降低评价尺度偏差:一项政治效能感测量的实验

Controling Rating ScaleBias:A Experiment on Political Efficacy Measurements

评价尺度偏差是政治学评价类题目中常见的测量误差,"虚拟情境锚定法"被认为是一种降低此类误差的方法。本文依托北京大学中国国情研究中心2009年全国公民意识调查,在其中加入了一组采用"虚拟情境锚定法"测量外部政治效能感的题目来跟传统的效能感测量方法进行比较,实验结果发现:(1)运用传统效能感测量方法时,高效能感的比例为30.64%,而使用"虚拟情境锚定法"后,高效能感的比例占到了46.91%,这就验证了评价尺度偏差的存在。(2)效度检验发现"虚拟情境锚定法"得到具有更好建构效度的测量结果,从而验证了评价尺度偏差对测量质量的影响。通过实验对题目设计提出建议:(1)控制评价尺度偏差的基本思路是遵循测量指标设计上的具体化、标准化和高可比性原则。(2)对于能采取具体方式提问的变量,就要避免抽象题目;对于能用事实性问题测量的变量,就要避免态度性测量。(3)在测量问题选项的尺度设计上,要避免评价方向不一致,应尽可能地理顺一致的量尺向度,规避尺度不等距的问题。

Rating scale bias is a common measurement error in political survey.The approach of anchoring vignettes is claimed to be effective in coping with this issue.This article tries to evaluate the measurement validity of the approach of anchoring vignettes and traditional approach with an experimental survey of measuring the political efficacy through the national survey 'Attitudes towards Citizenship in China(2009)' conducted by the Research Center of Contemporary China.The results show that:(1) When using traditional approach,the percentage of 'high level political efficacy' is 30.64%.When using the approach of anchoring vignettes,the percentage is 46.91%.The data result proves the existence of the rating scale bias.(2) The approach of anchoring vignettes is superior to the traditional approach in terms of construct validity assessment.It means that rating scale bias does affect the conclusions of the study.According to analytical results,the article puts forward a few advices on questionnaire design:(1) In order to control rating scale bias,the questions should be specific,standardized and comparable.(2) Using specific questions in place of abstract questions,and using factuality questions in place of attitudinal questions,whenever possible.(3) Avoiding using scales in a different direction or unequal interval.