This is an open access article distributed under the terms of the Creative Commons Attribution License (The rise of right-wing populism and authoritarianism has renewed focus on the stability and resilience of democracies. Right-wing populism is closely linked to attitudes such as racism and sexism (Mudde, 2007), making their empirical examination a key task for the social sciences. At the same time, the opposite view of democratic attitudes is also of crucial importance. Especially the internalization of democratic values as a stable form of diffuse support is crucial for the stability of democracies (Easton, 1975). There is a pressing need to better understand how such values develop. Although political socialization occurs throughout life, childhood is a key phase for internalizing values (Abendschön, 2010; Döring, 2018). However, empirical findings remain scarce, partly due to the challenges of conducting standardized studies with children (Abendschön, 2022).
In general, values can be defined as stable beliefs about desirable goals that influence individuals’ behaviour across different situations (Schwartz & Bilsky, 1987). They differ from attitudes particularly in their cross-situational consistency and their level of abstraction (Rokeach, 1973). Various types of values can be distinguished, but this study focuses specifically on the measurement of individually projective values (Klages, 1992). These are internalized at the individual level, but relate to perceptions of society (Vogelbacher, 2019).
The question of which values can be defined as democratic values is, in contrast, controversial. Given the length of this article, it is not possible to provide a detailed discussion of the definitions.1 As a minimum definition, it can be assumed that “basic principles of democratic regimes are commonly understood to include such values as freedom, participation, tolerance and moderation, respect for legal-institutional rights, and the rule of law” (Beetham, 1994; Simon, 1996, as cited in Norris, 1999, p. 11).
These values are internalized as part of political socialization. While the autonomous and voluntary internalization of social norms and values through rational reflection appears cognitively demanding, findings from cognitive and moral development research show that by primary school age—around six years and older—children already possess the ability to engage independently and flexibly with moral questions (Kohlberg, 1969; Vogelbacher, 2019). Empirical studies show that children encounter political values early on (van Deth et al., 2007). This implies that childhood is a formative period for the development of democratic value orientations (Abendschön, 2010).
Conceptualization and Operationalization of Democratic Values in Children
Measuring democratic values in childhood entails conceptual and methodological challenges, particularly in ensuring that the instruments used are both theoretically grounded and developmentally appropriate. Although numerous studies have emphasized the importance of early political socialization (Abendschön, 2017), the empirical study of democratic values in children remains underdeveloped and methodologically fragmented.
Several established studies have explored children’s political orientations, but often with limited theoretical consistency or measurement precision. For instance, van Deth et al. (2007) employed a broad set of items on citizenship, equality, and rules in the “Demokratie Leben Lernen” (DLL) study. Even if the items are thematically relevant, the theoretical reference to fundamental democratic values is worth discussing. Based on DLL-data Abendschön (2010) refers to helpfulness, hard work, gender equality, and law-abidingness as “good citizen orientations.” While some of these (e.g., law-abidingness, gender equality) align with democratic principles, the inclusion of hard work raises questions about conceptual clarity. For example, it is quite similar to the items measured in the PBVS-C developed by Döring et al. (2010), which however focuses on the measurement of basic values in children.
To address these limitations, the present study builds on Norris’s (1999) framework of core democratic values and uses various measurement instruments derived from previous empirical research to measure these values. Each value was defined conceptually and translated into short, school-related scenarios to ensure contextual relevance and cognitive accessibility for third- and fourth-grade pupils. The items were informed by both theoretical models and existing empirical instruments, while deliberately seeking to avoid their limitations.
Participation was operationalized via classroom-based decision scenarios (see Appendix 1.1). The items, adapted from Kinoshita (2006) and Helwig et al. (2003), confronted children with choices between majority voting, consensus decisions, and authority-based decisions. While the majority principle is operationalized as democratic, the latter are seen as non-democratic. Prior research has shown, that esp. younger children prefer consensus-based decisions due to its lower complexity (Helwig, 1998). Research has shown that children begin to develop differentiated views on the principle of majority rule from around the age of eight (Kinoshita, 2006). But, even younger children already tend to prefer decisions based on the majority principle (Helwig et al., 2003; Helwig & Kim, 1999; Helwig & Turiel, 2002).
