1. Introduction
Despite scholarly efforts to understand the effectiveness of propaganda in authoritarian countries, one question is often overlooked in the literature. How and why does propaganda distinctively affect people who are exposed to propaganda messages under various conditions? Scholars have long known that the effects of propaganda are heterogeneous. Research provides evidence that the effects differ by message content, message sources, and individual characteristics. However, less attention has been paid to varying contexts. This paper specifically addresses how and why propaganda affects propaganda recipients differently when they receive this propaganda information in two distinct situations, namely (1) forced exposure and (2) selective exposure.
Exposure to messages is the first step in the communication process in order for propaganda receivers to form and change attitudes. The existing literature on public opinion has paid considerable attention to the importance of exposure to political communications (Converse, Reference Converse1962; Shapiro, Reference Shapiro2011). However, what is being understudied is that recipients may be exposed to propaganda messages in significantly different ways. The type of exposure may, in turn, lead to distinct downstream effects such as changes in one’s level of support toward a policy or the incumbent regime. This study aims to address this lack of research in the literature on comparisons of forced versus selective exposure to messages within political communications in an authoritarian country.
Propaganda is ubiquitous in authoritarian systems and plays a large role in propagandists’ ability to retain power, regardless of the time in history or the political culture. Existing studies on the effects of authoritarian propaganda present rather inconclusive results. Some show that propaganda can affect the political attitudes of citizens in a way that is supportive of the authoritarian regime (Peisakhin and Rozenas, Reference Peisakhin and Rozenas2018; Weiss and Dafoe, Reference Weiss and Dafoe2019; Williamson and Malik, Reference Williamson and Malik2021), to show its strength and capacity (Huang, Reference Huang2015); lead public opinion (Jin, Reference Jin2019; Piechota et al., Reference Piechota2020; Wong and Liang, Reference Wong and Liang2021); manipulate emotions (Mattingly and Yao, Reference Mattingly and Yao2020); reduce the inclinations of citizens to protest (Huang and Cruz, Reference Huang and Cruz2021; Carter and Carter, Reference Carter and Carter2021a; Reference Carter and Carter2021b); make citizens more compliant (Little, Reference Little2017); and ultimately maintain the stability and legitimacy of the regime (Brady, Reference Brady2009). However, compelling research evidence has also shown that the public is aware of government efforts to influence them through such propaganda, and that citizens often resist propaganda messages (Chen and Shi, Reference Chen and Shi2001; Huang et al., Reference Huang, Wang and Shao2018), so that the mass media in authoritarian regimes cannot produce strong opinions on issues that are not important to individuals (Savin et al., Reference Savin, Kashirskikh and Mavletova2018). The policy approval rate of citizens, and the willingness to repost suspected propaganda messages, could decrease (Huang et al., Reference Huang, Wang and Shao2018). Citizens’ attitudes toward the state could also be worsened, and their desire to migrate abroad could grow stronger (Huang, Reference Huang2015).
Much of the research on the effects of propaganda relies on experimental designs in which individuals are passively and unintentionally exposed to pro-regime media content—that is, they encounter the messages without actively seeking them out (Wedeen, Reference Wedeen2015). These studies typically use randomly assigned, predetermined stimuli. However, this forced exposure raises concerns about external validity, since in real-world settings people often have the option to avoid such content. Indeed, when given the choice, individuals may deliberately steer clear of propaganda (Gaines and Kuklinski, Reference Gaines and Kuklinski2011; Prior, Reference Prior2013; Arceneaux and Johnson, Reference Arceneaux and Johnson2015; Huang and Yeh, Reference Huang and Yeh2019). While there is substantial theoretical work on the concept of selective exposure (Iyengar and Hahn, Reference Iyengar and Hahn2009; Stroud, Reference Stroud2011; Garrett et al., Reference Garrett, Gvirsman, Johnson, Tsfati, Neo and Dal2014), we still lack empirical evidence on whether the effects of propaganda differ depending on whether the exposure is voluntary or not.
In addition, selective exposure has been studied primarily within democracies, since citizens of democracies often have unrestricted access to a diverse range of ideological information (Iyengar and Hahn, Reference Iyengar and Hahn2009; Stroud, Reference Stroud2011). Authoritarian regimes, on the contrary, typically place limitations on what information citizens may access, such as restrictions on biased information. Citizens in authoritarian countries can also have their information consumption monitored by the government (Kalathil and Boas, Reference Kalathil and Boas2010; Morozov, Reference Morozov2011; Rø d and Weidmann, Reference Rod and Weidmann2015). Therefore, they usually only consume the information readily available to them (Stockmann, Reference Stockmann2013; King et al., Reference King, Pan and Roberts2017; Chen and Yang, Reference Chen and Yang2019), instead of consuming what may be unfavorable to the regime. However, although people are driven both by a need to keep their beliefs consistent and by a desire to seek out truthful, reliable information, they are not limited to specific social contexts. That explains why, even within restricted information environments, there is evidence that people still have information preferences and engage in selective exposure behaviors (Stockmann, Reference Stockmann2010; Huang and Yeh, Reference Huang and Yeh2019; Robertson, Reference Robertson2017; Wang and Westwood, Reference Wang and Westwood2024). That is likely quite similar to how people exhibit preferences and behaviors in open information environments. For example, Geddes and Zaller Reference Geddes and Zaller1989 have confirmed that the information process occurs in much the same way in authoritarian systems as it does within democracies, with the only difference being that the two populations have different levels of access to information from which to choose. Robertson Reference Robertson(2017) also reveals that citizens in authoritarian regimes exhibit selective exposure behavior as a function of their underlying political orientations, much like citizens in democracies, and citizens with different attitudes toward the regime access and process various kinds of information in different ways.
