Introduction

Methodology

The information used to test the hypotheses was taken from the Special Eurobarometer [5]. A hybrid survey, face-to-face and online, (N = 37.079) was carried out between April and May 2021. It covers the population of the respective nationalities of the European Union Member States, resident in each of the 27 Member States and aged 15 years and over. The survey has also been conducted in 11 other countries or territories outside the EU: five candidate countries (Albania, Montenegro, North Macedonia, Serbia and Turkey), as well as in Bosnia and Herzegovina, Iceland, Kosovo, Norway, Switzerland and the United Kingdom. To calculate the values of the respondents for each variable, East and West Germany were unified into one country using the “Patch: DE/UK WEIGHT” [67]. In this way, it was also possible to apply the appropriate weighting to the observations.

In Appendix A of the supporting information (see S1 File) the text of the items used to elaborate the necessary variables is provided. The measures and descriptive data are given in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Measurements and descriptive data for all variables.

https://doi.org/10.1371/journal.pone.0338384.t001

Results

Table 2 shows the results of the multilevel linear modelling. Based on the values shown in the empty model (model 0), the between-country variability is 0.559 and the within-country variability is 1.805. These results produce the aforementioned ICC value equal to 0.236, indicating that 23.6% of the total variance of CONSPI can be attributed to differences between countries.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Parameters of the multilevel regression analysis, standard error in parentheses.

https://doi.org/10.1371/journal.pone.0338384.t002

Examining the parameters of the fixed effects, we find that they remain virtually unchanged from the time the variable is added until the end of the hierarchical inclusion process. Model 1 includes all the individual variables related to each of the theoretical frameworks of public perception of science, and the control variables (age, sex and economic precariousness). Before analysing how each variable is influenced by the parameters obtained as fixed effects, it is interesting to know the size of the effect of each of the three separate perception models on the belief SCT. In Appendix B of the supporting information (see S1 File) it can be seen that the model that produces the largest reduction of intra-country variance is the Public Understanding model (10.2%), which includes variables related to attitudes towards science. The Science Literacy and Science in Society models produce a much smaller reduction (5.8% and 0.8% respectively).

The effect produced by the variable SCLITERACY is negative and significant, thus confirming the approach proposed in H1. In general, the formulations in H2 are also supported by the data: those who distrust scientists more (MISTRUST) have a higher level of conspiracy theory belief (H2.2). As will be seen below, this variable reduces a larger percentage of the intra-country variance, thus positioning itself as the science perception characteristic that contributes most to conspiracy thinking formation. The negative effect of INSTRUMENTAL and the positive effect of CRITICAL on the explained variable is also as expected (H2.1). Those people who are more likely to identify the harms of science and technology than their potential benefits are more likely to believe in SCT. Surprisingly, the BLINDFAITH variable indicates that these people have an overly optimistic (even naive) view of science in terms of what science is capable of achieving. With respect to the variables that make up the Science-in-Society model, the results show that those who are more involved in science and technology-related issues (ENGAGE) tend to believe less in SCT (H3.1). On the other hand, those who advocate the democratisation of science and consider it necessary for citizens to participate in decision-making in science (INVOLVEM) also seem more likely to believe in conspiracies theories within the same field (H3.2). Therefore, H3 is only partially confirmed.

In relation to the sociodemographic variables that we included as control variables, we found that the type of conspiracy theories studied in this study are not associated with the gender of the individuals (SEX). On the other hand, age (AGE) shows a significant and negative relationship with our dependent variable. Younger people tend to have a higher degree of conspiracy theory belief. Finally, the economic situation of households is also significant in the model. Those who are in a precarious economic situation, regardless of the degree (Highprec, Lowprec), are more likely to conspiracy theory beliefs of a scientific nature than those who do not have problems paying their bills (Noprec).

