Flip-flopping
in
a
Trump
era
A comparison between the American public’s evaluations of Trump and
Bush
Lotte Schrijver (s1827103)
Bachelor Project: Foreign Policy in a Trump Era
Supervisor: Dr. DiGiuseppe
Leiden University
2
“‘I could stand in the middle of 5th avenue and shoot somebody and I wouldn’t lose any
voters’- Donald Trump, 2016” (Donovan et al, 2019, p. 1). Recent findings show that this might also be true for Trump’s flip-flopping; American citizens seem to shrug their shoulders at Trump’s frequent inconsistencies (McCaul et al, 2019). However, in political campaigns, candidates often accuse their opponents of flip-flopping, which suggests that a politician’s inconsistencies electorally hurt him. The aim of this thesis is to solve this puzzle. Does flip-flopping hurt president Trump? Does polarization in the current political environment matter for how citizens punish politicians? And, as is often proposed in the media, does this make Trump’s presidency different from presidencies before? Finding the answers to these questions will have implications for understanding democratic accountability; if citizens are less inclined to punish president Trump based on his acts than they were for previous presidents, it would suggest that Trump is held accountable for his acts in a fundamentally different way than the presidents before him.
Thus, in this thesis, I will attempt to answer the following research question: “To what extent do American citizens’ evaluations of Trump’s flip-flopping differ from their evaluations of Bush’s flip-flopping?” With this question, I aim to fill a gap in the literature. Specifically, the literature is still inconclusive about if and when specific leaders are punished for flip-flopping. Moreover, it is unclear if some political leaders are punished more for their inconsistencies than others. This is particularly relevant for president Trump, who has often been accused of flip-flopping (e.g. Lee, 2017, April 13). By filling this gap, this project will make at least three contributions. Firstly, it will improve our understanding of the influence of contextual factors, such as polarization, on American citizens’ evaluations of their president. By comparing president Trump to president from an earlier era Bush, it will also further academic understanding of democratic accountability in the Trump era; this thesis will have implications for how we see Trump is punished by voters and for what. Finally, by improving understanding of under what conditions leaders are punished for flip-flopping, the answer to this question will contribute to the literature on flip-flopping.
In this paper I will first describe previous research on flip-flopping and citizens’ evaluations of presidents. The literature is still inconsistent about under what conditions a president is punished for flip-flopping. After identifying the gaps in the literature, I will develop a theoretical framework; I expect Trump to be punished less for flip-flopping than Bush due to increased motivated reasoning and polarization. My expectation is that this effect is especially present with strong partisans, who are less likely to take flip-flopping into account when
3
evaluating a president. With these expectations, I will describe my ideal research design. However, due to limitations to data availability, I will answer my research question more indirectly.
By examining individual evaluations of inconsistency and a president’s overall performance, I will find that the effect of individuals’ perception of a president’s trustworthiness, as a proxy variable for consistency, on overall evaluations of the president has decreased in the Trump era, while the effect of partisanship has increased. This suggests that consistency matters less for how citizens evaluate Trump than it did for Bush. Moreover, most likely because of increased polarization and motivated reasoning, citizens are more inclined to see their president through a partisan lens. That implies that citizens are also more likely to see flip-flopping through this lens and, thus, are more likely to condemn or rationalize flip-flopping based on their partisanship. This makes Trump’s presidency fundamentally different from those before.
Literature
Much research has been done on how American citizens evaluate their president (Gronke & Newman, 2010). Overall, this research identifies four main influences of presidential approval ratings: individual perceptions of economic conditions, war deaths, major events and individual assessment of the president’s character. While major events can have a rally-around-the-flag effect, improving presidential approval ratings, a negative perception of the economy and a high number of war deaths can have negative effects on presidential approval ratings. Similarly, if an individual sees the president as having positive personality traits, he or she is more likely to approve of the president. Overall, these factors explain most of the variance of presidential approval ratings.
However, seemingly more insignificant presidential policies and actions can also have a substantial impact on presidential approval ratings. For example, Collingwood et al (2018) demonstrate, by examining the effects of the Muslim ban on Trump’s approval ratings, that specific policies can have a significant effect on presidential approval ratings. Similarly, different strands of research show that a president’s inconsistent actions can have a negative effect on citizens’ evaluations of the president (Croco, 2016).
For example, as part of the crisis bargaining model, audience cost theory assumes that citizens evaluate leaders more negatively when they act inconsistently (Fearon, 1994). Specifically, this theory proposes that democratic leaders face audience costs when they back
4
down from a threat to a foreign country. Research has found empirical evidence for this; several authors conclude that, under specific conditions, leaders are punished for backing down from a threat (e.g. Tomz, 2007; Snyder & Borghard, 2011; Levendusky & Horowitz, 2012; Kertzer & Brutger, 2016) These findings suggest that citizens are inclined to use consistency as a standard to evaluate their leader.
These conclusions seem to hold in a broader context. The literature on flip-flopping shows that politicians are also punished for inconsistent actions in other contexts (Hummel, 2010). In other words, if a politician speaks or acts inconsistently, generally, citizens are more likely to evaluate him negatively. This most often results in more negative overall approval ratings of flip-flopping politicians (e.g. Allgeier et al, 1979; Sigelman & Sigelman, 1984; Hoffman & Carver, 1984; Carson & Dolan, 1985; Tomz & Van Houweling, 2010).
These lower approval ratings seem to stem from the attribution of negative characteristics, such as unreliableness and inconsistency, to flip-flopping politicians (Carlson & Dolan, 1985). In other words, flip-flopping might cause citizens to see a politician as more inconsistent or unreliable, which in turn leads them to be less likely to approve of this politician. Indeed, previous research suggests that citizens base candidate’s and presidential approval ratings on what personality traits they think the politician has (e.g. Funk, 1999; Gronke & Newman, 2010; Greene, 2001; Buchanan, 2016; Hardy, 2017; Clifford, 2018, p. 240). For example, Buchanan finds that citizens refer to specific personality traits when evaluating the president in over 75% of the time.
These personality traits have been divided into different typologies (Bittner, 2011). Most importantly, Kinder et al (1980) have organized traits for politicians’ personality assessments into four different dimensions, respectively competence, leadership, integrity and empathy. This typology has influenced later work on this topic; many authors have tested and incorporated these dimensions into their work (Bittner, p. 38). However, because there is high correlation between the dimensions competence and leadership and between integrity and empathy, later authors have collapsed these dimensions into the categories competence and character or warmth (e. g. Goodwin, 2015; Clifford, 2018; Bittner). The latter category refers to “perceptions of another's social intentions (e.g., friendly, honest), while competence represents their ability to bring about those intentions” (Clifford, p. 241).
There is strong internal cohesion between the characteristics within these categories. For example, Bittner (2011) shows that characteristics within the character dimension, such as
5
honesty, reliability or trustworthiness, strongly correlate. This means that when citizens see a political leader as unreliable, they are also very likely to see this politician as dishonest or untrustworthy. In other words, individual perceptions of these personality traits similarly predict positive and negative approval ratings.
