Wishful Thinking in the 2008 U.S. Presidential Election
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Research Article
Psychological Science
Wishful Thinking in the 2008 21(1) 140–146
© The Author(s) 2010
Reprints and permission: http://www
U.S. Presidential Election .sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797609356421
http://pss.sagepub.com
Zlatan Krizan, Jeffrey C. Miller, and Omesh Johar
Iowa State University
Abstract
In elections, political preferences are strongly linked to the expectations of the electoral winner—people usually expect their
favorite candidate to win. This link could be driven by wishful thinking (a biasing influence of preferences), driven by a biasing
influence of expectations on one’s wishes, or produced spuriously. To examine these competing possibilities in the 2008 U.S.
presidential election, a longitudinal study assessed uncommitted young voters’ electoral expectations and preferences over
four time points during the month before the election. The findings indicated clear support for wishful thinking: Over time,
people’s preferences shaped their expectations, but the reverse was not the case. Moreover, these relations were larger among
those more strongly identified with their political party and held even when perceptions of general candidate popularity were
taken into account. Finally, changes in electoral expectations were consequential, as they shaped disappointment in the electoral
results even after taking candidate preferences into account.
Keywords
optimism, predictions, election, motivated reasoning, likelihood judgment
Received 3/25/09; Revision accepted 5/21/09
My friends, we’ve got them just where we want them electoral ones (e.g., Babad, 1997). This conclusion is all the
more appealing given what we know about motivated reasoning;
—John McCain during the 2008 presidential campaign, there is overwhelming evidence that people’s beliefs and conclu
Virginia Beach, VA, October 13, 2008 (Mooney, 2008) sions can be influenced by their motives and goals in various
ways (e.g., Kunda, 1990). Moreover, at least some laboratory
Ask a voter during a political campaign who he or she thinks research has directly implicated desirability of outcomes as a
will win the election, and the answer should not be surprising. casual factor in shaping outcome expectations (e.g., Irwin, 1953).
People are usually confident their candidate will win! This Although seductive, the conclusion that wishful thinking is
observation was first documented by Hayes (1936) who, in the responsible for preference-expectation links is unwarranted
1932 U.S. presidential election, observed that 93% of Roose for two main reasons. First, the experimental support for desir
velt supporters predicted Roosevelt would win, whereas 73% ability bias in predictions has been somewhat elusive—it is
of Hoover supporters predicted Hoover would win. Clearly, confined mostly to outcome predictions in games of chance,
voters’ preferences were foretelling of their electoral predic and there is a lack of evidence for wishful thinking in predic
tions. Since then, links between preferences and expectations tions involving more naturalistic settings (see Krizan & Wind
have been documented with regard to a variety of social and schitl, 2007, for a review). In short, the experimental evidence
political events (e.g., Cantril, 1938; Granberg & Brent, 1983; on wishful thinking seems too equivocal for one to infer that
Olsen, 1997) and constitute one of the most robust findings in preference-expectation links are necessarily driven by the
social psychology. Electoral expectations in particular show biasing influence of wishes or desires.
strong links with preferences, with correlations usually Second, there is an obvious alternative to the notion that
exceeding .50 (see Granberg & Brent, 1983). preference-expectation links are formed by the causal
What can account for this close correspondence of outcome
preferences and relevant expectations? It might be tempting to
Corresponding Author:
conclude that people believe what they want to believe, mean Zlatan Krizan, Department of Psychology, W112 Lagomarcino Hall,
ing that their rose-colored glasses prevent them from forming Iowa State University, Ames, IA 50011
accurate expectations about relevant outcomes, in this case E-mail: zkrizan@iastate.eduWishful Thinking in the 2008 Election 141
influence of preferences. Namely, the reverse could be true: of winning to be higher. For such reasons, studies examining
People’s expectations could drive their preferences. In the political knowledge as a moderating factor should also assess
political arena, there are multiple ways in which this could knowledge content (e.g., how biased is it toward one’s candi
occur. Voters might shift their candidate preferences in order date) in addition to the amount of knowledge (cf. Dolan &
to support the candidate who is perceived most likely to win Holbrook, 2001), although this is typically not done (one rea
(i.e., get on the bandwagon; see Navazio, 1977; Simon, 1954). son being the complexity involved in such measurement).
