The above is the title of the talk I’m giving at noon next Tues, 30 Sep 2025 at Jerome Greene Annex on the Columbia campus (and also on zoom, so it seems). It’s part of a lecture series for emeritus professors at Columbia, but anyone at Columbia is welcome to come.
Last time I spoke to a group of retirees at Columbia was just about exactly 25 years ago, and it was a lot of fun. They asked great questions. Now I’m 25 years closer to retirement (gulp!), and I’m expecting some good questions again.
My selective history (or more accurately, my crude reconstruction) of political polling and election forecasting goes as follows:
1. Before 1900, political workers counted votes, it was hyper-local and labor-intensive
2. From 1900, mobile populations, more anonymity, lower voter turnout, motivation for thinking about public opinion rather than just counting votes
3. 1936, Literary Digest poll (and how it could’ve been fixed), Gallup poll
4. Commercial and opinion polling, cluster sampling, quota sampling, challenges of representativeness
5. From in-person polling to telephone surveys to internet panels: the mode of data collection informs the method of sampling
6. Sample adjustments, declining response rates, and different forms of nonresponse
7. Rationality of voting, rationality of responding to a poll, how these have changed over the decades
8. Accuracy of pre-election polling and exit polling since 1948
9. Variation of pre-election polls during the campaign
10. The state of our understanding in the 1970s: Jimmy the Greek in 1972, poll variation in 1976, anything could happen
11. Geographic polarization and the rise of the swing state
12. Political polarization and the decline of the swing voter
13. Political science in the 1970s/80s makes its way into the conventional wisdom of the 1990s: “It’s the economy, stupid”
14. The era of nothing matters: 1992 and 1996, and our fundamentals-based forecast
15. Close elections from 2000 onward: swing states, steady polls, and a fixed forecasting target
16. The rise of poll aggregation and probabilistic election forecasting
17. Differential nonreponse as an explanation of polling variation
18. High expectations and the 2016/2020/2024 polling errors
19. Off-year elections and party balancing
20. Changing demographic bases of support of the two parties
21. Who are the nonvoters and what do they want?
22. Information other than from horse-race polling
23. Primary elections, third parties, and other complexities
24. Looking forward
I must say that this post caused an itch I want to scratch concerning the number 25.
I mean, you write that you are giving a talk in september of the year 2025, and that it was 25 years ago that you last gave a talk to retired professors at Columbia, then you mention the number 25 shortly after that again, but then you subsequently come up with 24 (!?) points in your selective history or reconstruction overview.
Not on my watch!
So, should this list of 24 points be part of the talk itself, perhaps it’s okay for me to suggest point 25. “Looking back” (following point 24 “Looking forward”) where you could note what you mention in the blog post that the last time you talked to retired professors at Columbia was just about exactly 25 years ago and that they had very good questions which you will expect again now.
Phew, that’s a bit of a relieve I must say.
Enjoy the talk!
No Ray Fair?
Maurice:
Ray Fair is part of item 13, “Political science in the 1970s/80s makes its way into the conventional wisdom of the 1990s: ‘It’s the economy, stupid.'”
“Political science in the 1970s/80s” includes the work of Steven Rosenstone, Doug Hibbs, Ray Fair, and James Campbell, probably lots of others too of whom I’m unaware.
No mention of “psephology” and pebble counting?
Paul:
I’d not heard the term “psephology” . . . I looked it up and it’s defined as “the statistical study of elections and trends in voting.” That’s covered in items 13 and 14 on fundamentals-based forecasts.