In model 2 we add a country-level variable (DEMO) that expresses the level of democratic quality of a country. This factor has a clear negative effect, showing that the inhabitants of countries with higher levels of democracy have lower levels of conspiracy theory beliefs (H4). The rest of the variables in the model are not affected in their relationship with the dependent variable and maintain similar levels of significance and directionality in the rest of the models in the analysis. Fig 1 shows in detail the positioning of the countries included in the study in relation to the CONSPI and DEMO values, as well as a nearly linear relationship between both variables. The lower right corner is occupied by fully established European democracies such as those of the Scandinavian countries and the UK. In the middle of the graph we find Mediterranean countries and some former Soviet republics (those that became independent earlier) while in the upper left corner are fundamentally the republics that were part of the former Yugoslavia, plus Turkey and Hungary. It should be noted that SCT belief decreases sharply only when countries reach a real democratic status (above 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Relation between conspiracy theory beliefs and democracy index.

https://doi.org/10.1371/journal.pone.0338384.g001

The summary presented in Table 3, shows that as we move up the hierarchy followed in our analysis each model fits the data better than the previous one, as all differences in the deviance values, (−2LogL_{Reduced model}) -(−2LogL_{Full model}), are significant. It can also be seen that the inclusion of all individual variables, model 1, gives rise to a PRV of the residual variance of 15.1%, indicating that the respondent characteristics included in the equation contribute to explaining that percentage of the level of within-country conspiracy theory beliefs. This set of factors, on the other hand, generates more between-country variance. In contrast, when adding the country democracy index, model 2, the between-country variance decreases significantly (62.1%). This reduction indicates that, of the total CONSPI variance attributed to differences between countries, 14.7% (0.236x0.621) is explained by the DEMO variable.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multilevel effect size reporting.

https://doi.org/10.1371/journal.pone.0338384.t003

This suggests that the quality of democracy may moderate the relation between some individual variables, which justifies cross-level interactions [69]. As discussed in the theoretical framework, the socio-political context may determine the relationship between different elements of individuals’ positioning towards science. It is therefore of substantive interest to test such interactions with science perception variables included in the model [73]. SCLITERACY and MISTRUST were selected as lower-level variables because they turned out to be the two variables that, individually, produced the largest PRV in the within-country variance against the empty model, 5.8% and 9% respectively (see Appendix C in S1 File). The addition of the interaction (Model 3, Table 2) produced a significant improvement in model fit (χ²(2) =110.49; p < 0.001). Thus, although the value of the PRV was negligible, the two lower-level variables were included as random effects [73] separately [74: 833]. The latter two models (4a and 4b in Table 2) also yield a significantly better fit than the model without random slopes.

The results show a negative effect on the interaction between democracy and knowledge and a variance of the slopes that is significantly different from zero (Model 4a). This indicates that higher levels of democracy intensify the negative relationship between scientific literacy and belief in SCT. On the other hand, although no significant cross-level interaction was found between democracy and distrust, non-zero variance was also obtained when incorporating this attitudinal item as a random effect (Model 4b). These results allow us to partially confirm our H5, that is, the moderating effect of the level of democracy only occurs in the relationship between knowledge and belief in SCT.

In order to graphically visualise the nature and direction of the interactions and random effects described [73] the graphs presented in Figs 2 and 3 were produced. To do so, countries were grouped according to their level of democratic quality following the typology proposed by the Economist Intelligence Unit: Full democracies (8–10), Flawed democracies (6–7.99), Hybrid regimes (4–5.99), and Authoritarian regimes (0–3.99) (Appendix D in S1 File shows which group each country used in the study falls into).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Relationship between CONSPI and SCLITERACY according to the level of democracy.

https://doi.org/10.1371/journal.pone.0338384.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Relationship between CONSPI and MISTRUST according to the level of democracy.

https://doi.org/10.1371/journal.pone.0338384.g003

The first figure shows that Hybrid and Authoritarian regimes have a low slope, so these are countries where increased knowledge hardly reduces the level of conspiracy theory belief. The opposite is true for countries with high rates of democracy. In these cases, the slope is high and, in addition to starting with lower levels of conspiracy theory beliefs for the lowest values of SCLITERACY, it even reaches a zero value of conspiracy when the degree of knowledge reaches maximum values.

