Why Politicians Won't Apologize: Communication Effects in the Aftermath of Sex Scandals

Abstract

A startling feature of the countless recent sex scandals involving politicians has been the almost complete lack of public apologies. This note explores the electoral incentives politicians face when crafting communication strategies in the aftermath of sex scandals. We focus on two communication strategies – denials and apologies – and assess their impact on incumbent support across a wide range of scandals that vary in terms of the seriousness of the charges as well as the availability of evidence. Using data from a series of survey experiments, including over 10,000 respondents we find that citizens punish incumbents who apologize, even in the case of accusations that appear the least serious in the eyes of voters. Moreover, apologies fail to generate political support compared to denials, even in cases when voters are exposed to evidence. This suggests that in most cases apologies are simply not politically viable communication strategies.

Introduction

In the last few years, high-profile politicians have been accused of sexual misconduct of varying levels of transgressions, with different political ramifications. Politicians overwhelmingly deny charges, even in the face of seemingly strong evidence. Most notably, former presidents Donald Trump and Bill Clinton were able to do so without losing political support, as shown in our analysis in Supplementary Material S5.

By contrast, the few incumbents – such as Anthony Weiner and Mark Sanford – who apologized for their behaviour appeared to pay significant political costs. Such patterns present a normative problem beyond the transgressions themselves. First, the absence of public apologies is problematic because they prevent reconciliation and forgiveness (Fehr, Gelfand, and Nag 2010) that could be facilitated through the perpetrator's remorse (Hornsey et al. 2019). Moreover, denials fuel uncertainty (Schumann and Dragotta 2020), provoke divisive responses reflecting pre-existing dispositions (Fischle 2000), and therefore, increase polarization.

The goal of this note is to explore the possible role of voter preferences in this lack of such apologies and to consider the empirical attributes of public scandals that limit or enhance electoral incentives for honest communication. To measure the public incentives for apologies, our study compares their impact to denials.¹ These strategies have alternative rationales when managing the impressions of voters and, therefore, offer exclusive choices for the politician in their response strategy. On the one hand, denial creates doubt about the veracity of the accusations at the cost of creating the impression of hypocrisy (Scott and Lyman 1968). On the other hand, apologies maintain the honest image of the accused at the cost of revealing the guilt. We ask under what circumstances voters prefer apologies compared to denials when it comes to the incumbent's support in sex scandals.

To assess the causal effect of elite communication on voters' evaluation, we rely on a series of survey experiments in which we exposed over 10,000 American adults to fictional sex scandals.² The experiments varied (1) the severity of the accusations, going from a consensual extramarital affair to rape, (2) the partisanship of the incumbent (co-/out-partisan, or non-partisan), (3) the communication used by the politician (denial or apology), and (4) the availability of evidence.

Across our studies, voters punished incumbents for their apologies even in the case of scandals that they did not perceive as particularly serious. In addition, denials increased support for politicians to a greater degree in the case of more severe accusations. This suggests that voters place much greater emphasis on innocence rather than honesty. Furthermore, we demonstrate that these communication effects persist, even in circumstances we thought would favour honest communication, when the politician is a co-partisan and when the evidence is revealed to voters. First, while accused politicians garner more support from their co-partisans on the baseline, elite responses impact public opinion similarly across partisan aisles; therefore, apologies fail to sway even the likely voters of accused incumbents. Second, apologies proved ineffective even when we provided respondents with additional evidence corroborating the accusation. Compared to voters who only received the evidence, the additional, ex-ante, or ex-post apology did not increase voter support.

Our study demonstrates that, while political pundits often lament the lack of honest apologies, especially in the context of sex scandals, it seems that public opinion simply doesn't offer sufficient incentives for politicians to apologize instead of denying in any of the circumstances we study – even considering the possibility that evidence can refute these denials, and therefore, the politician is caught lying. Based on our findings, we conclude that politician apologies are unlikely because (1) scandals that become public are perceived as serious, (2) conclusive evidence for the allegations are rare, and (3) even in the availability of evidence, voters would not punish dishonesty.

