a significant
effect on whether voters in battleground states supported Gore or Bush. In battleground
states voters who disliked the President were more likely to support Bush than Gore
(odds ratio, 0.54). Evidently, Clinton’s image did not significantly affect voters on “safe”
ground. Three factors seemed to affect voters on “safe” ground—Gore’s image, Bush’s
image, and party identification (Kelley and Mirer 1974).
Clinton’s job performance, though significant in Democratic primaries, seemed to
have no effect in the presidential election (Weisberg and Hill 2001). In some
presidential primaries positive effects from Clinton’s performance tended to balance
negative effects from Clinton’s image. This “balancing” effect was not so evident in the
general election.
17
Predictor “Safe” Ground “Battle” Ground Democratic
Primaries
Gore Image 20.991* 7.029* 27.974*
Bush/Bradley Image .021* .009* .037*
Clinton Image .254 .054* .563*
Performance 1.642 1.256 1.748*
Party Identification 9.467* 7.491* 1.465*
Ideology 1.617 1.841 .621*
Electability 1.010 1.550* 7.685*
Correctly Predicted 92.9% 91.1% 80.1%
N 356 343 8298
As expected, candidate images are important decision criteria in both the spring
primaries and the fall presidential election. Party identification is also important in both
election seasons. Perhaps under “normal” political conditions, with more consistency in
performance and personal evaluations, performance would emerge as a general decision
criterion.
Discussion
The 2000 Democratic nomination contest appears to fit the consensual
nomination pattern (Keech and Matthews 1977). An early favorite (Al Gore) faced a
weak opponent (Bill Bradley), and the early favorite easily won the nomination. This
characterization only fits the primary and caucus results; it does not fit the political-
resources profile. Bradley was not necessarily a weak opponent. At the end of 1999
Gore was reported to have $5.7 million cash-on-hand; Bradley actually had more money
to spend, about $8.3 million. Bradley also had more cash-on-hand at the end of the third
quarter reporting period, $10.7 million to Gore’s $10.3 million.25 From August 1999 to
Table 3. The Clinton Factor in the General Election
Note: States were divided into “safe” and “battle” areas following Sides (2002), who
relied upon The Cook Political Report and CNN. The battleground states were: AR,
AZ, DE, FL, IL, IA, LA, ME, MI, MN, MO, NV, NH, NM, OH, OR, PA, TN, WA,
WV, and WI. Gore’s pollsters also defined these as battleground states (Jamieson and
Waldman 2001, 91-93). Results for 2000 Democratic primaries were created by
merging all cases into a single data file.
*p ≤ .05 testing Ho: Odds Ratio = 1.00.
18
December 1999, Gore was always the first choice of registered Democrats surveyed by
the Gallup Organization, but Gore’s lead seemed to be evaporating (Newport 2000).
Although general-election, trial-heat polls showed both Democrats losing to Bush,
Bradley seemed to be a stronger general-election candidate than Gore (Newport 2000).
By January 2000, after the news of Bradley’s heart problem, Gore’s lead in the first-
choice poll of registered Democrats widened considerably.
It seems that news of Bradley’s heart problem damaged his electability, making
him a weaker candidate. Some observers have suggested that Bradley was not a very
effective campaigner (Ceaser and Busch 2001; Mayer 2001). However, the analysis
reported in this paper has suggested that Bradley would have won the New Hampshire
primary had his health not become an issue. Bradley might have faired better in other
states, especially in New England primaries (Connecticut, Maine, Rhode Island, and
Vermont) where his image was as favorable as Gore’s. New Hampshire voters actually
had a more favorable image of Bradley, 80.3 percent favorable to Gore’s 72.5 percent.
Another political factor made Bradley’s bid for the nomination more difficult—Gore
switched from running as an incumbent to running as a non-incumbent.
Several signs indicated that Gore campaigned as a non-incumbent. Gore
distanced himself from the administration’s policy toward Elian Gonzalez (Alvarez 2000;
Seelye 2000). He moved the campaign headquarters from Washington, D.C., to
Nashville, Tennessee, making him a Washington outsider (Johnson 2001). His
announcement speech repeatedly emphasized “family,” creating a stark contrast to the
Clinton family (Gore 1999). His acceptance speech contained a politically significant
line: “This election is not a reward for past performance” (Gore 2000). Two of his
resume advertisements omitted his eight years as Vice President (Berry 2000; Richardson
2000). He selected Joseph Lieberman, a Clinton administration critic, as his vice-
presidential running mate. With few exceptions (e.g., a New York fundraiser), Gore and
President Clinton never campaigned together, a clear break from their teamwork in 1992
and 1996. In keeping with the maxim, “actions speak louder than words,” recall how
passionately the Gores kissed before the Democratic nominee delivered his acceptance
speech. Was this a sign that their marriage was a different kind of relationship than what
the country had witnessed the previous eight years? Finally, as Tseng (2002, 206) has
19
noted, “Gore did not mention Clinton’s name on the campaign trail or even in the
debates.”
