Abstract
Campaign promises are a central mechanism for voters to hold politicians accountable, and information about their breakage or fulfillment features prominently in the media during election campaigns. Despite the importance of campaign promises, previous research yields conflicting expectations regarding their influence on citizens. Some theories suggest citizens vote based on policy performance and, therefore, consistently penalize actors who break their promises. Other theoretical accounts, however, argue that exposure to information during election campaigns often has minimal effects on citizens due to strongly held prior beliefs and partisan motivations. The goal of the current study is to address these competing claims by systematically testing the conditions under which citizens penalize politicians for breaking promises. We conducted four experiments (total N = 7,030), three of them preregistered, in two countries under varying political conditions. We find that (1) broken promises decrease domain-specific evaluations of leaders but have little impact on evaluations of actors’ overall performance; (2) broken promises have limited effects when people have strong priors about the political actor who made the promise; and (3) citizens downplay and rationalize promises broken by ingroup, but not outgroup, leaders. These results suggest that even though information about broken promises is salient in campaign communications, its impact on citizens is context-dependent and often quite limited.
Campaign promises play a pivotal role in the electoral process, and information about them features prominently in elite communication and news coverage during election campaigns. Candidates signal their policy intentions to potential voters by making promises through various communicative channels, such as campaign advertisements, political debates, party platforms, and social media (Dobber and Vreese 2022; Mellon et al. 2021; Sulkin 2009; Thomson et al. 2017). These campaign promises are highly visible in the news (Kostadinova 2019; Müller 2020).
Campaign promises are considered a unique form of communication between citizens and representatives. First, promises often include explicit commitments by elites to an action or outcome. This is in contrast to other political statements that may signal politicians’ stances on issues but provide no measurable criteria for citizens to evaluate their representatives’ performance (Bøggild and Jensen 2024; Thomson et al. 2017; Tomz and Van Houweling 2009). Indeed, theories of democratic representation emphasize the role of promises as a critical mechanism through which citizens can hold representatives accountable (Mansbridge 2003; Naurin et al. 2019b). Second, journalists in recent decades have increasingly focused on non-substantive aspects of politics, such as candidates’ personality traits or who is winning in the polls (Aalberg et al. 2012; Patterson 2016). In contrast, many campaign promises deal with policy issues and still gain significant media coverage (Kostadinova 2019; Müller 2020), which can increase citizens’ attention to and awareness of candidates’ performance on policy-related matters. Policy-based voting is highly desirable in democracies, where political elites are expected to represent citizens’ preferences (Pitkin 1967).
Given the prominence of campaign promises in citizens’ political information environment, as well as their importance for democratic representation, examining how people respond when they learn that promises have been broken or fulfilled is crucial. Existing research, however, offers conflicting theoretical perspectives. On the one hand, some research suggests citizens vote based on policy performance and, therefore, consistently penalize actors who break their promises. Theories of prospective voting suggest citizens base their voting decisions on the level of congruence between their own policy preferences and what political actors promise to do in the future (Elinder et al. 2015), and literature on retrospective voting argues citizens punish parties and candidates for breaking their past promises (Matthieß 2022).
On the other hand, prominent theoretical approaches suggest that exposure to information during election campaigns often has minimal effects on citizens (Broockman and Kalla 2023). First, according to the Bayesian updating framework, when people hold strong prior opinions and possess much prior knowledge of the candidates or parties, they are unlikely to update their evaluations of them in response to new pieces of information (Green et al. 2002; Hill 2017; McDonald et al. 2019). Second, the extensive motivated reasoning literature suggests that citizens often process political information with the goal of casting their political ingroup in a positive light and the political outgroup in a negative light (Bisgaard 2019; Peterson and Iyengar 2021; Taber and Lodge 2006). Together, these perspectives suggest that exposure to information about campaign promises may play a more modest role in shaping citizens’ attitudes.
The goal of the current study is to address these competing theoretical expectations by systematically testing the conditions under which information about broken and fulfilled promises shapes citizens’ attitudes. We do this with four well-powered experiments (total N = 7,030), three of them preregistered, that cover a variety of political contexts. Our experiments were conducted in two countries (the U.S. and Israel), used real and fictitious politicians and promises, examined promises on bipartisan and partisan issues, and tested the effects of promises on various attitudinal outcomes. Together, this diverse set of contexts provides a broad account of how broken campaign promises affect citizens’ attitudes.
