How to Spot an Accurate Political Poll

By Adam Garrie, The Kennedy Beacon 

As the election draws nearer, millions of Americans will be presented with a dazzling array of presidential poll results, each trumped up as more accurate than the next. 

But what makes for a useful and valid poll as opposed to a misleading one?

A recent New York Times/Siena College poll found that in the key swing states of Michigan, Pennsylvania and Wisconsin, Vice President Harris is leading former president Trump by 4 percentage points. Although a 50% to 46% lead may appear to be an open and shut victory, it is no such thing in this case. 

In each of the three states surveyed, the margin of error is greater than Harris’ margin of victory. This indicates a statistical dead heat between the two candidates rather than a decisive win for either.

In other words, there’s no meaningful takeaway from this poll except that it’s misleading. The spun narrative is that Harris is beating Trump in key swing states. She is not.

To make matters more misleading, while actual ballots in November will include the names of multiple presidential candidates, the New York Times/Siena poll only lists two candidates, further skewing the poll's findings away from the official election results.

To understand the difference between an accurate, transparent, and useful poll versus one with little real-world significance, there are several key aspects of polling that you should know.

National polls vs. state polls 

While national polls aim to predict the popular vote, this is not a scientific way to predict who the next president will be. State-by-state polls are often the better indicator of the eventual winner because the US presidential election is not determined by the popular vote, but by the Electoral College system. This is why polls focusing on swing states are often more determinative than national polls when it comes to trying to predict who will win. 

Where is my candidate? 

One of the ways that polls about the presidential race can shape rather than reflect public opinion is through the exclusion of candidates on a survey. For instance, if an independent candidate like Robert F. Kennedy Jr. or a Green candidate like Jill Stein is excluded from a poll, the overall results will not reflect voter behavior when presented with a ballot on election day. This is the case because the actual ballot will give voters the option of voting for Kennedy, Stein, and Libertarian candidate Chase Oliver in addition to Republican Donald Trump and Democrat Kamala Harris. 

When such candidates are excluded, results will necessarily be different than those in the election. The exclusion of independent or third party candidates will often be requested by a commissioning organization intentionally. 

Horse race vs. favorability 

Many polls ask the respondent to choose between various candidates for a public office. But the question can be framed in several ways.

A traditional “horse race” question asks respondents to select which candidate they plan to vote for in the election. Other polls ask which candidate the voters favors most. Candidates with high favorability ratings sometimes have low numbers in the “horse race” question polls.  Some voters who prefer candidate C, for instance, may feel obliged to vote either candidate A or B for strategic reasons, if they’ve been told candidate C doesn’t have a chance of winning. 

Sample size 

A poll’s accuracy is only as good as its sample size. A sample size is the number of individuals polled. The larger the sample size, the more accurate the poll typically is. In a state like Michagin or Pennsylvania, a decent poll ought to have a sample size of 800 or more respondents. 

Sample selection 

It is also important to review just who is being polled. Some polls only record responses from registered voters. Others take responses from all adult citizens. Some polls only focus on likely voters. If some voters haven’t voted for a number of years, they will not be considered in these polls. 

Reading polling results 

Most of the time, the public will not view the raw data from a poll. Instead, they will read the results, as summarized by a journalist or individual on social media. Because of this, those in the media who report on a poll’s results have the opportunity to exclude certain data that may not fit with their agenda. 

Sometimes, a media outlet will include or link to the raw polling data. For the most accurate reading of a poll, you should look for the “crosstabs,” which offer detailed relationships between the respondent and their selection. For instance, a detailed crosstab will explain what people between the ages of 18-25 feel about a candidate while also giving you the opportunity to determine how this differs from other age groups. The same applies to breaking candidate selection down by gender, income, race, religion and other demographic factors. 

Weighted polls

Some pollsters apply formulas to their results known as a “weighted” polling methodology. The Kennedy Beacon’s Leah Watson wrote an extensive piece on this process. 

Polling companies vs. commissioning bodies 

Polls are conducted by multiple companies and academic institutions commissioned by media outlets, political campaigns, political action committees (PACs), and private businesses. 

A “Reuters/Ipsos” poll, for instance, is commissioned by Reuters;  and Ipsos is the polling company that conducts the field work. 

When a poll is commissioned, it is typically for a specific purpose, which may introduce the natural biases of the commissioning organization. Therefore, when determining whether you trust the results of the poll, one consideration should be your trust in the organization commissioning the poll. 

The Bottom Line

Not all polls are created equal. Some are created with the intention of manipulating the public. Understanding the various methodologies, nomenclature and structure of polls and surveys will help you to determine whether you are receiving accurate or inaccurate information. 

Last week, The Kennedy Beacon reported on a nationwide poll conducted by John Zogby Strategies which demonstrates that Robert F. Kennedy Jr. is the only candidate who can beat Trump. Details of the poll, including a link to the crosstabs, can be found here.