Gore’s campaign consultants also argued for non-incumbency. In Electing the
President 2000: An Insider’s View (Jamieson and Waldman 2001), Gore’s consultants
stated the following strategic consideration that applied to both the nomination and the
general-election campaigns:
We couldn’t just run on the Clinton record. We couldn’t say that this
election, to quote the convention speech, is “a reward for past
performance.” We had to make it about the future: what to do with
the prosperity, especially in a situation where voters had cognitive
dissonance about Clinton. (Bob Shrum quoted on page 57)
It seems that the 2000 Democratic nomination contest was fought under unusual
circumstances. Gore’s electability was limited by his ties to Clinton; Bradley’s
electability, by his health. It seems that a tacit agreement governed their contest: Gore
would not attack Bradley’s health, Bradley would not use Clinton’s scandals to attack
Gore, and they both would focus the debate on the few policy issues that divided them. It
was neither necessary to attack the Clinton administration, nor vital that it be defended.
No such implicit norm of electoral combat limited Republicans from launching
various versions of the Clinton-Factor attack. Bush’s campaign advisors were well aware
of the unique politics of the 2000 presidential campaign—a popular incumbent, with
whom the Democratic nominee was associated, was personally disliked by potential
swing voters (see Matthew Dowd’s remarks in Jamieson and Waldman 2001, 25).
Linking Clinton and Gore might be a way of moving some swing voters to Bush. Recall,
for example, Bush’s frequent refrain, “I pledge to uphold the dignity of the office,” made
with his right hand raised in the air. Two remarks by his running mate are worth quoting
(Cheney 2000):
On the first hour of the first day, [George W. Bush] will restore decency
and integrity to the Oval Office.
Mr. Gore will try to separate himself from his leader’s shadow. But
somehow we will never see one without thinking of the other.
The billboard (see text box on page 20) the Republican Party placed next to Vice
President Gore’s Nashville headquarters illustrates how Republicans made sure potential
20
voters would “never see one without thinking of the other.” The Republican attack
argued: “Vice President Gore can’t restore honesty and dignity to the Oval Office
because of his close (guilt by) association with President Clinton.”
It seems as if the Gore-Bradley contest de-activated the Clinton Factor. Its effect
3333333
It seems as if the Gore-Bradley contest de-activated the Clinton Factor. Its effect
was minimized, in part, by the Gore campaign acting as if there were no connection
whatsoever. Despite this attempt to attenuate its effect, the exit-poll results reported here
demonstrate that the Clinton Factor did have a significant effect on primary voters. The
general election contest seemed to activate the Clinton Factor. Arguably, both camps
helped this factor emerge (e.g., Gore’s debate performances and Republican attacks).
The effect of the Clinton Factor was concealed in the spring by Gore’s unbroken string of
victories in party caucuses and primaries. Its effect was revealed in the fall by Bush’s
According to the Republican Party press release issue on November 29, 1999:
The billboard will show Bill Clinton hugging Al Gore along with Gore's
famous quote, "One of our Greatest Presidents," made at Clinton's post-
impeachment pep rally. The advertisement is intended to emphasize the drag
that Al Gore's support of Bill Clinton has had on his presidential campaign.
Source: Tennessee Republican Party. Press release and image are available
from http://www.tngop.org/gallery/sign.html.
21
victory, a contest decided, in part, by voters in battleground states whose choice of Gore
or Bush the Clinton Factor affected. Based on the analyses reported here the answer to
Bob Edwards’s question to Gore about the Clinton Factor seems to be: Some spring
voters and many more fall voters blamed Gore for the scandals and they got back at
President Clinton through his Vice President.
22
References
Abramowitz, Alan I. 1987. Candidate Choice before the Convention. Political Behavior
9:49-61.
Abramowitz, Alan I. 1989. Viability, Electability, and Candidate Choice in a Primary
Election. Journal of Politics 51:977-992.
