ASSESSING THE PROGRAMMATIC EQUIVALENCE ASSUMPTION IN QUESTION WORDING EXPERIMENTS UNDERSTANDING WHY AMERICANS LIKE ASSISTANCE TO THE POOR MORE THAN WELFARE
ASSESSING THE PROGRAMMATIC EQUIVALENCE ASSUMPTION IN QUESTION WORDING EXPERIMENTS UNDERSTANDING WHY AMERICANS LIKE ASSISTANCE TO THE POOR MORE THAN WELFARE
复制标题
DOI：
10.1093/poq/nfs054
复制
发表时间：
2013-03-01
影响因子：
3.4
通讯作者：
Paris, Celia
中科院分区：
文献类型：
--
作者：
Huber, Gregory A.;Paris, Celia
在试图理解为什么美国人对福利表现出相对较高的反对程度时，学者们经常求助于本杂志报道的一个典型实验的分析(Smith 1987；Rasinski 1989)，在该实验中，受试者被问及他们对“福利”或“援助穷人”的支持程度。这一实验一贯表明，美国人对福利的支持远远少于对穷人的援助。这种差异被解读为简单地将相同的核心项目描述为福利而不是对穷人的援助会压低支持率。这些分析中的关键假设之一是方案等价性：相对于“援助穷人”一词，“福利”一词描述的是相同的方案，但在考虑哪些方面有所不同。这份研究报告检验了这一关键假设的有效性。对新的实验数据的分析表明，美国人认为哪些项目是福利，哪些项目他们认为是对穷人的援助，似乎有根本的区别。我们讨论了我们的研究对解释先前依赖于这个实验来检验框架理论的研究的影响，并提出了对调查实验设计的更广泛的影响。
In trying to understand why Americans display relatively high levels of opposition to welfare, scholars have frequently turned to the analysis of a canonical experiment reported in this journal (Smith 1987; Rasinski 1989) in which subjects were asked about their support for either "welfare" or "assistance to the poor." This experiment consistently shows that Americans are substantially less supportive of welfare than of assistance to the poor. This difference has been interpreted as evidence that simply describing the same core programs as welfare rather than assistance to the poor depresses support. The key assumption in these analyses is one of programmatic equivalence: relative to the words "assistance to the poor," the word "welfare" describes the same programs, but differs in which considerations it brings to mind. This research note examines the validity of this key assumption. Analyses of novel experimental data show that there appear to be basic differences in which programs Americans consider to be welfare and which they consider to be assistance to the poor. We discuss the implications of our research for interpreting prior studies that rely upon this experiment to test theories of framing, and we suggest broader implications for survey experimental designs.