This implies that, as citizens see a flip-flopping politician as more inconsistent and unreliable, they also see him or her as more dishonest and untrustworthy. That means that, overall, flip-flopping can hurt citizens’ perception of a president on the dimension of warmth or integrity. Moreover, because citizens have been shown to use this dimension as a standard for presidential evaluation, approval ratings decline as more negative characteristics are attributed to a president. Consequently, by making citizens’ impression of a politician on the morality dimension more negative, flip-flopping plays a substantial role in presidential approval ratings.
However, the findings on the consequences of flip-flopping on citizen evaluations of politicians are not conclusive; some authors argue that politicians are not uniformly punished for their inconsistencies (e.g. Croco, 2016; McCaul et al, 1995). These authors diverge from the idea that flip-flopping is always punished, or even noticed, by citizens. Specifically, a politician’s flip-flopping is sometimes punished by voters, but only when voters disagree with a politician’s current position after the flip-flop. In that way, it might be rational for a leader to flip-flop in some situations. Additionally, literature on memory and information processing raises the question if voters even notice or remember a leader’s flip-flopping (Putnam et al, 2014, p. 1202). This question becomes even more prominent when one knows that many Americans do not pay attention to politics (Gilens, 2001).
This implies that the logic of citizens’ punishment for flip-flopping is not unitary, but based on particular conditions (e.g. Croco, 2016; Sorek et al, 2018). For example, Doherty et al (2015) argue that voters might be more forgiving for flip-flopping under certain conditions. Specifically, when the time period since the flip-flop has occurred has been longer, citizens are less inclined to punish their leader for flip-flopping. Moreover, Doherty et al find that issue type matters for the level of punishment; if the issue of the flip-flop is more complex, as is the case in foreign policy, the negative effect of flip-flopping on individual evaluations of the politician is weaker. Finally, Robison (2017) adds another condition to our understanding of flip-flopping; when a politician provides the public with justifications for their flip-flop, the public is less inclined to punish him or her for this inconsistent act.
6
Individual psychological processes might also matter for how much punishment a leader receives for flip-flopping. Several authors show that voters reason on the basis of preconceived ideology, political preferences and partisanship when punishing a leader for flip-flopping (e.g. McCaul et al, 1995; Doherty et al, 2015; Croco, 2016; Sorek et al, 2018). For example, Croco (2016) shows that a citizen’s reaction to flip-flopping is conditioned on their agreement with the politician’s current position, because citizens are inclined to evaluate leaders on the basis of motivated reasoning. In other words, citizens evaluate this new information (the fact that a politician has flip-flopped) from a partisan or ideological lens. This creates a heterogeneous effect of flopping: the voters who disagree with his current position punish a leader for flip-flopping, but the citizens that agree with his current position condemn him less for his inconsistency.
The effect of motivated reasoning on citizens’ evaluations should not be underestimated: it has a profound effect on all political reasoning and behavior. For example, Taber and Lodge (2012, p. 176) show that motivated reasoning fundamentally shapes how all voters process political information. Moreover, the effect of motivated reasoning on political behavior increases in times of polarization.
Specifically, citizens are more likely to evaluate political policies and leaders on the basis of preconceived opinions when polarization is high (McDonald et al, 2019; Leeper & Slothuus, 2014). Research has found that polarization in the United States has increased since the 1980s, suggesting that motivated reasoning has increasing influence on Americans’ political reasoning and behavior (Lebo & Cassino, 2007; Iyengar & Krupenkin, 2018). Citizens from the same parties progressively have the same ideas, making ideology within the parties less diverse (Donovan et al, 2019, p. 6; Schultz, 2018, p. 9). Consequently, the partisan gap widens; as citizens become
increasingly more “Democratic” or “Republican”, they often also diverge from the other party. This can be seen in graph 1: the partisan gap has grown substantially over the last 25 years.
Graph 1. Polarization in the mass public over the last 25 years (Pew Research Center, 2017, October 4)
7
This high level of ideological correlation has also resulted in increased animosity towards the other party, also called affective polarization.
Polarization has major consequences. Overall, polarization leads citizens to be less likely to evaluate a president on the basis of accuracy motivations, such as perceptions of the economy (Donovan et al, 2019). Moreover, according to Iyengar and Krupenkin (2018, p. 215), polarization has a profound impact on democratic accountability. When polarization and motivated reasoning is high, leaders are less likely to be sanctioned for unethical behavior, incompetency and, most importantly, inconsistent behavior.
Thus, there are indications that, due to recent polarization, the effect of motivated reasoning on American citizens’ political reasoning and behavior has increased. Other work by McDonald et al (2019) also points in this direction. Specifically, these authors find that president Trump is not punished for flip-flopping. When citizens, prior to the flip-flop, approve of Trump, they are willing to ignore or rationalize his flip-flopping, but if they oppose him, they are more inclined to condemn him for the inconsistency (McDonald et al, p. 3). According to the authors, the current American political environment is highly polarized, which results in citizens being more likely to evaluate Trump’s behavior on the basis of prior partisanship. This suggests that flip-flopping does not matter for Trump’s approval ratings; citizens do not evaluate the president’s actions on the basis of the inconsistency itself, but of their prior beliefs about him and their current agreement with his position (McDonald et al, p. 4).
This work implies that due to motivated reasoning and polarization, contemporary presidents are better able to get away with flip-flopping. In other words, if Trump flip-flops, it would matter less for his approval ratings than it did for previous presidents (Croco et al, 2019, March 12). However, McDonald et al (2019) do not directly test this; they do not compare citizens’ punishment for flip-flopping under different presidents. Therefore, I intend to contribute to the literature by comparing president Trump with a previous president and examining the differences in the punishments they face for flip-flopping.
With this approach, I will further refine our understanding of the consequences of motivated reasoning and polarization. For example, polarization might help explain under what conditions leaders are punished for flip-flopping. As the work above shows,leaders are only punished for flip-flopping under certain conditions, but it is still largely unclear what these conditions are. By directly comparing flip-flopping in the Trump era to flip-flopping under Bush, I attempt to further test the validity of these conditions. Moreover, my approach will give
8
a tentative answer to an important question, often posed in the media and academic literature: how different is Trump? Do Americans live in a new era of presidency and presidential evaluation? Therefore, I will attempt to answer the following research question: “To what extent do American citizens’ evaluations of Trump’s flip-flopping differ from their evaluations of Bush’ flip-flopping?”
A number of critical points regarding this research question should be made. I will compare Bush to Trump to hold as many factors, specifically partisanship, constant as possible. Moreover, because I propose that polarization will influence the punishment a leader receives for flip-flopping, it is important that the two cases have some distance in time; during Obama’s presidency, polarization might have been equally strong as under Trump’s leadership. This lack of variation would inhibit the examination of the effect of polarization. Therefore, I will compare Trump to earlier president Bush.