Our experiments show that the impact of broken campaign promises is highly dependent on the political context and often quite limited. First, while broken promises consistently shape citizens’ evaluations of the way the promise-breaking politician handles the specific policy domain the promise deals with, they have little impact on evaluations of that politician’s overall performance. The limited effects on overall evaluations are particularly apparent in contexts where respondents have strong prior attitudes toward the candidates. Importantly, the effects of broken promises on overall evaluations are minimal regardless of (a) citizens’ pre-existing levels of support for the policy being discussed and (b) the personal importance people attach to the political issue. Overall, these results are consistent with a growing literature indicating that the impact of campaign information on citizens’ attitudes is highly contingent and often limited (e.g., Broockman and Kalla 2023; Druckman and McGrath 2019; Naurin et al. 2019a).
Second, our results demonstrate that citizens’ partisan attachments shape their responses to broken promises. When exposed to broken promises by ingroup (but not outgroup) politicians, citizens employ two reasoning strategies that enable them to continue supporting the politician despite the apparent transgression: (a) They acknowledge the problematic act of breaking a promise but separate it from evaluations of job performance (i.e., decoupling); and (b) they downplay the promise and justify the act of breaking it (i.e., rationalization). These results demonstrate that partisan-motivated reasoning shapes citizens’ responses to broken promises and points to the specific mental strategies driving people’s politically motivated responses.
Citizens’ Responses to Broken Promises
A campaign promise is a statement by a candidate or a party in which they commit to some future action or outcome (Naurin and Thomson 2020; Thomson et al. 2017). An important distinction is between narrow and broad promises (Naurin et al. 2019b). Narrow promises are those in which the meaning of the pledge and what constitutes fulfillment are unambiguous. For example, promising to reduce the standard rate of corporate tax to 12.5 percent in 2003 is a narrow pledge because it states a precise number (12.5%) and a specific time frame (2003). Broad promises, however, are more open to interpretation. For example, promising to implement a program to expand the number of school pupils taking Science is broad since the nature of the program and the desired number of pupils are not stated (Thomson et al. 2017).
A substantial literature suggests that campaign promises play a pivotal role in the electoral process. Research on prospective voting argues that citizens base their voting decisions on the level of congruence between their policy positions and the policy goals parties and candidates promise to pursue in the future (Elinder et al. 2015). The retrospective voting literature argues that citizens’ vote choices are based on their evaluation of political actors’ past performance: Candidates and parties who fulfill their campaign promises are perceived as more credible, trustworthy, and accountable (Matthieß 2022), and the fulfillment of previous promises is an important criterion voters use to decide whether to reward or penalize political actors (Aragonès et al. 2007; Naurin 2011).
Indeed, various studies find that breaking promises can harm political actors. Observational data from sixty-nine elections in fourteen countries shows that governing parties’ vote shares correlate positively with their pledge fulfillment record (Matthieß 2020). Broken promises are especially costly for parties when the promises deal with ideological issues that concern voters’ core beliefs (Tavits 2007). Additional observational studies indicate that voters react more strongly to parties’ campaign promises than to some of their past actions (Elinder et al. 2015). Experimental evidence further shows that the public uses campaign promises to update its prospective beliefs about new candidates and to retrospectively penalize promise-breaking politicians (Bonilla 2021; Born et al. 2018) and that voters penalize governments for broken promises more than they reward them for fulfilled ones (Naurin et al. 2019a).
The Conditional Effects of Campaign Communication
The studies presented above view promises as an important factor in shaping citizens’ attitudes and evaluations. However, two prominent theoretical approaches suggest that the impact of broken promises might be more context-dependent and, under some circumstances, even minimal (Broockman and Kalla 2023). First, the Bayesian updating approach posits that citizens consider a wide range of information in addition to promises when forming their attitudes (Bullock 2009; Hill 2017). Second, the motivated reasoning approach argues that citizens often process information about candidates and parties through a partisan lens (Bisgaard 2019; Peterson and Iyengar 2021).
Bayesian Updating and Broken Promises
Under the Bayesian updating framework, citizens formulate their evaluations of candidates based on various factors, such as past performance, political experience, social group membership, and more (Broockman and Kalla 2023; Rogowski and Sutherland 2016). In this framework, people assign less weight to new information when their prior beliefs are strong (Green et al. 2002; McDonald 2020). Moreover, having much prior knowledge about candidates reduces the likelihood that any new piece of information—whether positive or negative—will influence people’s evaluations of political actors (Broockman and Kalla 2023; DellaVigna and Gentzkow 2010).
According to the Bayesian updating approach, in political contexts where information about candidates is abundant, the impact of campaign communication is more limited because citizens have already established strong opinions about the relevant actors and are closely familiar with them (Kalla and Broockman 2018). In such contexts, citizens have already been exposed to and processed diverse messages from journalists, opinion leaders, and campaigns, leading to a reduced impact of new messages. One notable exception to this rule occurs when the new information is highly unusual and unexpected (Bullock 2009; Kalla and Broockman 2018). In other contexts, however, new information exerts limited effects, and citizens assimilate it into their existing judgment of the candidates (Hill 2017). According to the Bayesian updating perspective, the effects of encountering information about promise breaking on citizens’ attitudes should be limited when people have strong prior opinions and possess much prior knowledge of the candidates.