Equality was captured through items addressing task assignments (e.g., sweeping, playing soccer) based on gender (see Appendix 1.2). In contrast to the original formulation in the DLL survey (van Deth et al., 2007), the abstraction was reduced and the focus shifted from the parents to the children's everyday lives, thus aiming to minimize socially desirable responding related to family norms (Emde & Fuchs, 2012). This approach allows for the identification of potential stereotypical attributions regarding gender roles held by children and thus captures at least one dimension of equality.
In line with Norris’s (1999) definition of democratic values, particularly the emphasis on respect for legal-institutional rights and the rule of law, these principles were operationalized through two distinct but related dimensions. First, following rules was measured using items adapted from the DLL study (van Deth et al., 2007), which presented classroom-based scenarios asking whether children believe rules should always be followed (see Appendix 1.3). The aim was to capture both the general importance of rules and personal willingness to comply as expressions of respect for legal norms. Second, the renunciation of violence was included as a closely related value dimension (see Appendix 1.4). Although rarely addressed in previous studies, it reflects the expectation that democratic societies reject violence as a means of conflict resolution. This was assessed through two items measuring approval or disapproval of violence toward peers and authority figures. The conceptualization aligns with existing instruments such as the ALLBUS (GESIS, 2021), emphasizing that democratic norms favour non-violent, legally grounded solutions in interpersonal and institutional contexts.
Freedom of opinion was assessed through four items (see Appendix 1.5): the first addressing the general legitimacy of expressing one’s views, as similarly surveyed by Šerek and Lomičová (2020); the second and the third addressing the acceptance of dissenting opinions within the group and towards younger pupils (Solomon et al.,1972); the fourth item examines whether it is legitimate for school authorities to restrict pupils’ freedom of expression, which is also inspired by former surveys (Haerpfer et al., 2022; Helwig, 1998).
Particular attention was paid to ensuring that the questionnaire was designed to be as child-friendly as possible. When designing surveys for children, specific requirements must be met to support comprehension (Maschke & Stecher, 2012). First, this includes, administering the survey in classroom settings, which makes the situation more familiar and comfortable for children (Lang, 1985). Second, simplicity is also essential in the design of response options. Since children in third and fourth grade are already capable of reading, pictorial response formats (see for example the DLL study or the World Vision studies, Andresen & Hurrelmann, 2010) were deliberately not used. Nevertheless, clear and simple categorical answer formats were chosen to provide structure and simplicity of interpretation, as recommended in the literature on children's questionnaires (Maschke & Stecher, 2012). In addition, regular breaks were integrated into the questionnaire to keep all children on the same pace and to allow interviewers to read questions aloud to the class.
All items used categorical response options (e.g., yes/no, boys/girls/both), which were dummy-coded for analysis. One option consistently reflected a democratic orientation, the other a position contrary to democratic norms. This dichotomization was conceptually motivated, as the aim was to contrast acceptance vs. rejection of normative principles, rather than measure degrees of endorsement.
Method
Sample and Implementation of the Study
In this study, a total of 429 primary school pupils in 3rd and 4th grade in North Rhine-Westphalia in western Germany were surveyed between August and November 2023. Children were recruited from a total of 24 classes. 47,9% of the children stated that they were female, which meets the distribution of the representative IGLU study (McElvany et al., 2023, p. 46). In terms of socioeconomic status, the children surveyed tend to be distributed in the upper range of the “Family Affluence Scale (FAS)” (Hobza et al., 2017). The mean score on the 13-point scale is 9.58 (SD = 0.12), which means that the pupils surveyed tend to have higher social status.
Initially, the designed questionnaire was checked in two pre-tests. First, the questionnaire was applied in a focus group discussion with three 7 to 9-year-old children to test the wording, the understanding of the items and the relevance for children in this age (Kränzl-Nagl & Wilk, 2000). This was particularly relevant for those items which were translated into German or newly developed and is generally recommended for children studies (Maschke & Stecher, 2012). Second, a test survey with a complete class of fourth graders was conducted. After coordinating the research project with the school management, class teachers and parents, each survey was conducted by two students from a research seminar in the Master of Education. Interviews were conducted in class using a standardized test administration script.
