Opinion surveys and political representation.

Abstract

Public opinion surveys are a modern vehicle for political representation, a way for citizens to express themselves and a means for leaders to learn about citizens. Surveys influence the content of the media, structure political discussion, provide information that influences political decisions and calibrates formulas for the allocation of government funds. Surveys also dominate scientific research in the social sciences. For all the uses to which scientists and the government put surveys, there is a common flaw: survey nonresponse, the proportion of people selected to participate in the study but who, for various reasons, are not interviewed. The fraction is large, and it is growing. Despite the importance of surveys to politics and to science, we do not know either the causes or consequence of nonresponse. In Chapter II, "Who's Missing?" I examine how the sample of typical academic surveys differ from benchmark estimates of the population. In Chapter III, "Why do people participate in surveys?" I model the process by which people are drawn into surveys as comprised of three stages: contact, eligibility and compliance. Drawing on multiple methods introduced in this chapter, I develop the components of a theory of survey participation as a set of relationships between potential respondents and the process of survey participation. Chapter IV, "The structure of survey compliance," unites the ideas developed in Chapter III into a structural model of survey compliance. In Chapter V, "How survey nonresponse damages scientific research," I review the statistical literature on how nonresponse undermines univariate and multivariate analyses of survey data. The heart of this dissertation is Chapter VI, "Applications of corrections for nonresponse." In this chapter, I apply the model of survey participation developed in Chapters III and IV and the corrections explicated in Chapter V to a diverse set of models of politics. In Chapter VII, "What is response rates worsen?" I simulate the effects of declining response rates on univariate and multivariate analysis. Finally, in Chapter VIII, "Surveys and misrepresentation," I return to a discussion of the role of public opinion surveys in the political process.