In Fig 3 it is interesting to see that countries with low levels of democracy have a relationship between distrust in scientists and belief in SCT that is always higher than that observed in countries with greater democratic guarantees. In fact, we see that in these countries, even at lower levels of distrust, some level of conspiracy theory belief occurs and it reaches the highest level when distrust increases. In full democracies, however, the level of conspiracy thinking is only moderate for the most extreme scores of distrust towards scientists.

Discussion

The Science and Technology Studies approach can help to explain the public’s belief in SCT and provide some evidence to address the concern raised by Harambam [76] about the lack of discussion of this topic by scholars in this field. Of the various factors influencing public perception of science, a lack of trust in scientists is the most significant contributor to the acceptance of unwarranted beliefs, followed by the level of scientific knowledge and a critical attitude towards science. Therefore, of the three theories of public perception of science, Public Understanding contributes the most to explaining this fact.

This contrasts with the traditional emphasis in the field of science communication to combat misinformation or negative attitudes towards science (e.g., vaccines, health measures and technological innovations) by promoting scientific literacy. However, when it comes to controversial scientific issues, educational level or scientific knowledge does not significantly impact the likelihood of acting in accordance with scientific consensus [60,77]. The idea that making facts available to the public contributes to changing minds has been widely contested [78]. Indeed, in some cases a ‘boomerang effect’ has been observed, with public polarisation increasing in response to science-based messages related to climate change [79]. Nevertheless, our findings align with previous literature which indicates that education is negatively correlated with conspiracy theory beliefs, although it is not the most significant factor in explaining this phenomenon [80,57].

Attitudinal variables, especially distrust of scientists and a critical or sceptical view of how science works, play a key role in explaining belief in SCT. Our results are in line with other studies that have shown that the rise of this type of theories may be related to a problematic relationship with scientific authority in a context of growing institutional distrust [81–85]. Conspiracy theory beliefs, therefore, would have less to do with a public discussion about the accuracy of scientific facts than with a much deeper phenomenon: the dissolution of a shared truth and the questioning of epistemological sources on which there used to be consensus (the epistemic trust pointed by Vaupotič et al. [86]. Indeed, Nefes et al. [81] have recently shown how a spectrum of communities dedicated to the spread of conspiracy theories emerged in the public conversation on X about distrust of political and health institutions during the COVID-19 pandemic. Distrust was the catalyst for the spread of conspiracy theories and the “gateway” connecting one to the other.

Our results also are aligned with the theory proposed by Mede and Schafer [12] about scientific populism. This phenomenon is characterised by the transposition of the anti-elitist discourse to the field of science and the antagonism between a virtuous ordinary people and an unvirtuous academic elite represented by scientist and scholarly institutions. This discursive framework may have important consequences for the gradual loss of legitimacy of scientific institutions and their epistemic authority.

However, there is an important nuance to this breakdown in trust. Our results show that those who believe in SCT tend to distrust scientists and have a critical view of how science works. But, at the same time, they have very high — albeit misguided — expectations of what science can achieve. Indeed, some studies have shown that people who believe in conspiracy theories are at least partly pro-science [87]. Nevertheless, this kind of “naive” blind faith in science as a system of knowledge suggests that the trust at stake is not in the scientific method, but in the potential corruption of those who implement it, thus fostering the idea of the unvirtuous academic elite described by Mede and Schafer [12].

The relatively low explanatory power of the Science-in-Society model suggests that attitudes promoting the democratisation of science and public participation may not directly translate into lower conspiracy theory belief. These attitudes often represent normative views about how science should operate, rather than perceptions of how it actually works in practice. Consequently, their influence might be more indirect and context-dependent. For example, in contexts where institutional trust is low, calls for greater participation could coexist with scepticism or disillusionment, thereby limiting their protective effect against conspiracy theory beliefs.