Despite the fact that authoritarian governments engage in earnest efforts (censoring, information manipulation, and discourse or ideology cultivation) to create a one-sided information environment (Geddes and Zaller, Reference Geddes and Zaller1989) the decentralization and commercialization of the media industry and deepening internet penetration have, over time, significantly altered the influence of mass media within authoritarian regimes. That has made it impossible for contemporary authoritarian governments to exert absolute control over the media (Zhu et al., Reference Zhu, Jie and Shi2013; Price, Reference Price2015; Huang and Yeh, Reference Huang and Yeh2019; Kuhnhenn et al., Reference Kuhnhenn, Lee and Zhang2020). The central authorities in China lack full control over the nation’s media, because at different levels of government there are competing narratives and motivations regarding the media. This often results in an inconsistency in the coverage and content of publications of national parties compared to those of more localized provincial publications (Jaros and Pan, Reference Jaros and Pan2018). In addition, the Chinese government can allow a full spectrum of negative comments about the regime, its policies, and its leadership—instead of engaging in broad censorship—as long as the criticism is unlikely to result in collective action (King et al., Reference King, Pan and Roberts2013). Authorities also avoid widespread and strident acts of censorship because sometimes these have the potential to backfire (Hobbs and Roberts, Reference Hobbs and Roberts2018). Instead, they utilize a strategy of ‘conditional public opinion guidance instead of outright bans on unfavorable reports (Tai, Reference Tai2014).
Research on information consumption behavior in China shows that consumers have preferences that restrict the type of content they elect to consume, and their perceptions of the Chinese government are the main factor influencing their news preferences. Individuals with a more favorable view of the West and a more critical attitude toward their own government are more likely to engage with content that portrays other nations positively or highlights what is negative about China (Huang and Yeh, Reference Huang and Yeh2019). When the political stance of the state media diverges from public opinion, Chinese citizens are inclined to select commercialized, non-state news sources, believing those to be more trustworthy (Stockmann, Reference Stockmann2010). The selective exposure of citizens also varies based on news topics. For coverage of domestic news, Chinese netizens prefer articles that take the opposite position of the government, whereas for foreign news and information they prefer articles that align with the Chinese government’s viewpoint.
This paper contributes to recent studies that have adapted traditional experimental designs to allow participants some degree of choice regarding message content (Gaines and Kuklinski, Reference Gaines and Kuklinski2011; Arceneaux and Johnson, Reference Arceneaux and Johnson2013; Levendusky, Reference Levendusky2013b; Trilling et al., Reference Trilling, Van Klingeren and Tsfati2016; Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2018). Specifically, my goal was to identify distinct treatment effects of propaganda messages for selectors, meaning participants who choose to read propaganda news when given the option, and nonselectors, referring to those who choose not to engage with such content. To achieve this, I employed the design developed by Gaines and Kuklinski Reference Gaines and Kuklinski(2011), which integrates self-selection into the experimental setup. This approach maintains randomized treatment assignment and estimates the overall average treatment effect while also producing estimates of potentially different treatment effects for individuals most and least likely to receive the treatment outside the experiment.
This research chose China as the case study because of its massive and pervasive propaganda system. Citizens who work in public institutions in China are forced to learn propaganda information as part of their job obligation. It is very likely that their knowledge of propaganda messages will be evaluated explicitly and implicitly in ways that impact their career prospects. However, how effective are propaganda messages if people are forced to learn them? There is also a group of people who, motivated in varying ways, voluntarily participate in or contribute to these propaganda messages. Would propaganda effects be different for this group of people who self-select into propaganda exposure?
We may consider that people are more likely to resist information if they are forced to receive it. To the contrary, this article finds that (at least in the case of Chinese propaganda) when given the choice, citizens who select not to receive propaganda news reported higher average treatment effects on pro-regime attitudes than citizens who would have willingly chosen to read propaganda news (i.e., participants in the control group were assigned to read a non-propaganda news). That finding implies that propaganda can be effective even when people are not inclined to view it voluntarily. In addition, this study shows that participants who prefer not to read the propaganda news exhibited higher average treatment effects in rating the issue presented in the news as more understandable and important, compared to how those who preferred to view it rated that news. This phenomenon suggests a possible rationalization pathway, as people may tend to convince themselves that what they were forced to read was worthwhile.
As this research involves collecting data from citizens living in authoritarian regimes and exposing citizens to propaganda news articles, my first priority was to design and implement the study in a way that maximized the health, safety and well-being of those who participated in the research. A detailed description of the steps I took to protect the participants is presented in the Appendix A.1.
I consider this article to have four contributions. First, this study theoretically contributes to the literature on propaganda effects in authoritarian systems, and it also taps into the separate treatment effects on participants who are forced to view versus willingly viewing propaganda messages. Looking at the conditions under which people are exposed to propaganda, this paper adds another important dimension to understanding the effects of propaganda beyond the sources and content of the messages. Secondly, it contributes theoretically to the applicability of forced versus selective exposure theory to an authoritarian system and demonstrates a compelling contrary result that, in some cases, media may work effectively on involuntary viewers. Third, this paper is the first to adopt a non-traditional experimental design in gauging media effects in an authoritarian system. Using Gaines and Kuklinski Reference Gaines and Kuklinski(2011)’s design, which incorporates self-selection into an experiment, this research improves the external validity of the experimental results by taking into account how propaganda effects differ on people who are mostly and least likely to be exposed to the message outside of an experimental scenario. Finally, this study proposes a possible explanation for understanding why propaganda works even on involuntary viewers and thus speaks to the large body of scholarly work on how propaganda helps maintain regime stability in authoritarian countries. It highlights the potential role of psychological rationalization in people’s information processing, thus bringing political psychology theories into an understanding of propaganda effects.
2. Existing literature and hypotheses
2.1. Propaganda in authoritarian countries
Many scholars have analyzed propaganda effects by examining propaganda’s creation, content, distribution, and targeted recipients. Research shows differential effects based on whether or not the propagandist is affiliated with government or official channels (Zhu et al., Reference Zhu, Jie and Shi2013). For example, opinion leaders influence policy opinions and encourage information sharing when they are not perceived as affiliated with the propaganda campaign. If audiences believe these opinion leaders are propaganda agents, these effects disappear (Huang et al., Reference Huang, Wang and Shao2018).
