Changing environment, rising new media: Retesting the exposure-acceptance model in China

Abstract Based on the China data from the World Values Survey (WVS) 2000, Kennedy found a significant curvilinear relationship between education and regime support in the rural subsample but not in the urban subsample, partially verifying the exposure-acceptance model. He attributed this phenomenon to the “urban–rural gap” in educational opportunities in China. In the past two decades, China has narrowed this gap. Has this changed the patterns of regime support? To revisit this question, we replicate Kennedy’s analyses using the WVS 2018 data. The results fail to support the exposure-acceptance model in either the rural or the urban subsample. This inconsistency may stem from the explosive growth of new media in China over the past 20 years, which has made criticism more accessible. New media can affect people’s media exposure patterns in a way that goes beyond the scope of the exposure-acceptance model. Therefore, we introduce the measure of new media usage into Kennedy’s models and find a linear negative effect on regime support at the national level. We conclude that the applicability of the exposure-acceptance model may have declined in China due to the rise of new media. The long-term influence of new media on regime support in authoritarian nations should be explored in further research.


PUBLIC INTEREST STATEMENT
Traditionally, the Exposure-Acceptance Model was widely used to explain the regime support in non-democratic countries. Previous work had used data from the early 2000s and verified this model in China's rural samples. This study argues that the narrowing of the rural-urban gap in education and the rising new media may make the model no longer applicable to China. Based on the WVS2018 data, we reveal that the new media plays a significant impact beyond the scope of the Exposure-Acceptance Model. Thus, we suggest to develop new and more elaborate models to better explain and predict the changing trend of regime support in non-democratic countries.

Introduction
Previous research has demonstrated high public support for the regime in China (e.g., Shen & Guo, 2013;Stockmann, 2010), posing the question of how an authoritarian regime like that of China maintains high levels of public support. Kennedy (2009) argued that the exposure-acceptance model might provide the best answer. This model is based on two assumptions: 1) exposure to political communication is positively associated with the general level of political awareness and 2) uncritical acceptance of such messages is negatively associated with political awareness (Geddes & Zaller, 1989). Therefore, this model predicts that the relationship between education and regime support is patterned in an inverse U shape: people who have completed compulsory educationwho are politically aware but lack critical scrutiny-exhibit the highest support (Geddes & Zaller, 1989;McGuire, 1968).
Based on the Chinese national sample from the World Values Survey (WVS) 2000, Kennedy (2009) found a significant curvilinear relationship between education and regime support in the rural subsample but not in the urban subsample. He attributed this difference to unequal educational opportunities for rural and urban citizens. Unlike urban citizens, who could complete their entire education without leaving their familiar urban environments, villagers needed to leave home for high school. Thus, their high school experiences could influence their perceptions of society and the regime, thereby reducing their support for the regime (Kennedy, 2009).
However, this situation may have changed in China over the past 20 years. First, although unequal access to education between rural and urban areas is still notable, educational expansion has reduced inequalities to some extent (NBSPRC (National Bureau of Statistics of the People's Republic of China), 2001, NBSPRC (National Bureau of Statistics of the People's Republic of China), 2019). In terms of inequalities based on years of education, empirical results suggest that the urban-rural gap among younger age groups has narrowed (Liu & Li, 2010;Yang et al., 2014;Yiwen & Boran, 2021). This trend suggests an overall improvement over time. Second, alternative information sources, such as the Internet and new media, have grown exponentially over the past two decades (China Internet Network Information Center, 2001, 2019. According to the acceptanceexposure model, resistance to state propaganda depends on a combination of higher education and access to alternative political messages (Geddes & Zaller, 1989;Kennedy, 2009). However, with the rise of new media, public support for the regime is no longer affected only by education and mass media exposure. New media have become alternative sources of information for Chinese citizens, accounting for an important share of media consumption since the 2010s (Tong & Zuo, 2014). Moreover, new media have made criticism more accessible. Thus, people are frequently exposed to diverse information that deviates from the official discourse, which may eventually erode their support for the regime (Liu, 2020;Nip & Fu, 2016;Wong & Liang, 2021;Yang & Wu, 2022). This new phenomenon has gone beyond the scope that the exposure-acceptance model was originally designed to explore. Accordingly, a critical question arises: can the exposureacceptance model still explain the patterns of regime support in authoritarian states like China in the era of new media?
To answer this question, this study first used the China data from WVS 2018 and generally replicated Kennedy's first six models in 2009. That is, we tried to verify the potential effects of education and state media exposure on regime support, and observe whether the curvilinear relationship between education and regime support still exists in urban and rural samples. Then, due to the emergence of new media, we also constructed a variable of new media exposure and introduced it into Kennedy's models. And we expected to find the influence of the new variable that leads the models to produce different results.
Our findings advance our understanding of the long-term effect of new media on regime support in authoritarian states. We argue that the emergence of new media may significantly change the patterns of media exposure in these countries by offering people more diverse sources of information (Yang & Wu, 2022;Zhang & Guo, 2021). This marks a radical departure from the era of mass media dominated by newspapers and television, when the exposure-acceptance model was developed. Therefore, we may find that the model is less applicable to authoritarian states where new media have proliferated. At the same time, authoritarian regimes with a strong online influence, such as China and Russia, continue to increase their propaganda investment in new media (Gunitsky, 2015;Zhuravskaya et al., 2020). For these reasons, new and more elaborate models are needed to observe and predict the changing trends in regime support in these countries.

