Introduction

On March 18, 2021, during the strategic dialogue between China and the U.S., Yang Jiechi, the top Chinese foreign policy official at that time, stressed that “(t)he United States has its own model of democracy, and China has its own style” [54]. Furthermore, only days before the U.S. held the Summit for Democracy, the Chinese government published a white paper titled “China: Democracy That Works” to highlight its regime’s democratic elements while criticizing Western democracy as bourgeois and spurious democracy [76]. These statements indicate the Chinese government’s official stance on democracy and intention to strengthen its narratives.

However, official rhetoric does not always translate into genuine action. For instance, despite the People’s Congress being a symbol of representative democracy, its imbalanced functions have led to a notable deficit in representation [90]. Moreover, the broader population may not always align with the government’s narrative. In an ironic twist, in November 2022, several protests erupted across mainland China against the zero-COVID policy enforced by the authorities. Two months later, the residents of Wuhan and Dalian demonstrated against the reduction in the local government’s medical benefits. Participants in these movements generally believe that these policies infringe on their freedom and democratic rights. These incidents suggest that the discourse presented by high-level Chinese officials may not accurately or fully represent the views of the Chinese people. Thus, concerning the “D-word,” there could be a “two-skin” (liangzhangpi) issue. Additionally, the popular understanding of democracy as a concept is not always fixed. What type of democracy do the Chinese people desire, particularly as China enters Xi Jinping’s new era? This study aims to examine this question, bridging the gap between official stances and popular sentiments, and understanding the evolving nature of democratic perceptions in China.

Undoubtedly, democracy is a captivating yet controversial concept in political science today [27, 40, 50, 97]. Although numerous eminent political theorists have defined democracy in various ways [21, 34, 65, 66], and people across different countries comprehend the “D-word” differently [12, 13, 17, 47], as a value, democracy has been widely accepted worldwide [24]. In particular, since the third wave of democratization has spread globally, no leader from any regime can publicly reject democratic values, regardless of their true beliefs. In recent years, democracies have encountered unprecedented challenges. The failure of the Western demonstration effect, the rise of populism, and the emergence of radical right-wing parties pose significant threats to liberal democracy [41, 57, 60]. Simultaneously, the phenomenon of authoritarian durability has pushed social scientists to reconsider democratic theories in non-Western societies. As Nathan indicates, “[a]uthoritarian systems have been with us longer, have ruled more people, and for all we know may rule more people in the future than democratic system” [56]. In the case of the People’s Republic of China, a prime example of regime durability under long-term one-party rule, the nation has significantly challenged the existing theories and predictions related to democratization and attempted to establish its own democratic narratives, particularly socialist democracy with Chinese characteristics.

Several renowned indicators, including Freedom House, Polity IV/V, and Varieties of Democracy (V-Dem), consistently categorize China’s regime type as authoritarian or autocratic. However, multi-round representative academic surveys such as the Asian Barometer Survey (ABS) and the World Value Survey (WVS) indicate high levels of democratic support and satisfaction in China. How can this paradox be explained? Setting aside the “D-word” discourse controlled by Chinese propaganda, the most plausible explanation is that the Chinese popular understanding of the “D-word” is unique. Do Chinese people have a democratic perception of Chinese characteristics? Are there any cognitive differences between the Chinese official discourse and people’s understanding? At least two contributions can be made by addressing these questions. Theoretically, this study can update empirical democratic studies. Practically, this can help us better understand the popular Chinese understanding of democracy and the durability of China’s regime.

To answer the aforementioned questions, this study first reviews official democratic discourse and academic research and then adopts a novel approach as a supplement to traditional methods, which combines automated text analysis and semi-supervised machine learning, to analyze open-ended questions about democracy in a Chinese national survey. The chief findings reveal that, compared with traditional close-ended survey batteries, nearly half of the Chinese respondents have no idea about the direct meaning of democracy. For those who do understand the “D-word,” freedom of speech, minben zhuyi, and consultation with consent principle form the three pillars of democratic cognition in China while recognizing democracy as competitive elections or multi-party systems is not a mainstream conception among ordinary Chinese people yet. Moreover, based on the analysis of open- and closed-ended batteries, conclusions regarding the forms of democratic perceptions differ. To some extent, these distribution features suggest that ordinary Chinese people have a more complex understanding of democracy than was previously assumed. To some extent, this complexity reflects that the current Chinese communist regime suffers less pressure from competitive elections, which is often labeled as a key feature of democracy.

