Until the votes are counted after Nov. 5 there is no way to know for certain who will be the next president of the United States – Vice President Kamala Harris or former President Donald Trump. In the meantime, we resort to polls.
Pollsters survey Americans to get a sense of which candidate has support, but since it’s impossible to ask every American how they intend to vote, lots of polls are conducted. Average the results from these polls and we get a polling average, a one-number measure of a candidate’s election chances.
While it may be tempting to view polling averages as predictions of the election results, they are better understood as a snapshot of public sentiment for each candidate. Polling is not an exact science; those calculating polling averages contend with balancing factors like poll bias, quality, reliability, and recency.
Let’s take a look at FiveThirtyEight’s polling average, one of the polling modelers, for the presidential race in Pennsylvania and explain the caveats behind the numbers.
FiveThirtyEight’s current polling average for Trump is %. This means based on an estimated average of recent polls, % of surveyed Pennsylvanians expressed support for him.
Harris is currently receiving % in FiveThirtyEight's estimates. Giving a current lead in the polls by a % margin
You might have noticed that the numbers add up to %. This is because voters who responded with intention to vote for a third-party candidate or they’re still undecided.
The responses to polls are pretty straightforward but combining them into a single number isn’t an exact science. This is where poll modeling comes into play.
A poll model is an opinionated method of coming to an average for each candidate based on polling data. Modelers make mathematical assumptions about each polls’ biases and reliability based on the quality of the pollster and the sample of people surveyed. This means different models will produce different numbers.
So if you see a different number somewhere else, neither are wrong. They’re just different interpretations.
- Advertisement
Here’s a look at how much the FiveThirtyEight polling average has changed since Harris replaced President Joe Biden on the Democratic ticket. They are informed by at least different polls.
The only way to get a fully accurate read of the entire state would be to survey the entire state. This is unrealistic, so pollsters identify a cross-section of the population to be representative of the entire state.
However, with that also comes uncertainty in accuracy. This is represented by the margin of error.
As new polls are released — between now and the election — they’ll be updated on this page.
New polls are undertaken all the time, though not every single poll conducted will be included. FiveThirtyEight may choose to exclude a poll if they fail to meet standards of disclosure and methodology.
Some polls are conducted by pollsters or sponsored by groups who are partisan. Partisan pollsters tend to overestimate support and generally only publish polls that appear favorable for the pollster's preferred candidate. This is considered within the model producing the average.
Polls may also be funded by partisan sponsors such as super PACs and campaign committees, though sponsored polls are conducted by both partisan and non-partisan pollsters. The Philadelphia Inquirer, in partnership with the New York Times and Siena College, recently sponsored a set of polls, the results of which are analyzed here.
Keep an eye on whether polls are sampling likely voters or the larger universe of registered voters. Pollsters early in the cycle tend to survey registered voters and then as the election gets closer will narrow their focus to voters who say they are likely to cast a ballot.
List of individual Pa. polls
V: Voters
LV: Likely Voters
RV: Registered Voters
Data and Methodology
The Inquirer pulls data from FiveThirtyEight which curates a list of trusted pollsters. New polls are undertaken all the time, though not every single poll that is conducted is included in 538's models if they fail to meet standards of disclosure and methodology.
The 14-day polling average change is calculated using a comparison of the mean polling average in the past week and the mean polling average in the week prior to the last. This is done to reduce daily noise and the outsized impact of certain polls’ effect on the daily polling average.
Staff Contributors
- Design, Development, and Reporting: Sam Morris and Jasen Lo
- Data: FiveThirtyEight
- Editing: Sam Morris, Aseem Shukla and Julia Terruso
Subscribe to The Philadelphia Inquirer
Our reporting is directly supported by reader subscriptions. If you want more journalism like this story, please subscribe today