Partisan-Motivated Reasoning and Broken Promises
A second line of research suggesting a more limited impact of campaign promises focuses on the influence of partisan identities on citizens’ attitudes. According to the partisan-motivated reasoning literature, citizens process political information in ways that favor their political ingroup. This literature demonstrates that partisanship colors people’s evaluations of information in a wide range of contexts, including candidates’ policy performance and norm violations, causing them to process political information in biased ways that cast their political ingroup in a positive light and the political outgroup in a negative light (Anduiza et al. 2013; Bisgaard 2019; Peterson and Iyengar 2021).
Recent studies present some evidence that partisan identities may also play a role in the processing of information about campaign promises. Supporters of governing parties evaluate the government’s degree of pledge fulfillment more positively than supporters of the opposition, regardless of whether the pledges were actually fulfilled (e.g., Pétry and Duval 2017; Thomson 2011). Citizens are more likely to engage in pledge-based voting when pledges are made by a party they identify with (Bøggild and Jensen 2024), and out-partisans are less likely than in-partisans to reward politicians for keeping promises (Bonilla 2024). Building on this line of research and extending it to additional contexts, we expect the following:
H1: Citizens will be less inclined to penalize ingroup (compared to outgroup) politicians who break a campaign promise.
While numerous studies have demonstrated that people perceive political reality through a partisan lens, the ways in which individuals convince themselves that their predefined conclusions hold up in the face of contrary evidence remain less explored. Therefore, we examine not only whether partisan-motivated reasoning occurs in the context of processing information about broken promises, but also how. To this end, we propose two mental strategies citizens may employ to maintain their support for ingroup politicians who fail to keep their campaign promises.
The first reasoning strategy we study, decoupling, is a psychological separation process whereby people selectively dissociate their assessments of specific actions from judgments of the general performance of a person, organization, or firm (Bhattacharjee et al. 2013; Shulman et al. 2020). When engaging in decoupling, people acknowledge that the behavior under question—in our case, breaking a campaign promise—is negative, undesirable, or even immoral. However, they deliberately refrain from incorporating this negative information into their evaluation of the actor’s overall performance. Decoupling is considered an especially appealing reasoning strategy because it does not involve condoning immoral or otherwise problematic acts; with this strategy, the action is condemned but separated from overall performance judgments (Bhattacharjee et al. 2013). We thus posit the following:
H2: Promise breaking by ingroup politicians will lead to higher levels of decoupling among citizens than promise breaking by outgroup politicians.
The second reasoning strategy we explore here, rationalization, refers to cases where people reconstrue transgressions as less problematic (Shu et al. 2011). Unlike decoupling, in which individuals acknowledge the wrongdoing but do not incorporate it into their evaluations of general performance, the goal of rationalization is to downplay and justify the problematic act to make it seem less damaging and more acceptable (Bandura et al. 1996; Bhattacharjee et al. 2013; Bisgaard 2015). When information can potentially damage people’s political ingroup, they become highly motivated to justify it. However, when it portrays the outgroup unfavorably, they have little reason to do so (Bisgaard 2019). Applying this principle to our context, we expect the following:
H3: Citizens will rationalize promises broken by ingroup politicians more than promises broken by outgroup politicians.
Research Design and Data
We examine the effects of broken promises with four well-powered experiments, of which three were preregistered. 1 Table 1 provides an overview of our experimental studies, and Supplemental Information Section A presents the sample composition in each study. All four experiments sampled adults (eighteen or above) with quotas for gender, age, and education matched to U.S. and Israeli population benchmarks. The experiments examine promises whose source is an individual leader. While various studies have focused on promises made by parties (e.g., Thomson et al. 2017), in recent decades, many Western democracies have witnessed a notable process of political personalization, whereby the dominance of individual politicians is increasing both in the political arena and in the news coverage of politics (Balmas et al. 2014; Rahat and Kenig 2018). Citizens are increasingly exposed to statements by individual politicians in the media, engage with them directly on social media, and place growing emphasis on the actions of leaders when making political decisions (Nicholson 2012; Van Aelst et al. 2017; Zamir 2024).
Overview of Experiments.
Note. U.S. participants were recruited through Lucid, and Israeli participants were recruited through iPanel.
In Study 1, we presented to a sample of American participants (N = 1,821) either a broken or a fulfilled promise by either President Joe Biden or former President Donald Trump. In Study 2, we presented to a sample of Jewish Israeli participants (N = 1,354) either a broken or a fulfilled promise by the Prime Minister (PM), Benjamin Netanyahu. Participants in these two studies were shown real promises made by these leaders during an election campaign.