Authoritarian governments pay meticulously strategic attention to techniques of mass persuasion and ideological propaganda (Brady, Reference Brady2009; Ma, Reference Veronica2016). They try to convince the public that they are powerful enough to maintain social control and political order (Huang, Reference Huang2015; King et al., Reference King, Pan and Roberts2017). Therefore, they invest in “soft” state propaganda (Wong and Liang, Reference Wong and Liang2021) to influence public opinion (Jin, Reference Jin2019; Piechota et al., Reference Piechota2020; Liu, Reference Liu2021) and make themselves appear competent (Rozenas and Stukal, Reference Rozenas and Stukal2019).
Grasping the key content of the propaganda can also strengthen the propaganda effect, which can be proved by numerous evidence from China (Stockmann and Gallagher, Reference Stockmann and Gallagher2011; Weiss and Dafoe, Reference Weiss and Dafoe2019; Pan et al., Reference Pan, Shao and Yiqing2020). Brady and Juntao Reference Brady and Juntao(2009) explain that since 1989, the CCP has persuaded the Chinese population that the current political system is most appropriate for China. In doing so, the CCP found a replacement for continuous revolution through a source common to democratic societies: popular support.
Different audiences may also influence the strategies and impacts of the propaganda. For example, research shows that the effectiveness of propaganda may depend on the listener’s value orientation and psychological uncertainty (Rieger et al., Reference Rieger, Frischlich and Bente2017), predisposition to the message (Adena et al., Reference Adena, Enikolopov, Petrova, Santarosa and Zhuravskaya2015), level of political awareness (Geddes and Zaller, Reference Geddes and Zaller1989), and even their family ties to the regime (Kao, Reference Kao2021).
2.2. Forced versus selective exposure to information
Why forced versus selective exposure yields different experimental outcomes is explained in three primary ways in the literature.
Arceneaux and Johnson Reference Arceneaux and Johnson2013 provided an explanation based on the concepts of dilution and differential treatment effects. Dilution describes how the impact of media is weakened when audiences are presented with a wider variety of content options, as some individuals prefer media that simply entertains them. The differential treatment effect arises when individuals are exposed to content without having a choice; if given a choice, their reactions to the content might differ. Several studies by Arceneaux and Johnson Reference Arceneaux and Johnson2013 demonstrate that media effects are influenced by individuals’ preferences. However, it is important to note that this framework assumes that each participant is affected by the media in the same way regardless of whether they have the option to choose the content they consume.
One line of theory suggests that propaganda can sometimes be effective on involuntary viewers, particularly in contexts where repeated exposure is present. A body of literature supports the idea that forced exposure to propaganda, especially in authoritarian regimes, can lead to shifts in attitudes and perceptions. Stroud et al. Reference Stroud, Feldman, Wojcieszak and Bimber(2019) found that when individuals are repeatedly exposed to media content against their will, initial resistance may diminish over time as they gradually rationalize the messages being conveyed. Kim and Cao Reference Kim and Cao(2016) further illustrated this dynamic in the context of government-sponsored conspiracy theories, showing that while viewers may initially resist the content, forced exposure combined with repetition gradually reshapes their beliefs. It echoes the work of Peisakhin and Rozenas Reference Peisakhin and Rozenas(2018), who demonstrated that biased media, when continuously presented without options for avoidance, can influence electoral outcomes by reinforcing regime support. Wedeen Reference Wedeen2015 similarly observed that in authoritarian Syria, even unwelcome propaganda was able to mold public opinion by bolstering the regime’s image of strength and discouraging dissent. In the context of partisan media, Arceneaux and Johnson Reference Arceneaux and Johnson2013 found that repeated forced exposure could weaken opposition and facilitate partial persuasion, particularly in environments where alternative sources of information are scarce. Collectively, these studies suggest that, over time, forced exposure can reduce resistance and encourage internalization of propaganda, eventually leading even reluctant viewers to align with pro-regime narratives.Footnote 2
On the other hand, psychological reaction theory states that because people value freedom of choice, when it is removed it triggers psychological resistance and a desire to regain that freedom (Brehm, Reference Brehm1966). When exposure to media content is assigned to participants in experimental experiences, participants who are not given the choice to select it may experience heightened psychological reactance and, therefore, resist the propaganda messages within that media content. The first preregistered hypothesis states that propaganda will be less effective on participants who prefer not to view propaganda news if given a choice (H1). Psychological reactance is defined in the research literature as consisting of both negative emotional responses, known as affective reactance, and mental resistance in the form of counter-arguing, referred to as cognitive reactance (Dillard and Shen, Reference Dillard and Shen2005; Miron and Brehm, Reference Miron and Brehm2006; Rains, Reference Rains2013). One hypothesis that explains a possible mechanism behind the different effects of media on voluntary versus involuntary audiences is as follows: Individuals who would choose to avoid propaganda content when given the opportunity are likely to experience stronger negative emotional responses if they are exposed to it under conditions of choice (H2a Affective Reactance).
The other explanation for differences in forced versus selective exposure to messages is cognitive dissonance. The original theory of cognitive dissonance, offered by Festinger Reference Festinger1962, described that cognitive dissonance becomes aroused when people’s cognition and/or behaviors are in conflict. As it relates to forced versus selective propaganda exposure, greater cognitive dissonance and resistance to propaganda arises when one is forced to view content instead of being given a choice among several kinds of content. Festinger Reference Festinger1962 stated that when people encounter cognitive dissonance, they are motivated to reduce that conflict. Researchers have explored different methods individuals employ to manage cognitive dissonance, including trivialization and effort justification. Trivialization refers to downplaying the significance of conflicting cognitive elements (Festinger, Reference Festinger1962; Simon, Greenberg and Brehm, Reference Simon, Greenberg and Brehm1995). Effort justification describes the tendency for people who invest effort in an unpleasant task to perceive that task as more worthwhile (Aronson and Mills, Reference Aronson and Mills1959): (1) those who would prefer not to view propaganda news will rate the issues discussed in the propaganda news as less important than those who prefer to view it when given options. (H2b Trivialization); and (2) those who prefer not to view propaganda news will rate the issues discussed in the propaganda news as more understandable than those who prefer to view it when given options. (H2c effort justification)
3. Research design
3.1. Methods
Historically, research experiments on propaganda effects have been done by analyzing how those with different characteristics respond differently to identical information provided to them by researchers. Participants were not given the option of what information they were exposed to, which significantly limited the ability to measure propaganda effects. This traditional experimental design does not provide adequate insight regarding the effects on those who would be naturally inclined to consume that information if given a choice versus those who prefer not to be exposed to it. As Prior Reference Prior(2013) observes, in the most extreme scenario, experimental outcomes are observed solely among participants who would never naturally encounter the treatment, rendering the experimental effect purely hypothetical.
