Why Hasn’t Polling Embraced New Technology?

Polling is one of the great contradictions of politics. Everyone complains about it, but no one can live without it. Politicians value it both because it’s a scorecard for how they’re doing and it’s a useful data source to indicate how the public views any given issue, not so much because they’re worried about actually doing what the public wants but because it helps them analyze the impact of any political choice on their next primary. Reporters like polling because it’s an easy structure for a story. Consultants and lobbyists like polling because they can use it to advance their bill/ policy/ cause or kill something they don’t like.

However, polling has faced a lot of challenges over the past decade, in part because Trump’s popularity and support was hard to accurately measure and in part because the methods for surveying people changed (no one answers the phone if they don’t know the caller). Pollsters had to start deploying hybrid approaches of calls and online with seemingly every pollster using a different methodology (some far more rigorous than others). To me, polling is still important and always will be. But what I don’t understand is why modern day political polling hasn’t embraced and incorporated new tools that can help capture public perception and opinion.

Take prediction markets. Most polling had the Presidential race neck and neck with most reputable pollsters having Harris ahead. The prediction markets saw it differently. People betting on the outcome (technically, buying contracts) are assessing how they think other people will vote, not just how they feel about a particular candidate. That often tends to be a far more accurate assessment than just the amalgamation of individual opinion. But I’m not aware of any pollsters that take prediction market data into account when they provide their analysis and advice. In a lot of elections, there may not be a market or enough participation for the data to be useful, but sometimes there is and this could be applied not just to candidates but also to ballot referendums and pending legislation.

Take Google search. While the Google Search Index itself is not publicly available, tools like Google Trends and Google Keyword Planner are, as are a number of third party tools and browser extensions. What people search for online is a good window into what they’re interested in. Candidates or legislation getting lots of queries helps indicate enthusiasm, intensity, and potential for fundraising, grassroots and GOTV. And as AI platforms start to replace Google as the default search engines, tools like ChatGPT Enterprise Analytics will also help.

Finally, why not just ask AI what AI thinks people think about any given issue or candidate? This is not more complicated than conducting a search on the half a dozen leading AI platforms. The answers might be consistent, they might not be. But it’s good information to have and easy enough to access.

Pollsters will be a lot better than me at deciding how to weight each input. And I suspect many pollsters who have been at the job for decades will pooh pooh the idea of new inputs and different methodologies. But we know that polling is often flawed. We know new ways of accessing opinion and information now exist. And the additional sources of information are generally publicly available and free or inexpensive, so there’s no major barrier to entry.

Not all three of the new sources I mentioned will work on every question or survey. But prediction markets are only going to get bigger and more comprehensive. AI search is growing exponentially and AI itself gets smarter every day so asking it what it thinks others think is only going to become more useful.

To be clear, I don’t think adopting this approach is a significant business opportunity. A new startup using this approach would certainly not be a venture scale model. But it could make polling more accurate and comprehensive. I hope someone gives it a try.