While using real promises increases external validity, this approach may be prone to “pretreatment effects”: If respondents have already been exposed to these promises in the real world prior to their participation in the study, this may limit the ability to detect effects within the experiment (Slothuus 2016). For this reason, in Study 3, we presented to a sample of American participants (N = 1,913) a fixed fictitious promise we had crafted ourselves. This promise was randomly attributed to either Joe Biden or Donald Trump, and randomly presented as either having been broken or fulfilled. In Study 4, we again presented to a sample of American participants (N = 1,942) a fixed fictitious promise that was either broken or fulfilled. This time, however, the promise was attributed to a fictitious leader, of whom respondents could not have prior knowledge. Taken together, the variety of contexts our four studies investigate enables us to provide a broad account of when campaign promises matter.
To account for possible heterogeneous effects according to the promise source, the four experiments evaluate the effects of campaign promises by ingroup and outgroup politicians separately (Fernandez-Vazquez and Theodoridis 2020; Naurin et al. 2019a). We determined respondents’ political ingroup and outgroup based on their stable political identities. In the U.S., we used their party identification; in Israel, we used their identification with an ideological bloc (Shamir et al. 2017). In the three U.S. studies, pure independents were excluded as they cannot be classified as supporters of a political party (Druckman et al. 2013). Leaning independents were considered weak partisans. 2 In the Israeli study, no respondents were excluded, and those who indicated their political leaning as “center” were considered part of the center-left bloc.
Case Selection
We focus on the U.S. and Israel, two countries with similarities and differences. On the one hand, these countries are characterized by relatively high levels of polarization (Fiorina and Abrams 2008; Gidron et al. 2020), which can shape the way citizens process political information (Druckman et al. 2013). We note, however, that these countries are not extreme outliers in terms of polarization (Gidron et al. 2020; Wagner 2021). There is also no evidence that these countries have extreme cases of personalization in news coverage and on social media (Van Aelst et al. 2017; Zamir 2024). Thus, we believe our findings can be generalized to other political contexts.
On the other hand, significant differences exist between these countries that can affect how individuals evaluate political candidates. In the U.S., political divisions center primarily around partisanship (Iyengar and Westwood 2015), whereas in Israel, party loyalty is less stable than identification with ideological groups (Oshri et al. 2021). In addition, while the American plurality electoral system results in a “winner takes all” outcome, the Israeli proportional multiparty system often necessitates cooperation between ideological opponents to form a government (Lijphart 2012). Comparative research shows that electoral systems with dispersed power, such as Israel’s proportional system, pose distinct challenges regarding campaign promises: Elites are less likely to fulfill their campaign pledges (Thomson et al. 2017) and citizens face greater difficulty in holding representatives accountable due to the multiple actors involved in policymaking (Hobolt et al. 2013; Powell and Whitten 1993). Testing the effects of promise breaking on citizens in different political systems enables us to assess whether their impact remains consistent or varies across different institutional settings.
Procedure
The four experiments had the same basic structure. After providing their informed consent, participants answered a series of background questions measuring their demographic characteristics, political affiliation, ideology, political interest, the importance they attach to several political issues, and their level of support for various policies. Participants were then randomly assigned to an experimental condition, where they read about either a broken or a fulfilled campaign promise by a political leader. Next, we conducted a manipulation check, which confirmed that the promises were perceived as intended (i.e., as broken or fulfilled) in all cases. Participants then answered our outcome measures and were debriefed (see Supplemental Information Section B for manipulation and balance checks).
All four studies were reviewed and approved by the IRB at the second author’s university. For Studies 3 and 4, which involved deception, participants were immediately debriefed and informed that the presented information was fictitious. We elaborate on the ethical considerations involved in these experiments in Supplemental Information Section C.
Design and Stimulus Material
In all four experiments, we presented to participants statements by politicians in which they committed themselves to a future action or outcome. In Studies 1 and 2, the leaders promised a specific policy measure (e.g., deporting all illegal immigrants). In Studies 3 and 4, they promised to pass legislation addressing a specific problem and to do so within a specific time frame. Importantly, even though some promises we used were narrower (i.e., more specific) than others, we did not ask participants to judge themselves whether the promise was fulfilled. Instead, the stimulus text stated explicitly whether the promise was broken or fulfilled. Within each experiment, the stimuli were kept as similar as possible in terms of length, structure, and language (for full stimulus material, see Supplemental Information Section C). We also ensured the respondents considered the policy domains the promises dealt with as relatively important, on average (see Supplemental Information Section F). In the following, we discuss the unique features of each experimental study.