Abramowitz, Alan I. 2001. Why Joe Lieberman is not in a Secret Secure Location:
Explaining the Failure of the Presidential Election Forecasting Models. Paper read
at Annual Meeting, Southern Political Science Association, at Atlanta.
Abramson, Paul R., John H. Aldrich, Phil Paolino, and David W. Rohde. 1992.
"Sophisticated" Voting in the 1988 Presidential Primaries. American Political
Science Review 86:55-69.
Aldrich, John H. 1980. Before the Convention. Chicago: University of Chicago Press.
Aldrich, John H., and Forrest D. Nelson. 1984. Linear Probability, Logit, and Probit
Models. Beverly Hills: Sage Publications.
Altman, Lawrence K. 2000. Bradley's Doctors Say He is in Excellent Shape. New York
Times, January 30, 2000, A20.
Alvarez, Lizette. 2000. Senate Will Hold Hearings on Raid on House in Miama. New
York Times, April 26, 2000, A1, A16.
Bartels, Larry M. 1988. Presidential Primaries and the Dynamics of Public Choice.
Princeton: Princeton University Press.
Bartels, Larry M. 2000. Partisanship and Voting Behavior, 1952-1996. American Journal
of Political Science 44 (1):35-50.
Berry, Candice. 2001. Getting to Know Gore (campaign ad) [World Wide Web]. 2000
[cited October 1, 2001 2001]. Available from
http://cbsnews.cbs.com/now/story/0,1597,227267-412,00.shtml.
Butler, David, and Donald Stokes. 1976. Political Change in Britain. 2nd College Edition
ed. New York: St. Martin's Press.
Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960.
The American Voter. New York: John Wiley.
Ceaser, James W. , and Andrew E. Busch. 2001. The Perfect Tie. Lanham: Rowman &
Littlefield.
Cheney, Richard B. 2002. Richard B. Cheney Address Accepting the GOP Nomination
for Vice President 2000 [cited August 1 2002]. Available from
http://www.cnn.com/ELECTION/2000/conventions/republican/transcripts/cheney
.html.
Conover, P. J., and S. Feldman. 1981. The Origins and Meaning of Liberal/Conservative
Self-Identification. American Journal of Political Science 25:617-645.
Converse, Philip E., and Roy Pierce. 1986. Political Representation in France.
Cambridge: Belknap Press of Harvard University Press.
Conway, M. M. , and M. L. Wyckoff. 1980. The Kelley-Mirer Rule and Prediction of
Voter Choice in the 1974 Senate Elections". Journal of Politics 42:1146-1152.
Cook, Rhodes. 2000. Race for the Presidency: Winning the 2000 Nomination. Vol. CQ
Press. Washington.
23
Crotty, William, and John S. Jackson, III. 1985. Presidential Primaries and Nominations.
Washington: CQ Press.
Dao, James. 1999. Bradley Misses Event for Check of Heart Beat. New York Times,
December 11, 1999, A1, A14.
Dao, James. 2000. Bradley and Doctor Say Bouts Are No Worry. New York Times,
January 22, 2001, A11.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and
Row.
Edwards, Bob, and Al Gore. 2001. Al Gore Interview Transcript from Morning Edition
July 19, 2000 [World Wide Web]. National Public Radio, 2000 [cited October 1,
2001 2001]. Available from
http://www.npr.org/news/national/election2000/bios/gorepop.html.
Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New
Haven: Yale University Press.
Gallup, George. 1940. Is there a Bandwagon Vote? Public Opinion Quarterly 4 (2):244-
249.
Geer, John G. 1989. Nominating Presidents: An Evaluation of Voters and Primaries.
New York: Greenwood.
Germer, Mattie J. 2001. The Road to the Presidency: A Timeline of the 2000 Presidential
Campaign [World Wide Web]. 2001 [cited October 1, 2001 2001]. Available
from http://www.ksg.harvard.edu/iop/campaign-timeline.pdf.
Gopoian, J. David. 1982. Issue Preferences and Candidate Choice in Presidential
Primaries. American Journal of Political Science 26:523-546.
Gore, Al. 2001. Announcement Speech 1999 [cited October 1, 2001 2001]. Available
from
http://www1.cnn.com/ALLPOLITICS/stories/1999/06/16/president.2000/gore/tra
nscript.html.
Gore, Al. 2001. Acceptance Speech (CNN) 2000 [cited October 1, 2001 2001]. Available
from
http://www11.cnn.com/ELECTION/2000/conventions/democratic/transcripts/gore
.html.
