Sears and Lau (1983) presented evidence that apparent self-interest effects can be, and have been, generated in political surveys by question order artifacts. This evidence was based in part on a tabulation of published reports of self-interest effects in the NES series, specifically on the political effects of personal financial situation. From another analysis of the NES data, Lewis-Beck (1985) concluded, to the contrary, that personal finances have in fact had a consistent effect on voting preferences, without significant contamination from such artifacts. We here argue that his analysis inappropriately defines the conditions for possible contamination. We first lay out a theory of when such contamination effects might occur. We then repeat our analysis, taking into consideration both his observations and our own reappraisal of our procedures. We obtain results consistent with our original position, although the results are confounded by different types of questions appearing disproportionately in contaminated and uncontaminated conditions. However, the 1984 election appears to be a special case, in which self-interest effects were strong and relatively uncontaminated. We then report a split ballot experiment that is not confounded by item content, and find results consistent with our original position. However this methodological debate may be resolved, on the larger question of whether people's economic self-interest has major political implications, the evidence seems clear. In cases not contaminated by item order, which we would take to be the most appropriate test of self-interest effects, personal finances have on the average had only a small effect on political responses.
All Science Journal Classification (ASJC) codes
- Sociology and Political Science