Study 1: Real Promises by Two U.S. Presidents
Our first study tested the effects of real broken and fulfilled campaign promises by two U.S. Presidents, Joe Biden and Donald Trump. This study employed a 2 × 2 between-subject design manipulating the promise source (Biden or Trump) and its outcome (broken or fulfilled). A fifth (control) group did not read any text. Study 1 presented to subjects campaign promises relating to immigration policy, a salient and divisive issue in the U.S. (Amsalem 2019; see Supplemental Information Section D). For Trump, we relied on real promises he had made during the 2016 presidential election campaign, which were either broken or fulfilled during his presidency. For Biden, since Study 1 was conducted about four months after his inauguration, we focused on policies he promised to implement during his first 100 days in office.
Study 2: Real Promises by the Israeli PM
Our second study also tested the effects of real campaign promises, but this time in the multiparty context of Israel. In the last decades, Israeli citizens have increasingly perceived their political group identities in terms of ideological blocs (right vs. center-left) rather than party identification (Shamir et al. 2017). Therefore, we have tested the effects of promises by the leader of the right-wing bloc, PM Benjamin Netanyahu, on citizens from his in- and out-bloc (i.e., the right and center-left blocs, respectively). We used real promises Netanyahu had made during the March 2020 election campaign that were either broken or fulfilled by the time Study 2 was fielded. The promises focused on the Israeli-Palestinian conflict, a highly salient and divisive issue in Israel. The experiment included three conditions: Broken promise, fulfilled promise, and a control group who received no stimulus text.
Study 3: A Fictitious Promise by Two U.S. Presidents
Like our first study, Study 3 employed a 2 × 2 between-subject design manipulating the promise source (Biden or Trump) and the outcome (broken or fulfilled). A fifth (control) group received no stimulus text. However, in this study, we used a fictitious promise whose content was held constant across conditions. This design enables us to eliminate potential pretreatment effects (Slothuus 2016) and confounders related to the promise content, which varied between conditions in Studies 1 and 2. To create a single campaign promise that could credibly be presented as either having been broken or fulfilled by either Trump or Biden, we used a bipartisan issue. A pretest indicated that the country’s preparedness for natural disasters constitutes such an issue (see Supplemental Information Section D). Therefore, the promise all respondents read about was to pass legislation that would help communities across the country prepare for and rebuild from natural disasters. For Trump, participants were told he had made this promise during the 2016 campaign. For Biden, since Study 3 was conducted a little more than a year into his presidency, participants were told he promised to pass this legislation during his first year in office.
Study 4: A Fictitious Promise by a Fictitious Leader
Our final experiment also kept the content of a fictitious promise constant. However, in this experiment, we attributed the promise to a fictitious leader, Paul Miller, who was presented as an incumbent state governor elected in 2018. Study 4 employed a 2 × 2 between-subject design manipulating Miller’s partisanship (Republican or Democrat) and the outcome of the promise (broken or fulfilled). A fifth (control) group did not receive any stimulus text. Similar to Study 3, we used a promise on a bipartisan issue to credibly manipulate the partisanship of its source and whether it was broken or fulfilled. Our pretest (Supplemental Information Section D) revealed that state infrastructure constitutes such an issue. Hence, the promise presented to all respondents was to pass legislation that would ensure the state’s highways, bridges, roads, water pipes, and broadband networks were upgraded and rebuilt. Subjects were told Governor Miller had made this promise during the 2018 gubernatorial election campaign.
Measures
After reading about a campaign promise, respondents evaluated the politician who made the promise both specifically (i.e., regarding the policy domain the promise dealt with) and in general. This separation is important because performance evaluations on particular policies are more volatile than general job approval (McDonald et al. 2019). In the following sections, we describe how we measured these evaluations as well as other outcomes and control variables. All outcomes in all experiments were recoded to a scale of 0–1 to facilitate comparison across outcomes and studies (see Supplemental Information Sections E and F for question wording and descriptive statistics, respectively).
Policy Approval
We asked respondents to evaluate how they thought the politician they had read about handled the specific policy domain the promise dealt with (for instance, in Study 1, people were asked to evaluate how Biden/Trump handled U.S. immigration policy).
General Job Approval
We measured respondents’ evaluation of the promise source more generally by asking them how much they approved of the way the politician they had read about handled his job as president (Studies 1 and 3), PM (Study 2), or governor (Study 4).
Overall Warmth
This measure was constructed by averaging several items. First, respondents rated the leader they had read about on the following four traits: Competent, cares about people like me, selfish (reverse-coded), and honest. Second, respondents were asked to report their feelings toward the politician’s party using an eleven-point feeling thermometer (FT). Third, respondents reported their feelings toward the politician himself on a similar eleven-point FT. To create an overall warmth score for each politician, we averaged the above measures, with higher values indicating warmer feelings, that is, more positive evaluations of the politician and his party (Study 1: α = 0.89; Study 2: α = 0.87; Study 3: α = 0.88; Study 4: α = 0.81).