Gurian, Paul-Henri. 1986. Resource Allocation Strategies in Presidential Nomination
Campaigns. American Journal of Political Science 30:802-821.
Gurian, Paul-Henri, and Audrey A. Haynes. 1993. Campaign Strategy in Presidential
Primaries. American Journal of Political Science 37:335-341.
Harwood, John. 2000. Bradley Has Had Heartbeat Episodes Since Christmas. New York
Times, January 21, 2000, A4.
Hedlund, Ronald D. 1977-78. Cross-over Voting in a 1976 Open Presidential Primary.
Public Opinion Quarterly:498-514.
Hinckley, Barbara. 1980. The American Voter in Congressional Elections. American
Political Science Review 74 (3):641-650.
Hinckley, Barbara. 1981. Congressional Elections. Washington, D.C.: Congressional
Quarterly Press.
Holm, John S. , and John P. Robinson. 1978. Ideological Identification and the American
Voter. Public Opinion Quarterly 42:235-246.
24
Jamieson, Kathleen Hall, and Paul Waldman, eds. 2001. Electing the President 2000:
The Insider's View. Philadelphia: University of Pennsylvania Press.
Jewell, Malcolm E. 1974. A Caveat on the Expanding Use of Presidential Primaries.
Policy Studies 2:279-284.
Jewell, Malcolm E. 1983. Democrat or Republican?: Voters' Choice of a Primary. Paper
read at Annual Meeting, Southern Political Science Association, at Birmingham,
Alabama.
Johnson, Dennis W. 2001. No Place for Amateurs. New York: Routledge.
Keech, William R., and Donald R. Matthews. 1977. The Party's Choice. Washington:
Brookings Institution.
Keeter, Scott, and Cliff Zukin. 1983. Uniformed Choice. New York: Praeger.
Kelley, Jr., Stanley , and Thad W. Mirer. 1974. The Simple Act of Voting. American
Political Science Review 68:572-591.
Key, V. O. 1966. The Responsible Electorate. New York: Vintage Books.
Kritzer, H. M. 1980. The Representativeness of the 1972 Presidential Primaries. In The
Party Symbol, edited by W. Crotty. San Franciso: W. H. Freeman.
Lengle, J. I. 1981. Representation and Presidential Primaries. Westport, CT: Greenwood
Press.
Levitin, T. E. , and W. E. Miller. 1979. Ideological Interpretations of Presidential
Elections. American Political Science Review 73:751-771.
Marshall, Thomas R. 1984. Issues, Personalities, and Presidential Primary Voters. Social
Science Quarterly 65:750-760.
Mayer, William G., ed. 2001. The Presidential Nominations. Edited by G. M. Pomper,
The Election of 2000. New York: Chatham House.
Miller, Arthur H., and Martin P. Wattenberg. 1985. Throwing the Rascals Out: Policy
and Performance Evaluations of Presidential Candidates, 1952-1980. American
Political Science Review 79:359-372.
Miller, Warren E. 1956. One-Party Politics and the Voter. American Political Science
Review 50 (3):707-725.
Miller, Warren E. 1991. Party Identification, Realignment, and Party Voting: Back to the
Basics. American Political Science Review 85 (2):557-68.
Newport, Frank. 2001. Bush and Gore Maintain Leads for Their Party's Nomination
[World Wide Web]. Gallup News Service, 2000 [cited October 1, 2001 2001].
Available from http://www.gallup.com/poll/releases/pr000113.asp.
Nimmo, Dan , and Robert L. Savage. 1976. Candidates and Their Images. Pacific
Palisades: Goodyear.
Norrander, Barbara. 1986. Correlates of Vote Choice in the 1980 Presidential Primaries.
Journal of Politics 48:156-166.
Norrander, Barbara. 1988. The Role of Ideology in Presidential Nominations. Paper read
at Annual Meeting of Midwest Political Science Association, at Chicago.
Norrander, Barbara. 2000. The End Game in Post-Reform Presidential Nominations.
Journal of Politics 62 (4):999-1013.
Patterson, Thomas E. 1980. The Mass Media Election. New York: Praeger.
Patterson, Thomas E. 1993. Out of Order. New York: Vintage Books.
25
Polsby, Nelson W. 1981. Coalition and Faction in American Politics: An Institutional
View. In Party Coalitions in the 1980s, edited by S. M. Lipset. San Franciso:
Institute for Contemporary Studies.
Pomper, Gerald M. 2001. The 2000 Presidential Election: Why Gore Lost. Political
Science Quarterly 116 (2):201-223.
