It has been about ten years since the AAPOR pollster conference was first attended. Back then, the event was dominated by conventional pollsters who, while acknowledging that change was inevitable, were not comfortable with the sacrifices required to accommodate new people, methods, and ideas.
This crowd reminded me of the Republican Party at that time, which was figuring out how to handle demographic change and Hispanic voters after Barack Obama’s re-election. Despite the differences that exist, both faced comparable threats, such as long-term trends, status threat from newcomers, and the possibility that traditional values would be endangered by accommodating new ideas.
AAPOR is now completely transformed, thanks to Donald J. Trump, who proved that Republicans could win without supporting immigration reform, and therefore, pollsters would have to introduce new ideas. After two historically significant poor cycles, AAPOR is no longer the same. Though the traditional pollsters are still prominent, presentations based on current methods are now ubiquitous. In 2014, the year I first attended, the Innovators Award was bestowed on someone who referred to AAPOR as “Buggy-Whip Manufacturers.”
The shift toward innovation in the polling community is tangible, specifically in public political polling. Today, almost no pollsters use the methods they did a decade ago, with the ABC/Post poll being the only notable exception. Despite the introduction of innovations, the problems faced by the industry still persist, such as difficulty in reaching Trump supporters, the rise in costs, and the loss of faith in representative sample surveys. Having a well-designed survey is no longer enough, as the most theoretically sound surveys generated the worst outcomes in 2020.
The innovation in polling has so far taken place on two distinct tracks. The first aims to discover new ways of sampling voters in an age of low response rates, while the other track focuses on enhancing unrepresentative samples through statistical adjustments. If there is an underlying theory of the Times/Siena poll, it is to get the best of both worlds: high-quality sampling with sophisticated statistical adjustment. Public polls that can effectively make this claim are scarce, as some polls use bad sampling with sophisticated statistical modeling, while others have decent sampling but with simple demographic adjustment.
Due to the pandemic, it has been a few years since the author attended AAPOR. Nevertheless, from the vantage point, it would appear that this year marks the first time that these two parallel tracks appear to converge. They have yet to converge fully, as the traditional pollsters resist the adjustments required to increase their methods’ efficacy, while the new entrants struggle with the expenses related to sampling. Nonetheless, the researchers on both tracks appear to be acknowledging the inadequacy of their respective solutions and are dabbling in the other side’s ideas.
Early on, for instance, the researchers recognized that current survey designs struggle to reach less committed voters with lower education levels who may be more likely to support Mr. Trump. This problem may never be resolved flawlessly and therefore warrants the best traditional and non-traditional methods.
The author guarantees that they will have further information on their Wisconsin experiment that featured parallel telephone and high-incentive mail surveys before the 2022 election, in the upcoming weeks. Over the last week or so, we got the final data necessary to begin this analysis, and they have started to dig in. It’s too early to conclude, but there’s some intriguing information. Stay tuned.