Changing environment: reduced "urban-rural gap" and its potential influence
In the seven decades since the foundation of the People's Republic of China (PRC), the government has invested heavily in expanding the education system and improving the average education level in both rural and urban areas. However, as in most other developing countries, considerable education inequality between rural and urban areas, usually described as the "urban-rural divide" or the "urban-rural gap," was still widely observed in the early 2000s. Studies have shown that in China, this inequality can be traced back to the establishment of the hukou system in the late 1950s, according to which urban and rural households are registered separately. This has led people to seek education in the areas of their registered residences and has divided the population of mainland China into two distinct groups within an uneven two-track public education system (Golley & Kong, 2013, 2018Li et al., 2015;Qiao, 2008;Wang et al., 2020;Yi et al., 2012). For instance, in 2002, the college education enrollment rate for rural students was 2.37%, compared with 19.89% for urban students. Moreover, from 1990 to 1999, the admission rate in rural regions increased by only 4.33%, as opposed to 147.13% in urban areas (Gou, 2005).
Nevertheless, research has shown that policies promoting education in the last decade have led to a remarkable improvement in average educational attainment and a sharp decrease in education inequality in China (Golley & Kong, 2018;Yang et al., 2014). Specifically, the junior middle school enrollment rate in mainland China has risen to almost 100%, in effect eliminating the "urban-rural gap" in nine-year compulsory education. Furthermore, junior high school graduates in mainland China now have more opportunities to receive secondary education. Therefore, significant progress has been made in narrowing the regional gap in the past two decades, and the difference in average years of schooling between rural and urban residents has been reduced for younger people. For people aged 60 years or older, the difference is 4.5 years; for those aged 50-59 years, it is 3.2 years; for those aged 40-49 years, it is 3.1 years; for those aged 30-39 years, it is 4.3 years; for those aged 20-29 years, it is 2.5 years; and for those aged 15-19 years, it is 0.4 years (Golley & Kong, 2018;Yang et al., 2014).
However, considering the cumulative effect of the uneven distribution of educational resources between urban and rural areas over a long period, higher levels of educational development have been achieved in urban areas than in rural areas of China. Urban students have access to comparatively high-quality education and better chances of higher educational attainment. Consequently, education inequality between rural and urban areas is still notable. These trends should be reflected in the WVS 2000 and WVS 2018 datasets. Accordingly, we put forward the following proposition:

P1: Education inequality between urban and rural areas of China is lower than it was in the 2000s but remains substantial.
Kennedy (2009) adopted the exposure-acceptance model to test the level of public support for the regime in mainland China in 2000. Using the WVS 2000 data, he confirmed that education and media exposure had significant effects on regime support in rural China and found that the relationship between education and regime support was curvilinear. However, the urban subsample did not support a curvilinear relationship. Kennedy attributed this difference to differences in rural and urban higher education and educational experience at the high school level: in urban China, students received senior high school or college education in familiar urban environments.
Thus, the pattern of regime support did not differ between junior high and high school graduates. In contrast, villagers needed to leave home for senior high school. Therefore, their high school experiences resulting from a combination of new knowledge and new living spaces may have influenced their perceptions of pervasive media messages and social reality, thereby eroding their support for the regime (Kennedy, 2009).
However, because of the narrowing of the "urban-rural gap" in education, these differences have likely changed. If the senior high school and college enrollment rates of rural and urban residents have converged, the shape of the relationship between education and regime support in rural areas should increasingly resemble that of urban areas. Therefore, we put forward the following proposition: P2: Due to the decrease in education inequality between urban and rural areas, the relationship between education and regime support in rural areas of China has changed from curvilinear to linear.