Political Scandals

Political scandals are mediated events that feature transgressions of universally accepted norms by political actors leading to public outrage (Esser and Hartung 2004; Thompson 2000). The study of political scandals is inherently multidisciplinary. Sociologists and anthropologists have focused on scandals' cultural embeddedness, political scientists and psychologists have explored their political impact, and communication scholars have examined the role of mass media. After a short review of the relevant literature, we introduce our theoretical account of how politicians’ responses influence voters' reactions to scandals.³

A large body of the political science literature has established the damaging impact of scandals on election outcomes and the favourability of the politicians involved (Banducci and Karp 1994; Fernandez-Vazquez, Barbera, and Rivero 2016; Hirano and Snyder 2012; Welch and Hibbing 1997; Winters and Weitz-Shapiro 2016). However, studies also found a great deal of heterogeneity in voters' responses depending on the type (Carlson, Ganiel, and Hyde 2000; Doherty, Dowling, and Miller 2011); the context of the scandal (Carlin, Love, and Martinez-Gallardo 2015); the traits of the accused politicians, like gender or ethnicity (Berinsky et al. 2011; Bhatti, Hansen, and Leth Olsen 2013; Smith, Powers, and Suarez 2005); the predispositions of the voters such as partisanship (Costa et al. 2020; Klar and McCoy 2021; Sikorski, Heiss, and Matthes 2020); or sexism (Masuoka, Grose, and Junn 2021).

A necessary condition for a transgression to become a scandal is communication or mediation (Esser and Hartung 2004; Thompson 2000). Past research emphasized the central role of the mass media in covering scandals (Costas-Perez, Sole-Olle, and Sorribas-Navarro 2012; Nyhan 2015; Sikorski 2017), but much less is known about how politician responses influence voters' opinion about scandals (Sikorski 2018, 3,120).

Motivated by theories on political image restoration and blame management (Benoit 2014), we argue for the central role played by the politicians' communication on voters' reactions to scandals. We know that politicians can avoid blame by using strategic communication (Grose, Malhotra, and Van Houweling 2015; McGraw 1991). This might be especially true for scandals regarding sexual misconduct, in which case competing frames (Chong and Druckman 2007) might allow for more strategic discretion on the side of the politician.

Apologies and Denials

While some experimental studies have explored the impact of political communication in the aftermath of sex scandals, their results have been mixed (Sigal et al. 1988; Smith, Powers, and Suarez 2005). Specifically, the relative impact of two key, but fundamentally different communications – public apologies and denials – is unclear, possibly because researchers have tested these responses across different types of sex scandals that may moderate the communications' relative effects (Schumann and Dragotta 2020). In the following, we develop a theory about the conditions under which denials or apologies might be more effective in avoiding negative electoral consequences.

Building on existing work, we propose that two factors drive voter preferences during a scandal. First, voters consider the available evidence and elite messages to gauge whether the accused politician is guilty of the charges. Second, voters also consider the degree to which the politician's communication regarding the scandal is honest. While the question of why voters care about guilt is trivial, honesty might require more explanation. We argue that besides its moral value – facilitating reconciliation and repair (Fehr, Gelfand, and Nag 2010; Hornsey et al. 2019; Schumann 2014; Schumann and Dragotta 2020) – honesty also matters for voters because it signals the politician's performance in representing them in different domains concerning other democratic outcomes.

Voters' rationale presents a dilemma for the politician: denying can elicit doubts about the accusation, but also create an impression of hypocrisy (Scott and Lyman 1968). Conversely, apologies maximize the perceptions of honesty and necessarily reveal guilt as well. We argue that the choice will depend on the severity of the accusation and the likelihood of evidence. First, the more severe the charges are, the more importance voters might attach to innocence compared to honesty, leading incumbents to deny the charges. Second, when the politician anticipates that the wrongdoing will be likely proven, we expect them to apologize, given the chances of getting caught and appearing both guilty and dishonest.