Although voters might also shift their preferences to support Another possibility is that social influence from one’s peers
the candidate perceived as the underdog (e.g., Ceci & Kain, (e.g., friends and family) could drive both electoral prefer
1982), such a case would imply a negative relation between ences and inferences about the electorate as a whole (e.g.,
preferences and expectation rather than the positive one usu Fischer & Budescu, 1995). In short, given the potential for
ally observed. Finally, people’s expectations might influence reverse causality and third-variable problems, we should
their preferences outside awareness, as voters cope with the demand additional evidence before concluding that in a given
possibility of their candidate’s defeat by rationalizing in context, preference-expectation links solely (or even mostly)
advance why the unwanted candidate might not be so bad after reflect wishful thinking.
all (see Kay, Jimenez, & Jost, 2002). Together, this evidence
indicates there are numerous ways in which expectations
could influence preferences rather than only wishes and pref A Longitudinal Approach
erences influencing expectations. An effective way to examine causality in real-world settings
In sum, by themselves, preference-expectation links are that at least partially mitigates these problems is to use longi
insufficient to infer that wishful thinking is operating, even tudinal (i.e., cross-lagged panel) designs (see Finkel, 1995;
though such an assumption is often made. In fact, many analy Locascio, 1982). Such designs increase our confidence in
ses have examined additional factors that could suggest moti causal conclusions to the extent we can demonstrate one vari
vated reasoning is at work. For example, Babad (1997) able to predict another at a future point in time even after con
demonstrated a higher preference-expectation link among trolling for both variables at initial assessment, and we can do
those more strongly committed to a given candidate and lower so over multiple assessment intervals. As mentioned before,
links among those who were given monetary incentives for experimental findings do provide some support for wishful
accurate predictions or were able to accurately report current thinking in predictions (Krizan & Windschitl, 2007), but these
polls. Furthermore, Dolan and Holbrook (2001) examined findings speak to how variables interrelate in principle, rather
National Election Study data from 1984 to 1996 and showed than how they function in real-world settings (see Henshel,
that preferences are less predictive of expectations among 1980). The same can be said for the potential causal role of
those with more political knowledge, supporting their case that expectations—most research supporting bandwagon and
more accurate knowledge reduces the influence of one’s wishes. underdog effects has been experimental (e.g., Ceci & Kain,
Granberg and Brent (1983) examined similar national data for 1982; Navazio, 1977).
the elections of 1952 to 1980 and again found preference- Investigators have already used longitudinal approaches to
expectation links to be highest among those who were highly examine these causal issues. Lazarsfeld, Berelson, and Gaudet
invested but poorly informed. (1948) examined 1940 national election data that included
All of these findings are consistent with the wishful-thinking assessment of preferences (i.e., voting intentions) and expec
hypothesis and the general understanding of motivated tations (i.e., predictions of the winner) at two different time
cognition—desirability biases should be stronger among those points before the election. They concluded that both causal
more invested in the outcome (e.g., Krizan & Windschitl, pathways were operating; there was evidence both for wishful
2007), motivation to be accurate or accountable should reduce thinking and bandwagon effects. Granberg and Brent (1983)
judgment biases (e.g., Tetlock & Kim, 1987), and constraints reexamined these findings using more appropriate analyses
of reality should pose limits on motivated distortion of judg and reported additional data from the 1980 election that sug
ments (e.g., Kunda, 1990). However, the same findings can gested only a causal effect of preferences. However, they also
again be explained in alternative ways that do not necessitate used only two time points, and barring replication, the rele
a causal influence of preferences. The critical problem involves vance of this finding for contemporary political environments
the possibility that acquired knowledge about a candidate is unclear.