Ideal theory and research design
Overall, I expect the nature of presidential evaluation to have changed because of increased polarization and motivated reasoning. Most importantly, individual partisanship has been found to have a greater impact on how citizens evaluate Trump than Bush (Leeper & Slothuus, 2014; Iyengar & Krupenkin, 2018; Donovan et al, 2019). In other words, citizens in the Trump era are more likely to see Trump’s presidency from a partisan lens. Moreover, accuracy motivations, such as an individual’s perception of the economy, have a smaller effect on presidential approval ratings under Trump than under previous presidents. Similarly, because citizens increasingly perceive Trump’s presidency through a partisan lens, president Trump’s acts, such as flip-flopping, might matter less for how citizens evaluate him than it did for previous presidents.
Therefore, I expect that, after a flip-flop, evaluations of Bush to become significantly more negative, while evaluations of Trump change less. As shown above, I expect this difference to be explained by increased motivated reasoning and polarization. Because citizens increasingly see president Trump’s actions through a partisan lens compared to previous presidents, their opinions about him are less likely to change on the basis of new information, such as an inconsistent act. Therefore, I expect a flip-flop to have a small effect on Trump’s presidential approval ratings, while it would have a bigger negative effect on Bush’s approval ratings. As a consequence of polarization, I also expect that individual partisanship has a bigger effect on Trump’s approval ratings than on Bush’s. Additionally, I expect an interactive effect
9
between flip-flopping and partisan strength on approval ratings; because motivated reasoning has a stronger effect on individuals with strong partisanship, I expect them to be more likely to see a flip-flop from a partisan lens. Therefore, a flip-flop would have a smaller effect on strong partisans than on weak partisans.
In order to test these expectations, I need two datasets that include dates before, the date of, and dates after the flip-flop. Because the same questions and wordings are used for the two groups of respondents, this would create a research design similar to a natural experiment. This research design would allow me to compare the two groups with holding other factors, such as other political issues, wars and the state of the economy, relatively constant, making it the ideal way to test my hypotheses. Moreover, by finding similar flip-flops for both presidents, I could compare Trump and Bush in a similar context, for example regarding the topic of the flip-flop and the way it was framed in the media. Thus, this design would not only allow me to compare public opinion before and after a flip-flop, but also compare effects between presidents, while isolating the examined relationships as much as possible.
However, the amount of data on this topic is severely limited. Firstly, it seems that president Bush’s inconsistencies were rarely framed by the media as flip-flopping. This raises the question if Americans ever saw Bush’s changes as flip-flopping. Secondly, although American public opinion has been researched very often, there has not been a survey for every day for the past decade. Therefore, I did not have data for every date on which a flip-flop occurred. Moreover, many of these surveys were done in one day, which makes them inadequate for my research design. Due to these limitations, I could not find a data set that was suitable to test my expectations.
Theory
Therefore, I will use a different approach to answering my research question. By looking at how the impact of partisanship and individual evaluations of inconsistency on presidential approval ratings have changed between Bush’s and Trump’s presidency, I attempt to understand, although indirectly, if citizens’ evaluations are impacted by a president’s flip-flopping. As explained above, flip-flopping can cause negative approval ratings through the attribution of negative characteristics to flip-flopping politicians. By using the more indirect approach, instead of testing the effect of flip-flopping on these approval ratings, I will test the effect of a consequence of flip-flopping, the attribution of negative characteristics, on approval ratings. Consequently, finding that inconsistency has a smaller effect on Trump’s approval
10
ratings than on Bush’s might imply that citizens are less inclined to take a cause of the attribution of inconsistency, flip-flopping, into account when evaluating president Trump than they were when evaluating Bush. Similarly, finding evidence that partisanship has a stronger effect on individual approval ratings of Trump than of Bush would indicate that citizens are increasingly likely to ignore or rationalize flip-flopping, because they evaluate inconsistencies through a partisan lens.
Indeed, previous research suggests that for the contemporary American citizen, it does not matter so much what the president does and how he does it, but more if he is part of the same party. The research by Donovan et al (2019) on the declining impact of evaluations of the economy on presidential approval ratings shows that, in the Trump era, citizens are less inclined to evaluate the president on the basis of accuracy motivations, such as government policy or views of the economy. Although individuals’ perception of the economy still somewhat predicts their approval of the president, it has a smaller effect on Trump’s approval ratings than on previous presidents’ ratings. I can apply this to my research; similar to their perception of the economy, Americans might be less inclined to use personality standards, such as consistency, to evaluate their leader (Kinder et al, 1980; Gronke & Newman, 2010; Buchanan, 2016). In this way, I might find that how individuals see a president’s inconsistencies has a smaller effect on Trump’s approval ratings than on Bush’s.
Moreover, Donovan et al (2019) found that partisanship has a bigger effect on Trump’s approval ratings than on the approval ratings of previous presidents. Because of high polarization in the Trump era, motivated, partisan, reasoning increasingly shapes presidential approval ratings. This explains why accuracy motivations have a lower effect on Trump’s approval ratings; because these accuracy motivations are increasingly seen through a partisan lens, they fail to have an independent effect on Trump’s approval ratings. This does not mean, however, that partisanship had no effect on the approval ratings of previous presidents. Because, as shown above, motivated reasoning has an effect on all political behavior by all individuals, partisanship has always had an effect on presidential approval ratings (Taber & Lodge, 2012). However, although motivated reasoning is present in all individuals, it is stronger in individuals with a high level of partisanship (Leeper & Slothuus, 2014). These individuals, strong partisans, are more likely to examine new information through a partisan lens. Therefore, I expect that inconsistency has a smaller effect on presidential approval ratings in these individuals than in individuals with a low level of partisanship. Because strong partisans evaluate the president through their partisan lens, they might be less inclined to use personality
11
standards, such as consistency, as a criterium. For example, even though a strong Republican might see Trump as inconsistent, he still sees him in a positive light based on mutual partisanship. Thus, the expected relationship between perceptions of consistency and approval ratings might be weaker in individuals with strong partisanship, while it might be stronger in weak partisans.
In short, I expect to find increased polarization between the Bush and Trump presidency. Therefore, I expect that, even though consistency might still have a positive effect on both presidential approval ratings, because of polarization, it has a much smaller effect on Trump’s approval ratings. Similarly, I suggest that partisanship has a strong effect on Trump’s approval ratings. Even though I expect partisanship to predict Bush’s approval ratings reasonably well, I expect this variable to predict Trump’s approval ratings better. In order to refine my understanding of the effects of motivated reasoning on presidential approval ratings, I will also examine the relationships for individuals with different levels of partisanship. I expect a heterogenous effect of the perception of consistency on individual presidential approval ratings: perception of consistency might have a weaker effect on individual presidential approval ratings of strong Democrats and Republicans than on those of weak Democrats, weak Republicans or people with no partisan preference.Thus, I will test the following hypotheses:
H1: Individuals are more likely to evaluate Bush on the basis of consistency than
Trump.
H2: Individuals are more likely to evaluate Trump on the basis of partisanship than
Bush.
H3: Weak partisans are more likely to evaluate presidents on the basis of consistency
than strong partisans.
