Description
Survey research is in a state of crisis. People have become less willing to respond to polls and recent misses in critical elections have undermined the field's credibility. Pollsters have developed many tools for dealing with the new environment, an increasing number of which rely on risky opt-in samples. Virtually all of these tools require that respondents in each demographic category are a representative sample of all people in each demographic category, something that is unlikely to be reliably true. Polling at a Crossroads moves beyond such strong limitations, providing tools that work even when survey respondents are unrepresentative in complex ways. This book provides case studies that show how to avoid underestimating Trump support and how conventional polls exaggerate partisan differences. This book also helps us think in clear and sometimes counterintuitive ways and points toward simple, low-cost changes that can better address contemporary polling challenges.
Table of Contents
Part I. Polling in Context: 1. Modern Polling: Challenges and Opportunities; 2. The Story of Polling in 2.5 Fiascos; 3. Weighting: An Appreciation and Critique; 4. The Wild West of Contemporary Polling; Part II. A Framework for Modern Polling: 5. Non-Ignorable Non-Response; 6. Contemporary Polling and Non-Ignorable Non-Response; Part III. Fighting Non-Ignorable Non-Response: 7. Bounding Damage from Non-Ignorable Non-Response; 8. Selection Models for Non-Ignorable Non-Response; 9. Next Generation Selection Models; 10. Randomized Response Instruments; 11. Putting it Together; Part IV Applications: 12. Non-Ignorable Non-Response in Political Surveys; 13. Non-Ignorable Non-Response in Public Health; 14. Conclusion.