drives both preferences and expectations regarding that candi
date, effectively producing preference-expectation links (i.e.,
a third-variable explanation; see Krizan & Windschitl, 2007). Purpose
For example, if you were a young, uncommitted voter who To test for presence of wishful and expectation-driven effects
learned that Barack Obama was the first African American in the 2008 U.S. presidential election, we conducted a longitu
president of the Harvard Law Review (a very prestigious legal dinal investigation involving uncommitted student voters in
publication), it would make sense for you both to develop a which preferences and expectations were assessed at four dif
more positive attitude toward him and to anticipate his chances ferent time points over 1 month before the election. Although142 Krizan et al.
these young adults were not necessarily representative of the Table 1. Characteristics of the Sample
general voting population, there were several key advantages
Variable Percentage
of the approach we used. First, young voters are an increasingly
important demographic in political elections, playing a critical Gender (n = 148)
role in the election of Barack Obama (Tufts University, 2008). Male 32.4
Second, by focusing on uncommitted voters, we were able to Female 67.6
examine dynamics in preferences and expectations that would Party affiliation (n = 152)
Democrat 32.2
be difficult to capture among more typical voters whose politi Republican 21.7
cal preferences tend to be relatively fixed. Third, focusing on Independent, other 46.1
the month before the election enabled capturing those last shifts Ideological orientation (n = 152)
in opinions that typify uncommitted voters considered very Liberal 45.3
important in shaping election outcomes (e.g., Fenwick, Wise Moderate 25.0
man, Becker, & Heiman, 1982). Fourth, we assessed both pref Conservative 23.7
erences and expectations as continuous variables, enabling us Don’t know 5.8
to examine more fine-grained changes in preferences or expec
tations not possible with American National Election Studies
data (www.electionstudies.org), where both are measured as
binary variables. Finally, given that Iowa holds the first caucus October 16. To assess basic background information, we asked
in the election season, students in this state were constantly participants to indicate their party affiliation (Democrat,
exposed to information about the election and had multiple Republican, or Independent/Other) and general political lean
opportunities to learn about the candidates. ings (rated on a scale ranging from 1, extremely liberal, to 7,
We also assessed individuals’ perceptions of poll results extremely conservative). They also indicated social identifica
and favorability of candidates’ media coverage at each time tion with the more preferred party on a social identity measure
point. To the extent that preferences are found to shape expec developed by Greene (1999). This measure requires responses
tations over time even after taking into account perceptions of to 10 statements (e.g., “This group’s successes are my suc
candidates’ general popularity, arguments about such variables cesses” or “The limitations associated with this group apply to
driving preference-expectation dynamics over time become me also”), rated on a 4 point scale, ranging from disagree
less compelling. In addition, we examined to what extent iden strongly, 1, to agree strongly, 4 (α = .87). Finally, participants
tification with a preferred party might moderate wishful think reported their age, gender, ethnicity, and ACT score.
ing, that is, the influence of preferences on expectations that Participants’ mean age was 21 years (SD = 2.5), and their
occurs over time. Although this approach is a conceptual par mean ACT score was 24.7 (SD = 5.0). As is typical of college-
allel of examining political investment as a moderating factor age populations in the Midwest, most individuals were Cauca
of preference-expectation links mentioned earlier (e.g., Gran sian (87%) and expressed liberal leanings (see Table 1).
berg & Brent, 1983), in this case it would allow for a firmer Importantly, after the election, 83% of participants indicated
test of wishful thinking given the longitudinal nature of the they had voted, with 67% of those having voted for Barack
design. Finally, to establish the importance of changes in elec Obama.