Alternatively, I might also find no differences between how individuals evaluate Trump and Bush. Specifically, polarization might not have occurred or the effect of polarization and motivated reasoning on evaluations of the president might not be as strong as I expect it to be. In that case, I will not find that individual partisanship has a bigger effect on evaluations of Trump than on evaluations of Bush. I would also find that evaluations of Trump are as much affected by evaluations of his consistency as evaluations of Bush. Similarly, strong partisans might not be more inclined to evaluate the president on the basis of prior beliefs than weak partisans. Then, motivated reasoning would not have a stronger effect on strong partisans and
12
the expected heterogeneous effect would not occur. In short, if I find these results, these hypotheses should be rejected.
There might also be alternative explanations for finding results that support my hypotheses. As shown above, presidential approval ratings are also affected by other variables, such as views of the economy, major events or wars (Gronke & Newman, 2010; Donovan et al, 2019). Thus, in this research, I have to take the effect of these variables on citizens’ evaluations of the president into account. Similarly, individual characteristics can also have an effect on individual approval ratings of the president. Because gender and income have been found to have an influence on individual evaluations of political leaders, I will also include these variables (Argyle et al, 2016).
In order to test these expectations, I will define the concepts in the research question and explain how they are related to the theoretical expectations. Firstly, I define inconsistency as unstable behavior, that changes situationally and temporally (Winter et al, 1998, p. 233). When evaluating a politician’s consistency, individuals might take several factors into account, such as policy, party organization, speaking style, and so on. I define flip-flopping as a particular type of inconsistency; it is an instance when a president publicly changes a policy position (Doherty et al, 2015, p. 470). Because this thesis examines public opinion, it should be taken into account that an individual might interpret these concepts differently; it is possible that an individual sees all inconsistencies as flip-flopping or sees the terms as unrelated. Secondly, American citizens’ evaluations describe their overall view of the president. In this research, this concept designates individuals’ presidential approval ratings. Because I examine the effect of consistency evaluations on evaluations of presidential overall performance, I will avoid circular reasoning by not including character traits or other more specific views of the presidency in this latter concept.
Polarization means, in this case, that citizens’ ideological views have become more consistent, while views of the out-party have become more negative and shaped by stereotypes (Iyengar & Krupenkin, 2018, p. 201). In other words, polarization is the “ ‘sorting’ of the mass public into more homogeneous parties” (Schultz, 2018, p. 9). The process of polarization implies that people and their views are increasingly in line with either the Democrats or Republicans. In this environment, individuals are more likely to evaluate presidents on the basis of motivated reasoning. Motivated reasoning occurs when an individual evaluates new information on the basis of prior beliefs. In the political environment, this often means that an individual views new information about a politician through a partisan lens (Donovan et al,
13
2019, p. 5). Thus, motivated reasoning is related to partisanship, which is the alignment of an individual with a particular political party. Partisanship is a scale; it can be strong, when an individual is strongly in favor of either the Republican or Democratic party, or it can be weak, when an individual has no clear preference for either party.
Research design
Ideally, as described above, the expectations are tested by experiments, either with a classical or natural design. Indeed, previous research on flip-flopping has often used experiments to test hypotheses. An experiment is adequate for this topic because of its high internal validity and quantitative strategy, which is suitable for comparing the presidents and testing the effect of individual partisanship on presidential evaluations. However, as my ideal research design shows, the disadvantage of experiments is that they require time and other resources which are not available to me within the scope of this research project. Therefore, other methods will be considered.
Another way of testing the hypotheses is a cross-sectional study. The advantage of this design is that, using statistics and quantified variables, presidential approval ratings can be compared in a more standardized and systematic way than with a case study. Especially public opinion can be measured and compared more easily with quantitative than with qualitative data. Moreover, several control variables can be included to isolate the examined relationship. Finally, although a cross-sectional study is lower in internal validity than an experiment, it is higher in external validity, because the effect will be measured with real-world events and evaluations. Because of these advantages, I will use this research design to test the hypotheses.
I will test the hypotheses using two datasets from the Pew Research Center. These datasets consist of survey answers of American citizens to questions about politics and the news. The first data set was conducted from October 6 to 10 in 2005 and includes the evaluations of Bush from 1,500 respondents (Pew Research Center, 2005). The second data set was conducted from September 18 to 24 in 2018 and includes the evaluations of Trump from 1,754 respondents (Pew Research Center, 2018). However, because I only have answers to some questions from part of the sample, the number of respondents is reduced to around 500 for each sample. In both data sets, I will use binary logistic regression analysis to examine the effect of evaluations of inconsistency and partisanship on presidential approval ratings. The first advantage of this data set is that, besides presidential approval, it includes many variables that specify individuals’ view on politics and specifically the president, allowing me to control
14
for important variables that might affect the examined relationship. Secondly, because both surveys have mostly used the same questions and wording, I can include the same independent and dependent variables in both analyses. Finally, these surveys were conducted at a considerably similar point in time in both presidencies. Therefore, there is no confounding effect of, for example, upcoming presidential elections.
Because I expect motivated reasoning to have a bigger effect on Trump’s approval ratings than on Bush’s, I will first test if polarization has occurred. Increased polarization with imply that motivated reasoning has an increased effect on citizens’ political evaluation. Ideally, I would test if citizens’ ideology within the Democratic and Republican party has become more homogenous or if perceptions of the other party have become increasingly negative, as polarization is defined in the relevant literature (Schultz, 2018, p. 9). However, with this dataset, this test is not available to me. Therefore, I will measure polarization by testing if the number of strong partisans has grown. A higher number of strong partisans would mean that more people have moved to either one of the partisan ‘poles’, implying that more people see the world and politics through a partisan lens. However, this does not necessarily imply that views within the party have become less diverse. Therefore, this test should be interpreted with caution.
Then, I will run two binary logistic regression analyses for both presidents. In the first analysis I will examine the effect of trust and partisanship on individual presidential approval. This produces the following equation: 𝑦𝑖 = 𝑏0+ 𝑏1𝑡𝑟𝑢𝑠𝑡𝑤𝑜𝑟𝑡ℎ𝑖𝑛𝑒𝑠𝑠𝑖+ 𝑏2𝑝𝑎𝑟𝑡𝑖𝑠𝑎𝑛𝑠ℎ𝑖𝑝𝑖+
𝜀𝑖. Thus, by testing the effect of these variables, I will test hypothesis 1 and 2. However, because
I use different data sets, I cannot produce relevant confidence intervals. Consequently, I will not be able to say with certainty that there is a difference between the presidents. In the second analysis, I will zoom in on how this relationship varies between different groups by testing the heterogeneous effect of trust on individuals presidential approval ratings for groups with different levels of partisanship. This will create the following equation: 𝑌𝑖 = 𝑏0+ 𝑏1𝑡𝑟𝑢𝑠𝑡𝑤𝑜𝑟𝑡ℎ𝑖𝑛𝑒𝑠𝑠𝑖 + 𝑏2𝑝𝑎𝑟𝑡𝑖𝑠𝑎𝑛 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ𝑖 + 𝑏3𝑡𝑟𝑢𝑠𝑡𝑤𝑜𝑟𝑡ℎ𝑖𝑛𝑒𝑠𝑠 𝑖×
𝑝𝑎𝑟𝑡𝑖𝑠𝑎𝑛 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ𝑖 + 𝜀𝑖. This analysis will test hypothesis 3.