The remainder of this paper is organized as follows. The second section reviews the distinct but interlacing discourses about democracy in China from official and academic perspectives, and based on this, identifies the typical types of democratic discourse. The third section introduces the data, core research questions, and methods. Following this, the results of the analysis and explanations are presented. The last section reviews the findings and their contributions to the literature.

Understanding China’s Democratic Discourse: From Official and Academic Perspectives

The Dynamic Chinese Official Democratic Discourse

Existing studies reveal that compared with modern European state formation and long-run democratic development, China’s long history of united state-building has led it to adopt an alternative system of centralized bureaucracy and authoritarian rule instead of representative democracy [25, 73]. However, limited experience in democratic construction does not necessarily imply that the Chinese Communist Party (CCP) leaders and the Chinese people have not recognized the value of democracy. There was a popular saying in China, “democracy is a good thing” [91], which serves as evidence for this realization. Chinese leaders have long understood the functions of democracy from the perspective of instrumental rationality. As for ordinary Chinese people, Li Junru, the former vice president of the CPC Central Party School, commented that “[e]very Chinese knows that replacing autocracy with democracy is a historical process” [44]. In this regard, Hu Yue indicates, the term “democracy” rarely disappears from the political languages of authoritarian regimes [33].

From an official perspective, the discourse on democracy within the CCP has undergone significant shifts and exhibited various characteristics at different times. As early as the Revolution and Civil War era, the CCP adopted a democratic discourse similar to the Western context to oppose the dictatorship of the Chinese Nationalist Party (Kuomintang, KMT) under Chiang Kai-shek’s rule [32, 51]. After the CCP assumed power and established the People’s Republic of China on the mainland, elements in democratic discourse such as constitutionalism and multi-party systems were replaced by slogans such as “people’s democratic dictatorship” (renmin minzhu zhuanzheng), “people are the masters of their own country” (renmin dangjia zuozhu), and “socialist democracy” (shehui zhuyi minzhu). During Mao’s era, although the top leadership made limited effort to promote comprehensive elections and established an unprecedented centralized authority, they never abandoned democracy in their political discourse. They attempted to implement their subjective understanding of democracy [3], which ultimately resulted in the Cultural Revolution.Footnote 1

After ending the Cultural Revolution and adopting the reform and opening-up (gaige kaifang) policy, China embraced the globalization wave. Concurrently, as China opened its doors, it cannot entirely avoid outside influences such as the third wave of democratization. Observing the upheaval in Eastern Europe, the disintegration of the Soviet Union, and widespread social unrest in the late 1980s, the CCP came to understand the significance of advancing democracy to strengthen its ruling legitimacy. Consequently, for the past 40 years, each post-Mao CCP leader, from Deng Xiaoping to Jiang Zemin and Hu Jintao, has implemented various political reforms characterized by democratic elements [4]; since the onset of its reform and the opening-up, the CCP has conducted elections that incorporate democratic principles at the local level [30]. Additionally, party leaders have come to appreciate the importance of developing their own democratic narrative systems to withstand encroachment by Western nations. For example, the concept of “democratic voting” as defined by the CCP emphasizes the obligation to vote, contrasting with the Western emphasis on voting as a fundamental right [55].

Since President Xi Jinping’s rise to power in 2012, China’s domestic political landscape has experienced significant changes, with a notable tightening of political control [59, 71]. However, there is no evidence to suggest that the Chinese leadership is moving away from the democratic rhetoric. Contrarily, the political decline of Western society and pressure from ideological competition have prompted the Chinese authorities to adopt a strategy that frames their own ideology using democratic discourse to strengthen their ruling legitimacy [33]. For example, the Chinese government openly integrated democracy into its Core Socialist Values, which bear some resemblance to those of its Western counterparts.Footnote 2 Furthermore, while diminishing the role of electoral competition, the Chinese government has emphasized representation in its congress, transitioning from a liberal to a “socialist democracy” [80].