toral expectations, we tested whether these expectations
uniquely predicted elation or disappointment with the election
outcome after taking candidate preferences into account. Procedure and measures
Critical measures in the initial survey were as follows. Expec
tations were assessed first with the question, “Who do you
Method think will win the U.S. presidential election this year?” which
Participants was rated on a 7-point scale ranging from definitely Barack
Obama, 1, to definitely John McCain, 7, with the additional
One hundred fifty-three students from Iowa State University option of someone else (which no one chose). Preferences
served as participants. Students were encouraged to partici were assessed with the question, “Who will you vote for in the
pate if they were interested in the election and were not fully upcoming presidential election?” rated on a 7-point scale
committed to a candidate; 57% did so in exchange for $20, ranging from I will definitely vote for Obama, 1, through Don’t
whereas others participated as one means to earn extra credit know who will I vote for, or if I will vote at all, 4, to I will defi-
for a psychology course. Retention (87% at completion) was nitely vote for McCain, 7. Next, to measure perceptions of can
facilitated by making the full reward contingent on continued didates’ general popularity—which served as covariates in the
participation. All participants were recruited during Septem main analyses—we asked participants, “Which of the presi
ber 2008. dential candidates do you think is currently leading in the
Participants were e-mailed a link to an initial Web survey national polls?” (response options ranged from Obama to
during early October of 2008, which they all completed before They are tied to McCain) and “According to what you haveWishful Thinking in the 2008 Election 143
personally experienced, how would you characterize recent 4.5 Democrat Republican
(i.e., over the last week) media coverage of the presidential Independent
candidates?” which participants rated from strongly favoring 4.0
Electoral Expectation
Obama, 1, to strongly favoring McCain, 7.
3.5
These measures were administered via the Web again over
three time points in roughly weekly intervals, with the vast 3.0
majority of the responses being made during the weekends
of October 18 and October 25 and after the weekend of 2.5
November 1. We used partial counterbalancing to control for
2.0
possible order effects. Half of the respondents first provided
expectations on the first follow-up, with the order switched on 1.5
each subsequent assessment relative to the other half of
respondents. Perceptions of polls and media coverage were 1.0
always assessed last. Finally, the day after the election, we Time 1 Time 2 Time 3 Time 4
assessed whether participants actually voted and asked them Assessment Point
to rate how satisfied they were with the election result on a Fig. 1. Democrats’, Republicans’, and Independents’ predictions of the
scale ranging from very elated, 1, to very disappointed, 7. winner of the 2008 U.S. presidential election as a function of time. Participants
were asked, “Who do you think will win the U.S. presidential election this
year?” They responded using a 7-point scale ranging from definitely Barack
Analytic strategy Obama, 1, to definitely John McCain, 7. The value of 4 denotes indifference. The
assessment points were approximately early October (Time 1), October 18
We used structural equation modeling by means of Mplus (Time 2), October 25 (Time 3), and November 1 (Time 4).
(Version5.0; Muthén & Muthén, 2007) to test our main hypoth
eses. Specifically, we fitted a path model to the longitudinal
data where (a) all repeated measures were allowed to correlate Preference-expectation dynamics
with one another at each time point, (b) all were regressed on
themselves at the previous time point (i.e., stability relations), Figure 3 presents the path model of relations among prefer
and (c) preferences and expectations were regressed on each ences, expectations, and perceptions of general candidate
other at the previous time point (i.e., cross-lagged relations). favorability over time. The model fit was acceptable, χ2(77, N =
As this model did not quite reach acceptable standards of fit, 153) = 144.70, p < .001, CFI = .94, RMSEA = .076, 90% con
comparative fit index (CFI) < .90, root-mean-square error of fidence interval (CI) = [.057, .095], standardized root-mean-
approximation (RMSEA) > .10, we examined modification square residual (SRMR) = .086. Excluding initial assessments,
indexes, which suggested that freeing a single regression path
from preferences at Time 2 to preferences at Time 4 would
lead to considerable improvement in model fit. The model pre Democrat Republican
sented next includes this modification. Note that none of the 4.5 Independent
substantive results differed between the prespecified and the
modified model. 4.0