In other words, one of the most challenging problems confronting the concept of average treatment effect (ATE) is partial compliance. This may, in turn, lead to a misleading causal effect by only incorporating the effect of those who are intended to be treated (i.e., ITT, intent-to-treat effect). The implication behind this problem is that there are two potentially distinct groups of people. One group would have chosen the treatment given the choice (selectors), while the other group would not have chosen the treatment if given the option (non-selectors). The effects of the treatment of these two groups of people should be different.
Hence, accounting for individuals’ choices is essential for accurately assessing the impact of information exposure in real-world contexts. Nonetheless, one might question whether permitting self-selection in an experiment undermines the ability to identify causal effects, as it conflicts with the traditional principle of random assignment. In response, researchers have recently taken steps to confront this challenge. Arceneaux and Johnson (Reference Arceneaux and Johnson2013); (Reference Arceneaux and Johnson2015) and Levendusky Reference Levendusky(2013b) have adopted the “Participant Preference Design” in their studies. They assess participants’ media exposure preferences prior to randomly assigning them to various treatment groups, enabling an analysis of how the same media content affects individuals differently based on their preexisting preferences. Gaines and Kuklinski Reference Gaines and Kuklinski(2011) do not assess participants’ exposure preferences in advance; instead, they randomly assign individuals to either a traditional randomization condition or a self-selection condition, where participants are free to choose whether or not to receive the treatment. Knox et al. Reference Knox, Yamamoto, Baum and Berinsky2019 integrate the two methods by first evaluating participants’ media exposure preferences and then randomly assigning them to either a forced exposure group or a self-selection group. Huang and Yeh Reference Huang and Yeh2019 introduce an innovative design termed “Randomized Realization of Self-Selected Treatments.” In this approach, participants first state their preferred content, and then, among those who express interest in a particular article, a subset is randomly selected to receive the article, while the remaining individuals do not receive it. de Benedictis-Kessner et al. Reference de Benedictis-Kessner, Justin, Berinsky and Yamamoto(2019) employ a novel experimental approach called the Preference-Incorporating Choice and Assignment (PICA) design to evaluate the persuasive impact of partisan media. This design randomly places participants into either a forced exposure group or a free choice group, with each condition mirroring methods previously used independently by other researchers (Arceneaux et al., Reference Arceneaux, Johnson and Murphy2012; Arceneaux and Johnson, Reference Arceneaux and Johnson2013; Levendusky, Reference Levendusky2013a; Reference Levendusky2013b).
This study adopts Gaines and Kuklinski Reference Gaines and Kuklinski(2011)’s design because it best suits the purpose of this project, which is to analyze the treatment (propaganda news article) effects of selectors (citizens who prefer to read propaganda news if given a choice) and non-selectors (citizens who prefer not to read propaganda news when given a choice).Footnote 3
Gaines and Kuklinski Reference Gaines and Kuklinski(2011)’s experimental design randomly assigns participants into three groups: (1) a forced treatment group, (2) a forced control group, and (3) a group where people are given the choice of whether or not they want to receive the treatment (self-selection group). Participants in the forced treatment group were not given the option to choose what news articles to view, but were assigned a propaganda news article to read. Participants in the forced control group were also not given an option to choose which news articles to view, but were assigned a non-propaganda news article to read. Participants in the self-selection group were given the option to choose between a propaganda or a non-propaganda news article to view.
The fundamental problem for causal inference is that for any individual we cannot observe both potential outcomes simultaneously. Therefore, in Gaines and Kuklinski Reference Gaines and Kuklinski(2011)’s design, at the individual level, we cannot observe either selectors’ or non-selectors’ propaganda effects because only one potential outcome can be realized. But random assignment ensures that the mixtures of self-selectors and non-selectors are identical, in expectation, in all three randomly formed groups. In other words, the percentage of selectors (or non-selectors) in the three groups should be the same, which can be estimated by calculating the fraction of people who have chosen the propaganda (or non-propaganda) news article from the self-selection group. Therefore, we were able to derive selectors’ treatment effects by using the difference of means between the self-selection group and the forced control group, and the estimated parameter of the selectors’ fraction, with the following formula: 
$\hat{t}_{\mathrm{s}}=\frac{\bar{Y}_{S}-\bar{Y}_{C}}{\hat{\alpha}}$. And the treatment effect of non-selectors can also be similarly identified by 
$\hat{t}_{n}=\frac{\bar{Y}_{T}-\bar{Y}_{S}}{1-\hat{\alpha}}$, where the uppercase subscripts denote the three randomly assigned groups, C(ontrol), T(reatment), and S(elf-selection), and α is the proportion of the population that self-selects into a treatment when given the choice.
The hypotheses were evaluated using data from six online survey experiments carried out in China between 2019 and 2021. These experiments aimed to examine the effects of forced versus selective exposure to propaganda messages, with different propaganda content featured in each survey round. The study was conducted by a Chinese research firm,Footnote 4 which posted the survey link on their website for registered members to participate.Footnote 5
Each survey in this study consisted of four parts. A flowchart detailing the experimental procedure is shown in Figure 1. After providing informed consent, participants first answered questions regarding their political ideology, political interest, consumption of political news, and political knowledge. Then, respondents were randomly assigned to either a self-selection condition or a forced exposure group. In the self-selection condition, participants were presented with two titles for a propaganda news article and one for a non-propaganda article, from which they chose one to read (see Figure 2). Those in the forced group were randomly assigned again to read either propaganda or non-propaganda news.Footnote 6 Overall, participants were randomly placed into one of three conditions: (a) forced exposure to propaganda news, (b) forced exposure to non-propaganda news, or (c) self-selection, allowing them to pick between propaganda and non-propaganda news titles. After reading the selected article, participants reported their reactions and answered questions about their political attitudes. Finally, the study concluded with a demographic questionnaire. All survey materials are available upon request (see Appendix A.2.2).