President Xi Jinping has repeatedly used the democratic rhetoric in his speeches. For instance, in 2014, President Xi first equated people’s democracy with consultative democracy (xieshang minzhu) during a ceremony commemorating the 65th anniversary of the Chinese People’s Political Consultative Conference. He stated that “the essence of the people’s democracy is that the people get to discuss their own affairs” [85]. In 2017, the same sentence was included in a report from the 19th CCP National Congress [86].Footnote 3 In 2019, he even declared that “China’s people’s democracy is a type of whole-process democracy” [87], which was subsequently incorporated into the report of the 20th CPC National Congress [84]. Admittedly, “process” is different from “procedure,” and whole-process people’s democracy in the Chinese context is not the same as procedural democracy regarding competitive elections and checks and balances in the Western democratic context. However, these two words share some similarities. For example, “process” in the Chinese context contains elements of procedure. Specifically, according to an official statement, whole-process people’s democracy includes democratic elections, consultations, decision-making, management, and oversight, which are also important procedural components of liberal democracies [76]. Therefore, the emphasis on “whole-process” may continue to indicate the importance of procedural aspects. For some international observers, these statements may be perceived as propaganda aimed at maintaining and legitimizing the ruling party’s base [48]. Thus, incorporating the democratic rhetoric into official discourse could simply be a strategic move for propaganda purposes rather than reflecting a true commitment to fostering democratic principles. Beyond official declarations, do the Chinese people share the same views as these Chinese discourses? Are there any discrepancies between the popular understanding and official narratives? If not, what type of democracy do ordinary Chinese citizens want?

An Ongoing Controversy Debate

Scholars have significantly contributed over the past two decades in answering these questions, resulting in a series of representative explanations. These conclusions can be divided into two closely related categories. One is to define democracy based on its content. Using this definition of democracy, numerous scholars have devised a range of questions for interviewees in various survey types. These questions cover different dimensions of democratic content, such as competitive elections, freedom of speech, the rule of law, social equality, and poverty eradication. Another aspect focuses on categorizing these contents according to their patterns or forms, which include substantive and procedural democracy. For instance, elections, liberty, freedom, and the rule of law are aspects of procedural democracy, whereas good governance elements such as social equality and poverty reduction fall under substantive democracy. Based on these classification methods, scholars have drawn two contrasting conclusions when examining the popular Chinese understanding of democracy.

One camp contends that the predominant Chinese democratic perception is minben democracy, in terms of content in an outcome-oriented form. Furthermore, the interpretation of minben democracy can be traced back to the culturalist approach, which may explain why the democratic perception in the Chinese context differs from that of people in other countries [15, 68, 69]. Specifically, minben, as a political concept, has deep roots in Confucianism and the Mandate of Heaven (tianmingguan) in China’s extensive history dating back to the Western Zhou dynasty [95,96,97]. This notion establishes a unique state-society relationship. The minben doctrine mandates that rulers acknowledge people as the foundation of the state (minwei bangben), cater to their material interests, and improve their living standards. In the minben context, leaders’ responsibilities closely resemble the concept of guardianship introduced by Dahl [22]. However, the masses expect the government to provide good policy outcomes and governance, such as meeting basic necessities and needs, narrowing the wealth gap, and achieving social impartiality. In contrast, aspects typically associated with liberal and procedural democracy, such as multiparty systems, competitive elections, checks and balances, and the rule of law, are not considered crucial. Moreover, the culturalist approach posits that traditional ideology can mitigate the effects of modernization. As such, Confucianism does not propel Chinese people to pursue modern democratic values, as it has a negative correlation with liberal democracy [38, 70, 92]. Consequently, as long as Confucianism’s influence persists, Chinese people’s democratic perceptions will continue to exhibit a substantive orientation, and Confucianism will cast a shadow over China’s future democratization [69, 70]. Moreover, minben democracy aligns well with political meritocracy [7, 10], regime legitimacy [15, 16], and CCP ideology [38]. Consequently, this narrative is widely accepted in China, prompting leading experts to use these symbolic findings to defend and interpret official discourse [88, 89]. Considering this, as one typical conclusion asserts, “[d]emocracy in China does not reflect a tension between the masses and authority” [94].

Interestingly, some scholars have presented different conclusions using similar national survey data, attributing their findings to both procedural and liberal democracy in terms of the content and pattern of the “D-word.” These scholars base their views on modernization theory, which posits that as long as modernization and political socialization advance, people’s democratic cognition will gradually shift from substantive to procedural, regardless of cultural influences on the economy [23, 35]. Consequently, traditional Confucianism does not inherently conflict with modern liberal democracies. For instance, South Korea and Taiwan—both traditional Confucian societies—have successfully established liberal democracies. In this regard, since mainland China has undergone rapid economic growth and urbanization over the past 40 years, it should not be an exception, and there are indications that its case follows this logic. For instance, Hu Peng categorized democracy into three distinct types: freedom, political participation, and good socioeconomic performance. Combining the street-interviewing method and survey data analysis, Hu Peng found that most Chinese people view democracy as political participation that falls under procedural and liberal democracies [32]. Moreover, by comparing four waves of ABS survey data, Zhai Yida revealed that, between 2008 and 2016, the proportion of Chinese individuals who perceived democracy procedurally increased and surpassed the rate of the substantive type [94]. Additionally, Wang Heng, drawing on multiple rounds of national sampling surveys, demonstrated that contemporary Chinese democratic perceptions exhibit mixed characteristics and that the Chinese public’s democratic understanding aligns well with modernization theory during the transitional period [79].