In the first analysis, I will use the question ‘Do you approve or disapprove of the way [president] is handling his job as president?’ as the dependent variable. This variable is measured on a binary scale with disapproval as the baseline category. For the independent variable partisanship, I will use the question ‘In politics today, do you consider yourself a Republican, Democrat or Independent?’. I recoded the answers to this question into a variable
15
with three categories in which Republicans and Democrats are compared against the baseline category independent/no preference. I will also run this analysis a second time with control variables. These control variables and the wordings of all used questions can be found in the appendix.
The ideal question to measure individual views of a president’s inconsistency, was not present in the data set. Therefore, I rely on a different measure of consistency; I will use the question ‘Does [President] impress you as trustworthy or not trustworthy?’ As shown above, reliability, consistency and trustworthiness are part of the morality dimension of political traits. However, from the many definitions of trust, such as “the willingness to place its interests under the control of others”, it becomes clear that consistency and trustworthiness do not describe the exact same phenomenon (Hoffman, 2002, p. 394). Nevertheless, Bittner (2011) has shown that traits within the morality dimension have strong internal cohesion. This means that how individuals attribute these traits to a president strongly correlates. Therefore, it can be expected that these traits have a similar effect on presidential evaluations. Thus, trustworthiness is the best proxy variable to measure the effects of perceptions of consistency on presidential evaluations.
In the second analysis, I will also use the variable for trustworthiness. Moreover, I will create an interaction variable with trustworthiness and partisanship. In this analysis, individuals who initially described themselves as Republican or Democrat are strong partisans, while independents and individuals with no preference as weak partisans. Although this categorization does not distinguish between Republicans and Democrats, it does create categories of a considerable size, making it more suitable for analysis. This point will be further examined in the discussion. After creating dummy variables, I will interact this dummy variable with trustworthiness, comparing it with the baseline category weak partisans. In this way, I will see both the interacting effect of partisan strength on the relationship between trust and presidential approval.
By using this design, I will avoid some of the major risks of research on public opinion and flip-flopping. Firstly, by making inferences about individuals on the basis of aggregate approval ratings research on aggregate public opinion can sometimes risk ecological fallacy. However, I will avoid this fallacy by looking at the effect of individual perceptions on individual presidential approval ratings. Moreover, by testing the relationship between perceptions of trustworthiness and individual presidential approval ratings on different levels of partisanship, I will further tease out individual relationships.
16
Secondly, research on flip-flopping and audience costs runs should be aware of the strategic selection effect. This effect has been named as a methodological problem within audience cost theory (Schultz, 2003ta). Because leaders avoid cases in which audience costs would be high, researchers only see cases in which audience costs are relatively low (Snyder & Borghard, 2011). This problem is similar for flip-flopping; because leaders think citizens condemn flip-flopping, they avoid flip-flops that the public would be very negative about. Therefore, researchers are likely to underestimate the negative consequences of flip-flopping. However, because I do not use cases in which flip-flopping has occurred, but examine citizens’ responses to inconsistency as a general concept, I will avoid this problem.
Results
Firstly, I have looked at the numbers of strong and weak partisans to see if there is evidence of increased polarization. However, although I expected that the number of strong partisans had grown, it seems that polarization has not occurred. In the Bush dataset, the percentage of strong partisans is %60.6, while in the Trump dataset the percentage of strong partisans is %60.3. Since these percentages are not significantly different, I conclude that polarization has not occurred. These findings indicate that, overall, motivated reasoning does not have an increased impact on aggregate presidential approval ratings, because the number of people who are highly influenced by motivated reasoning does not appear to have increased. However, this does not mean that strength of partisan reasoning has not increased within individuals. Therefore, I will zoom in on the examined relationships and look at the individual impact of partisanship and trustworthiness on individual approval ratings.
I expected the effect of trustworthiness on presidential approval ratings to have weakened under the Trump presidency. At first sight, this seems to have occurred; the effect of trustworthiness has declined significantly (See table 1, columns 1a). After adding the control variables, this result holds, although the effect on both approval ratings declines significantly (see columns 1b). Specifically, the odds that someone approves of Bush when he or she finds him trustworthy are over 41 times higher than when someone finds him untrustworthy, while for Trump the odds are only 22 times higher. This implies that, while trustworthiness is still an important predictor of approval, its importance has declined significantly under the Trump presidency. Consequently, it seems that citizens are increasingly less likely to use trustworthiness as a standard when evaluating their president.
17
These findings are also reflected in the results on the impact of partisanship on individual approval ratings. I expected this effect to have increased, which, before adding the control variables, seems to be the case: an increased effect of both Republican and Democratic partisanship on individual presidential approval ratings under the Trump presidency is found. This effect holds for Democratic partisanship in model 1b, although the significance of the effect of Republican partisanship under the Trump presidency has disappeared. Specifically, the odds for a Democrat to approve of Bush were 2.7 times lower than for an Independent, but the odds that a Democrat approved of Trump were almost 7 times lower. These results suggest that, although the findings on Republicans are not conclusive, the effect of partisanship has increased since the Bush presidency. These findings are in line with the results on trustworthiness; if citizens increasingly evaluate a president on the basis of partisanship, they might be less inclined to use personality trait standards as a criterium for evaluation.
Surprisingly, the effect of perceptions of the economy are higher for Trump’s approval ratings than for Bush’s. Looking at the work of Donovan et al (2019), I would expect to find similar findings for perceptions of the economy as for trustworthiness. Since individuals increasingly see their president through a partisan lens, they would be less inclined to use standards such as the economy to evaluate their president. However, this does not seem to be the case; the effect of perception of the economy has increased.
Finally, I attempt to refine these results by looking at a heterogeneous effect of trustworthiness on individual presidential approval ratings. This will test the effect of motivated reasoning more precisely and therefore provide an explanation for the results above. I expected to find that trustworthiness has a bigger effect on the approval ratings of weak partisans. However, it is not clear from the results if this effect has occurred. Firstly, the interaction effect on Bush’s approval ratings is not significant (see column 2a). The effect is significant for Trump’s approval ratings. Specifically, when a weak partisan finds Trump trustworthy, the odds that he or she approves of Trump are 13 times higher than when he or she find him untrustworthy, while for a strong partisan the odds are over 23 times higher. Consequently, the interaction effect is in the opposite direction of what was expected. Adding control variables to this model does not significantly change this effect. These results suggest that strong partisans are more inclined to use trustworthiness as a standard to evaluate president Trump. This implies that the effect of motivated reasoning on individual presidential evaluation is different than expected.
18
Table 1. Binary logistic regression analysis of presidential approval ratings.
Note: binary logistic regression odds ratios with standard errors in brackets.