Rising new media: a "negative" alternative to state media in China
The tone of news content has been cited as a source of systematic bias in what citizens know about their governments and the world around them Moy & Scheufele, 2000;Soroka et al., 2019). As China is still considered one of the nations with the lowest degree of freedom of speech and press (Fu et al., 2013) and the primary function of mass communication is to consolidate the Communist Party of China (CPC)'s rule (Hachten, 1987) by covering more positive than negative news stories (bao xi bu bao you; Xiao, 2012), coverage bias in Chinese mass media embodies a set of known features that may best be described as "positivity bias" (Shen & Guo, 2013). With censorship systematically imposed on all kinds of traditional media content, such as newspaper articles, radio programs, and television transcripts (Gu, 2014), leaving almost no room for dissenting opinions or alternative explanations, Chinese people immersed in this news environment are chronically exposed (e.g., Higgins, 1996) to interpretations that favor the Party (Stockmann & Gallagher, 2011).
However, although studies have found that the mass media can change public trust in governments, and news with a negative tone appears to be more attractive and pervasive (e.g., Miller et al., 1979;Robinson & Levy, 1985;Soroka, 2014). Experimental results related to "negativity bias" in human behavior confirm that negative information produces considerably stronger psychophysiological responses than positive information (Trussler & Soroka, 2014). This has also been observed in political behavior and communication (e.g., Diagnault et al., 2012;Patterson, 1996;Soroka, 2012Soroka, , 2014Soroka & McAdams, 2015).
Although censorship in China is very strict, the government allows public discussions online to some extent, and negative information, such as direct public complaints and criticism of the government and its policies, may appear in new media (Jiang, 2010;Yang, 2011). Moreover, Internet use in China is growing rapidly. Internet users increased from 21 million in 2000the year of the fourth WVS wave in China-to 829 million by the end of 2018 (when the seventh wave was conducted), accounting for 59.6% of the population (CNNIC, 2001, China Internet Network Information Center, 2019. Furthermore, "the Internet . . . has become a platform upon which various new media appear" (Lee et al., 2013, p. 35). In 2010s, 91% of Chinese survey respondents reported visiting social networking sites in the previous six months, compared with 30% in Japan, 67% in the U.S., and 70% in South Korea. Like people around the world, Chinese users flock to these platforms to socialize and share content (Zhou & Wang, 2014).
Thus, many researchers (e.g., Marmura, 2008) believe that the Internet and new media may serve as alternative channels for citizens in China to voice their complaints which are rare in the mass media. News on the Internet is disseminated rapidly to diverse audiences and can affect the government's image and reliability (Pan, 2019). The sheer volume of "fast-paced" (Boyd et al., 2010, p. 10) information dissemination in new media and dissent expressed visually make it more difficult for the state to control them. Therefore, the Internet can help Chinese people break the information asymmetry by offering alternative channels (Nip & Fu, 2016;Shao et al., 2012). New media can play an important role as platforms for breaking news and political commentary that censorship authorities generally discourage and are therefore not available in state-sanctioned mass media (Hassid, 2012). In cases of breaking news, distrust in traditional media, which regularly cover food scandals and human rights violations, leads many people to turn to new media for information (Nip & Fu, 2016;Yang & Wu, 2022;Zhang & Guo, 2021). Therefore, the influence of new media on regime support may differ from that of mass media, which are characterized by positivity bias aimed at consolidating support for the regime (Shen & Guo, 2013). Users of new media are likely to use the Internet to express dissatisfaction or post negative information (Lee & Cude, 2012;Poell et al., 2014). By providing more space for negative content, new media may allow more room for negativity bias, thereby reducing trust in the regime (Lei, 2011;Soroka & McAdams, 2015). Thus, we put forward the following propositions: P3: Exposure to mass media positively correlates with regime support in China.

P4: Exposure to new media negatively correlates with regime support in China.
Furthermore, since research has shown that news with a negative tone seems more attractive and pervasive (e.g., Skowronski & Carlston, 1989;Soroka, 2014Soroka, , 2014Trussler & Soroka, 2014).), we put forward the following proposition: P5: Exposure to new media has a stronger effect on regime support than exposure to mass media in China.