Based on the above leaders’ statements and academic research findings, we draw several conclusions that can guide improvements in this study. First, regarding this controversial topic, no consensus has been reached within academia, and both perspectives hold merit. Second, most existing studies employ closed-ended survey items to investigate Chinese perceptions of democracy. Although these options are carefully designed and regularly updated to measure democratic perceptions in different dimensions, they may inadvertently restrict respondents’ choices based on their first impressions. For instance, if a respondent initially selects democracy as a competitive election and then considers another option, such as meeting necessities, they may hesitate to make a choice. Third, the intersection between academic research and official discourse remains limited, despite efforts from the academic community, and officialdom and academia appear to be talking at cross purposes. Related to this, some official discourses such as consultative democracy may have been overlooked in the survey batteries. Finally, democratic perceptions are not always static, whether in advanced liberal democracies or the developing world. However, most existing survey data may be outdated and may not promptly reflect public opinion during Xi Jinping’s new era. Consequently, it is essential for scholars to reexamine the popular understanding of democracy using updated data and methods. In this context, this study aims to integrate official discourse and research findings to capture popular Chinese perceptions of democracy, using a combination of new and traditional methodologies.

Data

To answer the research questions, we draw upon data from two iterations of the Chinese Social Governance Survey (CSGS). These comprehensive surveys were conducted via in-person interviews and were the result of collaborative efforts among leading Chinese institutions. Using a two-step scientific sampling process involving GIS/GPS sampling [42] and subsequent probability proportional to size (PPS) sampling, the research team collected data from 125 county-level units across 26 provincial areas. Ultimately, we obtained 4200 valid responses in 2015, and 4941 valid responses in 2019. The CSGS dataset has at least two advantages. First, compared with existing studies based on potentially outdated survey data [32, 68, 69, 79], the CSGS 2015 and 2019 data capture and reflect the latest trends in public opinion, particularly as China enters Xi’s new era. Second, considering the Chinese government’s internet censorship [46] and incomplete coverage of internet usage among the Chinese population [19, 20], national sampling may be more effective and representative than online surveys, data obtained through web scraping, or other non-probability sampling methods.

Furthermore, most mainstream surveys employ closed-ended items to gauge the respondents’ perceptions of democracy, which may influence their first impressions. The following section provides a detailed illustration of the limitations inherent in traditional survey methodologies that predominantly use closed-ended questions to measure perceptions of democracy. This is followed by an introduction to the need to integrate alternative approaches to enhance and refine our current understanding. These methods aimed to overcome the bias often encountered in closed-ended surveys.

The Necessity of Integrating Closed-Ended and Open-Ended Questions

In both the CSGS 2015 and 2019 surveys, four multiple-choice questions on the meaning of democracy were presented. Specifically, the interviewers queried the respondents on four separate occasions about the most important characteristic of democracy, offering four options each time. Notably, the combinations of options differed between the two surveys. Figure 1 presents the distribution of the answers to these closed-ended questions.

Fig. 1
figure 1

Meaning of Democracy Based on Closed-Ended Batteries

Full size image

The above figures indicate two limitations of using closed-ended questions in the surveys. First, the distribution of responses changed with variations in the options provided. In Dimensions 1 and 4 of the CSGS 2015, the proportion of respondents who view democracy as a series of procedures and norms is higher than those who treat democracy from the substantive side. Conversely, when focusing on Dimensions 2 and 3, the pattern of democratic conception between procedural and substantive reverses, meaning that the number of people who believe that democracy’s nature meets people’s basic needs and realizes good governance outweighs the proportion of people who view democracy as a series of procedures and norms. This discrepancy can be attributed to various factors such as the order in which the questions are presented, which can create an anchoring effect. Second, when the combination of options is altered, the results change significantly. For instance, in Dimension 1 of the CSGS 2015, most respondents selected “people choose the government leaders in free and fair election” as the most important characteristic of democracy relative to the other three provided options. However, when the points of comparison change, for example, “Government narrows the gap between the rich and the poor,” and “Government does not waste any public money” have been replaced by “Basic necessities, like food, clothes and shelter, are provided for all” and “Politics is clean and free of corruption,” as demonstrated in Dimension 1 of the CSGS 2019, this choice becomes significantly less popular. In contrast, Dimension 3 of the survey battery in both the CGSS 2015 and 2019 has the same items; thus, the total proportion of procedural and substantive democratic perceptions has not changed. This phenomenon indicates that even the same option can yield different results when juxtaposed with different sets of alternatives, requiring the introduction of other methods for a comprehensive assessment.