Ranney, Austin. 1968. The Representativeness of Primary Electorates. Midwest Journal
of Political Science 12:224-238.
Ranney, Austin. 1972. Turnout and Representation in Presidential Primary Elections.
American Political Science Review 66:21-37.
Ranney, Austin. 1977. Participation in American Presidential Nominations, 1976.
Washington: American Enterprise Institute.
Ranney, Austin, and Leon D. Epstein. 1966. The Two Electorates: Voters and Non-
Voters in a Wisconsin Primary. Journal of Politics 28:598-616.
Reeves, Richard. 1984. The Campaign for Page One. In Frontline: PBS Video.
Richardson, Sean. 2001. Gore Trots Out His Resume (campaign ad) [World Wide Web].
2000 [cited October 1, 2001 2001]. Available from
http://cbsnews.cbs.com/now/story/0,1597,144543-412,00.shtml.
Schmitt-Beck, R. 1996. Mass media, the electorate, and the bandwagon. A study of
communication effects on vote choice in Germany. International Journal of
Public Opinion Research 8 (3):266-291.
Seelye, Katharine Q. 2000. Gore Supporting Residency Status for Cuban Child. New
York Times, March 31, 2000, A1, A20.
Sides, John. 2002. The Presidential Campaigns of 2000. Paper read at Annual Meeting of
the Midwest Political Science Association, at Chicago.
Stone, Walter J., Ronald B. Rapoport, and Lonna Rae Atkeson. 1995. A Simulation
Model of Presidential Nomination Choice. American Journal of Political Science
39 (1):135-161.
Tseng, Margaret. 2002. The Clinton Effect: How a Lame-Duck President Impacted His
Vice President's Election Prospects. In The Elecion of the Century, edited by S. J.
Wayne and C. Wilcox. Armonk, New York: M. E. Sharpe.
Wattier, Mark J. 1983a. Ideological Voting in 1980 Republican Presidential Primaries.
Journal of Politics 45:1016-1026.
Wattier, Mark J. 1983b. The Simple Act of Voting in 1980 Democratic Presidential
Primaries. American Politics Quarterly 11:267-291.
Wattier, Mark J. 1990. Retrospective Voting in Presidential Primaries. Midsouth Political
Science Journal 11:57-80.
Weisberg, Herbert, and Timothy Hill. 2001. The Succession Presidential Election of
2000: The Battle of the Legacies. Paper read at Annual Meeting, American
Political Science Association, at San Francisco.
Williams, Daniel C., Stephen J. Weber, Gordon A. Haaland, Ronald H. Mueller, and
Robert E. Craig. 1976. Voter Decisionmaking in a Primary Election: An
Evaluation of Three Models of Choice. American Journal of Political Science
20:37-49.
26
Figure 1. Predicting Choice
50
55
60
65
70
75
80
85
90
95
NH CT GA ME MD MA MO NY OH RI VT
Democratic Primary
Percent Correctly Predicted
Clinton Image Performance Seven-Variable Model
Note: The variables for the Seven-Variable Model are: Gore Image, Bradley Image, Clinton
Image, Clinton Popularity, Party Identification, Ideology, and Electability Criterion (see
Appendix B for measures and Appendix D for the multivariate statistics). The model for New
Hampshire also included Bradley’s Health.
27
Figure 2. The Clinton Factor
40
50
60
70
80
90
100
NH CT GA ME MD MA MO NY OH RI VT
Democratic Primary
Gore Percent
DisLike Clinton Like Clinton
28
Figure 3. Relative Effects
.00
.05
.10
.15
.20
.25
.30
.35
.40
.45
.50
NH CT GA ME MD MA MO NY OH RI VT
Democratic Primary
Absolute Value [Beta Weight]
Gore Image Performance Bradley Image Clinton Image
29
Appendix A. Exit Poll Methodology
HOW THE VNS EXIT POLL IS CONDUCTED
PLEASE NOTE: The first two paragraphs are a short description of the methodology suitable for
broadcast or newspaper reports.
The VNS exit polls were developed and conducted by Voter News Service. VNS is an
association of ABC News, CNN, CBS News, FOX News, NBC News and the Associated Press. The
exit poll results are based on interviews with a probability sample of voters exiting polling places around
the state on Election Day.