Voting Intentions
3.5
Results
3.0
Mean-level changes
2.5
Although not the focus of the present analysis, it is also
informative to consider how the overall level of people’s 2.0
preferences and expectations changed over time. As seen in 1.5
Figure 1, most people were aware that Obama was more
likely to win and recognized that his chances improved even 1.0
further as Election Day drew nearer (which was widely Time 1 Time 2 Time 3 Time 4
reported in the media at the time). However, it is also appar Assessment Point
ent that sympathizers of both parties grew somewhat more Fig. 2. Democrats’, Republicans’, and Independents’ voting intentions for
optimistic a couple of weeks before the election, although in the 2008 U.S. presidential election as a function of time. Participants were
the case of those identified as Republicans this still meant asked, “Who will you vote for in the upcoming presidential election?” They
responded using a 7-point scale ranging from I will definitely vote for Obama, 1,
that Obama was perceived as slightly ahead. In terms of pref to I will definitely vote for McCain, 7. The assessment points were approximately
erences, there did not seem to be any substantial changes early October (Time 1), October 18 (Time 2), October 25 (Time 3), and
over time (see Fig. 2). November 1 (Time 4).144 Krizan et al.
Time 1 Time 2 Time 3 Time 4
.50*
Candidate .88* Candidate .89* Candidate .47* Candidate
Preference Preference Preference Preference
.16† .25* .27*
.05 .11 .05 .18*.19 −.04 .09 .07 .12 .10 .03
.37 .35 .45 .88
Media Media Media Media
.56* Perception Perception Perception Perception
Poll .18 Poll .27 Poll .40 Poll
.14 Perception Perception Perception .26 Perception
.32 .00 .12 .22 .32 .19
.20 .00 .25 .01 .28 .03 .27
Electoral .33* Electoral .52* Electoral .51* Electoral
Expectation Expectation Expectation Expectation
Fig. 3. Path model showing the relations among candidate preferences, electoral expectations, and media and poll perceptions of general candidate
favorability over time. Standardized path coefficients are shown. Critical paths are shown in black; paths involving covariates are shown in gray. The
assessment points were approximately early October (Time 1), October 18 (Time 2), October 25 (Time 3), and November 1 (Time 4). Significant and
marginally significant paths are marked, †p < .10, *p < .001.
the R2 for preferences and expectations ranged from .20 to .91. Δχ2(3, N = 151) = 5.25, p = .15, there was clear evidence for
Inspection of paths in Figure 3 reveals the standard finding moderation. As seen in Figure 4, cross-lagged wishful-thinking
that electoral expectations show more fluctuations than do paths were substantially greater among those more identified
voting intentions, although both showed substantial stability with their preferred party across the first (z = 1.90, one-tailed
over time. Also as would be expected, expectations tended to p = .03) and third intervals (z = 1.90, one-tailed p = .04), with
show higher relations with poll and media perceptions than did the exception of the middle interval (z = .22). Taken together
preferences. with the results from the general model including covariates,
There was clear evidence for wishful thinking—candidate these findings support the conclusion that voter’s candidate
preferences consistently predicted electoral expectations at preferences biased their expectations.
subsequent assessment, even after taking into account previ
ous expectations and perceptions of general candidate favor
ability. Furthermore, these relations became stronger as Consequences of electoral expectations
Election Day approached, despite the fact that electoral expec To examine whether electoral expectations had unique utility in
tations became more stable. predicting reactions to the election outcome, we simultaneously
To test whether wishful thinking was more robust among regressed postelection disappointment on preferences and
those who more strongly identified with their preferred party, expectations assessed just before the election (at Time 4). As
we split the sample into two subsamples based on whether typically observed, preferences were a strong predictor of post
their average score on the social identity measure (Greene, election disappointment (β = 0.77, p < .001), such that those
1999) was above the minimum scale value reflecting disagree who favored McCain were much more likely to indicate disap
ment regarding identification. Those scoring at this value (2.1) pointment with the election outcome. There was also a modest
or higher were considered to identify highly (n = 91), whereas tendency for expectations to exacerbate disappointment (β =
those scoring lower than this value were considered to identify 0.11, p = .07), so that those more optimistic about McCain’s
low (n = 60). Next, using multiple-groups analyses, we fitted chances tended to experience marginally more disappointment.