Further, the CSGS complements its conventional multiple-choice format with open-ended questions, allowing participants to share their initial views on democracy without the limitations of predefined options. Therefore, scholars could further explore these open responses using computer-assisted text analysis. Specifically, CSGS asks each respondent, “What does democracy mean to you?” in an open-ended format, thereby enabling them to define democracy independently. After receiving the respondents’ answers, the interviewers asked the same question twice and recorded the answers in the questionnaire. Admittedly, some may argue that respondents in authoritarian regimes may exhibit self-censorship tendencies and not provide honest answers [14, 63, 67]. In response to this concern, other scholars have used experimental methods and reached differing conclusions, with the majority of Chinese people not concealing their preferences because of political pressure [43, 74]. Consequently, the issue of self-censorship and its impact remains inconclusive and debatable.

This study posits that open-ended questions primarily gauge knowledge rather than value judgments based on the following observations. The CSGS dataset comprises a series of closed-ended questions that probe an individual’s understanding of democracy from multiple perspectives. Interestingly, despite these guiding options that could aid respondents in formulating an answer, around 20% of the feedback was either categorized as “do not understand the question” or was left blank. In contrast, for the open-ended question, after collating all the responses from each participant, we observed that in the absence of predefined choices, nearly half of the respondents attempted to answer. This suggests that, when directly querying the concept of democracy, half of the respondents were uncertain about its definition. This pattern implies that when faced with a question centered on knowledge, respondents are less likely to mask their genuine views because of potential political repercussions.

To Answer, or Not to Answer

To examine which individuals are more likely to answer open-ended questions on democracy, we employed a logistic regression model using Bayesian estimation. The outcome variable was a dichotomous measure indicating whether the interviewee provided at least one answer.Footnote 4 Based on previous studies on democratization, we selected three explanatory variables: educational level [1, 64], internet usage [18, 26, 53], and household income [2, 77]. These variables were queried during the survey. Additionally, we controlled for respondents’ demographic attributes, including gender, age, residential setting (rural or urban), and CPC membership.

There are several reasons for selecting the Bayesian approach rather than OLS to estimate our model. First, in Bayesian modeling, the quantities of interest are treated as probabilistic distributions rather than fixed values. This corresponds better to how people really think, and alleviates the overreliance on “statistical significance” of the frequentist approach [72]. Second, Bayesian methods are more appropriate for political science because assumptions of frequentism, such as i.i.d., are seldom met in this discipline [28]. Third, the Bayesian approach does not test a point null hypothesis, but directly provides posterior probabilities for the research hypothesis, which is more logical and intuitive for our understanding [37].

Figure 2 presents the predicted probabilities of answering the open-ended questions on democracy. Specifically, we generated some “typical” cases and multiplied them to 2000 simulated model parameters based on the posterior distribution. By “typical,” we mean except the quantities of interest, all other variables are held to their median. Therefore, all the cases should take observed values, which is more meaningful than the “average” case approach. The figure reveals that, in both the CSGS 2015 and 2019 surveys, individuals with higher levels of education, more frequent internet use, or greater household income were more inclined to respond to open-ended questions about the meaning of democracy. For instance, the probability that a typical Chinese individual with a university degree answers an open-ended question is greater than 80%, whereas for a typical high school graduate, the probability falls below 65%.

Fig. 2
figure 2

Predicted Probability of Answering the Open-Ended Question

Full size image

Topic Classification of the Open-Ended Questions

This section examines the content of respondents’ answers to the open-ended questions. Some representative surveys, such as ABS Wave I, employed open-ended questions to explore people’s democratic conceptions, coding the answers into 922 categories. These coding methods are thorough, but also complex and time-consuming to organize prior to analysis. Nevertheless, some scholars have attempted to categorize the open-ended answers through manual sorting [17, 68, 69]. These investigations reveal new characteristics compared with the analysis of closed-ended questions and serve as a reference for our analysis.

