The lead up to Election Day on November 5th will see political opinion polls released with increasing speed. The sheer number of polls, and the language that surrounds them, can be confusing. As a qualitative researcher who follows the polls, Peter Finn shares some tips to be informed rather than overwhelmed.
- This article is part of ‘The 2024 Elections’ series curated by Peter Finn (Kingston University). Ahead of the 2024 election, this series is exploring US elections at the state and national level. If you are interested in contributing to the series, contact Peter Finn (p.finn@kingston.ac.uk).
In the lead up to the US election, even fair-weather political observers will be bombarded with news about polls. Many will focus on the top of the ticket race between Vice President Kamala Harris and former, and potentially future, President Donald Trump. Yet, there will also be polls on key state races, as well as polls exploring particular issues. Below, drawing on the experience of following polling in previous election cycles I highlight some key themes and terms that may help you better understand and interpret polls and related coverage and analysis: even if, like me, you have no intention of carrying out such polls yourself and are not schooled in the quantitative techniques that are required to carry out, and fully interpret, large polls.
What makes a political poll?
An opinion poll – about politics or anything else – is essentially a survey that aims to tell us what the public thinks about a product, issue or political candidate (such as whether they are likely to vote for them). A population, such as a state or the country is sampled randomly to produce a ‘panel’ of results. Depending on who is doing the polling, the number of people surveyed can vary from about 1,000 to up to 10,000. Polls are conducted by universities, market research, and dedicated polling firms. Historically polls were conducted by phone, but in the online era, this has mostly been supplanted by smartphones and desktops.
Crosstabs
Any in-depth discussion of a poll will inevitably feature a discussion of the ‘crosstabs’. This term is not quite as mystifying as it may first appear but is a shorthand for digging into the details of who was interviewed or completed an online poll, how they answered the various questions and how they fit into certain categories. Think of them like different tables, lines, and sheets on an Excel spreadsheet (which is often exactly what they are) that allow you to see how the top lines of a poll that have made it into the media have arisen, as well as to explore the data for other insights and trends.
Weightings
It is not always possible for a pollster to obtain a representative sample of the general population, or any subset they are trying to reach. There are multiple reasons why this might be the case such as a particular group being harder to reach in the aggregate, but it may also just occur randomly (especially if anyone can fill in a survey). In such instances polling organisations might add more weight to a smaller portion of their data (or, if thought about inversely, a smaller weight to a larger part) to try and most accurately reflect reality. If a pollster (or anyone using their data) is open about this, weighting poll data is not in and of itself problematic, though it will obviously impact how the data is analysed.
Focus on averages
Especially in the lead up to an election, there is an endless stream of polling related to various aspects of US politics (though, there are generally far more national than state-level polls). In areas, such as a presidential race or a high-profile senate race, rather than focus on the noise generated by any one poll, it is likely easier (and, I find, certainly better for your sanity) to focus on polling averages. Perhaps the best known is put together by 538, which continues to be influential even after the departure of its founder Nate Silver in 2023. But there are also many others.
Photo by Towfiqu barbhuiya on Unsplash
Interestingly, putting together an average is as much an art as it is a science (though some will disagree at this claim, especially from someone who does not do polling), as decisions need to be made about, for instance, how much weight to give to different polls and pollsters, and whether, or how, polls taken closer to an election should be given greater weighting.
Outlier polls
As well as making it easier to get an overall picture, such averages can also help with the interpretation of polls that appear to present a different picture from others, so called outlier polls. Outlier polls may arise due to differences between, or tweaks to, methodologies, as a statistical quirk (again, despite what some might say, thinking of polling as much of an art as it is a science is useful here, as sometimes even great artists can get it wrong on occasion), or they might signal a change in public sentiment that, as yet, other polls have yet to capture.
If a poll is an outlier due to differing methodology, but is shown to have more accuracy by an election result then it is likely that others will try and adopt similar methods moving forward; if it is a quirk of who responded then the noise of a particular outlier poll will be lost within poll averages; but if it does represent a change in sentiment, then this should be picked up by further polls, and poll averages will shift in the direction of the initial outlier.
Margin of error
Another term that will inevitably arise in any conversation about polling that has any depth. As with crosstabs, the term ‘margin of error’ can seem slightly intimidating to the uninitiated. The great news is that it is actually a very useful and simple idea. In short, it is an acknowledgement that the data collected in polls is not perfect, and, as a result, a certain degree of uncertainty needs to be built into the statistics one draws from the data. As such, if Harris or Trump are shown to be slightly ahead within a race, but the difference is said to be within the margin of error, then they might be further ahead than shown, not actually be ahead, the two could be tied, or (if the lead is very small) the other candidate might actually be leading the race by a similarly small amount. For example, a recent Suffolk University poll showed Kamala Harris leading Donald Trump 48 to 43 percent. But, with a 3.1 percent margin of error, the true figures for Harris and Trump could be anywhere between 45 and 51 percent and 40 and 46 percent, respectively.
Pollsters try to capture reality, but polls are not reality
Polls are not perfect, and pollsters are only human, and there is no single type of poll, nor any single focus of polling (though be prepared to hear lots about polls focused on the many intricacies of the Harris-Trump race!) Polls are not reality but simply attempts to capture data that can be used to construct a picture that is as close to reality as possible. If you are particularly interested in understanding what might happen in November at the top of the ticket, think about tracking polls that are taken in swing states, but do not forget that there will be a wealth of interesting details to tease out further down the ballot. If, come election day, polls appear to show any race (or the race) leaning in a particular direction, keep in mind that this does not guarantee the outcome. What following polls can do, however, is provide frames for understanding how complex races are evolving and where they may reflect and differ from those in previous cycles.
- Please read our comments policy before commenting.
- Note: This article gives the views of the author, and not the position of USAPP – American Politics and Policy, nor the London School of Economics.
- Shortened URL for this post: https://wp.me/p3I2YF-edR
