Economic Survey Response Behavior: Experiments Using A Virtual Laboratory

丹尼尔·L·麦克法登加州大学伯克利分校SBR-9601149经济和其他调查一直被试图获取受访者真实行为信息的问题所困扰。调查问题必须仔细构思，以达到获取准确信息的目的。在调查设计中出现的具体问题包括锚定和焦点。锚定是指一种被广泛观察和有充分记录的家庭效应，在这种效应中，问题提供的可能不相关的线索或背景会影响反应。例如，受试者被要求判断一个不确定的数量，如亚马逊河的长度是否超过了某个特定的值(锚)，然后要求估计，通常会给出一个被拉向锚的回答。焦点指的是四舍五入的、常规的反应。例如，被问及到达工作所需时间的受试者通常会给出5分钟、10分钟、15分钟、30分钟、60分钟这样的答案，而不是中间值，所以即使真实值有一个连续的分布，人们也会观察到在这些焦点值下的反应尖峰。这项研究将使用计量经济学方法来推断这些偏差的程度，以及如何根据使用互联网(万维网)作为虚拟实验室的问卷设计实验来调整这些偏差。

Daniel L. McFadden University of California - Berkeley SBR-9601149 Economic and other surveys are plagued by the continuing problem of trying to elicit information about respondents' true behavior. Survey questions must be carefully framed to accomplish the aim of eliciting accurate information. Among the specific problems arising in survey design to be investigated are anchoring and focal points. Anchoring refers to a widely observed and well- documented family effects in which presumably irrelevant cues or settings provided by questions influence responses. For example, subjects asked to judge whether an uncertain quantity, such as the length of the Amazon, exceeds a certain value (the anchor) and then asked for an estimate, will typically give a response that is pulled toward the anchor. Focal point refers to a rounded off, conventional response. For example, subjects asked for the time it takes them to get to work will frequently give answers such as 5, 10, 15, 30, 60 minutes, rather than intermediate values, so that one observes spikes of responses at these focal values even though the true values have a continuous distribution. The research will use econometric methods to infer the extent of these biases and how they might be adjusted for based on experiments with questionnaire design using the internet (world-wide web) as a virtual laboratory.