Data source: the World Values Survey
The data used in this study were from the WVS, an international research project devoted to analyzing people's values, beliefs, and norms from a comparative cross-national and longitudinal perspective. The project was launched in 1981 and has been running in more than 120 countries. Its extensive geographic and thematic scope and the public availability of its data and findings have made the WVS the largest ever noncommercial cross-national empirical time-series investigation of human beliefs and values and one of the most authoritative and widely used surveys in the social sciences.
China joined the project in 1990, which coincided with the second WVS. Due to the uniform quality criteria for all participating countries, WVS data from China have been widely used, and studies based on these data have been published in leading social science journals since the early 2000s (e.g., Delhey et al., 2011;Kennedy, 2009;Silver & Dowley, 2000).
The 2000 and 2018 surveys were used in this analysis. WVS 2000 was conducted by the Research Center for Contemporary China at Peking University. WVS 2018 was conducted by a consortium of 12 universities led by the Public Opinion Research Center of the School of International and Public Affairs at Shanghai Jiao Tong University. The 2001 survey used multistage probability-proportional-to-size sampling based on household registration to obtain a random sample from a master national sample. To meet the requirement of universal coverage of adults, including immigrants, the 2018 survey employed stratified multistage probability-proportional-tosize sampling based on GPS/GIS-assisted area sampling to avoid issues arising from traditional sampling based on household registration, such as outdated household registration information, excessive vacant households, and exclusion of immigrant populations (Landry & Shen, 2005). In both waves, qualifying respondents were selected from dwellings using the Kish grid method (Kish, 1965). Both surveys considered factors such as the outcomes of multistage sampling (design effect), and nonresponses (for reasons such as unqualified individuals, vacant residential units, interview refusals, and language barriers) to satisfy a confidence level of 95%. The response rates in the two surveys were 72.2% (2000) and 61.7% (2018). Eventually, 1000 completed and valid face-to-face interview-based questionnaires were collected during the 2000 survey, and 3036 were collected in the 2018 survey. Geddes and Zaller (1989) used the dependent variable "satisfaction with government policy," whereas Kennedy used the WVS 2000 item "satisfaction with national leadership" as a measure of regime support. However, this item was removed from the 2018 survey. Tellingly, research (e.g., Chen & Shi, 2001;Li, 2004) suggests that as citizens of a one-party state, people in China do not make a clear distinction between the government and the Party. Moreover, official media tend to conflate the two using terms such as dang he zheng fu ("party and government"). Therefore, the measure of regime support in the 2018 survey consisted of two items (71 and 72): "Confidence: The government (in your nation's capital)" and "Confidence: Political parties." Both items were rated on a 4-point Likert scale (1 = not at all, 4 = a great deal). In this study, these two items were selected based on two considerations. First, they are conceptually close to the dependent variables used in Kennedy's (2009) work. Second, many other studies (e.g., Wang, 2011) have used these WVS 2000 items to measure regime support in China. Since the Party commands full control of the central government, the two items show a high correlation and were thus combined into a single dependent variable (M = 3.23, SD = .64; Cronbach's alpha = .858). The do not know and no answer responses were not included in the analysis. Thus, 3018 of the 3036 samples were used. Like Kennedy's study (Kennedy, 2009), this study adopted a dichotomous treatment of the dependent variable. Scores above the average were recorded as 1, representing a high level of support (n = 1516), whereas scores below the average were recorded as 0, representing a low level of support (n = 1502).

Media exposure
Media exposure is typically measured as the frequency of reading or watching political news. However, the two surveys contained different sets of news consumption measures. The 2000 survey used only one generic item related to political news consumption rated on a 5-point Likert scale (1 = never, 5 = every day; M = 3.78, SD = 1.35). Moreover, the 2000 survey referred only to traditional mass media, such as newspapers, magazines, television, and the radio. In the 2018 survey, the measure of media exposure was refined to specify media types. Therefore, items related to the use of newspapers, magazines, television, and the radio were combined into the variable of mass media exposure instead of the 2000 media exposure variable (M = 2.77, SD = 0.96; Cronbach's alpha = .664), whereas items related to the use of cell phones, email, and the Internet were combined into the variable of new media exposure (M = 2.27, SD = 1.41; Cronbach's alpha = .854).

