Interpreting poll data as a prediction of election outcomes is a practice as old as the field, rooted in a fundamental misunderstanding of what poll data means.
By first understanding how polls work at a fundamental level, this book gives readers the ability to discern flaws in the current methods. Then, through specific political examples from both the US and UK, it is shown how polls famously derided as "wrong" were, in fact, accurate. While polls are not always accurate, the reasons we can and can't (rightly) call them "wrong" are explained in this book.
This book will equip readers with the tools to navigate the mismatch of expectations. It is not intended to replace more technical applications of statistics but is accessible to anyone interested to learn more about how poll data should be understood, compared to how it's currently misunderstood.
Preface
1. Public Consumption of Data: Some Historical Perspective
2. Polls
3. What Makes a Poll “Wrong” Part 1
4. Introducing: Ideal Polls
5. Throw it in the Average
6. What Makes a Poll “Wrong” Part 2
7. Introducing: Simultaneous Census, Present Polls, and Plan Polls
8. What’s For Lunch?
9. The Fallacy of Margin (Spread) Analysis
10. The Fallacy of Proportion Analysis
11. Instrument Error: Weighted Results and Literal Weight
12. What’s (actually) For Lunch?
13. Real Polls + Bad Math = Fake Errors
14. My Simultaneous Census
15. Introducing: Adjusted Poll Values
16. Compensating Errors and Poll Masking
17. Welcome to Mintucky
18. Mintucky Results
19. Mintucky Poll Error & Jacob Bernoulli
20. The Point Spread Problem
21. Remembering Nick Panagakis (1937-2018)
22. Finding the Base of Support
23. Trump-Clinton 2016
24. The Law of 50% + 1
25. The Clintons’ Lessons
26. Trump-Clinton-3rd Party 2016
27. We* Don’t Talk About Utah
28. Informing A Prediction
29. One Number Can Tell You A Lot
30. Don’t Call It a “Rule” & How To Report Polls
31. UK Elections & Brexit
32. The Polls Weren’t Wrong About Brexit
33. UK General Election Polls: Not Wrong Either
34. The Future, and “Try It”
Biography
Carl Allen is a researcher and analyst of sports and political data. His background includes quantitative analysis for University of Louisville Baseball, as well as MLB data for Stats Perform. Currently, Carl is the owner of Triple Digit Speed Pitch, a company that partners with schools and youth sports organizations for sports-themed fundraising events in Ohio. Carl's current research interests include inefficiencies in political betting markets, and improving polling methods. When he's not analyzing political data, Carl teaches English to adults, and he still finds time for MLB and NFL analysis. Carl received his BS from the University of Louisville in 2012 (Exercise Physiology, Spanish) and MS in 2013 (Sport Administration). You can find Carl on social media and Substack: @RealCarlAllen.