A flowchart of the Experimental Procedure
Screenshot of the Choosing Titles of Two News Displayed to Participants
To increase the external validity of the experiments, all news articles that served as experimental treatments were chosen from publicly accessible news outlets in China. For the research purpose of my studies, I need multiple propaganda and non-propaganda newsFootnote 7.
In the series of six online survey experiments, I shuffled around the propaganda and/or non-propaganda news presented to participants in the treatment and control groups. The news varied in content. As this series of online survey experiments contains six surveys, the replication of the differential propaganda effects between forced versus selective exposure serves as a robustness check, ensuring to a large extent that the significant propaganda effects we see on selectors versus non-selectors exist across news items with various topics and content. A brief description of the experimental treatments is presented in Table 1.
A brief description of the experimental treatments
3.2. Measurements
3.2.1. Propaganda effectiveness
The effectiveness of the propaganda is the main dependent variable. In this study, I used several measures to calculate the effectiveness of propaganda in terms of pro-regime attitudes, including measures of trust in government (central and local); satisfaction with government (central and local); evaluation of China’s current and future development; and confidence in China’s development, political efficacy, and national pride.
3.2.2. Psychological reactance
To explore the potential mechanisms through which propaganda affects participants differently when they are forced or selected to view propaganda, I used three measurements that examine the potential mechanisms. First, after operationalizing the negative affect of the reaction in the literature (Brehm, Reference Brehm1966; Dillard and Shen, Reference Dillard and Shen2005; Miron and Brehm, Reference Miron and Brehm2006; Rains, Reference Rains2013; Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2018), I asked the respondents in my research to report their experience of negative emotions when reading a propaganda article they were assigned to read.
3.2.3. Cognitive dissonance
Building on established operationalization methods in the literature (Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2018), I used attitude importance to represent trivialization. Participants rated how personally important the issue discussed in the propaganda news was to them. To approximate effort justification, I measured issue understanding by asking participants whether they believed they had a fairly good grasp of the topic covered in the propaganda article. While these measures are not flawless representations of trivialization and effort justification, they are widely regarded as the best theoretical proxies for these concepts (Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2018).
4. Results
Approximately 800 respondents completed each survey, with about 50% female and 80% participants at or below age 30. Although the samples studied in this article are not nationally representative, they map and align with the demographics of Chinese netizens in several demographic characteristics. An overall statistics summary of the sample is shown in Appendix A.3.1.
Appendix A.3.2 shows the balance plots for the six surveys and the combined data from the six surveys. It shows that randomization among participants is successful in all three conditions in all surveys. Appendix A.3.3. shows manipulation checks of the experimental interventions. Respondents showed reactionsFootnote 8 that align with what each propaganda news item was intended to arouse in all studies.
I used 12 individual dependent variables to measure participants’ pro-regime attitudes. To maximize the precision of the estimation, I conducted principal component analysis (PCA), restricting several pro-regime attitude variables (8 out of 12) onto one component. This single component explains around 70%-80% of the variance, and the loading of each variable suggests that they all reasonably conform to a single dimension. In the following analyses, the results are shown using the extracted PCA scores as the dependent variable representing a compound index of pro-regime attitudes. The variables and loading in PCA are shown in Appendix A.3.4. The results for individual dependent variables are presented in the Appendix A.3.5.
In general, I found evidence showing significant positive propaganda effects on participants who prefer not to read propaganda news if given a choice (non-selectors). Meanwhile, there are much less significant positive propaganda effects on participants who prefer to read propaganda news if given a choice (selectors) in all six surveys, and within the pooled data with all six surveys combined. Figure 3 reports the effects of propaganda messages on PCA scores of pro-regime attitudes in all studies.Footnote 9
Propaganda Effects on PCA Scores of Pro-regime Attitudes
First of all, we see from ATE (that is, the average treatment effect of propaganda in the forced exposure arm) that all rounds of surveys (including the pooled dataset), with the exception of Survey 6, show that (similar to the traditional experimental setup of media effects), forcing participants to view propaganda news led to significantly more favorable pro-regime attitudes than was the case with participants in the control group. Except for Survey 6,Footnote 10 it appears that propaganda news did result in significantly higher regime support than non-propaganda news in the forced exposure arm.
Secondly, and what is more interesting, is that we saw significantly stronger propaganda effects on non-selectors (participants who prefer not to read propaganda news when given a choice) compared to effects on selectors (participants who prefer to read propaganda news when given a choice) across all rounds of surveys (including the pooled dataset).
More specifically, let us look at citizens’ pro-regime attitudes in the pooled results with all six surveys combined. If we only look at the average effects of the treatment, the propaganda news seems to significantly elevate participants’ pro-regime attitudes by about 0.171 points. But only when examining the heterogeneous treatment effects among selectors and non-selectors do we find that all the positive treatment effects come from non-selectors. For non-selectors (participants who preferred not to view propaganda if given a choice), the propaganda increased their pro-regime attitudes by 0.460 points—more than four times greater than the overall average treatment effect (p < 0.01). In contrast, selectors (participants who chose to view propaganda) showed virtually no significant change, with an effect of -0.054 points. This indicates that the propaganda had a much stronger positive influence on non-selectors, with a difference of approximately 0.514 points between the two groups, suggesting that non-selectors were far more susceptible to the effects of propaganda compared to selectors. In percentage terms, non-selectors’ pro-regime attitudes increased by about 46%, while selectors exhibited no substantial change, reinforcing the idea that those initially resistant to propaganda are the most significantly affected by it.