Message acceptance (education, respect for authority, and interest in politics)
Following Kennedy's conceptualization, education level, level of respect for authority, and interest in politics were considered factors influencing message acceptance. Education is one of the most important measures in the exposure-acceptance model. It is widely thought to increase the probability of exposure to political communication and the capacity for critical scrutiny to resist propaganda. In the two surveys, education was assessed as a 5-point ordinal variable ranging from lower than primary school to university level with a degree. The median was secondary school completed.
The cultural explanation (Geddes & Zaller, 1989;Shi, 2001) suggests that a high level of respect for authority may be a source of support for an authoritarian regime like that of China. In the WVS questionnaires, the level of predisposition to authority was measured using the following question rated on a 3-point scale: "Would it be good, neutral, or bad to have greater respect for authority?" According to Geddes and Zaller (1989), interest in politics reflects a greater capacity for critical scrutiny. It has also been used as a measure of political awareness (Zaller, 1990). The two WVS questionnaires asked, "How interested would you say you are in politics?" This item was rated on a 4-point scale ranging from very interested to not at all interested (

Hukou status as a categorical variable
Many studies have shown that the main cause of educational inequalities between urban and rural areas in China is the rigid hukou or huji (household registration) system. Students are divided into two groups with markedly different access to the public education system (Golley & Kong, 2013, 2018Knight, 2014;Luo et al., 2012;Qiao, 2008;Yi et al., 2012), which disfavors students in rural areas. However, hukou status was not measured in the WVS. Kennedy (2009) divided the sample into urban and rural subsamples according to the item "occupation/employment" in the 2000 survey, considering respondents who answered agriculture to be rural residents. The occupation/ employment variable was not included in WVS 2018. Therefore, in this study, the 2018 sample was divided according to the geographic distribution of the questionnaire. Hukou status was treated as a dichotomous variable (0 = rural, n = 1134; 1 = urban, n = 1884).

Control variables
For consistency with Kennedy's (2009) study, three demographic variables were used as control variables: gender, age, and family income. Gender was a dichotomous variable (0 = female, 1 = male), with an approximately balanced male-to-female ratio (

Overview
Logistic regressions were the main statistical tools employed in this study. The modeling process followed two steps. First, the WVS 2018 dataset (N = 3018) was used, and Kennedy's (2009) first six models were generally replicated (full models 1 and 2, urban models 3 and 4, and rural models 5 and 6). Specifically, regime support was linked to education level, mass media exposure, level of respect for authority, and interest in politics, controlling for demographics. Because the statistical test for a curvilinear relationship uses the quadratic equation y = ax + bx 2 , a squared education variable was also included in the logistic regression models. Second, new media exposure was added to the logistic regressions as a regime support predictor.

Descriptive statistics
Empirical studies in Western democratic societies suggest that demographic characteristics have independent effects on regime support (e.g., Abramson, 1983;Cole, 1973). We therefore examined bivariate correlations in the 2018 dataset prior to the formal analyses (Table 1). The results were mixed. First, regime support was associated with age but not with gender or income. Second, all three variables were negatively associated with interest in politics. Perhaps due to historical memory since the foundation of the PRC, older and richer people are more likely to abstain from politics. Third, in the 2018 dataset, education was a significant negative predictor of regime support.