Decoupling
Respondents rated their agreement with three statements measuring their support for dissociating judgments of promise breaking from judgments of job performance. The statements were based on Bhattacharjee et al.’s (2013) validated decoupling scale but modified to fit the context of campaign promises. Each statement was rated on a five-point scale (1 = Strongly disagree; 5 = Strongly agree). For example, respondents reported the extent to which they agree or disagree that “Judgments of politicians’ performance should remain separate from whether they keep their word or not.” We averaged the scores of the three items to a decoupling scale (Study 1: α = 0.53; Study 2: α = 0.72; Study 3: α = 0.52; Study 4: α = 0.5).
Rationalization
Respondents rated, on a five-point scale, their agreement with seven statements describing different ways to rationalize that it is acceptable for politicians to break their promises (e.g., “Politicians should not be at fault for breaking promises because all politicians do it”). Here, too, we relied on statements from Bhattacharjee et al.’s (2013) validated rationalization scale that was modified to fit the context of campaign promises. We averaged the scores on these seven statements to create a rationalization scale (Study 1: α = 0.81; Study 2: α = 0.77; Study 3: α = 0.83; Study 4: α = 0.83).
Control Variables
We measured demographic characteristics (gender, age, and education) as well as other variables that may predict citizens’ evaluations of politicians (partisanship, partisan strength, ideology, and political interest). We also measured pretreatment support for the policy the promise deals with, and the importance people attach to the policy domain the promise deals with. In the Israeli context (Study 2), we measured political bloc identification (instead of partisanship) by asking respondents to indicate their political leaning on the following scale: right, moderate right, center, moderate left, and left. Those identifying with the first two categories were considered part of the right-wing bloc (Netanyahu’s in-bloc), while all others were part of the center-left bloc (Netanyahu’s out-bloc; Oshri et al. 2021). We also measured additional covariates relevant to the Israeli context only: Which party respondents intended to vote for in the upcoming election and respondents’ strength of attachment to their political bloc (Bankert et al. 2017).
Results
Citizens’ Evaluations of Promise-Breaking Politicians
How do citizens evaluate politicians who break their campaign promises? Figure 1 shows the results from our four experiments (for full results, see Supplemental Information Section G). We begin with the results for ingroup politicians (left panel). In the first two studies, which exposed participants to real promises by real leaders in the U.S. (Study 1) and Israel (Study 2), citizens evaluated the performance of promise breaking (as compared to promise-fulfilling) politicians in the specific policy domain more negatively (policy approval; Study 1: b = −0.12, p < .001; Study 2: b = −0.17, p < .001). However, the effects of broken promises on general job approval and overall warmth toward ingroup politicians are very small and mostly statistically insignificant. 3 A similar pattern emerges in Study 3, which presented to American participants a fictitious promise by real leaders (Biden or Trump). Here, too, a broken promise by ingroup politicians decreases policy approval (b = −0.25, p < .001), but the effects on general approval and overall warmth are, in both cases, indistinguishable from zero. In sum, while a broken promise can decrease citizens’ approval of the way real ingroup politicians handle the specific issue, this does not translate into changes in citizens’ general evaluations of those politicians.
Effects of broken, compared to fulfilled, promises on citizens’ evaluations of ingroup (left) and outgroup politicians (right). Error bars represent 95 percent confidence intervals. The graphs are based on the unstandardized regression coefficients presented in Supplemental Tables G1–G8.
In contrast to the first three studies, which measured evaluations of real leaders, we observe a different pattern in Study 4, which exposed participants to promises by a fictitious politician of whom they could not have any prior opinion or knowledge. When exposed to a fictitious ingroup politician who breaks a promise, citizens evaluate him more negatively in regards to the specific policy domain (b = −0.44, p < .001) and in general (general approval: b = −0.34, p < .001; warmth: b = −0.25, p < .001). As Figure 1 shows, in this scenario, the penalty for broken promises is much larger for all three outcomes than in the previous three studies.
Together, these findings indicate that when a real-world ingroup politician breaks a promise, citizens update their evaluation of that politician in the specific policy domain but do not update their overall evaluations (Studies 1–3). However, when citizens have no prior opinion or knowledge of the politician (Study 4), both domain-specific and overall evaluations become more negative.
We now turn to the right panel of Figure 1, which presents the effects of promises broken by outgroup politicians. Focusing first on the policy approval outcome (first row), results show that, except for the Israeli case (Study 2), citizens evaluate outgroup politicians’ performance in the specific policy domain more negatively in the broken promise condition. However, the results of Studies 1–3 again show little evidence for an effect on overall evaluations of outgroup politicians. Study 4 is the only case where broken promises by a (fictitious) outgroup leader decrease citizens’ general evaluations of him (general approval: b = −0.33, p < .001; warmth: b = −0.28, p < .001). This indicates that when citizens have no prior opinion about outgroup politicians, new information suggesting that they have broken a promise negatively influences the way people evaluate them. 4 Formal interaction models testing whether citizens are less likely to penalize ingroup than outgroup politicians show no significant differences in the impact of promise breaking in almost all cases (see Supplemental Information Section G). 5 H1 is, therefore, rejected.