However, it is interesting to discuss the apparently contradictory direction taken by the two variables included in this model. On the one hand, greater engagement in science-related activities — such as visiting museums or attending public lectures — is associated with lower conspiracy theory belief. On the other hand, individuals who endorse conspiracy theories express stronger support for public participation in science and technology decision-making. While counterintuitive at first, this finding may reflect a broader tension between epistemic distrust and demands for participatory inclusion.

This paradox resonates with the growing literature on science-related populism and anti-expert sentiment, which interprets such attitudes as a reaction to perceived exclusion by scientific elites rather than a rejection of science itself [12,88,89]. People who believe in conspiracy theories may distrust scientists and hold critical views about how science operates, while simultaneously maintaining high — albeit sometimes unrealistic — expectations of what science can achieve. Previous studies have indeed shown that conspiracy believers are, at least partly, pro-science [87], suggesting that their distrust lies not in the epistemic system but in those who administer it. This helps explain their desire for more citizen involvement in science governance — an attempt to “reclaim” control from perceived corrupt or self-serving elites.

Recent scholarship also provides mixed evidence on the link between conspiracy theory beliefs and civic or political engagement. While Jolley and Douglas [90] speculate that conspiracy theory beliefs may suppress collective action and participation, Kim [67] finds that they can, in some contexts, stimulate engagement. In fact, in recent years some findings have clarified the role of activism in relation to conspiracy theories. People who believe in such theories are actually more prone to nonnormative political engagement [91] and online activism [92]. Bordeleau et al. [75], in turn, note that this relationship between activism and conspiracy theory beliefs occurs only in consolidated democracies, but not necessarily in developing democracies or more authoritarian regimes. Our findings contribute to this debate by suggesting that, in the scientific domain, conspiracy theory beliefs may foster a reactive form of engagement — one driven less by democratic ideals than by perceived alienation from expert decision-making. This interpretation situates our results within broader discussions about science populism, participatory democracy, and the social consequences of declining epistemic trust.

According to our results, people who believe in SCT tend to be critical of the system, particularly the way science is conducted. This attitude of dissatisfaction is often accompanied by a more generalised resentment, which can lead to a sense of alienation due to the perceived lack of opportunities to participate in public life [93,94]. When this social discontent is combined with economic disadvantage, it fosters a climate in which conspiracy theories are more likely to be embraced [49]. Therefore, it is not surprising that economic precariousness, among the control variables included in our study, has the greatest influence on the likelihood of believing in SCT.

Many of these questions suggest that a mismatch between the subject and their social, political and economic environment could trigger conspiracy theory beliefs. In line with these findings, recent work has emphasised the importance of considering contextual factors when exploring the extent to which these may contribute to an increase in conspiracies acceptance. In recent years, aspects such as wealth and its distribution, as well as the quality of democratic systems, have become increasingly important. However, Hornsey and Pearson [47] suggest that the results may still differ depending on how conspiracy theories were measured, with the potential for inconsistencies in the findings. Our analysis revealed a strong correlation between a country’s democratic quality and the prevalence of belief in SCT among its population. In line with the findings of Hornsey and Pearson’s review [47], conspiracy theorizing tended to be higher among more authoritarian nations, although this effect did not emerge reliably in the largest of their datasets. However, this association must be interpreted cautiously, given the strong collinearity between democracy and other structural factors such as economic development, institutional effectiveness, and corruption control. Future research using disaggregated or multi-indicator models could clarify the extent to which these effects are distinct or mutually reinforcing. We emphasise that more research is needed to confirm this initial impression of the relationship between SCT and trust-sensitive political realities.

This approach has become to shift the paradigm by moving conspiracy theory beliefs away from the mindset of individuals, which is influenced by personal characteristics, towards a more complex, context-dependent understanding of the phenomenon. This does not mean disregarding individual variables, but rather seeking correlations with social variables at the country level in order to gain a more comprehensive understanding of the SCT phenomenon. As Hornsey et al. [56] points out, the challenge in this area is to test for potential ‘ecological fallacies’ that could be affecting the evidence about conspiracy theory beliefs provided by individual variables so far.