Similar patterns occurred in the six individual surveys, with different propaganda and non-propaganda news being presented to various participants. This series of six online survey experiments, with the pooled results combining the six surveys, repeatedly demonstrated the strong, significantly positive propaganda effects on non-selectors.
On individual dependent variables (as shown in the Appendix A.3.5.), we see that non-selectors report significantly positive attitudes on trust toward, and satisfaction with, the central and local governments across different rounds of the six surveys. In ATE results showing significantly positive attitudes on the above variables, it was always the case that non-selectors contributed the significantly positive results. These findings suggest that propaganda is most effective in influencing individuals who are initially resistant to it, while those already inclined to consume propaganda are not further swayed.
4.1. Who are the selectors?
To examine the factors that shape participants’ preferences for different types of news content within the selective exposure group, logistic regression models were employed. In these models, the dependent variable captured participants’ reading choices, with a preference for propaganda articles coded as 1 and a preference for non-propaganda articles coded as 0. Figure 4 displays the marginal effects of several independent variables on reading preferences, based on data pooled from six studies. Since all independent variables were standardized to range from 0 to 1, each estimate in the figure reflects the change in the likelihood of preferring either type of article when the value of a given independent variable increases from its lowest to highest level, while holding other variables constant. The results show that Chinese men, older respondents, those with greater political interest and more frequent news consumption, Chinese Communist Party members, and individuals with lower income are more inclined to choose propaganda news content.
Impact of predictor variables on reading choices (Combined: propaganda vs. non-propaganda)
Similar patterns emerge across the six individual surveys. For instance, Survey 1 reveals that male Chinese citizens with greater political interest are more inclined to read propaganda news articles. Specifically, respondents became approximately 40.8 percent more likely to choose propaganda news as their political interest increased from the lowest to the highest level. Additionally, increasing age from the youngest to the oldest corresponded to about a 78.2 percent higher likelihood of selecting propaganda news. The gender variable also followed this trend, with males more likely than females to prefer propaganda news. However, membership in the Chinese Communist Party did not significantly influence exposure preferences in this survey. In Survey 2, similar findings appeared: respondents with higher political interest were more prone to select propaganda news when given the option. Survey 3 showed that lower educational attainment was associated with a greater likelihood of choosing propaganda news. In Survey 4, males and those with higher political interest again demonstrated a stronger preference for propaganda news. Survey 5 found that only males were more likely to opt for propaganda news. Lastly, Survey 6 indicated that CCP membership was linked to a higher probability of initially selecting propaganda news to read.
In summary, across the six surveys and the pooled results of the six surveys, the most robust variables indicating a higher likelihood of choosing to read propaganda news when given a choice were males and participants who held a higher-than-average level of political interest.
4.2. Potential mechanisms
This paper examined three potential mechanisms discussed in prior research: trivialization, effort justification, and affective reactance. Other alternative mechanisms are explored in the “Discussion” section. Following established operationalization methods (Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2019), attitude importance was used as a measure of trivialization, with respondents indicating how personally significant the issue presented in the propaganda news was to them. Effort justification was approximated by assessing issue understanding, where participants reported whether they considered themselves to have a fairly good grasp of the topic covered in the propaganda article. Although these measures are not perfect representations of trivialization and effort justification, they are regarded as the closest theoretical proxies (Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2019). Additionally, participants’ emotional responses after exposure to either propaganda or non-propaganda messages were assessed using the I-PANAS-SF scale, a concise and internationally validated version of the Positive and Negative Affect Schedule (PANAS) (Thompson, Reference Thompson2007).
According to the existing literature, we should expect participants who choose not to read a propaganda news item (non-selectors) to exhibit average treatment effects in rating the issues discussed in the news as more understandable but less important, compared to those who willingly choose to read propaganda news (selectors). More specifically, effort justification refers to a particular type of dissonance reduction strategy that people use as a way to confer value on a task that requires them to exert effort. If the reward or perceived reward for that effort does not meet their expectations, a dissonance occurs. To try to decrease that dissonance, the reward is reframed and viewed in a more positive way. Regarding news consumption, experiencing heightened cognitive dissonance may occur when individuals are compelled to read a news article instead of having the freedom to decide whether or not to engage with it. To alleviate the dissonance, one may assign a more positive value to the task in order to justify the effort of forced reading of the news article. The primary motivations for consuming the news are to learn and to monitor the information (David, Reference David2009). Therefore, a positive reward for undertaking the task of reading a news item is that the reader gains an understanding of its content. To evaluate that task as more valuable, one may conclude that after exerting the effort to complete the task, they actually have a better understanding of the content (Stroud et al., Reference Stroud, Feldman, Wojcieszak and Bimber2019). Conversely, trivialization operates in the opposite manner by diminishing the significance of conflicting cognitive elements (Festinger, Reference Festinger1962; Simon, Greenberg and Brehm, Reference Simon, Greenberg and Brehm1995). In this study, participants who were compelled to read an article they did not select themselves may have experienced cognitive dissonance. According to trivialization, to alleviate this discomfort, those forcibly exposed to such content would tend to rate its importance lower than they would have if they had chosen to engage with it voluntarily.
Figure 5 and Figure 6 show the combined results of the six surveys. We see that non-selectors reported significantly higher average treatment effects in scoring than selectors did, on how understandable the issue discussed in the news was(i.e., effort justification). This finding aligns with existing scholarly literature, which frequently observes that individuals compelled to view news media content tend to report higher average ratings of issue understanding compared to those who have the option to choose whether or not to engage with the content. However, contrary to the existing literature, non-selectors scored significantly higher than selectors did on the average treatment effects in rating the extent of importance of the issue discussed in the news (i.e., trivialization). It is possible that when people feel they have to do certain things, they convince themselves that what they are forced to do is nevertheless worthwhile or important. That may then lead people to place higher understanding and importance scores on the things they have no choice but to do. However, I did not see consistently robust results for affect reactance. In other words, I did not find significantly negative emotions reported after non-selectors viewed the news.Footnote 11 System justification theory has a lot of use in understanding propaganda, as it explains the underlying psychological need to justify the social system, even when this need outweighs self-interest and group interest (Jost, Reference Jost2020). We see a possible rationalization pathway being expressed in non-selectors of propaganda as well—as people are motivated to justify that what they are forced to do is actually worthwhile. This implies a potential for the psychological needs of people to profoundly impact the effectiveness of regime propaganda, such that those who unwillingly consume propaganda may justify and even develop pro-regime attitudes due to their own individual psychological makeups.