To conclude this section, our four experiments indicate that while broken promises can decrease citizens’ evaluations of politicians’ performance in the specific policy domain the promise deals with, their general evaluations of real leaders from both their ingroup and outgroup who break their promises are hardly affected. In the next section, we further explore potential differences in the ways people process information about ingroup and outgroup politicians who break their promises.
Reasoning Strategies for Coping With Broken Promises
We now turn to examine the mental strategies individuals employ in response to information about broken promises. Looking at the left panel of Figure 2, we find that when citizens are exposed to a broken promise by real ingroup politicians (Studies 1–3), they show higher levels of decoupling (Study 1: b = 0.07; p < .001; Study 2: b = 0.05; p = .004; Study 3: b = 0.06; p < .001). Only in the case of a fictitious political leader (Study 4) do we not observe any evidence for more decoupling (b = 0.006; p = .679). In the case of outgroup politicians (the right panel of Figure 2), however, none of the effects on decoupling are significant at the 95 percent confidence level. Taken together, these results generally support H2 by demonstrating that citizens dissociate judgments of promise breaking from evaluations of overall performance when a real politician from their political ingroup (but not outgroup) breaks a promise.
Effects of broken, compared to fulfilled, promises on citizens’ reasoning strategies after exposure to promises by ingroup (left panel) and outgroup politicians (right panel). Error bars represent 95 percent confidence intervals. The graphs are based on the unstandardized regression coefficients presented in Supplemental Tables G1–G8.
Moving to the second mental strategy, rationalization, the left panel of Figure 2 shows that broken promises by an ingroup politician increase rationalization when these promises are made by real leaders in the U.S. (Study 1: b = 0.07; p < .001; Study 3: b = 0.06; p < .001). However, we find no evidence for increased rationalization in the Israeli case (Study 2: b = 0.008; p < .587) and when a fictitious politician makes the promise (Study 4: b = 0.002; p = .909). When examining outgroup politicians (right panel), the effects of breaking a promise on rationalization are statistically insignificant in all cases. In sum, our experiments provide partial support for H3, showing some evidence that citizens rationalize broken promises by politicians from their own party, and no evidence at all that they do so for promises broken by outgroup politicians. 6
Additional Tests
We conducted several preregistered secondary analyses. First, we tested whether the effects of broken promises are moderated by prior support for the policy the promise deals with and by the importance respondents attach to the policy domain. We did not find consistent interaction effects for these moderators (Supplemental Information Section K). Second, a comparison of the broken and fulfilled promise conditions to the control group (Supplemental Information Section H) suggests that ingroup supporters respond more strongly to broken than fulfilled promises when the outcome is policy approval. Additional non-preregistered analyses found no evidence that the effects are driven by specific leaders (Supplemental Information Section I).
Discussion
Campaign promises are highly salient in elite communication and media coverage, and they constitute a central aspect of citizens’ political information environment. While some theories argue that voters penalize political actors who break their promises (Born et al. 2018; Markwat 2023; Matthieß 2020, 2022; Tavits 2007), other theoretical accounts suggest that the impact of information in campaigns is limited due to strong prior opinions and partisan loyalties (e.g., Bonilla 2024; Green et al. 2002; Naurin et al. 2019a). The current study addresses these competing theoretical expectations by systematically testing the conditions under which information about broken and fulfilled promises shapes citizens’ attitudes.
Based on four well-powered experiments that were conducted in two countries, involved real and fictitious leaders and promises, and covered both partisan and bipartisan issues, we find that the impact of broken campaign promises strongly depends on the political context. First, across experiments, broken promises shape citizens’ evaluations of how well politicians handle the specific policy domain related to the promise. However, broken promises exert a much smaller (in fact, in most cases, close to nonexistent) effect on people’s evaluations of politicians’ overall performance. Importantly, the effects of broken promises on overall evaluations are limited regardless of citizens’ prior policy support and how personally important they consider the policy issue to be.
A second key finding is that upon exposure to broken promises by ingroup (but not outgroup) politicians, citizens employ two reasoning strategies that enable them to maintain their support for the politician despite the apparent transgression: They alter the standards they use to judge ingroup (but not outgroup) politicians by engaging in decoupling, and they justify the transgressions of ingroup (but not outgroup) politicians by engaging in rationalization.