***p<0.001, **p<0.01, *p<0.05. Bush (1a) Trump (1a) Bush (1b) Trump (1b) Bush (2a) Trump (2a) Bush (2b) Trump (2b) (Constant) 0.045 0.177 0.030 0.051 0.032 0.200 0.017 0.043 (0.353) (0.184) (0.471) (0.589) (0.587) (0.191) (0.642) (0.593) Trust (ref. = ‘untrustworthy’) Trustworthy 81.831*** 51.882*** 41.405*** 22.315*** 97.769*** 45.556*** 65.530*** 13.174*** (0.360) (0.305) (0.392) (0.457) (0.628) (0.400) (0.645) (0.618) Party (Ref. = ‘independent/ no preference’) Republican 2.613** 4.245*** 2.228* 1.892 (0.319) (0.336) (0.352) (0.505) Democrat 0.359** 0.183*** 0.365* 0.147** (0.343) (0.374) (0.390) (0.564) Economy (Ref. = ‘negatively’) Positively 2.403* 3.831** 2.390** 4.512*** (0.345) (0.427) (0.335) (0.421)
Iraq war (Ref. = ‘negatively’) Positively 3.973*** 4.676*** (0.301) (0.288) Investigation (Ref. = ‘negatively’) Positively 23.357*** 28.807*** (0.413) (0.411) Trustworthiness x Partisanship (Ref. = ‘weak partisans’) Strong Partisans 1.469 9.597*** 1.063 10.534*** (0.777) (0.623) (0.798) (0.899) -2LL 371.865 403.020 305.757 203.516 374.748 428.997 332.794 208.972
Cox and Snell’s R2
0.553 0.573 0.564 0.644 0.514 0.554 0.548 0.639
Nagelkerke R2 0.741 0.769 0.758 0.865 0.690 0.743 0.735 0.858
19
Discussion
These findings show that, as expected, the effect of trustworthiness on individual presidential approval ratings has decreased, while the effect of the partisanship has increased. Therefore, I can confirm the first and second hypotheses. Thus, while the impact of personality standards on presidential evaluations is declining, the effect of partisanship grows. Tentatively, I can conclude from this that approval ratings are increasingly shaped by partisanship at the cost of standards such as personality traits. Moreover, from these findings it seems that the effect of motivated reasoning on political evaluation has grown. In other words, citizens are more inclined to evaluate their president through a partisan lens in the Trump era, while former standards matter less for their evaluations.
However, the indicators of polarization and perceptions of the economy point in a different direction. Since the number of high partisans has not grown, it should be expected that the number of people who are highly influenced by motivated reasoning also has not grown. Although this research indicates that partisanship has a stronger effect on individual political evaluation, since I could not find evidence of polarization, I have not found an explanation for this effect. Secondly, the effect of perceptions of the economy on individual presidential approval ratings has grown in the Trump era. This suggests that, contrary to Donovan et al’s (2019) findings, accuracy motivations have an increased impact on presidential approval ratings. In other words, approval ratings in the Trump era are not fully shaped by motivated reasoning, but also by other standards such as the economy.
Nevertheless, because I have not tested motivated reasoning directly, I cannot confirm that motivated reasoning has not grown among the American public over the last few decades. The findings on the influence of trustworthiness and partisanship on presidential approval ratings certainly point in this direction. Future research should look at which accuracy motivations are still important for presidential approval ratings. Additionally, our understanding of polarization and motivated reasoning should be improved in order to explain why partisanship has a higher effect on presidential approval ratings in the Trump era.
In these analyses, I have attempted to separate the effects of partisanship, trust and evaluations of the economy, wars and scandals. However, these variables are highly related and affected by each other in multiple directions. Not only do evaluations of trustworthiness influence how people evaluate Trump’s presidency, but also do prior beliefs such as presidential approval influence people’s view of a president’s trustworthiness. For example, the increased
20
effect of perceptions of the economy on presidential approval ratings might be explained by an increased effect of motivated reasoning. Thus, there are multiple explanations for why I have found these effects. Most importantly, this shows that it is very hard, if not impossible, for researchers to separate the effects of partisanship, other prior beliefs and evaluations of real-world events and political leaders. Future research might be more able to avoid these problems by using an experimental design to isolate causal relationships and by directly measuring motivated reasoning and its effects on every variable used in the research.
Based on the results above, I reject the third hypothesis. I could not find significant results for Bush’s approval ratings and I found effects for Trump’s approval ratings in a direction opposite than expected. From the research, it is not directly clear how to explain these findings. Perhaps, motivated reasoning is not stronger in strong partisans, making strong partisans less inclined to evaluate a president through a partisan lens. Alternatively, strong partisans’ perception of a president’s trustworthiness might be affected more strongly by motivated reasoning than weak partisans’ perception. That would mean that there is stronger correlation between perception of trustworthiness and presidential evaluation for strong partisans, because they are both seen from the same partisan lens. This point is interconnected with the difficulty, as described above, of identifying causal direction and isolating effects in a cross-sectional design. Future research should attempt to replicate these results and find explanations for the causal direction.
Possibly, these findings can be explained by the categorization of partisanship. I have attempted to run analyses with more categories of partisanship, dividing between both partisan strength and party. In these analyses, the results were even more insignificant. This might be explained by the small size of the categories, which could have inflated standard errors. Moreover, there were indications of high multicollinearity, which also could have led to high standard errors. Future research might avoid these problems by using a larger dataset. Most importantly, these findings suggest that the categorization used in this research did not make a valid distinction between strong and weak partisans. If the used questions and categorization are not measuring the underlying concept of partisanship, it might explain why the results were opposite from what was expected.
Consequently, these problems might be solved by using a more precise measure of partisanship and polarization. Although the question in the dataset gives an indication of an individual’s partisanship, it does not directly measure the strength of this partisanship. For example, someone who initially identifies as a Republican might do so because he or she feels
21
pressured to give a straightforward answer. Future research can use multiple questions, for example about feelings towards a party, agreement with its ideology and political activism, in order to measure partisanship in a more refined way. This issue is interconnected with the problems surrounding my measure of polarization, which is based on level of partisanship as well. Because I do not measure the correlation of ideology within the parties, I do not measure the concept polarization directly, according to the used definition of polarization. This might explain why my results disconfirm earlier research, which has conclusively argued that polarization has occurred. Future research should rely on more refined measures of polarization. For example, researchers could use negative perceptions of the out-party as a valid indicator or could look at the alignment of an individual’s ideology with party ideology.
Finally, by using a proxy variable for inconsistency, I do not measure the relationships described in my hypotheses directly. Although in earlier research, both concepts have been categorized in the same dimension, individuals might still see these concepts as very different. For example, an individual might not only see a president as untrustworthy when he is inconsistent, but also when he lies, has conflicting interests or covers up unethical behavior. Thus, an individual might see Trump or Bush as untrustworthy, but not as inconsistent and vice versa. Possibly, this leads to different findings using the proxy variable than using a more direct measure of the underlying concept. Moreover, trust in government and politicians has been associated with support (Bowler & Karp, 2004). Consequently, this research risks circular reasoning by examining the relationship between trust and approval, which can both measure support. Therefore, future research should rely on a more direct measure of consistency.
