The blind scorekeepers: Journalism, polling, and the battle to define public opinion in American politics

Survey research has been touted as an important tool for measuring citizens' attitudes, facilitating deliberation, and ultimately achieving more responsive government. Yet the reality of how polling data is used in American politics falls short of these goals. While political scientists have studied elite use of opinion data for message-testing and position-taking, little evidence has been gathered about the quantity and quality of survey data in the news or how these aggregated snapshots of mass opinion shape and reshape our understanding of collective preferences.;This dissertation offers new clarity on these dynamics and explores why poll results in the news rarely fulfill their function as a "public utility." Instead of helping the public hear each other's perspectives better or empowering citizens' voices in policymaking, most references to survey data in the news end at tracking the standing of politicians in the partisan "score," with advocacy groups actually sponsoring a growing share of data that assesses support for policies and other attitudes. Audience demand may only be partly responsible. Vast changes in both the news and opinion research industries are undermining journalists' abilities to adjudicate between competing opinion claims while a mismatch between journalistic norms and the constraints of conventional research practices limit the usefulness of many surveys as effective reporting tools. By averaging across all Americans, polls may obscure the full spectrum of citizens' attitudes, resulting in the marginalization of public opinion except in cases with clear electoral implications. Additionally, the dissertation presents results from two survey experiments that indicate how perceptions of collective opinions, and poll reporting in particular, exert powerful (if conditional) effects on attitudes, synthesizing theories of impersonal and elite cueing---with significant implications for the study of micro and mass opinion. Rather than informing the public about itself, coverage of polls may enflame existing identities, induce conformity, and engender less accurate perceptions of others' opinions. The dissertation employs a range of methodological approaches including large-scale content analysis and 41 in-depth interviews with prominent journalists and practitioners.