The error due to sampling for most state exit poll questions is approximately plus or minus 4
percentage points (see the attached chart). This means 95 times out of 100, samples like this one will have
results within 4 points of what a complete count of all voters would show. Results based on part of the
sample will have larger margins of error, as will groups of voters--such as Blacks--who are found
disproportionately only in some precincts.
Sampling
The samples were selected in two stages. First, a probability sample of voting precincts within
each state was selected that represents the different geographic areas across the state and the vote by party.
Precincts were selected with a probability proportionate to the number of voters in each precinct. Each
voter in a state had the same chance to have his or her precinct selected. There is one exception. In some
states, precincts that have large minority populations were sampled at a higher rate than other precincts.
The sample weighting (described below) adjusts the representation of these precincts to their correct share
of the total vote. Second, within each precinct, voters were sampled systematically throughout the voting
day at a rate that gives all voters in a precinct the same chance of being interviewed.
Weighting
The exit poll results are weighted to reflect the complexity of the sampling design. That is, the
weighting takes into account the different probabilities of selecting a precinct and of selecting a voter
within each precinct. For example, minority precincts that were selected at a higher rate receive a smaller
weight than other precincts of the same size. There is also an adjustment for voters who were missed or
refused to be interviewed, which is based on their observed age, race and sex.
NOTE: The exit poll results are only meaningful if they are weighted. Unweighted tabulations may
be seriously misleading and should not be used for any reason.
30
Appendix A. (continued)
Sampling Error
The error due to sampling depends on, among other things, the number of respondents in each
group.
The sampling errors below are for most voter groups. There are 95 chances in 100 that a sample
result is within this many percentage points of the result that we would have obtained by trying to
interview every voter.
The sampling error also depends on how many precincts have voters with the characteristic of
interest. For example, black or high income voters may be found clustered in only a few sample precincts.
A characteristic that is clustered in a few precincts has a larger sampling error than characteristics that are
not clustered such as age or sex. Sampling error may be up to three times larger for clustered
characteristics.
Example: If there are a total of 1,000 respondents in the sample, the sampling error on the
percentage of men voting for a candidate depends on the number of men in the poll. In
this example, if there are 500 men, the table below shows that the sampling error is plus
or minus 6.0%.
Respondents In Base of
Percentage Error Due To Sampling (+/-)
100 13.0%
250 8.0
500 6.0
750 5.0
1000 4.0
1500 3.5
2000 3.0
2500 2.5
5,000 2.0
7,500 1.5
10,000 1.3
12,500 1.1
15,000 1.1
31
Source: Voter News Service.
32
Appendix B: Concepts, VNS Questions, and Variable Codes
Concept VNS Question Variable Code
Candidate Choice In today’s Democratic presidential primary, 1= Gore; 0 = Bradley
did you just vote for:
G Bill Bradley G Al Gore
Gore Image Is your opinion of Al Gore: 1 = favorable; 0 = unfavorable
G Favorable G Unfavorable
Bradley Image Is your opinion of Bill Bradley: 1 = favorable; 0 = unfavorable
G Favorable G Unfavorable
Clinton Image Is your opinion of Bill Clinton as a person: 1 = unfavorable; 0 = favorable
G Favorable G Unfavorable
Clinton Performance Do you approve or disapprove of the way 1 = approve; 0 = disapprove
Bill Clinton is handling his job as President?
G Approve G Disapprove
Party Identification No matter how you voted today, do 1 = Democrat; 0 = all other
you usually think of yourself as a:
G Democrat G Independent
G Republican G Something else
Ideology On most political matters, do you 1 = liberal; 0 = all other
consider yourself:
G Very liberal
G Somewhat liberal
G Moderate
G Somewhat conservative
G Very conservative
Electability Criterion Which one candidate quality mattered most 1 = best chance; 0 = all other
in deciding how you voted? (Check only one)
G He is a strong and decisive leader
G He has new ideas
G He is not a typical politician
G He has the right experience
G He has the best chance to win in November
G He stands up for what he believes
G He is a loyal Democrat
Bradley’s Health Regardless of how you voted today, are you 1 = yes; 0 = no
concerned that Bill Bradley’s health would
interfere with his ability to serve effectively as
President?
