Who’s winning the presidential race? It depends whom you ask. [column] | Local Voices

With two months to go until the Nov. 5 election, it appears Vice President Kamala Harris has opened a narrow lead in national surveys over former President Donald Trump.

“Appears” is the key word, since public opinion polling is an educated guessing game.

In a presidential election this close, any prediction employing September polls is mere speculation.

This doesn’t prevent campaigners, journalists, academics and, yes, pundits and gamblers from investing in polls and prognostications.

Political consultants make daily decisions based on models of what the electorate looks like and what turns voters on and off. Horse-race assessments draw attention to media outlets. And, a lot of money is at stake, not only for bettors but also powerful interests in politics.

The object of all this is to persuade the public to turn out (or not), choose candidates and parties, and shape the direction of the country.

I hope the following gives readers a better understanding of how polling and election forecasting work. Nothing should disrupt your plans to get politically involved this fall.

Polling challenges

The accuracy of a poll depends on the ability of the researcher to draw a representative sample of the population. For many reasons, it has become increasingly difficult for pollsters to do so.

Simply finding respondents is a big problem. Landlines have given way to smartphones, which provide users multiple communication avenues, from email to texting to direct messaging.

Furthermore, fewer people are picking up the phone or responding to electronic messages, likely signs of feeling overwhelmed by advertisers’ attempts to get our attention. In 2020, it was estimated that it took 15,000 calls to get 950 responses to a telephone public opinion survey. The COVID-19 pandemic exacerbated the challenge.

In addition, there may be systemic over- or under-sampling of particular groups. Elderly and educated people are more likely to reply to surveys. Republicans are more suspicious than are Democrats of institutions conducting polls, a finding that may help account for underestimation of the Trump vote in 2016 and 2020.

Pollsters adjust to these and other biases by imposing a model of the population on the sample. For example, if we know that the population is 12% Black and our sample contains only 8%, a researcher will assign a greater weight to responses of the Black people who were surveyed. High-quality polls will adjust on eight to 12 variables, including age, education and gender.

Even if the statistical model of the population is accurate, there are subjective factors that produce differences among polls. Social scientists disagree on how to predict a likely voter, for instance.

If pollsters can overcome these hurdles, remember that an accurate poll is at best a snapshot of what people think at a specific time within a reported margin of error.

It is difficult for polls to measure attitudes during unsettled times, such as a foreign policy crisis. Similarly, the impact of Tuesday’s ABC News debate between Harris and Trump won’t be known for at least 48 hours after the event, once people have talked to each other, sifted through news and social media accounts, and endured spin doctoring from the campaigns.

Look for patterns

Expert forecasters attempt to account for variations in polling mechanics by ranking pollsters for accuracy, use of responsible methods, and partisan bias. The latter characteristic is more likely to occur if the poll is sponsored by a partisan organization.

For instance, the 538 website (popularized by the statistician Nate Silver and now owned by ABC News) issues grades for national and state pollsters, denotes which polls have partisan or ideological sponsors, and adjusts findings accordingly.

Rather than rely on one poll, readers should consult polling averages and watch for patterns or trends over time. Tracking averages are provided by sites such as 538 and RealClearPolitics. Note that polling averages are affected by decisions of which polls to include or exclude, so it is not surprising that averages vary across organizations.

As for trends, nearly all polls and national polling averages documented a Trump lead over President Joe Biden of around 3 percentage points before Biden withdrew from the race July 21.

Now the tables have turned. As of Thursday, the 538 and New York Times averages showed Harris with a 3-point advantage over Trump nationally. RealClearPolitics and The Washington Post indicated that Harris was ahead by 2 points.

What Harris has done is motivate the Democratic base, especially young adults, people of color and women. Demoralized after the disastrous June 27 Trump-Biden debate, Democrats now are as enthusiastic about the election, if not more than, Republicans.

None of this has discouraged Trump’s devoted followers or Trump himself, who seeks to define Harris as inauthentic, dishonest and alien.

Additionally, Trump wants to tie Harris to President Biden and the sour mood of Americans. In polls this year, at least 60% of those polled nationally believe the nation is moving on the wrong track, while only about 25% see it going in the right direction. Likewise, evaluations of Biden’s job performance for most of 2024 hovered around 40% approval.

Interestingly, polls show Biden’s withdrawal from the race improved his standing. Gallup’s August 2024 survey revealed 43% Biden approval, up from his lowest rating of 36% the previous month.

A Suffolk University/USA Today poll conducted in late August put the president’s approval rating at 48%. Attaining that kind of number six weeks ago probably would have kept Biden in the race.

Some predictions

Prognosticators of election results blend polling data with economic indicators. The 538 model, recently revised, and The Economist forecast connect poll results with an index of economic conditions such as jobs, inflation, consumer sentiment and real personal income. Currently, 538 gives Harris a small edge, while The Economist says the election is a 50-50 proposition.

At the annual meeting of the American Political Science Association, held last week in Philadelphia, one research paper predicted a Trump victory based on state presidential approval and economic conditions. Though the Biden-Harris administration boasts a strong record on job growth, the slowing of the economy in an election year works against the Democrats.