Potential Mechanisms (Effort justification)
Potential Mechanisms (Trivialization).
5. Discussion
5.1. Preference falsification
Preference falsification, the act of misrepresenting a preference under perceived public pressures (Kuran, Reference Kuran1997), is a plausible alternative explanation to the results obtained in this study. Despite obtaining data anonymously through a socially embedded local survey company online, one may wonder if participants self-select into propaganda news and falsify their pro-regime attitudes out of fear. Although the existing literature on Chinese public opinion has increasingly shown that preference falsification is less of a concern than previously thought (Lei and Lu, Reference Lei and Jie2017; Stockmann et al., Reference Stockmann, Esarey and Zhang2018; Huang et al., Reference Huang, Intawan and Nicholson2022; Nicholson and Huang, Reference Nicholson and Huang2022), I took the following steps to alleviate this concern.
Preference falsification may first operate at the news article selection stage, which may result in higher likelihood of choosing propaganda news among participants, such as CCP members and ethnic minorities. To see who is more likely to opt for propaganda news in the selective exposure group, I conducted logistic regression analyses using respondents’ reading preferences as the outcome variables. Across the six surveys and the pooled results of the six surveys, the most robust variables indicating a higher likelihood of choosing to read propaganda news were males and participants who have a higher than average level of political interest. Being a CCP member only showed a higher likelihood of choosing propaganda news to read in one round of the surveys. In all survey rounds, being an ethnic minority was not correlated with a higher likelihood of opting for propaganda news. Therefore, I did not see sufficient evidence to suggest self-selection in propaganda news out of fear.
However, preference falsification may still operate at the second level, where participants could falsify their pro-regime attitudes out of fear. I have used multiple ways to reduce this concern. First, if preference falsification did exist, non-selectors (i.e. participants who prefer not to consume propaganda news if given a choice) would have, on average, reported higher levels of negative emotions than selectors (i.e. participants who prefer to consume propaganda news when given an option), even after taking into account the possible negative emotions that could arise from the news article for both selectors and non-selectors. As presented in A.3.9, I compared the self-reported emotions of selectors and non-selectors post-treatment. Across the six surveys and the pooled results of the six surveys, I did not see a clear and consistent pattern showing that non-selectors had more severe negative emotions than selectors. In some instances, non-selectors even self-reported significantly higher positive emotions than selectors. Secondly, I conducted a subgroup analysis among participants who reported more or less CCP-aligned political ideology pre-treatment, to see how participants in the two subgroups reacted to the propaganda news post-treatment. We may assume preference falsification is less likely to exist among participants who already indicated that they aligned less with CCP’s political ideology pre-treatment. Therefore, if we observed significantly positive pro-regime attitudes among this subgroup of participants post-treatment, and not only among the subgroup of participants who self-reported to have more CCP-aligned political ideology, we would have additional support for the result that the significant increase of pro-regime attitudes come from propaganda news, rather than preference falsification. As shown in Appendix A.3.10, participants who self-reported having less CCP-aligned political ideology did show significantly positive pro-regime attitudes post-treatment, and sometimes even to a higher extent than those who self-reported to have a more CCP-aligned political ideology before treatment. Third, 12 dependent variables were included to measure pro-regime attitudes. If a non-selector falsified their pro-regime attitudes due to fear of retribution, they would likely have falsified at least multiple dependent variables. For example, an individual who falsified her pro-regime attitudes was likely to rate not just one, but both trust toward the central government and satisfaction with the central government as high. As demonstrated in Appendix A.3.5, we see that this was not the case in all six rounds of surveys. For instance, non-selectors reported differential levels of trust and satisfaction toward and with the central and local governments, indicating substantial meanings other than falsifying pro-regime attitudes out of fear. Fourth, the experimental model Gaines and Kuklinski Reference Gaines and Kuklinski(2011) adopted that was used in this study included randomization, which helped to even out the proportion of participants who may have engaged in preference falsification and left the inference unharmed.
5.2. Alternative explanations
Despite the significantly more positive propaganda effects that I obtained through all rounds of surveys for people who prefer not to read propaganda news than those who prefer to read it, I acknowledge the possibility that this finding may be limited to certain conditions. I also considered the following alternative explanations for my findings.
First, there is a possibility that the main finding may be limited to certain types of propaganda messages. In particular, the first two propaganda news articles (that is, US sanctions on Chinese officials and postponement of Hong Kong’s elections) adopted as treatments in this study may be perceived as more of the nationalistic type of propaganda. Future research on the effect of propaganda may differentiate the types of propaganda, particularly along the line of nationalism, given its prominence in China, to see if and how different types of propaganda affect my main findings.
The alternative explanations of my main findings are discussed in the following. First, because as Geddes and Zaller Reference Geddes and Zaller1989 argued, “⋯ highly aware people are more heavily exposed to government-dominated communication media, but are also better able to resist the propaganda they encounter”, it could be that people who prefer to view propaganda are less affected by propaganda due to their high levels of political sophistication. I found that people who have higher levels of political interests were more likely to be “selectors”, so people who prefer to view propaganda news may be more politically sophisticated in coming up with counter-arguments, leading to less effectiveness. To rule out this explanation, I divided participants into two samples, one including those who have high political interests and the other including those with low political interests. As shown in Appendix A.3.11, in the majority cases, no matter whether participants had high or low political interests, non-selectors showed significantly higher pro-regime attitudes than did selectors.
Secondly, it could be that propaganda is effective even on involuntary viewers because propaganda serves as a signaling function (Huang, Reference Huang2015; Wedeen, Reference Wedeen2015) that deters citizens from engaging in anti-regime activities. However, this study only examined propaganda effects on changing people’s pro-regime attitudes. Future research may look into how propaganda affects voluntary and involuntary viewers’ pro-regime behaviors.