G Yes G No
33
Appendix C: Concepts, NES Questions, and Variable Codes
Concept NES Question Variable Code
Candidate Choice Who did you vote for? (v001249) 1= Gore; 0 = Bush;
all other = missing
Gore Image Gore Feeling Thermometer (v000360) 1 = favorable (51 - 100);
Note: Valid responses range from 0 degrees 0 = unfavorable (0 – 49);
to 100 degrees; 50 is neutral all other = missing
Bush Image Bush Feeling Thermometer (v000361) 1 = favorable (51 - 100);
Note: Valid responses range from 0 degrees 0 = unfavorable (0 – 49);
to 100 degrees; 50 is neutral all other = missing
Clinton Image Clinton Feeling Thermometer (v000359) 1 = unfavorable (0 - 49);
Note: Valid responses range from 0 degrees 0 = favorable (51 – 100);
to 100 degrees; 50 is neutral all other = missing
Clinton Performance Do you approve or disapprove of the way 1 = approve; 0 = disapprove;
Bill Clinton is handling his job as President? DK/NA/RF = missing
(v000339)
Party Identification Generally speaking, do you think of 1 = Democrat;
yourself as a Republican, a Democrat, 0 = other valid responses;
an independent, or what? (v000519) DK/NA/RF = missing
Ideology Where would you place yourself on this 1 = liberal;
scale, or haven’t you thought much about 0 = moderate and conservative;
this? (v000447) DK/NA/RF = missing
Electability Who do you think will be elected 1 = Gore; 0 = Bush;
President in November? (v000485) all other = missing
Source: The 2000 National Election Study, Study Number 3131, available at
http://www.umich.edu/~nes/studyres/nes2000/nes2000.htm
34
Appendix D. Multivariate Analyses for 2000 Democratic Presidential Primaries
NH CT GA ME MD MA MO NY OH RI VT Pooled
Gore Image .43* .44* .40* .37* .42* .46* .29* .43* .49* .35* .43* .42*
Bradley Image -.27* -.27* -.25* -.36* -.29* -.25* -.40* -.28* -.32* -.35* -.31* -.31*
Clinton Image -.07* -.10* -.07 -.06 -.11* -.03 -.13* -.09* -.06* -.11* -.07* -.08*
Clinton Performance .06* .06* .05 .04 .04 .09* .00 .03 .02 .07 .04 .04*
Party Identification -.03 .04 .10* .04 .07* .01 .05 .07* .07* .07* .00 .04*
Ideology -.03 -.03 -.06 -.06 -.06* -.11* -.01 -.06* .01 -.12* -.04 -.06*
Electability .12* .17* .14* .17* .10* .22* .10* .16* .03 .17* .13* .14*
Bradley’s Health .16* NA NA NA NA NA NA NA NA NA NA NA
R2 .46 .44 .35 .43 .44 .46 .34 .45 .52 .48 .42 .44
S.E. est. .37 .37 .30 .38 .35 .36 .39 .36 .31 .36 .38 .36
N 732 860 604 581 881 883 566 1317 704 538 621 8298
Percent Predicted 79.4 78.3 88.8 78.6 82.5 81.7 79.2 80.3 87.3 83.6 77.5 80.1
Notes: NA = variable not available. Pooled results were created by merging all cases into a single data file.
*p ≤ .05
35
1 President Ford was probably hurt by his pardon of Richard Nixon.
2 See the March 2001 symposium on “Election 2000 Coverage” in PS: Political Science
and Politics 34 (1). Available from http://www.apsa.net.org/ps/march01/.
3Specific policy issues are not examined. Policy issues have not mattered to primary
voters (cf., Wattier 1990). The exit-poll questionnaires include only a few policy issues.
Ideology might capture, to some extent, the effects of policy issues. See Weisberg and
Hill (2001) for the effects of policy issues in the general election.
4 See Parent, Jillson, and Weber (1987) for a study of other kinds of contextual factors.
5 Primary voters also had a choice—whether, and to what extent, to give Gore credit for
the accomplishments of the Clinton administration.
6 Vice Presidents sometimes run as surrogate incumbents, as did George Herbert Walker
Bush in 1988; but they sometimes attempt to run as non-incumbents, especially under
unfavorable political conditions (e.g., Hubert H. Humphrey in 1968). Some analysts
have suggested that surrogate incumbency is not a realistic possibility—the electoral
fortunes of Vice Presidents are not enriched by a good economy (Norpoth 2001). Other
analysts have asserted that Gore would have won the general election if only he had
campaigned more effectively as a surrogate incumbent (Ceaser and Busch 2001; Pomper
2001). Electing the President 2000: The Insider’s View indicates that Gore’s consultants
were quite concerned with how to spin the “Clinton Factor” (i.e., popular evaluations of
Clinton as a person) so as to gain Gore some tactical advantages. Presenting Gore as
only a surrogate incumbent, with nothing else to offer, was not a viable option (Jamieson
and Waldman 2001).
