Key Takeaways: A Brief History of Polling and Political Measurement

1. Polling has evolved through distinct eras, each defined by its dominant technology and methodology. From straw polls (1824-1930s) to quota sampling (1930s-1950s) to probability-based telephone surveys (1960s-2000s) to online panels and mixed-mode designs (2010s-present), each era solved the problems of the previous one while creating new challenges.

2. The 1936 Literary Digest disaster established the most important principle in survey research: representativeness matters more than sample size. The Digest's 2.4 million biased respondents produced a worse result than Gallup's few thousand carefully selected respondents. This lesson applies today: a large non-probability sample is not inherently superior to a smaller probability-based one.

3. The 1948 Dewey-Truman upset revealed the limitations of quota sampling and the importance of late-campaign measurement. The errors---interviewer selection bias, early cessation of polling, and poor handling of undecided voters---triggered the transition to probability sampling, which became the gold standard for decades.

4. The golden age of telephone polling (1970s-2000s) depended on conditions that no longer exist: near-universal landline penetration, high response rates, social norms around answering the phone, and relatively high institutional trust. Understanding why the golden age worked helps explain why it ended.

5. The cell phone crisis undermined the sampling frame that made telephone polling possible. As cell-only households grew from 5% to over 55% of the population, landline-only polls became systematically biased. The industry's response---dual-frame sampling---was more expensive and complex but necessary.

6. Response rates have collapsed from over 35% to under 5%. This collapse is driven by caller ID, robocalls, survey fatigue, time pressure, and declining institutional trust. Low response rates make all polls more vulnerable to nonresponse bias, regardless of how the initial sample is selected.

7. Online panels offer dramatic cost and speed advantages but raise fundamental questions about sampling methodology. The shift from probability to non-probability sampling is the most consequential methodological debate in modern polling. The answer is not yet settled.

8. Every major innovation in polling creates new blind spots. Quota sampling fixed selection bias but introduced interviewer bias. Probability sampling fixed interviewer bias but required telephone access. Online panels fix cost and access problems but reintroduce self-selection concerns. No method is permanently reliable.

9. Failures drive progress. The Literary Digest disaster led to scientific sampling. The 1948 debacle led to probability sampling. The cell phone crisis led to dual-frame and mixed-mode designs. Post-2016 reforms included education weighting. The pattern suggests that current errors will eventually produce the next methodological advance.

10. Measurement shapes reality: polling methodology determines whose voices count. At every historical stage, the people excluded from measurement---by poverty, by technology, by language, by distrust---were rendered politically invisible. Expanding the circle of who is counted has required not just political will but methodological innovation.