Our results align with a growing literature suggesting that the impact of campaign information on citizens’ attitudes is context-dependent and often limited (e.g., Broockman and Kalla 2023; Druckman and McGrath 2019; Naurin et al. 2019a). We lean on two theoretical approaches to interpret our results. First, the Bayesian updating approach would suggest that general evaluations of performance are less sensitive to new information about broken campaign promises because citizens’ evaluations of real-world leaders are based on a variety of factors in addition to promise breakage or fulfillment. These additional factors include leaders’ personality traits, political experience, performance in other domains, social group membership, and many more (Amsalem and Zoizner 2024; Broockman and Kalla 2023; Dai and Kustov 2022; Hill 2017).
Second, our results can also be interpreted through the lens of partisan-motivated reasoning. While many prior studies have shown that motivated reasoning shapes the processing of political information, very few have delved into the specific mechanisms that enable this tendency. Our findings demonstrate not only that partisan identity can condition the impact of broken promises, but also highlight how citizens keep supporting leaders from their political ingroup in the face of clear evidence of wrongdoing. We demonstrate that citizens employ two mental strategies: They disconnect evaluations of promise breaking from evaluations of job performance, and they downplay and rationalize transgressions in ways that cast their group in a positive light. These results broaden our understanding of how politically motivated reasoning operates. They also open the door for additional research examining (a) whether these two strategies play a role in the processing of other types of political information and (b) what reasoning strategies other than decoupling and rationalization may drive politically motivated responses to information.
Our finding that partisanship plays a role in the processing of information about campaign promises is consistent with research showing that motivated reasoning shapes the way citizens view the political world in general (Taber and Lodge 2006) and how they evaluate potentially problematic behaviors by politicians in particular (Anduiza et al. 2013; Bonilla 2024; Naurin et al. 2019a). However, most studies on the role of partisanship in citizens’ responses to broken promises were conducted in one context. We build on and extend the insights from these studies by investigating the impact of partisanship across a wide variety of contexts and conditions. We show that partisan identity matters for promises across two political systems (a two-party plurality system and a multiparty proportional system), for promises related to both partisan and bipartisan issues, and for real as well as fictitious promises and leaders.
The relative consistency of our findings across the two countries sheds light on debates about how electoral systems shape citizens’ evaluations of politicians’ performance. On the one hand, studies show that electoral systems with dispersed power make it harder for citizens to attribute responsibility and hold representatives accountable (Hobolt et al. 2013; Powell and Whitten 1993), which may explain why elites in such systems are less likely to fulfill their pledges (Thomson et al. 2017). On the other hand, recent comparative work finds no evidence that such institutional differences affect how citizens evaluate promise-breaking representatives (Matthieß 2020). Our findings support the latter perspective, suggesting that strong prior opinions and partisan loyalties can limit the impact of promise breaking on citizens in different institutional contexts.
This research has several limitations. First, our experimental manipulations consist of a single campaign promise. In the real world, however, the effects of promises on individuals may be cumulative, and broken promises may have stronger effects when citizens are continuously exposed to multiple broken promises. We encourage future research to examine long-term effects by using longitudinal experiments or multiwave panels. A second limitation of our study is that it focuses exclusively on policy-based campaign promises, addressing topics such as immigration, national security, and local affairs. However, not all promises are policy-oriented. For example, in multiparty systems, politicians and parties often state during the campaign which parties they are willing (or not) to form a coalition with (Borges et al. 2021). Breaking such promises can lead to different outcomes than those reported here. Third, some of the promises we used were narrower (i.e., more specific) than others. While our manipulation checks suggest people interpreted pledge fulfillment correctly in all cases, further research is needed to systematically examine whether citizens respond differently to narrow versus broad promises. Fourth, we have tested the influence of commitments by political actors to a measurable action or outcome. In the real world, however, politicians often make ambiguous statements that evade straightforward verification or measurement (Tomz and Van Houweling 2009). Future research can clarify whether such statements lead to the same results as those reported here. While we encourage researchers to explore these questions, we believe our findings shed important new light on the impact of campaign promises on the way citizens evaluate politicians.
Supplemental Material
sj-docx-1-hij-10.1177_19401612251317261 – Supplemental material for When Do Broken Campaign Promises Matter? Evidence From Four Experiments
Supplemental material, sj-docx-1-hij-10.1177_19401612251317261 for When Do Broken Campaign Promises Matter? Evidence From Four Experiments by Alon Zoizner and Eran Amsalem in The International Journal of Press/Politics
Footnotes
Acknowledgements
We wish to thank Lior Sheffer, Omer Yair, Stanford’s Political Psychology Research Group (PPRG), and Stanford’s Polarization and Social Change Lab (PASCL) for their helpful and thorough comments and suggestions.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Smart Institute of Communication at the Hebrew University of Jerusalem.
Supplemental Material
Supplemental material for this article is available online.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
