Measurement Context Effects in Telephone-Survey-Based Tests of Causal Models

The purpose of this research is to examine the issue of measurement context effects in survey-based tests of attitudinal and related models. The specific issue examined concerns the degree to which the measurement process affects the objects of measurement (i.e., various attitudinal and related concepts). Based upon the memory accessibility-diagnosticity theory specified by Feldman and Lynch (1988) and the concept of spreading activation (Tourangeau and Rasinski 1988; Anderson 1978, 1983; Collins and Loftus 1975), the effects of context questionnaire items on answers to, and estimated relationships among, target questionnaire items in a study involving measures of antecedents and consequences of attitudes are examined. The findings indicate some measurement context effects in an equation predicting blood donation intentions. First, the findings suggest that measuring expectations prior to measuring intentions, when compared to measuring expectations after measuring intentions, caused the intentions scores to be higher. Since the respondents had generally favorable attitudes toward blood donation, this supports the Feldman and Lynch (1988) argument that context survey questions can result in activated beliefs that are diagnostic for answers to subsequent questions. Second, the findings suggest that measuring expectations prior to (after) measuring intentions resulted in a statistically significant increase (decrease) in the association between expectations and intentions. Third, the findings produced evidence that buffer questions that separate expectations (i.e., context) measures from other target measures reduced the context effects associated with the expectations measurement. An implication of this finding is that the use of intervening questionnaire items to separate questions that are likely to be influenced by measurement context effects may reduce context effects.