Due to the relatively small sample of 429 children and with approximately 20% missing data in the non-imputed regression models, it seems reasonable to multiple impute these missing values. Fully conditional specifications (fcs) with 20 imputations and 100 burn-in cycles were used to calculate missing values for all variables surveyed, regardless of their scale level (von Hippel, 2020). The following calculations refer to the imputed data set with all cases.
Validation Approach
The validation strategy applied in this study primarily relies on confirmatory factor analyses (CFA), which are mainly used to assess the internal structural validity of the developed scales. These appear to be well suited for validating the present measurement instrument, as the theoretical construct of democratic values is predetermined as well as the number of factors (Reinecke, 2014). At the same time, it should be emphasized that CFA addresses only one of several dimensions of validity. In accordance to the theoretical basis of the questionnaire, it is expected that the respective items as manifest variables measure the underlying value as a latent construct and accordingly load positively and significantly on the corresponding factor. At the same time, all factors are expected to be interrelated as they together represent the underlying concept of democratic values (see Figure 1).
Figure 1
Simplified Model of the Confirmatory Factor Analysis (CFA)
Accordingly, it must first be checked to what extent the items used measure the latent constructs. Since variables were collected categorically in order to reduce complexity when surveying children, dummy variables must be coded first. Even if CFAs are designed for metric scale levels, the use of nominal variables is possible (Reinecke & Pöge, 2010).
For each scale, the response category that corresponds to the underlying democratic value is selected. For the value of participation, agreement with the majority decision is coded as 1. Preferences for consensus or teacher decisions are coded 0 (for an overview of the exact response categories, see Appendix 1). The internal reliability of the scale measuring the democratic value of participation is 0.422, thus falling just below the commonly accepted threshold of 0.5 (Verma & Abdel-Salam, 2019). Low reliability coefficients are a well-documented challenge in research with children (Emde & Fuchs, 2012). Even in comparable studies, such as the previously discussed DLL study, reliability values must be regarded as marginal (Abendschön, 2010, p. 129). Given the significant research gap in the political socialization of democratic values in childhood, it appears justifiable to use these initial findings as a starting point for further development and refinement. For the value equality, the answer “both”, is coded as 1. A preference for one of the two genders is coded as 0. The internal consistency of this scale is 0.42. For the value following rules, the answer “important” is coded as 1 and, for renunciation of violence, the rejection of both forms of violence. For the former, the bivariate correlation between the two items is 0.36 (p ≤ .05), while for the latter, it is 0.21 (p ≤ .05). For the value of freedom, general agreement, acceptance of minority opinions and rejection of authoritarian decisions are coded as 1. All other answer options are coded with 0. In this case, the internal consistency is an unacceptable -0.023.
The coefficients in the CFA are to be interpreted as correlations between the manifest variables and the respective factor. The crucial criteria are the direction and significance. As usual with correlations, a coefficient from approx. 0.3 is considered a medium correlation and from 0.6 as satisfactory (Reinecke & Pöge, 2010). In addition, the fit indices indicate the model fit, with a CFI of 0.96 or greater and an RMSEA of 0.05 or less indicating that the model can be accepted (Hu & Bentler, 1998, 1999). To compare different models, the AIC is reported, with a smaller value indicating a better model fit in comparison (Bühner, 2021; Reinecke & Pöge, 2010).
Results
First, it is worth taking a look at the overall model of the CFA (see Appendix 2.1). With regard to the participation factor (F_MR), the items 3 and 7 stand out, as their factor loadings miss the significance level of 5%. With a p-value of .063 Item 1 is also just above the required 5% level. Furthermore, it can be seen that the correlations with the participation factor are quite small. Only Item 6 shows a low to medium correlation. All four items load significantly and positively on the factor equality (F_EQ). In particular, Items 3 and 4 demonstrate moderate correlations with the factor. This pattern also applies to the factors following rules (F_FR) and renunciation of violence (F_RV), where significantly positive factor loadings between 0.12 and 0.25 are observed. In contrast, a heterogeneous pattern emerges for the factor freedom, characterized by opposing factor loadings. Items 1 and 4 load positively, whereas Items 2 and 3 load negatively. Despite the recoding of the items as dummy variables in line with theoretical considerations, they appear to measure opposing latent constructs. This is especially evident from the significant correlations of Items 1, 2, and 3, which clearly contradict the hypothesized structure.