Conclusion
To conclude, despite these limitations, some tentative implications for the evaluation of flip-flopping can be inferred from the findings. The evidence indicates that Trump is punished less for flip-flopping than his predecessor. From the smaller effect of trustworthiness evaluations on Trump’s approval ratings than on Bush’s approval ratings, it can be concluded that, if this variable has the same effects as its proxy variable, evaluations of inconsistency have a smaller effect on individual approval ratings of Trump than of Bush. This implies that, nowadays, citizens are less likely to take a leadership’s inconsistency into account when evaluating a president’s overall performance. Considering that flip-flopping is a form of being inconsistent, I conclude that citizens are less inclined to take flip-flopping into account when evaluating president Trump. This would mean that citizens are less likely to punish Trump for flip-flopping by giving him lower approval ratings. Thus, within the scope of this research
22
project, the answer to the research question is that the evaluations of Trump deteriorate less after a flip-flop than evaluations of Bush did. However, because of the severe limitations to the analysis and the indirect measurement of citizens’ reaction to flip-flopping, this answer should be taken with caution. As described above, future research should rely on a more direct examination of the research puzzle.
The possible difference in public response to flip-flopping might be explained by increased motivated reasoning. Due to recent polarization, individuals are more likely to evaluate their leaders on the basis of prior beliefs, such as partisanship. Therefore, individuals might be more likely to ignore, rationalize or, conversely, demonize a flip-flop. However, although I have found that partisanship has a bigger effect on Trump’s approval ratings than on Bush’s, I have not found direct evidence of motivated reasoning or polarization. Additionally, the unexpected effect of strong partisanship suggests that motivated reasoning works in different ways than expected. Thus, from this research it is unclear why partisanship has a stronger effect on individual presidential approval ratings. However, because previous research has emphasized the increased importance of polarization over the last few decades, future research should also look in this direction to explain why Trump is punished differently for flip-flopping.
These conclusions have several implications for the literature on flip-flopping. Most importantly, this research shows that punishment for flip-flopping is not the same in all situations and for all leaders. Although previous research has formulated many conditions to explain when citizens condemn flip-flopping, the inconsistent findings about the effect of these inconsistencies show that the literature is still inconclusive about this. This research project proposes aggregate- level explanations, such as polarization, and psychological explanations, such as motivated reasoning, in order to improve our understanding of when and how leaders are punished for flip-flopping. Further research should attempt to find additional conditions of punishment for flip-flopping.
Moreover, this project fits into new research on Trump’s leadership and its idiosyncrasies. With the surprise at Trump’s election and recent shock about Trump’s presidential decisions, research, in addition to the media, should also ask the question how different Trump’s presidency is. Furthermore, researchers might ask if changes are a consequence of Trump’s acts and personality or of other developments, such as polarization. For example, Trump’s limited punishment for flip-flopping might be explained by how he explains his flip-flops or treats the media, but also by increased motivated reasoning, over which
23
he has less control. Additionally, the way the media frame Trump as opposed to previous presidents or the frequency in which Trump displays inconsistencies might also explain why Trump’s flip-flops are evaluated differently. Finding an answer to these questions could explain if Trump’s presidency is only a short period in American history, or the product of long-term developments.
Finally, these findings have important implications for how we understand democratic accountability (Iyengar & Krupenkin, 2018). This research indicates that partisanship is becoming more important for how citizens evaluate their president, while characteristics such as trustworthiness and inconsistency are becoming less important. The importance of these developments should not be underestimated; if voters are less likely to change their opinion about their political leaders on the basis of these politicians’ acts and policies, it might lead citizens to be more tolerant towards unethical or undesirable political behavior. However, citizens’ role as controls for government and politicians is of profound importance, since we expect citizens in democracies to punish and reward their politicians. Thus, the gap between our understanding and expectations of democracies, and the way in which democracies actually function might grow larger. In other words, a future impairment of citizens’ democratic function might fundamentally change how we look at democracy.
24
Appendix
Questions used from the Pew Research Center October 2005 News Interest Index (Pew Research Center, 2005):
Q.1 Do you approve or disapprove of the way George W. Bush is handling his job as president?
1 Approve
2 Disapprove
Q.10F1 As I read some pairs of opposite phrases, tell me which one best reflects your impression of George W. Bush so far. (First,) does George W. Bush impress you as…
a.F1 Trustworthy or NOT trustworthy?
1 Trustworthy
2 Not trustworthy
Q.49 How would you rate economic conditions in this country today… as excellent, good, only fair, or poor?
1 Excellent
2 Good
3 Only fair
4 Poor
Q.61 How well is the U.S. military effort in Iraq going?
1 Very well
2 Fairly well
3 Not too well
4 Not at all well SEX
1 Male
2 Female
INCOME Last year, that is in 2004, what was your total family income from all sources, before taxes? Just stop me when I get to the right category.
25 2 10 to under $20,000 3 20 to under $30,000 4 30 to under $40,000 5 40 to under $50,000 6 50 to under $75,000 7 75 to under $100,000 8 100 to under $150,000 9 $150,000 or more
PARTY In politics TODAY, do you consider yourself a Republican, Democrat, or Independent? 1 Republican 2 Democrat 3 Independent 4 No preference 5 Other party
Questions from the Pew Research Center September 2018 Political Survey (Pew Research Center, 2018):
Q.2 Do you approve or disapprove of the way Donald Trump is handling his job as President? 1 Approve
2 Disapprove
Q.33 As I read some pairs of opposite phrases, please tell me which one best reflects your impression of Donald Trump. (First,) does Donald Trump impress you as..
a.F1 Trustworthy or NOT trustworthy? 1 Trustworthy
2 Not trustworthy
Thinking about the nation’s economy …
Q.44a How would you rate economic conditions in this country today… as excellent, good, only fair, or poor?
26 1 Excellent
2 Good
3 Only fair 4 Poor
Q.97 How confident, if at all, are you that Donald Trump is handling matters related to the special counsel investigation appropriately? [CLARIFY IF NECESSARY: special counsel Mueller’s investigation into Russian involvement in the 2016 election]
1 Very confident 2 Somewhat confident 3 Not too confident 4 Not at all confident
Now, just a few questions for statistical purposes only. SEX 1 Male
2 Female
PARTY In politics TODAY, do you consider yourself a Republican, Democrat, or independent? 1 Republican 2 Democrat 3 Independent 4 No preference 5 Other party
INCOME Last year, that is in 2017, what was your total family income from all sources, before taxes? Just stop me when I get to the right category.
1 Less than $10,000 2 10 to under $20,000 3 20 to under $30,000
27 4 30 to under $40,000 5 40 to under $50,000 6 50 to under $75,000 7 75 to under $100,000 8 100 to under $150,000 9 $150,000 or more
28
Literature
Allgeier, A. R., Byrne, D., Brooks, B. & Revnes, D. (1979). The Waffle Phenomenon: Negative Evaluations of Those Who Shift Attitudinally. Journal of Applied Social
Psychology, 9(2), 170-182.
Argyle, L. P., Arrajj, M., Covich, S., Garay, E.G., Gottlieb, J., Hodges, H.E & Smith, E.R.A.N. (2016). Economic performance and presidential trait evaluations: A longitudinal analysis. Electoral studies, 43, 52-62.