7 Mayer, to cite another authority, labeled Vice President Gore “a candidate . . .
thoroughly identified with the incumbent administration” (Mayer 2001, 33).
8 Other Democrats mentioned as possible candidates were Warren Beatty, Dick Gephardt,
Jesse Jackson, Bob Kerrey, John Kerry, and Paul Wellstone. All but Beatty announced in
late 1998 or early 1999 that they would not seek the nomination. Beatty’s candidacy was
probably just another way to promote his movie career and his recent movie, Bullworth
(Germer 2001).
9 Visit http://www.pollingreport.com/nh.htm for these tracking-poll results.
10 Our analysis of this question is limited by the fact that Voter News Service asked about
Bradley’s heart problem in only two states, New Hampshire and Delaware.
11This view of electability also presumes that primary voters, perhaps with the aid of
Gallup’s trial-heat polls, can judge how well their preferred candidate would do in the
general election against the viable candidates in the other party.
36
12Students of nomination politics have used the term electability to refer to voting for the
expected winner by primary voters. Electability is similar to the bandwagon effect, a
phrase commonly used to refer to voting for the expected winner by general-election
voters (e.g., Schmitt-Beck 1996).
13 A previous study indicated that ideology had significant effects even when other
factors such as electability, candidate images, party identification, and incumbent
popularity were controlled (Wattier 1990).
14 An ideological identification might represent a mutual bond of identification among
leaders and followers of a particular group (Conover and Feldman 1981). An ideological
vote may simply indicate group support; the electoral outcome, which group should rule.
15 Stone, Rapoport, and Atkeson (1995) have developed a two-stage model in which party
identification is one of the criteria used in the first stage to reduce the large, initial pool of
candidates down to a smaller set that is evaluated in terms of utility. It would seem that
the first stage they have proposed is unnecessary when winnowing of candidates occurs
before Iowa and New Hampshire.
16Although made during a debate in the 1980 general-election campaign, Reagan’s quip
to John Anderson (“John, do you really prefer Carter to me?”) is, perhaps, the best
example of a partisan-loyalty appeal.
17 Comments made by William Gardner, Secretary of State for New Hampshire since 1976, to a
reporter (Wickham 2000) reveal how much mischief is possible:
New Hampshire’s own voter registration process is a mix of open and closed systems,
Gardner explained. “We require the person to declare a party affiliation, but then we
allow the person on the way out to undeclare,” he said.
Technically, he said, when someone declares a party affiliation, you are agreeing to
support the “principals and candidates” of that party. “But no one really asks that
question,” Garner admitted.
So independents are permitted to vote in either primary, by walking in on election day and
asking to be switched from “undeclared” to a particular party. “Then on the way out you can say,
“`I’d like to go back to being an independent; I’ve been a party member long enough’,” Gardner
said.
Of the 64,000 independents who voted in the 1996 Republican primary, he said, “41,000 on the
way out filed a card and said ‘I want to go back’.”
18 Visit http://www.mich.edu/~nes/studyres/nes2000/nes2000.htm for a detailed
description of this national, pre-and-post-election study.
19Exit-poll questionnaires have a dynamic quality unlike NES or GSS questionnaires.
The content of the VNS questionnaire seems to respond to breaking events in the
campaign. For example, after Bradley and McCain ended their nomination campaigns on
March 9, measures of voters’ images of them were dropped from the questionnaire.
37
20Logistic regression has its own limitations, especially the inherent difficulty of
interpreting its coefficients. Since the logistic regression model is nonlinear, to quote
Long (1997, 61), “no single approach to interpretation can fully describe the relationship
between a variable and the outcome probability.”
21 This expectation is based on prior research and the fact that ninety-two percent (91.6%)
of partisans (strongs, weaks, and leaners) who approved of their party’s incumbent
president voted for the incumbent in presidential elections from 1972 to 2000 (Author’s
analysis, American National Election Studies Cumulative Data File, 1948-2000, Study
No. 8475).
22 The complete results of the logistic-regression analyses are available from the author
upon request.
23 This analysis assumes that the relative magnitudes of the beta weights indicate the
relative effects of the predictor variables. The results for ideology, electability criterion,
and party identification are reported in Appendix D.
24CLARIFY is described in King, Tomz, and Wittenberg (2000) and is available at
King’s web site, http://Gking.Harvard.Edu.
25 These figures are also from the Political Money Line web site.