The covariances between the factors, also reported in Appendix 2.1, are noteworthy. The factor participation exhibits non-significant negative correlations with the other factors. Moreover, the factor freedom correlates negatively and even significantly with the factors equality, following rules, and renunciation of violence. While those three show highly significant, positive, and satisfactory intercorrelations, the negative correlations with participation and freedom challenge the assumption that all factors can be subsumed under democratic values.
Additionally, the global model fit indices indicate that the model demonstrates limited fit to the data. The CFI of 0.808 is below the threshold of 0.96, but the RMSEA, at 0.038, is under the recommended cut-off of 0.05 (Hu & Bentler, 1998, 1999). In light of these findings, it seems necessary to further modify the model.
In the next step, only the significant variables were included in the CFA. Due to the small sample size, Item 1 was retained for the factor participation in subsequent calculations, while Items 3 and 7 were removed. For the factor freedom, the non-significant Item 4 was removed, as well as Item 1, which loaded inversely compared to the other items on the factor. Removing Item 1 resulted in positive loadings for Items 2 and 3 on the factor freedom. Additionally, the descriptive distribution of responses (see Appendix 1.5) revealed minimal variance in Item 1, with approximately 98% of children agreeing with the statement. Similarly, item 4 exhibited a high agreement rate of about 89%. Furthermore, this item required a considerable amount of missing data to be imputed, as roughly 14% of children did not respond to it.
In the adjusted CFA-model, the factor loading of Item 1 on the factor participation remains significant only at the 10% level (see Appendix 2.2). All other variables are significant at least at the 5% level. Correlations range between 0.07 and 0.30. Notably, the instruments measuring equality, following rules, and renunciation of violence demonstrate satisfactory psychometric performance in the present study. Specifically, these scales showed consistently significant factor loadings with clear theoretical alignment, indicating reliable measurement of these democratic values. For the factor freedom, two significant items remain, with correlations ranging from 0.07 to 0.25. All reported correlations remain highly significant in this model as well.
The pattern of covariances remains consistent, with the factor participation continuing to correlate negatively and non-significantly with the other factors. The remaining factors, however, correlate significantly and positively with one another, with correlations ranging between 0.5 and 0.8. Additionally, the now-significant correlation of the factor freedom with the other factors supports the decision to remove the first item (free_gen) from the model, as this adjustment establishes a relationship with the other democratic values. When comparing the two models, the adjusted model is preferable due to its lower AIC. The CFI improves to 0.912, falling just below the required threshold. Furthermore, the RMSEA value of 0.038 supports acceptance of the model.
Discussion
The present study aims to develop and validate a standardized instrument for assessing democratic values in elementary school children. Building on existing literature on political socialization and value development, the study examines the extent to which the underlying democratic values (participation, equality, following rules, renunciation of violence, and freedom) can be measured.
The results of the CFA suggest that the theoretically proposed structure of democratic values can be tentatively confirmed, though some limitations must be addressed. For the values equality, following rules, and renunciation of violence, the items measured the latent constructs significantly and in the expected direction. Participation and freedom items showed weak or inconsistent factor loadings. This clearly indicates that these two scales do not work as expected. The overall model fit improved after the removal of discrepant items, and the RMSEA now suggests an acceptable model. However, further work is needed to optimize the instrument. The overall model fit for the child sample and particularly the lower CFI reflects the challenges of capturing abstract value constructs in a young age group. While the RMSEA indicates acceptable fit, the CFI remains just below conventional thresholds, even after model adjustment. This may be due to the limited variance in responses and the simplified item formats, which were necessary to ensure child comprehension but may reduce measurement precision. These limitations underline the need for more age-sensitive item formats and further refinement of the instrument. By contrast, the values of equality, following rules, and renunciation of violence were measured with stable and interpretable factor structures. Thus, they provide a basis for further research and practical applications in the context of children's political socialization.
The CFA model was adjusted post hoc to improve fit by excluding items with low loadings. These results should be seen as preliminary and require future replication. Future research should therefore aim to replicate the identified factor structure through cross-validation in new datasets and conduct additional tests such as measurement invariance analyses across different groups (e.g., age, gender, or school context). These steps are essential to ensure that the developed factor structure is robust, reliable, and generalizable, and not merely an artefact of the present sample. Moreover, future studies should consider applying ordinal CFA techniques to account more precisely for the categorical nature of the response formats.