Taken together, we view these factors – scandal severity and the difficulty of collecting evidence – as crucial, though not comprehensive dimensions of sex scandal communication. Other, individual-level factors such as voters' partisan predispositions could also influence the relative effect of apologies and denials: co-partisans could perceive transgressions as less serious, the accusations or evidence less credible, and the politicians' denial more honest, in line with motivated reasoning (Leeper and Slothuus 2014). This could further complicate the optimal response strategy in different scandal conditions. Mapping the impact of two communication strategies across these dimensions will thus provide rich evidence regarding the type of scandals where apologies can be expected in the real world. In what follows, we explain how we operationalized these dimensions and how we tested the impact of elite communication.

Experimental Design

We designed a series of three experiments to study the effect of denials and apologies in the aftermath of political scandals. The surveys were fielded through Lucid (Coppock and McClellan 2019), a marketplace for online survey panels.⁴ In each experiment, subjects were asked to read a vignette about a fictional sex scandal featuring a sitting member of the US Congress, John Miller, in the form of a news story. They were then randomly assigned to read one of two communication strategies: either an apology in which Miller admitted the accusations and apologized for his behaviour or a denial where Miller denied all the accusations and called the news story ridiculous.

Finally, at various points in the experiment, they were asked their opinion about (1) the seriousness of the accusation, (2) their belief in the allegations, and (3) whether they thought Miller should resign.⁵ Across the three experiments, however, we executed some modifications in the design to address some conditions under which the communication effects are observed. We offer more details about the experimental design and procedure in Supplementary Material S2–S4.

Experiment 1

The goal of the first experiment (n = 7,880) was to compare the impact of these communication strategies across a broad range of sex scandals with varying severity.⁶ In the ‘least serious’ condition, Miller was accused of having a long extramarital affair, while in the most serious one, Miller was accused of rape. The in-between conditions included harassment, sexting, and sexual assault, resulting in five ‘seriousness’ conditions.⁷ We also randomly assign subjects to one of the two response strategies by Miller (apology or denial).

Finally, we randomized Miller's partisanship to either match respondents' party affiliation (measured pre-treatment), describe him as an out-partisan, or leave his party unspecified. On the one hand, this makes our vignettes more realistic and provides a more conservative test of the effect of communications, as partisans could simply disregard the politicians' communication and express preferences in line with partisanship (Broockman and Butler 2017). On the other hand, this feature of the design also allows us to test if the possible effect of communication varies across partisan sub-groups.

Experiment 2

The second experiment (n = 998) served as a replication of our first one with two important modifications. First, we utilized a repeated measures design in which the respondents' support for Miller, as well as their belief in the allegation, was elicited both before and after the incumbent's response. We use this first measure as a control group so that we can compare subjects' attitudes after each of the two kinds of responses to a baseline, which allows us to observe their absolute effects on resignation support. Second, in this experiment, we only included a subset of the scandal types.⁸ This decision was driven by a consideration of statistical power due to the lower sample size.

Experiment 3

In the third experiment (n = 1,244), we provided respondents with new information in addition to the scandal allegation and Miller's communication – a piece of evidence supporting the charges. The purpose of this modification was to tease out the effect of apologies and denials in a setting where voters are quite sure that the accusations are true, and thus should not update their beliefs about the incumbent's guiltiness based on the apology or denial. Subjects were randomly assigned to receive evidence of the accusation,⁹ elicited either before or after Miller's apology or denial.

In this experiment, we did not vary the scandal's seriousness.¹⁰ Moreover, we used a repeated measures design and asked each dependent variable after each of the three information stages: one after the accusation, one after the communication (or evidence), and one after the evidence (or communication). This yields comparisons of the measures under an allegation, evidence, communications, and evidence and communications (in a permuted order) conditions.

Results

The Relative Impact of Apology vs. Denial

Figure 1 presents our findings from the first experiment. As for the manipulation checks, the left panel shows that we successfully induced variation in the perceived seriousness of the scandal (55 to 75, on a 100-point scale). Second, we show that compared to their prior beliefs – elicited before the politician's response – subjects' posterior beliefs vary according to the response types.¹¹ Across each of the five scenarios, denials reduced while an apology increased voters' beliefs that the charges were true, though compared to denial, apologies' absolute effect on beliefs was much larger. This suggests that apologies are much more indicative of guilt, while denials are less effective in building uncertainty about the accusations – especially in the case of the most serious accusations.

