Stretched Rating Scales Cause Guided Responding

Decision making by policymakers, public health professionals, and health care providers is often guided by the extent to which individuals feel at risk for certain adverse health events. Such health risk perceptions can be assessed in surveys using different types of probability rating scales. It has recently been suggested that rating scales that offer decomposed numeric values at the lower end of the scale (stretched scales) improve the accuracy of estimates of small risks. However, the authors suggest that respondents use the differentiated small numeric values as cues to guide them to the correct response. Study 1 supports this proposition by showing that response distributions are substantially skewed toward the lower end of stretched rating scales and have restricted variances as compared with equal-interval scales. Study 2 provides experimental evidence that scores on the stretched scale are a result of guided responding. The results show that scores on stretched rating scales are not a valid reflection of respondents' risk perceptions, but, instead, guide responses to the end of the scale that has been stretched. The findings suggest that stretched rating scales result in biased risk estimates, which may hinder effective communication about health risks between decision- and policymakers as well as between individuals and their health care providers.