Recent advances in automated content analysis methods have facilitated the investigation of textual data in political science [29, 49, 83]. Scholars have developed various text classification algorithms, such as Latent Dirichlet Allocation (LDA) [11], naive Bayes [52], and support vector machines (SVMs) [39]. In this study, we applied a semi-supervised text classification model called newsmap to open-ended questions about democracy. The newsmap model was originally created to classify short news summaries according to their geographic focus. However, it can also be used to classify documents into predefined topics [81, 82]. The chief advantage of semi-supervised learning methods is that they can incorporate the scholar’s theoretical knowledge into an algorithm without incurring high costs. The newsmap model requires only a small dictionary containing a few seed words for each topic, whereas supervised learning methods may require a large human coding training set.

For our analysis, we first used the R package quanteda [8] to transform the original answer text into a document-term-matrix (DTM). The preprocessing of the Chinese textual data is mostly the same as that of the English data. However, attention should be paid to certain points because of the grammar and structure of Chinese language. For instance, word segmentation must be conducted because there are no natural boundaries, such as spaces, between words in Chinese. To achieve high segmentation accuracy, we used the jiebaR package [62] to conduct initial segmentation and then convert the segmented words to quanteda tokens. After tokenization, we removed the punctuation remarks, numbers, and Baidu stop words [9].

Before applying the newsmap model to the data, we investigated the most frequent words that appeared in the answers to help us select appropriate seed words for our dictionary. Figure 3 presents the top 100 open-ended responses to the CSGS 2019 survey.Footnote 5

Fig. 3
figure 3

Wordcloud of the Top 100 Words in CSGS 2019

Full size image

Figure 3 demonstrates that the frequency of some words such as people (人民) lie at the top of the list, and apart from the abstract words, the words with specific meanings such as freedom (自由), freedom of speech (言论自由), fair (公平), election (选举), opinion (意见), equality (平等) appear multiple times. When combining the above results and discourse review in the second part, we found that in contemporary China, the expressions of democratic discourse, whether from the official side or academic circle, were distributed regularly; hence, we identified five categories of democratic understanding: minben, liberty, consultation, election, and authoritarianism. We set a category called authoritarianism because some respondents provide answers such as “democracy is the system of democratic centralism,” “democracy is democratic centralism,” “democracy is compliance,” and so on. It is not surprising that the respondents provided these answers because democracy in the Chinese context is often used along with the slogan, “system of democratic centralism” (minzhu jizhongzhi), and people are somewhat affected by the ideological propaganda. This type of word usage originated from Leninism, which is the organizational principle of a revolutionary party that emphasizes centralization over democratic values. In addition, we found that some people made negative comments about democracy. These people answered that democracy was corrupt and useless, indicating that they potentially held authoritarian values. Therefore, we created an independent category labeled authoritarianism to capture these answers.

Table 1 lists the categories and seed words used in the model. Seed words are selected based on the method proposed by Watanabe and Zhou [82]. To validate the automated classification results, 200 randomly selected answers were manually coded as the gold standard. It is noteworthy that even human coders find it difficult to classify short answers. Nevertheless, the F1 score of the classification results reached 0.68, indicating that the outcome of machine learning was fine and acceptable.Footnote 6

Table 1 Seed words for newsmap model
Full size table

Figure 4 presents the classification results for the answers based on semi-supervised machine learning. Data from 2015 and 2019 revealed that nearly half of the participants were unfamiliar with the concept of democracy (47.7% in 2015 and 50.2% in 2019). Among those who provided answers, consultation, liberty, and minben were the predominant perceptions of democracy among Chinese. Notably, only a small fraction associated democracy with the hallmark feature of elections. This trend remained relatively consistent between 2015 and 2019.

Fig. 4
figure 4

Classification Results

Full size image

These results can be interpreted from two perspectives. To some extent, this feature distribution may indicate that, compared with people who live in liberal democratic societies, ordinary Chinese citizens have a complex attitude toward democracy because the proportion of people who have a democratic understanding versus those who do not is about equal. However, Chinese population hold democratic understanding with its characteristics, particularly as few of them define the “D-word” as competitive election. For example, it is surprising that liberty, particularly freedom of speech, occupied the most important position in respondents’ cognition. This phenomenon is unique, and should not be ignored. In this category, freedom of association, the multi-party system, and other typical dimensions (e.g., the right of conscience) of freedom have not been mentioned much among all answers. Ordinary people pursuing free expression also reflect that they have trouble expressing themselves. This situation is not incompatible with contemporary China because the authorities have recently imposed numerous restrictions on freedom of expression.