Figure 1. Scandal type, incumbent response and citizen attitudes (Study 1).

Note: The left panel shows the change in the mean perceived seriousness of the allegations by scandal type, obtained from the non-partisan group responses (measured prior to the politician response). The middle panel shows respondents' mean reported belief that the allegations are true, by scandal type. Prior beliefs were asked before the politician's communication. Apology and denial beliefs were asked after the communication. The right panel shows the mean resignation support for each scandal type and by response strategies. Error bars denote 95 per cent CIs. The regression results are in Supplementary Material S6.

We present our main finding in the right panel of Fig. 1, which plots average support for the accused incumbent as a function of scandal severity and the type of response (apology vs. denial). We observe a clear trend: the relative effectiveness of denial increases dramatically for allegations seen as more serious. For instance, the relative effect of denial is −5.5 points in the case of an extra-marital affair but −18.6 points in the case of rape. This pattern suggests that the relative weight that voters place on honesty and innocence varies as a function of the seriousness of the scandal. The cost of lying is comparable to the cost of the wrongdoing in the case of ‘light’ scandals; however, the returns of the strategies diverge dramatically in more serious scandals. At the same time, even for the scandals that are seen as less serious by our respondents, apologies fare worse. Therefore, honesty is not valued higher than innocence by voters. This suggests that politicians are always better off denying it, independent of the seriousness of the accusations.

The Absolute Impact of Incumbent Communication

While our findings are quite clear regarding the relative effectiveness of the two communication strategies, the question remains whether apologies are just less effective modes of incumbent communication, or whether they hurt politicians in the eyes of voters. To explore this question, we conducted a replication experiment in which we measured the outcomes (that is belief in the allegation and incumbent support) for a subset of allegation conditions both before and after the response. The pre-communication measures serve as a baseline to which we can compare post-communication levels of disapproval. The second experiment's findings demonstrate that denials are steadily effective compared to the pre-communication baseline, while apologies are not rewarded in low-seriousness conditions; furthermore, they hurt incumbents as the severity of the scandal increases (see Fig. 2).

Figure 2. Scandal type, incumbent response and citizen attitudes (Study 2).

Note: The zero line depicts the mean resignation support for subjects who only received the allegation, but no politician response. The left panel shows the coefficients for the mean reported belief in the allegation by response types and by scandal type compared to the reference group. The right panel shows the coefficient for the mean resignation support by response types and scandal types compared to the reference group. Ninety-five per cent of CIs are adjusted for the repeated measures by respondent IDs. The regression results are in Supplementary Material S9.

The Role of Further Evidence

The other limitation of our first experiment is that it could not parse out two potentially counter-veiling effects of apologies: revealing guilt and demonstrating remorse and honesty. By incorporating a phase in our design in which subjects receive additional evidence supporting the accusations, we can test the impact of communication in a setting where voters have a high degree of confidence that the accusations are indeed true, or, in other words, we can adjust for the ‘guilt revealing effect’ of apologies. As a result, we can see the potential electoral reward for honesty without it being masked by the effect of guilt. In addition, we can also consider the possibility that voters might punish a denial if lying is revealed, and this could change the relative electoral effectiveness of apologies and denials in the aftermath of scandals and, therefore, politicians’ behaviour.

We first estimate the impact of apologies compared to a baseline where the subjects are exposed to additional evidence confirming the charges, but no communication (observed from the respondents who received the evidence before the incumbent message). As Fig. 3 shows (right panel), an apology has a modest and statistically insignificant mitigating effect. Voters do not seem to appreciate apologies either before or after evidence is revealed. These results are surprising as the information environment making the accusations likely creates the most favourable conditions for apologies, according to our account.

Figure 3. Incumbent response, evidence and citizen attitudes (Study 3).

Note: The zero line depicts resignation support for subjects who only received evidence of the accusation, but no communication. Left panel: Mean reported belief in the allegation by treatment groups. Right panel: Mean resignation support by treatment groups. Ninety-five per cent of CIs are adjusted for the repeated measures by respondent IDs. The regression results are in Supplementary Material S10.