Bittner, A. (2011). Platform or Personality? The Role of Party Leaders in Elections. Oxford, United Kingdom: Oxford University Press.
Bowler, S. & Karp J. A. (2004). Politicians, Scandals and Trust in Government. Political
Behavior, 26(3), 271-287.
Buchanan, B. (2016). Citizen Oversight of Presidential Performance. Presidential Studies
Quarterly, 46(2), 234-263.
Carson, J. & Dolan, K. (1995). The Waffle Phenomenon and Candidates’ Image.
Psychological Reports, 57(3), 795-798.
Clifford, S. (2018). Reassessing the structure of presidential character. Electoral Studies, 54, 240-247.
Collingwood, L., Lajevardi, N. & Oskooii, K. (2018). A Change of Heart? Why Individual level Public Opinion Shifted Against Trump’s “Muslim ban”. Political Behavior, 40(4), 1035-1072.
Croco, S. E. (2016). The Flipside of Flip-Flopping: Leader Inconsistency, Citizen Preferences and the War in Iraq. Foreign Policy Analysis, 12(3), 237-257.
Croco, S. E., McDonald, J. A. & Turitto, C. (2019, March 12). Here’s another way Trump is different: Flip-flopping doesn’t affect his public approval. The Washington Post Doherty, D., Dowling, C. & Miller, M. (2016). When is Changing Policy Positions Costly for
Politicians? Experimental Evidence. Political Behavior, 38(2), 455-484.
Donovan, K., Kellstedt, P., Key, E. & Lebo, M. (2019). Motivated Reasoning, Public Opinion and Presidential Approval. Political Behavior, 41(1), 1-21.
29
Fearon, J. D. (1994). Rationalist Explanations for War. In K. A. Mingst & J. L. Snyder (Eds.),
Essential readings in world politics (Sixth edition) (pp. 374-397). New York, United
States; London, United Kingdom: W.W. Norton & Company.
Funk, C. L. (1999). Bringing the Candidate into Models of Candidate Evaluation. The journal
of politics, 61(3), 700-720.
Goodwin, G. P. (2015). Moral Character in Person Perception. Current Directions in
Psychological Science, 24(1), 38-44.
Greene, S. (2001). The role of character assessments in presidential approval. American politics
research, 29(2), 196-210.
Gronke, P. & Newman, B. (2010). Public Evaluations of Presidents. In G. C. Edwards III & W. G. Howell (Eds.), The Oxford Handbook of the American Presidency. Oxford, United Kingdom: Oxford University Press.
Hardy, B. W. (2017). Candidate Traits and Political Choice. In K. Kenski & K. H. Jamieson (Eds.), The Oxford Handbook of Political Communication. Oxford, United Kingdom: Oxford University Press.
Hoffman, A. M. (2002). A Conceptualization of Trust in International Relations. European
Journal of International Relations, 8(3), 375-401.
Hoffman, H. S. & Carver, C. S. (1984). Political Waffling: Its Effects on the Evaluations of Observers. Journal of Applied Social Psychology, 14(4), 375-385.
Hummel, P. (2010). Flip-flopping from primaries to general elections. Journal of Public
Economics, 94(11), 1020-1027.
Iyengar, S. & Krupenkin, M. (2018). The Strengthening of Partisan Affect. Political
Psychology, 39(1), 201-218.
Kertzer, J. D. & Brutger, R. (2016). Decomposing Audience Costs: Bringing the Audience Back into Audience Cost Theory. American Journal of Political Science, 60(1), 234-249. Kinder, D. R., Peters, M. D., Abelson, R. P. & Fiske, S. T. (1980). Presidential prototypes.
Political Behavior, 2(4), 315-337.
Lebo, M. J. & Cassino, D. (2007). The Aggregated Consequences of Motivated Reasoning and the Dynamics of Partisan Presidential Approval. Political Psychology, 28(6), 719-746
30
Lee, M. Y. H. (2017, April 13). President Trump, king of flip-flops (continued). The
Washington Post
Leeper, T. J. & Slothuus, R. (2014). Political Parties, Motivated Reasoning and Public Opinion Formation. Political Psychology, 35(1), 129-156.
Levendusky, M. S. & Horowitz, M. C. (2012). When Backing Down is the Right Decision: Partisanship, New Information, and Audience Costs. The Journal of Politics, 74(2), 323-338.
McCaul, K. D., Ployhart, R. E., Hinsz, V. B. & McCaul, H. S. (1995). Appraisals of a` Consistent versus a Similar Politician: Voter Preferences and Intuitive Judgments.
Journal of Personality and Social Psychology, 68(2), 292-299.
McDonald, J., Croco, S. E. & Turitto, C. (2019). Teflon Don or Politics as Usual? An Examination of Foreign Policy Flip-Flops in the Age of Trump. The Journal of
Politics, 81(2), 1-10.
Pew Research Center. (2005). October 2005 News Interest Index [Data Set]. Retrieved from https://www.people-press.org/dataset/october-2005-news-interest-index/
Pew Research Center. (2017, October 4). Partisan gap in political values continues to grow. Retrieved from https://www.people-press.org/2017/10/05/1-partisan-divides-over-political-values-widen/1_2-16/
Pew Research Center. (2018). September 2018 Political Survey [Data Set]. Retrieved from https://www.people-press.org/dataset/september-2018-political-survey/
Putnam, A., Wahlheim, C., Jacoby L. (2014). Memory for flip-flopping: Detection and recollection of political contradictions. Memory & Cognition, 42(7), 1198-1210. Robison, J. (2017). The Role of Elite Accounts in Mitigating the Negative Effects of
Repositioning. Political Behavior, 39(3), 609-628.
Schultz, K. A. (2003). Looking for Audience Costs. Journal of Conflict Resolution, 45(1), 32-60.
Schultz, K. A. (2018). Perils of Polarization for U.S. Foreign Policy. Washington Quarterly,
31
Sigelman , L. & Sigelman, C. (1986). Shattered Expectations: Public response to “out-of-character” presidential actions. Political Behavior, 8(3), 262-286.
Snyder, J. & Borghard, E. D. (2011). The Costs of Empty Threats: A Penny, Not a Pound.
American Political Science Review, 105(3), 737-456.
Sorek, A., Haglin, K. & Geva, N. (2018). In Capable Hands: An Experimental Study of the Effects of Competence and Consistency on Leadership Approval. Political Behavior,
40(3), 659-679.
Taber, C. S. & Lodge, M. (2006). Motivated Skepticism in the Evaluation of Political Beliefs.
American Journal of Political Science, 50(3), 755-769.
Tomz, M. (2007). Domestic Audience Costs in International Relations: An Experimental Approach. International Organization, 61(4), 821-840.
Tomz, M. & Van Houweling, R. P. (2010). Candidate Repositioning. Working paper
Winter, D. G. , John, O. P. , Stewart, A. J. , Klohnen, E.C. & Duncan, L. E. (1998). Traits and Motives: Toward an Integration of Two Traditions in Personality Research.