Academic modelers claim remarkable accuracy in predicting presidential election outcomes. Maybe the most famous — but not universally accepted — model is the “13 Keys to the White House” approach presented by Allan Lichtman, professor of history at American University. On Thursday, Lichtman predicted that Harris will win, based largely on positive economic factors, Biden’s impressive legislative record, and the absence of scandal or social unrest engulfing the White House.

To the extent that election forecast models are successful, the relevance of campaigns can be questioned. It is true that partisan voters make up their minds well in advance, many deciding even before the major parties choose their presidential nominees. And, if both parties run competent campaigns, the messages will cancel each other out.

Even the basic issue in any election — change versus the status quo — is complicated, especially in 2024. Vice President Harris represents the incumbent party, but she is a fresh face, having replaced Biden at the top of the ticket. A Trump victory would by definition represent change; however, as the three-time Republican nominee, he is hardly the new kid in town.

In other words, party identification is the primary determinant of the vote, regardless of whether things are going well nationally or for voters personally, what campaigns are doing, or what issues may arise. Most self-described independents vote nearly the same as do stronger party identifiers.

Of course, there is always the possibility that a dramatic campaign event will break through. As mentioned, the 2024 Trump-Biden debate changed the configuration of the race.

Allowing for rare exceptions, then, campaigns and the actions of candidates change relatively few minds. But in an election year as closely contested as 2024, the small percentage of undecideds in swing states will decide the outcome.

Paradoxically for American democracy, swing voters tend to be least informed and engaged politically, waiting until late in the campaign to decide.

In 2016 and 2020, late deciders favored Donald Trump over the Democratic candidate.

The Pennsylvania factor

As we all know, national popular votes do not determine the outcome of presidential elections — Electoral College votes do. Ask Presidents Al Gore and Hillary Clinton.

It is the distribution of the popular vote that matters. Though Biden won the national popular vote by 7 million votes in 2020, a shift of only 44,000 votes in three states would have given Trump a second term in the White House.

In 2024, seven swing states — Arizona, Georgia, Michigan, Nevada, North Carolina, Pennsylvania and Wisconsin — will determine the next president. Most experts designate Pennsylvania as the state most likely to decide the election.

That is almost certainly former President Trump’s assumption. The Keystone State is receiving more Trump visits, ads and campaign money than any other.

The Harris campaign agrees that Pennsylvania is crucial but, due to the vice president’s rise in the polls, believes she has more pathways to victory.

Polling in the swing states is more problematic than it is nationally. Surveys of state electorates usually feature smaller sample sizes and have larger error margins than do national polls, making definitive judgments of who is winning more difficult.

Each state’s political and election structure has its own peculiarities and Pennsylvania is no exception.

For example, the presidential election in Pennsylvania will be principally a Trump versus Harris contest. Robert F. Kennedy Jr., who suspended his third-party campaign when it appeared to take more votes away from Trump, is not on the ballot. Despite his endorsement of the former president, RFK Jr. is stuck on the ballot in Michigan, North Carolina and Wisconsin.

Interpretation of election laws can favor one party or the other. Democrats, who have taken greater advantage of early voting opportunities than have Republicans, have a majority on the Pennsylvania Supreme Court.

Another factor to consider is whether a state is holding a referendum to legalize abortion. Pennsylvania is not, but 10 states, including Arizona, Florida and Nevada, are, according to the KFF ballot tracker. Abortion politics may help the pro-choice Harris-Walz ticket, as voters have approved most pro-reproductive freedom and health referendums since the U.S. Supreme Court struck down Roe v. Wade in 2022.

Even without an abortion issue on the ballot, voter turnout in Pennsylvania should be high because of not only the intensity of the presidential election but also important down-ballot races, including those for a U.S. Senate seat (Republican challenger David McCormick hopes to oust incumbent Democratic Sen. Bob Casey) and the offices of state attorney general, treasurer and auditor general. Moreover, control of both houses of the state Legislature is at stake.

At this point in the race, Pennsylvania is stubbornly undecided over who should be the next president.

If Trump wins the commonwealth and the White House, there will be a lot of Monday-morning quarterbacking over whether Harris should have picked Gov. Josh Shapiro as her running mate instead of Minnesota Gov. Tim Walz, who visited Lancaster County last week.

October surprises?

Much can happen in the next two months to shift public opinion. The upcoming presidential and vice presidential debates. Developments in the economy or the wars in Ukraine and Gaza. Unexpected but sadly predictable tragedies such as Wednesday’s deadly shooting at Apalachee High School in Georgia.

Campaign events come and go — for example, Trump’s controversial visit to Arlington National Cemetery. Who knows which will stick in the minds of the remaining voters who are undecided?

And then there are the proverbial “October surprises,” some of which can happen in September. Already we have seen the stunning federal indictment of Russian media executives accused of funneling nearly $10 million to right-wing producers of videos and other social media content infused with Russian propaganda and disinformation.

Whatever happens, the polls leading up to Election Day 2024 are likely to say the outcome of the presidential election is too close to call. And they will be right — within the margin of error.

E. Fletcher McClellan, Ph.D., is professor emeritus of political science at Elizabethtown College. X (formerly Twitter): @mcclelef.