The value participation exhibits significant deviations. Several items had to be removed due to non-significant loadings. Additionally, the age effects on decision-making preferences reported in previous studies (e.g., Kinoshita, 2006) could not be replicated in this study. Also, there is no evidence, that the children in this study distinguish between the different domains of decision-making. This aligns with prior findings for the present group of 8–9 year-olds (Kinoshita, 1989, 2006). All this may indicate that the scenarios presented were difficult for children to evaluate or that the concepts underlying the decision modes are not yet fully understood at this age. The results underline the need to further improve the operationalization of this value. For example, less text-heavy items or picture-based scenarios could be used to better illustrate the decision-making situations to the children, similar to the PBVS-C (Döring et al., 2010).
The scale to measure freedom also shows methodological challenges. Due to uncorrelated and opposite factor loadings, two items had to be removed. The low variance of the first item indicates that the children surveyed are either very unanimous or that they may have already internalized social desirability. A further elaboration of the items based less on general principles and more on concrete examples that are closer to the reality of the children could help here. In addition, the situation in which the school administration restricts pupils’ freedom of expression may have been too abstract.
The covariances between the factors show that equality, following rules and renunciation of violence correlate positively and significantly with each other, whereas participation consistently shows a negative and non-significant relationship. This pattern indicates that, from the children's perspective, participation is seen as an independent value that is less closely linked to the other democratic values. Alternatively, this could be a methodological artefact based on the described measurement problems. In any case, it contradicts the theoretical assumption shown in Figure 1.
The results of the present study must be considered in the context of several methodological limitations that influence both the interpretation of the findings and their generalizability. First, the sample size of 429 children is sufficient to conduct confirmatory factor analyses, but too small to conduct more complex analyses such as invariance tests, especially for subgroups such as gender or grade level. Second, the use of a purely text-based questionnaire is a potential barrier, especially for children with lower reading skills. Although pre-tests and a comprehension question ensure comprehensibility, issues with more complex or highly contextualized items may have led to response bias. In addition to the reliance on text-heavy items, the varied design of response formats may also have contributed to reduced comprehension among the children. Although nominal response categories were deliberately chosen to facilitate understanding, this approach may have had the opposite effect. A promising example for potential revision is the PBVS-C by Döring et al. (2010), which was also used in this study to assess basic values. Furthermore, findings by Gerbeshi et al. (2024) suggest that children can respond to simple Likert scales. Future work should explore whether visual analogues or adapted Likert formats might offer a more nuanced balance between simplicity and measurement precision. Third, some items, particularly in the area of participation, seem difficult for the age group to understand. Despite the focus on school situations and the explanation of the decision-making modes, some items may have remained incomprehensible. The low variance in the responses for certain values, such as general freedom of opinion, also indicates that the questions were too general to actually measure the children's underlying values. It should also be noted that neither dropout rates nor the amount of missing data were unusually high. Future studies could incorporate external validation criteria, such as complementary behavioural observations or qualitative interviews. Especially the former could help to assess whether children have actually internalized the respective values and act in accordance with them, as implied by the definition of values (Schwartz, 2012).
Conclusion
The present study provides initial empirical evidence that children show interest in political issues reflected in the questionnaire, such as equality, participation, and freedom of opinion. The results of the study show, on the one hand, the difficulties of measuring democratic values in children, but also how promising this is. Esp. the values of equality, following rules and renunciation of violence can be measured nearly accurately. Future research should employ more comprehensive validation strategies to further substantiate and differentiate these findings.
This study offers several implications for future research. First, picture-based or interactive items may improve comprehension and cognitive accessibility. Second, realistic, child-relevant scenarios could enhance item validity. Third, larger, representative samples and qualitative approaches (e.g., interviews, group discussions) may deepen insights and support the refinement of item design and value operationalization.
In sum, this study provides a foundational step toward capturing democratic orientations in children and highlights the potential for early civic education based on robust empirical instruments. The results provide important impulses for the further development of survey instruments and lay a foundation for future methodological and content-related studies in this area. The long-term objective should be to establish democratic values as a central component of civic education and to make them measurable from an early age on.