the same path model described earlier, with the exception of
covariates (given the low sample size, we focused only on the
fundamental part of the model). This model also fit the data Discussion
reasonably well, χ2(22, N = 151) = 37.25, p = .022, CFI = .98, In the end, how can we account for links between electoral
RMSEA = .096, 90% CI = [.036, .15], SRMR = .032. To deter preferences and expectations? Experimental literature sug
mine whether party identification moderated the strength of gests potential for multiple causal influences, but both correla
wishful thinking, we tested an alternative model where the tional and experimental research to date is inconclusive
cross-lagged wishful-thinking paths were fixed across both regarding which influences are the most powerful or the most
groups, χ2(25, N = 151) = 42.50, p = .016, CFI = .98, RMSEA prevalent. Among young voters that we followed over the
= .096, 90% CI = [.042, .14], SRMR = .06, and then examined month preceding the 2008 U.S. presidential election, there was
difference in chi-squares and strengths of wishful-thinking clear support for wishful thinking—electoral preferences con
paths. Although the inferential test did not reach significance, sistently influenced expectations across all the three timeWishful Thinking in the 2008 Election 145
0.5 High (n = 91) Low (n = 60) The present investigation produced compelling evidence for
wishful thinking in a consequential, real-world setting and
0.4
Standardized Path Coefficient (β)
should thus be encouraging to researchers examining desir
0.36
ability biases in forecasts.
0.3 0.27 We also need to ask how candidate preferences influenced
0.24 electoral expectations. On one hand, preferences might have
0.2 0.2
influenced voters’ expectations at the time of assessment,
0.1 0.09 for example, through selective accessibility of preference-
0 consistent information or differential evaluation of favorable
and unfavorable information (see Krizan & Windschitl, 2007,
−0.1 for discussion of mediating mechanisms). Given the brief
−0.15 nature of laboratory studies, these types of accounts are gener
−0.2
ally considered most relevant for experimental research on
−0.3 desirability bias. On the other hand, in complex dynamic envi
ronments such as election campaigns, it is likely that candidate
−0.4 preference influenced how voters chose, attended to, and
Interval 1 Interval 2 Interval 3
remembered relevant information over time. Although it was
Assessment Interval initially thought that partisan preferences influenced relevant
Fig. 4. Results of the cross-lagged analysis testing the effect of wishful beliefs mainly through selective exposure to congenial infor
thinking over time. The graph shows standardized coefficients representing mation, empirical evidence suggests that to be a minor factor;
the relation between candidate preference at one assessment point and rather, political preferences mainly bias interpretation of infor
electoral expectations at the subsequent assessment point. Interval 1 refers
mation (see Taber, 2003). For example, individuals tend to
to paths from Time 1 (early October) to Time 2 (October 18), Interval 2
refers to paths from Time 2 to Time 3 (October 25), and Interval 3 refers hold higher standards of proof for arguments they deem unpal
to paths from Time 3 to Time 4 (November 1). Results are shown separately atable (e.g., Edwards & Smith, 1996). In sum, it is the iterative
for individuals who reported high and low social identification with their nature of information processing in real-world environments
preferred party. Error bars indicate standard errors.
that is likely responsible for why preferences influence expec
tations over time, as voters perceive relevant information and
intervals. On the other hand, there was no evidence that expec filter it through their partisan lenses (as those more identified
tations influenced electoral preferences. Although bandwagon, with their preferred party seemed to do in the current data). It
underdog, and rationalization effects have empirical support will rest on future research to conduct more fine-grained anal
from field and laboratory experiments (e.g., Ceci & Kain, yses of such processes in real-world environments.
