對于作為集合性概念的社區社會資本,以往研究通行的做法是對采自個體的數據以社區為單位取均值作為相應高水平構念,再進行因子分析。然而,由于缺乏必要的評估,這種聚合處理存在信效度問題。在對以往有關研究進行總結和反思的基礎之上,立足于某大型城市社區調查數據,本文檢驗了社區社會資本相關題項的ICCs、Rwg等指標,發現總體來講各指標組內一致性和信度較低。結合相關方法的進展,本文建議(認知性)社區社會資本構建應置于多水平潛變量模型的框架之下,并使用多水平驗證性因子分析進行建模。最后,文章對相關發現的理論意義和方法存在的局限進行了簡要探討。

For constructing the measurement index of the community social capital as a collective concept,the previous studies have conducted factor analysis based on the cluster means aggregated from the individual to form the corresponding construct at the community level. However,this method lacks reliability and validity due to the absence of necessary evaluation. This paper examines the ICCs and Rwg of relevant items of community social capital from a large survey of urban communities,and finds that the internal consistency and reliability of indicators are low. The author suggests that the measurement of( cognitive) community social capital should be placed under the framework of multilevel latent variable model,and finish with a multilevel confirmatory factor analysis. Theoretical implications of the findings and limitations of the method are also discussed.