Nonetheless, there is still a long way to go in efforts to capture cross-national differences in conspiracy theory beliefs. One of the most promising lines of analysis consist in the articulation of individual and contextual variables for a better understanding of SCT acceptance process. For example, de Boer and Aiking [48] found that the positive attitude toward vaccines had a stronger effect on opposing conspiracy beliefs in higher-income countries than in lower-income countries. In this regard, one of our most notable achievements in this study is demonstrating the moderating effect of the democracy index on individual variables acting as protective factor against SCT. Thus, citizens of countries with high levels of democratic quality have lower levels of SCT beliefs even among individuals with low scientific literacy or high distrust of scientists. Moreover, the negative relationship between scientific literacy and SCT beliefs is especially pronounced in countries with high democratic quality. In contrast, in countries classified as Hybrid and Authoritarian regimes, the proportion of citizens who believe in SCT remains high, with only a slight decrease even among those with the highest levels of scientific literacy. This fact may be related to the exception found by Alper et al. [54] in the relationship between knowledge and conspiracy theory beliefs: this relationship is always negative except for those individuals residing in countries with high levels of corruption.

One possible explanation for this asymmetric moderating effect lies in the different nature of the two predictors. Scientific literacy depends heavily on contextual factors such as educational systems, transparency, and access to reliable information — all of which tend to be more robust in consolidated democracies. Therefore, democratic quality can amplify the protective role of scientific knowledge by fostering environments where citizens are better equipped to evaluate evidence, contrast sources, and resist misinformation. In contrast, distrust of scientists operates through a different mechanism. It is often driven by affective, ideological, or identity-based orientations that are less sensitive to institutional conditions and more associated with cultural or populist narratives about expertise and power. Such distrust can persist, or even intensify, in democratic contexts, where pluralism and freedom of expression allow anti-elitist and conspiratorial discourses to circulate more openly. Consequently, democracy moderates the effect of knowledge but not that of distrust, since the former is shaped by structural and informational resources, whereas the latter reflects deeper cultural and value-based worldviews.

Taken together, these findings show that the determinants of science-related conspiracy theory beliefs operate at multiple levels. Individual predispositions—such as knowledge and trust—shape the likelihood of endorsing these beliefs, but their influence depends on the broader institutional context. The quality of democracy contributes to defining this context by establishing the levels of transparency, pluralism, and accountability within which citizens evaluate scientific information. Thus, democratic environments not only strengthen public trust in science but also enhance the capacity of individual factors to counter misinformation and unfounded beliefs.

Due to the embryonic state of the study of the relationship between structural variables and predisposition towards conspiracy theory beliefs, future research in this area is promising. More research effort is needed to clarify the role played by economic, political, and cultural factors in individual SCT acceptance. However, this approach must face an important limitation because such national variables often exhibit high collinearity which will require careful statistical analysis. To address this limitation, we specified a parsimonious Level-2 model including only the Democracy Index as the country-level predictor. The observed negative association between democracy and belief in science-related conspiracy theories should therefore be interpreted as a robust cross-national correlation that may partly reflect other correlated macro characteristics (e.g., GDP per capita, corruption, or government effectiveness), rather than an isolated causal effect of democracy.

Moreover, our findings regarding blind faith in science highlight the need to refine how different forms of trust in science are conceptualised and measured. Future studies could differentiate between trust in scientific methods, institutions, and actors, as well as explore how naive or uncritical faith in science manifests behaviorally — for instance, through the unreserved acceptance of scientific claims or the expectation that science can provide solutions to all societal challenges.

Finally, this new line of research on SCT could help to improve the effectiveness of European public policies aimed at tackling misinformation. Traditionally, these types of measures have focused on the importance of knowledge and scientific literacy in addressing unfounded beliefs. However, initial evidence suggests that there is no universal ‘recipe’ for reducing conspiracy theory beliefs, and that these types of policies must be adapted to each context to be impactful.

Supporting information

Acknowledgments

We are grateful to Professor Alberto del Rey for his valuable comments and suggestions, which substantially improved the analysis of this work.