Our last analyses assess the impact of denial in a context where subjects read the evidence after a denial and the politician is caught lying. Voters do not seem to punish this when compared to a simple revelation of guilt. As Fig. 3 shows (right panel), voters who received a denial and then the evidence, only support the resignation of the politician approx. 1 percentage point more than voters who only received the evidence. Furthermore, the left panel of Fig. 3 shows that this is not the case because, even with evidence arising later, denial still keeps the mean reported belief in the allegation lower compared to the evidence-only group. The difference is effectively zero between these groups' reported outcomes.

Even more devastating are the effects we observed for the group for whom the denial followed the evidence. Denial can reduce resignation support by approx. 4 percentage points compared to evidence only, and this comes from the denials' ability to fuel uncertainty (the mean reported belief in the allegation is 7 percentage points lower for this group than the evidence-only group). Altogether, apologies fail to garner support compared to denials even in the ‘ideal case’ – when further evidence is revealed.

One explanation for this result is that the additional evidence failed to fully convince respondents that the incumbent was guilty.¹² Beyond a seeming limitation of our experimental design, these patterns may also reflect a key feature of mediatized scandals where the media or even authorities are seen as controversial and political. In such cases, denials can capitalize on the lingering doubts about the evidence. In any case, even if evidence was uncontroversial, voters might still not reward apologies because of what they signal about the politician's character, as we attempt to explain in more detail in the conclusion.

The Role of Partisanship

Finally, how do partisan contexts change these findings? Prior to communication, co-partisan respondents perceived the allegations as less serious and less credible than out-partisan respondents. Consequently, co-partisan politicians, on average, were less likely to be sanctioned for the scandals than out-partisan politicians. Notwithstanding these absolute differences, the relative impact of denial compared to apology was not significantly different for co-partisans compared to out-partisans. On the one hand, denials resulted in lower post-treatment beliefs in the accusations, and with similar magnitude across the partisan contexts.¹³ On the other hand, it appears that co-partisans did not reward apologies vis-a-vis denials more compared to out-partisans, not even for less serious transgressions (see Supplementary Material S7).

Conclusion

Our results on voters' reactions to scandals and communications demonstrate how politicians involved in sex scandals can garner support by denying charges rather than apologizing. While both victims and voters at large would benefit from apologies, public opinion seems to generate incentives for denials instead, even in the case of wrongdoings the public itself does not view too serious, minimizing the problem of guilt, and even in an ideal world where guilty politicians denying charges would be caught lying. All in all, our study paints a dark picture of accountability in the era of the #MeToo movement: even increasing media scrutiny and public outrage fail to force incumbents to take responsibility for their actions.

There are two limitations of our study that future research should address. First, while our experimental approach addresses challenges that would make it difficult to study the same question relying on observational data (see Supplementary Material S5), studying voters' reactions in hypothetical situations could deviate from real-life effects because partisan-motivated reasoning is expected to be stronger towards well-known politicians or in cases with actual stakes.

Second, while our study offers some hints as to why voters do not appreciate apologies, it does not produce direct evidence on the matter. Perhaps, the politician's response strategy also provides an indirect signal about their leadership style. ‘Masculine defaults’ in politics (Lombard, Azpeitia, and Cheryan 2021) and misogyny in political communication (Boatright and Sperling 2019) might provide an advantage for male politicians' denials over apologies in sex scandals, as the earlier meets gender-normative expectations more about strength, boldness, and aggression, while apologies could be connected to notions of weakness.¹⁴ On the other side of the debate, voters might think apologies are fabricated (Lepore 2022; Schumann and Dragotta 2020) and only used pragmatically for political survival.

Given these limitations, future research could combine the approach presented here with both descriptive and causal studies of elite communication in the aftermath of real scandals using observational data. Moreover, future research should probe the impact of incumbent responses across a broader set of communication strategies, in other types of scandals such as corruption cases and test theories on how apologies influence voters' evaluation of the candidate in other important aspects.