As a mathematician who tries to keep up with politics regularly, it may come as a surprise that I am not a huge fan of political polling. Granted, there are tons of exciting mathematical questions at play in polling a group of people on their thoughts about candidates, policy proposals, and various other aspects of the political landscape. For instance, how does a pollster obtain a representative sample size or measure other statistics like confidence intervals and margins of error?
I won’t go into the answers to these questions (though I should point to at least one online resource, a fun webpage by the Pew Research Center explaining how polling works). Instead, I’d like to comment on the essay “How to Lie with (Political) Statistics” by Lily Hu, which was recently published in The Boston Review. This piece makes a powerful case to treat statistics from polls with caution and, at times, strong skepticism.
More precisely, Hu correctly points out that generating a poll is not just doing math and obtaining indisputable facts. In fact, there are always going to be some assumptions baked into polling models, and some interpretation required by polling analysts or political pundits that necessarily introduces more opinion into the process. Claiming “the mantle of science” in this context, Hu argues, not only “so much of our political life as fixed and preordained, even as so much of it is so clearly rapidly changing;” moreover, this “data-centered approach to politics” vastly limits political strategy to technocratic decision-makers who know a lot about math calling all the shots.
I do think that Hu, at other times in the essay, is overly harsh on pollsters, who on the whole collect and analyze data in good faith to elucidate trends in public opinion. In particular, polling conducted by universities can help understand campus life in a vast array of areas besides political issues, and is a vital tool for researchers in the social sciences.
However, Hu’s arguments about the reductionist and gatekeeping aspects of polling are precisely my qualms with the enterprise. Politics should not be boiled down to numbers — even if a pollster claims to have a good model of the political landscape, that model is at best telling the rest of us something about right now, which could look very different from how that something might be even in the near future.
This aspect of mathematical modeling, more generally, is crucial. For example, one potential application of my PhD research is in the construction of digital twins, a topic I’ve written about a while back and also won’t go into much detail about here. Very briefly, a digital twin (in my understanding — there are many definitions) is a computational model of a real-world physical system that can accurately adapt on its own to changes in the system; in other words, a digital counterpart responsive to unpredictable real-world events. It’s very hard to actually develop a digital twin, because mathematical models are not the most malleable objects, and even expert mathematicians can struggle to amend these complex models in light of new data that make the real world look very different from previous experiences.
This is to say that politicians or political activists can have a potentially enormous impact on public opinion — as Hu notes, this “presumably is the point of politics, after all.” So why should polls be treated as what is, as is so often done in reporting on political polls, in effect limiting imagination for what could be?
This brings in the gatekeeping aspect: the political establishment has seen many recent failures of ruling out popular candidates, or limiting the platforms of campaigns, based on what they think will “poll well,” or “what the polls are telling us.” This has led to losing campaigns and trust in politicians, who lose authenticity by claiming to represent what is popular while not seeming to truly believe in what they’re running on. The logic of mathematics can help a lot in many contexts, but in the case of human behavior, a stat-ification of core beliefs, values, and opinions is, in my view, not at all the way to go.