Regarding potential mechanisms, this project proposes one plausible mechanism, rationalization, that explains the significantly greater positive propaganda effects on non-selectors versus selectors. However, I acknowledge that there might be other mechanisms at work that future research can explore.
For example, one alternative explanation of the potential mechanisms that I presented is Zaller et al. Reference Zaller(1992)’s Receive-Accept-Sample (RAS) model. It argues that an individual forms an opinion according to the messages they have received (which is contingent on their level of political awareness), accepted (which is contingent on the consistency with her prior beliefs), and sampled (which is contingent on what issues hold priority at that moment). Individuals who are more politically aware are therefore more likely to receive messages, but less likely to accept messages that are inconsistent with their prior attitudes. As I found that non-selectors tend to be people who have low levels of political interest, one could argue that non-selectors’ lack of political awareness led to the larger propaganda effects on their scores of understandability and importance of the issues discussed in the news. To rule out this explanation, I divided participants into two samples, where one sample included those who had high political interests and the other included those with low political interests. As shown in Appendix A.3.11, no matter whether participants had high or low political interests, non-selectors showed significantly higher scores on understandability and importance of the issues discussed in the news than did selectors.
To alleviate the concern that non-selectors showed significantly more positive treatment effects than selectors due to selectors already being regime loyalists (i.e., ceiling effect), I first conducted a subgroup analysis among participants who reported more or less CCP-aligned political ideology pre-treatment. That enabled me to see how participants in the two distinct subgroups reacted to the propaganda news post-treatment. If the ceiling effect existed, we might assumed that we would observe significant treatment effects only among the subgroup of participants who self-reported to have less CCP-aligned political ideology. Therefore, if we also observed significantly positive pro-regime attitudes among the subgroup of participants who self-reported to have more CCP-aligned political ideology post-treatment, but not only among the subgroup of participants who self-reported to have less CCP-aligned political ideology, we would have additional support to alleviate our concern regarding the ceiling effect. As shown in Appendix A.3.10, we did not see a pattern throughout the rounds of surveys that indicated the treatment effects could only be observed among participants who self-reported less CCP-aligned political ideology. In particular, the data from the “combined” survey showed that there were significantly positive treatment effects for both subgroups of participants who reported more or less CCP-aligned political ideology. Additionally, non-selectors in both subgroups showed more significantly positive treatment effects than selectors, regardless whether or not they are regime loyalties.
Similarly, to rule out the potential explanation that higher treatment effects among non-selectors than selectors were due to lower levels of prior news consumption, I conducted another subgroup analysis among participants who reported higher or lower levels of political news consumption. As presented in Appendix A.3.12, when I combined six rounds of survey data for both subgroups (where respondents reported higher and lower political news consumption levels), it was non-selectors who showed significantly positive treatment effects in both situations.
Finally, I addressed the concern that the higher treatment effects observed among non-selectors compared to selectors may have been due to differences in how much attention each group paid to the news article. To do so, I examined whether forcing participants to read a specific news article resulted in a significantly higher degree of accuracy than when participants were given an option to choose which news article they wanted to read. As demonstrated in Appendix A.3.14, the results are mixed and inconclusive, with an affirmative yes to the above question in the first two rounds of surveys and no in the rest four rounds of surveys.
6. Conclusion
Scholars examining propaganda effects in authoritarian systems have long focused on the intention, content, and ways in which propaganda messages are created and disseminated, while overlooking the various conditions under which people are exposed to propaganda. Future research is encouraged that can focus in greater depth on answering the question of why propaganda yields differential effects on people whose viewing of it is forced (non-selectors) versus those who willingly select to view it (selectors). This study attempts to contribute to this area of research by focusing on the two conditions of forced versus selective exposure in the case of China.
We might assume, by following the discussions and conclusions drawn from the existing literature, that citizens who are forced to be exposed to propaganda messages in authoritarian countries will develop negative attitudes toward the propagandist government. That seems like a plausible response, since people tend to value personal freedom and the right to make their own choices. However, at least in the case of China and its propaganda, that assumption may not be completely valid. Across all six rounds of my surveys and the combined results of the six surveys, the outcomes instead revealed that non-selectors (citizens who prefer not to read propaganda news when given an option) report higher average treatment effects on pro-regime attitudes than selectors (citizens who prefer to read propaganda news articles when given a choice), with the control group reading an assigned non-propaganda news. That indicates that propaganda is effective even for people who are not inclined to view it voluntarily. Given how repetitive propaganda is in China, we would also expect the propaganda machine of that government to be more effective than the one-shot scenario that I deployed in these survey experiments.
In addition, the examinations into potential mechanisms on why non-selectors reported higher average treatment effects on pro-regime attitudes than selectors were revealing. This indicates that non-selectors placed higher average treatment effects in scores based on how understandable (i.e., effort justification) the issue discussed in the news was to them and how important (i.e., trivialization) the issues discussed in the news were. The result for effort justification aligns with what most scholars have found, while the result for trivialization is the opposite of what the existing literature finds. Both results suggest a possible rationalization pathway within this phenomenon in which people tend to convince themselves that what they were forced to read was worthwhile, and that is worthy of more scholarly investigation.
This study indicates that even when Chinese citizens consume propaganda involuntarily, they report greater average treatment effects on pro-regime attitudes than their counterparts who willingly consume the same propaganda. Although I acknowledge that rationalization might be only one possible mechanism that could explain the main result, the practical implication of this kind of psychological rationalization is significant in terms of the rather unexpected way in which it supports the effectiveness of authoritarian propaganda in affecting public opinions (at least in China). The fact that this phenomenon, related to psychological needs, is under-reported or less examined in the scholarly literature is also significant. More investigation into how and why the impact of propaganda is not always based on top-down dominance but is instead influenced by a person’s own unique psychological needs and rationalizations would contribute to a better scholarly understanding within this area of important research.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2025.10040. To obtain replication material for this article, https://doi.org/10.7910/DVN/E3LCFC.
Open access