1982; Kay et al., 2002; Navazio, 1977), there was no evidence
for these tendencies in our sample. Moreover, wishful thinking
over time was obvious even after taking into account voters’ Implications for political behavior
perceptions of how well candidates fared in the current polls Note that our results reflect the psychology of a younger group
and media coverage. Although these perceptions do not of uncommitted voters from a state (Iowa) very much informed
exhaust all possible exogenous factors that could produce about political campaigns. In this sense, there are clear uncer
preference-expectation links, they do reflect important sources tainties about the extent to which dynamics observed here
of information for voters. Finally, wishful thinking was espe characterize the voting public in general or other young voters
cially likely among those who strongly identified with their in particular. However, these findings do speak about political
preferred party. Taken together, these findings make for the perceptions of a very important demographic (young adults;
most compelling case to date that electoral expectations are see Tufts University, 2008) during a critical time in an elec
indeed driven by political preferences. These expectations, in tion, namely that final month (e.g., Fenwick et al., 1982). Note
turn, influence reactions to the election, so that above and also that these data mirrored several patterns generally
beyond the influence of political preferences, optimism about observed in national polling data—there was a clear realiza
the losing candidate can also exacerbate disappointment. tion that Obama was more likely to win as Election Day drew
nearer, and partisans generally became more optimistic regard
ing their candidate regardless of which one they supported (cf.
Implications for psychological theory Gallup Poll, 2009). In this sense, the sample seemed psycho
Whereas the notion that people engage in wishful thinking logically representative of a typical young voter, also reflected
about the future has been largely accepted as a truism, empiri in the fact that they favored Obama at about 2 to 1.
cal evidence to date has been modest. As we have argued, Also important was the observation that electoral expecta
preference-expectation links in themselves are unsatisfactory tions played a unique role in exacerbating disappointment with
evidence, and laboratory research has been limited and often the electoral results among McCain supporters. Although this
confined to artificial settings (Krizan & Windschitl, 2007). effect was modest, it suggests that campaigns’ efforts to boost146 Krizan et al.
optimism during the final preelection weeks might have negative Granberg, D., & Brent, E. (1983). When prophecy bends: The preference–
consequences down the line. In other words, rousing overopti expectation link in U.S. presidential elections, 1952–1980. Jour-
mism among the candidates’ supporters can exacerbate negative nal of Personality and Social Psychology, 45, 477–491.
reactions should the candidate lose. Such reactions could have Greene, S. (1999). Understanding party identification: A social iden
negative consequences on civic behavior, such as promoting tity approach. Political Psychology, 20, 393–403.
unwillingness to support the ultimate winner or increasing disen Hayes, S.P., Jr. (1936). The predictive ability of voters. Journal of
gagement from the electoral process (cf. Tykocinski, 2001). Social Psychology, 7, 183–191.
In sum, the results suggest that wishful thinking during Henshel, R. (1980). The purpose of laboratory experimentation and
elections is likely to be a double-edged sword. Although the virtues of deliberate artificiality. Journal of Experimental
increased optimism regarding one’s candidate might motivate Social Psychology, 16, 466–478.
behavior that will help get that individual elected (a primary Irwin, F.W. (1953). Stated expectations as a function of probability
reason why campaigns generally foster optimism among their and desirability of outcomes. Journal of Personality, 21, 329–
supporters), it also might lead to complacency before the elec 335.
tion (e.g., “my candidate will win anyway, so there is no need Kay, A.C., Jimenez, M.C., & Jost, J.T. (2002). Sour grapes, sweet
to vote”) or negative reactions after elections that could lead to lemons, and the anticipatory rationalization of the status quo.
antagonism toward the elected leader (e.g., “he bought the Personality and Social Psychology Bulletin, 28, 1300–1312.
election”) or the civic process itself. Krizan, Z., & Windschitl, P.D. (2007). The influence of outcome
desirability on optimism. Psychological Bulletin, 133, 95–121.