In addition to consultation and freedom of speech, there was a portion of respondents who had a minben understanding of democracy (12.4% in the 2015 survey and 14% in the 2019 data), which implies that a certain part of the population cared about government performance and policy outcomes. This result is consistent with previous findings. Thus, the transmission of Confucian values is stable and continuous, and continues to play an important role in contemporary China.

It is noteworthy that nearly identical proportions of people understood democracy as consultation. Consultative democracy in the Chinese discourse is not completely equal to deliberative democracy, which originated in Western civil society. Owing to its unequal forms of participation as well as the bottom line under the leadership of the ruling party [45], China is often described as consultative authoritarianism [75, 78]. Despite this, consultative democracy in China entails a process in which decision makers are expected to seek and consider public feedback, reflecting a preference for some level of negotiation over none. Viewing democracy as consultation is in accordance with President Xi Jinping’s statement on democracy, which presents a lower political risk.

Moreover, a certain percentage (approximately 4.6% in 2015 and 3.8% in 2019) of the people view democracy from an authoritarian perspective, such as democratic centralism and unity, which may be correlated with the propaganda influence of the CCP.

This study categorizes perceptions of democracy into distinct topics. These topics can be classified into procedural and substantive democracies. Consultation, freedom of speech, and elections emphasize specific procedures, and thus fall under procedural democracy. Notably, the percentage of individuals who perceived democracy as procedural surged to over 30% (32% in 2019 and 35.8% in 2015). This suggests that a significant proportion of people understand democracy as a procedure. Conversely, democratic notions rooted in minben zhuyi and good governance, traditionally linked to substantive democracy, represented slightly above 10% (12.4% in 2015 and 14% in 2019). This challenges previous studies in which the Chinese predominantly understood democracy in substantive terms.

Explaining Chinese Perceptions of Democracy

To explore the determinants shaping varied perceptions of democracy, we employed a Bayesian multinomial logit model, designating the estimated topic of each response as our dependent variable. Mirroring our previous analysis on whether to answer open-ended questions, we used educational level, internet usage, and household income as explanatory variables. Moreover, we controlled for respondents’ demographic attributes, including gender, age, residential setting (rural or urban), and CPC membership.

Figures 5, 6 and 7 illustrate the predicted probability distributions for the responses to each topic in the CSGS 2019 survey.Footnote 7 Figure 3 highlights that as the education level rises, there is an increased likelihood of respondents associating democracy with liberty. Conversely, the association between authoritarianism and minben decreases. Education does not appear to significantly influence perceptions of democracy in terms of consultations or elections. Figure 4 presents that internet usage is negatively associated with authoritarianism. Interestingly, as people’s internet usage increases, they tend to link democracy more with liberty but less with elections. Figure 5 demonstrates that higher household incomes bolster liberal-democratic perceptions while diminishing associations with consultation. These findings largely align with the modernization theory, which posits that education and economic development cause political development [5, 35, 58].

Fig. 5
figure 5

Predicted Probability of Answering Each Topic by Education Level

Full size image
Fig. 6
figure 6

Predicted Probability of Answering Each Topic by Internet Usage

Full size image
Fig. 7
figure 7

Predicted Probability of Answering Each Topic by Household Income

Full size image

Conclusion

Democratic backsliding occurs worldwide, particularly in Western societies. Simultaneously, authoritarian regimes are attempting to create their own interpretation of the “D-word.” Consequently, democracy faces unparalleled difficulties, and it is crucial for scholars to re-examine democratic notions in non-Western contexts. The global understanding of democracy varies. For instance, regions such as East Asia and North Africa tend to view democracy through instrumental values, in contrast to the intrinsic values placed on it in North America and Latin America [47]. China, with its distinct cultural and political system, presents an essential case study for exploring diverse democratic understandings.

Existing studies have made numerous significant contributions to our understanding of popular Chinese democratic perceptions. Although some studies have observed an increase in procedural democratic perceptions over the past 15 years, the majority maintains that Confucian minben democracy forms the mainstream content of democratic perceptions among ordinary people, resulting in a substantive pattern of the “D-word.” Additionally, previous studies typically used closed-ended survey batteries to assess democratic understanding from various perspectives; however, this approach may overlook the official discourse embedded within people’s cognition.