Results
Initially, this study analyzed the changes in education inequality between urban and rural areas in China over the past two decades by comparing the WVS 2000 and WVS 2018 datasets. The independent-samples t-test was used to test P1 (education inequality between urban and rural areas of China is lower than it was in the 2000s but remains substantial). Both datasets showed significant differences between the urban and rural subsamples (2000: t = 17.01, p = .000, Cohen's d = 1.119; 2018: t = 9.103, p = .000, Cohen's d = .467; Table 2).
A z-test was used to determine whether the change in the "urban-rural gap" between 2000 and 2018 was statistically significant. The sampling variance of Cohen's d is approximately equal to Therefore, to test H 0 : δ 1 = δ 2 (where δ 1 and δ 2 denote the true d values of the two education "urban-rural gaps"), the following equation was used: which followed a largely normal distribution under H 0 . Therefore, |z| = 3.645 ≥ 1.96, which means that H 0 was rejected at α = .05 (two-sided). Accordingly, Proposition 1 was fully supported. Table 3 presents the logistic regression results for the WVS 2018 dataset. Controlling for demographics, the estimates of models 1-6 indicated no association between mass media exposure and regime support in the expected direction, and no association was significant (full model 1: β mass = −.137, n.s.; full model 2: β mass = −.137, n.s.; urban model 3: β mass = −.085, n.s.; urban model 4: β mass = −.085, n.s.; rural model 5: β mass = −.176, n.s.; rural model 6: β mass = −.176, n.s.). Thus, P3 (exposure to mass media positively correlates with regime support in China) was not supported by the WVS 2018 dataset. These results are not in line with Kennedy's (2009) and Li's (2004) findings showing that state-controlled mass media have a consistent positive effect on regime support.
To test P4 (exposure to new media negatively correlates with regime support in China), the variable of mass media exposure was replaced with the variable of new media exposure in the models. Table 4  To test P5 (exposure to new media has a stronger effect on regime support than exposure to mass media in China), two independent variables were used simultaneously in the models to compare their effects on the dependent variables. The comparison results are shown in Table 5. Neither variable had a significant effect on regime support. Nevertheless, exposure to new media  Finally, we analyzed P2 according to the above results. In Kennedy's (2009) study, education and media exposure had significant effects on regime support in rural China, and there was a significant curvilinear relationship between education and regime support. As shown in Tables 3-5, education had a linear negative effect in all models in this study, except for urban models 4, 10, 15, and 16. Moreover, full models 2, 8, and 14, urban models 4, 10, and 16, and rural models 6, 12, and 18 indicated no significant curvilinear relationship based on the squared education variable. This suggests that the curvilinear effect of education on regime support may no longer exist. Due to the decrease in education inequality between urban and rural areas, the effect of education on regime support also changed in the rural subsample. Therefore, P2 (due to the decrease in education inequality between urban and rural areas, the relationship between education and regime support in rural areas of China has changed from curvilinear to linear) was supported.

