A chain of associations
The survey found that three measured factors were positively associated with community education satisfaction: individual needs and participation motivation, perceived service quality, and sense of community belonging. In the researchers’ structural equation model, the standardized paths to satisfaction were 0.271, 0.314 and 0.310, respectively, with each p-value below 0.001. In plain terms, people reporting stronger needs and motivation, better views of service quality, or a stronger sense of belonging also tended to report higher satisfaction. The pattern is a statistical association, not evidence that any of those factors caused the result.
The model also tested a sequence rather than treating the factors as unrelated. It linked needs and motivation with perceived service quality, service quality with belonging, and belonging with satisfaction. Using 5,000 bootstrap samples, the estimated indirect association through service quality was 0.202; through belonging, 0.093; and through the full chain, 0.089. The reported 95% confidence intervals for all three estimates stayed above zero, meaning the data were consistent with nonzero indirect paths under the model. They do not, however, establish that the sequence unfolds in that order in real life.
Across the model, the total indirect estimate was 0.384, while the direct estimate was 0.301 and the total estimate was 0.684. The service-quality-to-belonging chain accounted for 13.0% of the total estimate, with a reported interval of 9.0% to 18.4%. These figures describe how the model split an overall statistical association into components. They are not a percentage of residents who were satisfied, nor a measure of how much satisfaction would change after an intervention.
Where education level entered
Education level showed up as a moderator only at the first step, the link between needs and motivation and perceived service quality. Adding the interaction changed the model’s explained variance from 0.325 to 0.331, an increase of 0.006. The interaction coefficient was 0.275, with p = 0.031 and a 95% confidence interval from 0.032 to 0.538. The conditional slope was 0.355 among respondents with high school education or below, compared with 0.629 among those with college education or above.
The result did not extend across the model. The authors explicitly described the education-level analysis as exploratory. They did not report supported moderation for the direct satisfaction path or later mediation paths, and education level had no significant direct association with belonging or satisfaction.
What was measured
The evidence came from an anonymous, low-risk questionnaire completed by 494 Beijing residents. Institutional protocol approval and electronic informed consent were reported, although the study did not hold an ethics approval number. The questionnaire contained 54 items, with 29 items from four predefined dimensions used in the main analysis.
Residents volunteered through community workers and WeChat, using convenience sampling. Previous participation in community education was not required, so the survey was open to any volunteering community resident. The researchers used descriptive statistics, Pearson and Spearman correlations, a test for whether one general response pattern dominated, and structural equation modeling for the proposed chain. They also used 5,000 bootstrap samples to estimate uncertainty.
On the measurement side, the four-factor confirmatory model had strong reported overall fit: the chi-square-to-degrees-of-freedom ratio was 1.11, the comparative fit index was 0.977, the Tucker-Lewis index was 0.975, the root mean square error of approximation was 0.015, and the standardized root mean square residual was 0.035. But standardized item loadings ranged from 0.205 to 0.497, and the analysis also reported low average variance extracted and overlap between latent constructs. A good overall fit therefore did not remove all questions about how cleanly the questionnaire separated the concepts.
The common-method checks were more reassuring. Harman’s first factor accounted for 20.267% of variance against the paper’s stated 50% threshold, and adding an unmeasured latent method factor moved the comparative fit index from 0.977 to 0.987, a change of 0.010. The authors judged common method variance not to be a serious threat, but that check does not establish temporal order or causal direction.
What the survey cannot settle
The main caution is the same one that applies to the associations throughout the study. This was a single-region convenience sample of 494 residents, gathered in a cross-sectional design. The authors say that limits generalizability and prevents the study from establishing temporal order or causal mechanisms; the sample may also be biased toward people who had participated in community education. The findings therefore should not be read as proof that improving perceived service quality or belonging would raise satisfaction.
The model comparison narrowed one question but did not resolve it. A parallel mediation model fit worse than the chain model, with a comparison p-value below 0.001. But the forward and reverse chain models were statistically equivalent because both were saturated, meaning the fit statistics could not distinguish them. The forward order was therefore chosen partly on theoretical grounds, not because the data clearly ruled out the reverse order.
Education-level moderation also needs restraint. It was exploratory, and its interaction contribution was small. The result is best treated as a lead for future testing, not a confirmed difference in how community education works across education groups.
The authors declared that no financial support was received for the work or its publication.
Paper data and sources
Original title: Research on the influencing factors and mechanism of community education satisfaction
Authors: Nie L, Niu S
Journal/Repository: Frontiers in psychology
Status: Peer-reviewed
First online: 2026-10-08
DOI: Not available
Original paper · Full text