Although our results echo those of some existing studies, our empirical approach provides fresh evidence to reinforce these findings simultaneously. We offer a nuanced perspective on popular Chinese democratic perceptions by employing automated text analysis on open-ended questions. First, regarding the meaning of democracy, the Chinese understanding appears more multifaceted and intricate than previously recognized. When lacking reference options, approximately 50% of the respondents were not acquainted with the concept of democracy, which is much larger than the proportion of any group in the closed-ended batteries that went unanswered for various reasons. This finding indicates that half of the Chinese are confused about the intrinsic meaning of democracy, regardless of the reason. Of those who are, many define it in terms of consultation, freedom of speech, and Confucian minben democracy, rather than purely elections. This underscores the fact that the democratic ideals of ordinary Chinese citizens have deviated from Western norms. Second, regarding democratic patterns, elements such as consultations, liberty, freedom, and elections collectively represent procedural democracies. The insights derived from our quantitative text analysis indicate a Chinese preference for procedural and normative democracy over its substantive counterpart, which may continue to align with President Xi Jinping’s statement on strengthening the procedural building of socialist democracy, although the latter is not perceived in the same procedural manner in the common view. This methodology is more effective than relying solely on traditional survey methods to capture the breadth and depth of democratic perceptions among Chinese citizens.

The demonstrated effectiveness of our methodology in capturing diverse democratic perceptions indicates its potential for broader application in fields requiring an intricate understanding of complex social or political phenomena. Specifically, the newsmap model is exceptionally versatile, able to be customized to various political texts simply by adjusting its categories and seed words to suit different study contexts [82]. This flexibility makes it an ideal tool not only for probing public opinion but also for analyzing the policy positions of politicians, thereby enriching both public opinion research and comparative political analyses. For instance, Ivanusch successfully employed the newsmap model to categorize over 56,700 speeches delivered by Austrian legislators into 20 distinct issue categories, demonstrating its efficacy in handling large datasets and extracting meaningful thematic insights [36].

Furthermore, our study provides new evidence on the dynamic shaping perceptions of democracy in China. Specifically, there is a positive correlation between educational level, internet usage, and household income, with a more liberal and procedural interpretation of democracy. This indicates that as access to education expands, internet penetration increases, and household incomes rise, there could be an increasing alignment with Western democratic ideals. Such observations resonate with modernization theory, which is rooted in the Western context.

In addition to its academic significance, the empirical evidence has substantial implications. Considering that half of the population has no direct understanding of democracy, and a few Chinese people view democracy in the same way as their Western counterparts, the Chinese government is likely to face less pressure to open up electoral participation and democratization for some time compared with the situation in the late 1980s. This finding reinforces the current understanding of China’s regime stability. However, ordinary people in China value performance, policy outcomes, and other aspects of the substantive side of democracy, and are inclined to understand the “D-word” from a normative perspective. This is reflected in the CCP’s emphasis on procedural democracy in its democratic discourse. For instance, President Xi Jinping stated at the 19th CPC National Congress that, “(w)e must uphold long term and steadily strengthen China’s socialist democracy, make active and prudent efforts to advance the reform of the political system, and improve the institutions, standards, and procedures of socialist democracy” [86]. In the absence of electoral accountability, people’s expectations for consultation and freedom of speech necessitate that the CCP place more emphasis on advancing procedural democracy and acquiring public consent. Some practices of deliberation and consultation across various Chinese regions have been termed “authoritarian empowerment,” as they have enhanced the rights of the involved residents [31, 61]. In the long run, maintaining formal requirements will embody democratic principles in China’s future political development. Finally, as public opinion and cognition are not always static, Chinese people’s democratic perceptions are no exception. In addition to China’s domestic modernization and economic transformation, its popular democratic understanding may be influenced by the ever-changing global political landscape. In particular, protests against the COVID-19 lockdowns in December 2022 indicated a shifting situation. As Western democracy faces challenges and tensions between the two ideologies represented by China and the United States, it is essential to continue monitoring the evolution of democratic conceptions among the Chinese.

This study is faced with several limitations that warrant consideration. First, the responses may have been influenced by social desirability bias owing to the political nature of the topic, potentially leading participants to provide answers that they considered socially or politically acceptable. Although focusing on understanding rather than supporting democracy in open-ended questions may lessen this bias, it remains a concern. Second, the brevity of responses to open-ended questions poses challenges in categorization, even for human coders, which complicates the analysis. Despite achieving acceptable levels of classification accuracy, future research could benefit from incorporating advanced text analysis techniques such as generative AI tools [6]. These tools, including large language models, provide sophisticated capabilities for text generation and comprehension, enhancing the analysis of open-ended responses. Finally, there is a potential overrepresentation of liberal views of democracy in the results, as those with higher education, income, and internet usage–who are more likely to respond–tend to associate democracy with liberty. Future studies should use diverse data sources and methodologies to substantiate the findings of the present study.