Acknowledgments Kunda, Z. (1990). The case for motivated reasoning. Psychological
We thank Doug Bonett, Fred Lorenz, Meifen Wei, and Gary Wells for Bulletin, 108, 480–498.
their helpful comments on the manuscript. Lazarsfeld, P., Berelson, B., & Gaudet, H. (1948). The people’s
choice: How the voter makes up his mind in a presidential cam-
Declaration of Conflicting Interests paign. New York: Columbia University Pres.
The authors declared that they had no conflicts of interests with Locascio, J.L. (1982). The cross-lagged correlation technique:
respect to their authorship and/or the publication of this article. Reconsideration in terms of exploratory utility, assumption spec
ification and robustness. Educational and Psychological Mea-
References surement, 42, 1023–1036.
Babad, E. (1997). Wishful thinking among voters: Motivational and Mooney, M. (2008). Obama takes double-digit lead over McCain.
cognitive influences. International Journal of Public Opinion ABC News. Retrieved March 17, 2009, from http://abcnews.go
Research, 9, 105–125. .com
Cantril, H. (1938). The prediction of social events. Journal of Abnor- Muthén, L.K., & Muthén, B.O. (2007). Mplus (Version 5.0) [Soft
mal and Social Psychology, 33, 364–389. ware]. Retrieved from http://www.statmodel.com/
Ceci, S.J., & Kain, E.L. (1982). Jumping on the bandwagon with the Navazio, R. (1977). An experimental approach to bandwagon
underdog: The impact of attitude polls on polling behavior. The research. The Public Opinion Quarterly, 41, 217–225.
Public Opinion Quarterly, 46, 228–242. Olsen, R.A. (1997). Desirability bias among professional investment
Dolan, K.A., & Holbrook, T.M. (2001). Knowing versus caring: The managers: Some evidence from experts. Journal of Behavioral
role of affect and cognition in political perceptions. Political Psy- Decision Making, 10, 65–72.
chology, 22, 27–44. Simon, H.A. (1954). Bandwagon and underdog effects and the pos
Edwards, K., & Smith, E.E. (1996). A disconfirmation bias in evalua sibility of election predictions. The Public Opinion Quarterly, 18,
tion of arguments. Journal of Personality and Social Psychology, 245–253.
71, 5–24. Taber, C.S. (2003). Information processing and public opinion. In
Fenwick, I., Wiseman, F., Becker, J.F., & Heiman, J.R. (1982). Clas D.O. Sears, L. Huddy, & R. Jarvis (Eds.), Oxford handbook of
sifying undecided voters in pre-election polls. The Public Opin- political psychology (pp. 433–476). Oxford, England: Oxford
ion Quarterly, 46, 383–391. University Press.
Finkel, S.E. (1995). Causal analysis with panel data. Thousand Tetlock, P.E., & Kim, J.I. (1987). Accountability and judgment pro
Oaks, CA: Sage. cesses in a personality prediction task. Journal of Personality and
Fischer, I., & Budescu, D. (1995). Desirability and hindsight biases Social Psychology, 52, 700–709.
in predicting results of a multiparty elections. In J.-P. Caverni, Tufts University, Jonathan M. Tisch College of Citizenship and Public
M. Bar-Hillel, F.H. Barron, & H. Jungermann (Eds.), Contribu- Service. (2008). Youth turnout rate rises to at least 52%. Retrieved
tions to decision making—I (pp. 193–211). Amsterdam: Elsevier. February 28, 2009, from http://www.civicyouth.org
Gallup Poll, Inc. (2009). Election 2008. Retrieved March 2, 2009, from Tykocinski, O.E. (2001). I never had a chance: Using hindsight tac
http://www.statmodel.com/www.gallup.com/poll/election2008 tics to mitigate disappointments. Personality and Social Psychol-
.aspx ogy Bulletin, 27, 376–382.You can also read