Discussion
This study first used the China data from WVS 2018 and generally replicated Kennedy's first six models. But the results were inconsistent with Kennedy's findings (Kennedy, 2009). Logistic estimates indicated that the significant positive effects of education and state media exposure The "interest in politics" variable is dropped in the full model and urban model because it is highly correlated with "mass media exposure". However, the two variables are not highly correlated in the rural model. This tends to follow the relationship predicted in the Geddes and Zaller (1989) model. *p < 0.05; **p < 0.01.
Liang & Li, Cogent Social Sciences (2023) on regime support that Kennedy (2009) observed in the national sample were not reproduced in this study. Meanwhile, the curvilinear relationship predicted by the exposure-acceptance model was observed neither in the urban nor in the rural subsample. By contrast, Kennedy (2009) found a significant curvilinear relationship in the rural subsample but not in the urban subsample. Additionally, our findings also implied the decrease in high school and college education inequality between rural and urban areas. Thus, we suggested that the exposure-acceptance model might no longer apply to current China. We argued that the different results might be due to the emergence of new media in China over the past 20 years. Therefore, we constructed a variable of new media exposure and introduced it into Kennedy's models (Kennedy, 2009). The results of this study suggested that new media exposure had an independent linear negative effect on public support for the regime in China at the national level, whereas mass media exposure showed no significant effect. Kennedy's (2009) study on the relationships between mass media exposure, education, and regime support in China was the first to compare rural and urban regime support patterns. Prior to that study, education was typically treated as a control variable rather than as a variable of political awareness. Kennedy's comparison was based on two assumptions: 1) there is a significant difference in the distribution of educational opportunities between urban and rural areas and 2) high school and college education can provide sources of information alternative to state media by offering people the experience of living environments that differ from their communities of origin. Accordingly, with the sharp reduction in education inequality between rural and urban areas and the rise of new media as The "interest in politics" variable is dropped in the full model and urban model because it is highly correlated with "new media exposure". However, the two variables are not highly correlated in the rural model. This tends to follow the relationship predicted in the Geddes and Zaller (1989) model. *p < 0.05; **p < 0.01.
alternative sources of information, the differences in the effects of education on regime support patterns between rural and urban areas observed by Kennedy have diminished.
The change in the relationship between education and regime support from curvilinear to linear can alter the way in which we explore the sources of regime support in China. Previous research has already suggested that education has a strong linear negative effect on regime support, which may lead to "the failure of the CPC in socializing people through education" (Chen & Shi, 2001, p. 105). Similarly, the results of this study suggest that although education inequality between urban and rural areas still exists, the narrowing of the gap and the emergence of new media offer alternative sources of information and create a new space for political discussion for both urban and rural residents.
Revisiting the exposure-acceptance model was not meant to disprove Kennedy's (2009) findings but to test the model's current applicability. The exposure-acceptance model is based on two assumptions: 1) exposure to political communication is positively associated with general levels of political awareness and 2) uncritical acceptance of such messages is negatively associated with political awareness (McGuire, 1968). Since both media exposure and education level are directly related to the level of political awareness, the disappearance of notable differences in regime support between urban and rural areas is not surprising, considering the rise of new media and the narrowing of the "urban-rural gap." The The "interest in politics" variable is dropped in the full model and urban model because it is highly correlated with "new media exposure" and "mass media exposure". However, the two variables are not highly correlated in the rural model. This tends to follow the relationship predicted in the Geddes and Zaller (1989) model. *p < 0.05; **p < 0.01. exposure-acceptance model used to be applicable to any authoritarian regime (Geddes & Zaller, 1989). However, as new communication technology spreads, new uncertainties need to be taken into account. The model's diminished applicability to China may reflect a global trend due to the potential of new media to create spaces of freedom (Jost et al., 2018;King et al., 2017), especially in the absence of open media and civil society in authoritarian regimes, such as that of the PRC. In the 2010s, social media played an important role in the "Arab Spring" revolutions in the Middle East and North Africa (Bruns et al., 2013;Wolfsfeld et al., 2013) by providing ways to bypass restrictions on the freedoms of expression and association (Khondker, 2011).
However, even in authoritarian regimes, due to economic or other interests, governments are the main facilitators of the promotion of new media (Gerbaudo, 2018) and the expansion of education systems. Ironically, due to the significant positive correlation between education and the acceptance of new communication technologies (Eltantawy & Wiest, 2011), government efforts to raise the average education level and promote the development of new media inadvertently create a space for civic activism and resistance to media control (Wang & Kobayashi, 2021). Thus, one of the main functions of compulsory education, which, according to Key (1961, p. 340), is "indoctrination with more-or-less official political values of the culture in the classroom", may gradually be eroded by the rising new media. The modern era of media exposure based on new media differs sharply from the era of mass media dominated by newspapers and television, when the exposure-acceptance model was developed. Therefore, it would not be surprising to find that the model's applicability to authoritarian countries where new media have flourished has declined. Nevertheless, our findings do not necessarily mean that the linear negative relationship between new media and political support in authoritarian countries will persist. In fact, authoritarian regimes, especially those with a massive online influence (such as China, Russia, and Saudi Arabia), have continued to increase their propaganda in new media (Mattingly & Yao, 2022;Wang et al., 2020;Zhuravskaya et al., 2020). Especially in China, the government has successfully created a pro-government online atmosphere through various soft and hard means (Wang & Kobayashi, 2021;Wong & Liang,). As in the case of traditional mass media, the Ministry of Propaganda and Cyberspace Administration actively engages in the surveillance of and interference with Chinese social networks (Creemers, 2017;Miao & Lei, 2016;Shambaugh, 2007). With the strengthening of the government's ability to control and censor new media and the establishment of new state-run media, the negative effect of new media on regime support needs to be continuously tested. Moreover, we believe that a new version of the exposure-acceptance model is needed to observe and predict regime support trends in authoritarian countries in the complex and changing new media landscape. This research has certain limitations. First, we could not replicate Kennedy's (2009) procedure and measurements exactly, which inevitably reduced the comparability between the two studies. Second, regarding the significant positive correlations between respect for authority and regime support, it is conceivable that trust in the government among Chinese people may stem from a deeply rooted cultural and social tradition (e.g., Chen & Shi, 2001;Inglehart, 1990;Schlesinger, 1991;Shi, 2001;Tu, 2014). However, this study did not examine how a cultural predisposition to authority may have influenced the patterns of support in the WVS 2000 and WVS 2018 datasets. This could be explored in future research.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Declarations
The work described has not been submitted elsewhere for publication, in whole or in part, and all the authors listed have approved the manuscript and response to reviewers that are enclosed.

Citation information
Cite this article as: Changing environment, rising new media: Retesting the exposure-acceptance model in China, Zhehao Liang & You Li, Cogent Social Sciences (2023), 9: 2175432.