A poll that had seemed to show a shifting race really didnât show much change at all.
A new look from U.S.C. tracking. On Monday, the U.S.C.-Dornsife tracking poll released a new cut of its data. The results were eye-popping: It turns out that its previous poll results didnât mean what many people â including me â thought they meant. The new results are good news for Joe Biden.
This is not an ordinary tracking poll. Itâs a panel survey, which means respondents are contacted repeatedly. The benefit of a panel survey is potentially significant: If the results change, itâs because the attitudes of the respondents change, not because of changes in the composition of the sample. This was an ominous sign for Mr. Biden: His lead in the survey fell to just seven points, down from 12 points, from Sept. 11 to 17, suggesting a tightening race.
But it turns out that the U.S.C. pollsters were doing something odd. They showed their last seven days of results, but interviewed each respondent only once every two weeks. The problem is that one week of respondents was relatively favorable to Mr. Biden, just by chance, while another was relatively favorable to President Trump. As a result, U.S.C.âs results oscillated between a wide Biden lead and a tighter race, depending on whether the last week of interviews was the good or bad week for Mr. Biden. It turned the pollâs potential biggest advantage â the ability to track change over time â into a liability.
Today, U.S.C. released the results over two weeks, not just one, and they tell a totally different story. The poll now shows a fundamentally stable race, with Mr. Biden maintaining something like a 10-point lead, rather than a volatile race that has swung from Biden plus-13 to Biden plus-7 and back. With U.S.C. showing less change, an important point in the case for substantial tightening can be crossed off the list. Mr. Bidenâs lead bounced back to seven points in our average as a result.
Nope, still no shy Trump voters. There has been no shortage of articles about the possibility that President Trumpâs supporters are part of a silent majority that, as a group, do not reveal their true vote intention to pollsters. Morning Consult decided to put the theory to the test, and today released the results of a new experiment.
It found a group of voters online, then conducted the survey with half of the respondents over the phone and the other half online. If voters were afraid to divulge their support for the president, the theory goes, perhaps Mr. Trump would have more support online, where supporters wouldnât have to express their views to a person on the phone.
The survey found that it didnât make much of a difference. (It did make a difference for some socially awkward or possibly embarrassing questions, like those about peopleâs personal finances or attitudes about discrimination.)
Itâs still possible that such voters exist. Even if they donât, the polls can be wrong for all sorts of other reasons. But thereâs really just no serious evidence to support the idea that these voters exist in meaningful numbers, and there was no such evidence in 2016, either.
A look ahead to the bottom of the table. So far, weâve mainly focused on the states likely to decide the race. This week, the polls seem poised to focus on a different group of states: those that are highly competitive and yet unlikely to tip the election, like Georgia, Iowa, Ohio and Texas. The reason is simple: If Mr. Biden were to win them, he has probably won already. He has probably won Arizona if heâs won Texas; he has probably won Wisconsin if heâs won Iowa; he has probably won North Carolina if heâs won Georgia.
You can find these states at the bottom of our poll average table on this page. As you can see, Mr. Biden could easily win these states. And if he did, it would be a blowout.
So far this week, weâve received two polls of these states: a YouGov poll showing Mr. Trump up by just two points in Texas and a Democratic GBAO poll showing Biden up by three in Georgia.
On Tuesday, weâll get an Atlanta Journal-Constitution poll of Georgia. At some point soon, weâll get the presidential results from the vaunted Ann Selzer poll of Iowa.
To this point, polling in these states has been highly competitive. Weâll see what the results suggest.
State of the race at the end of the day: It still seems like a comfortable Biden lead.
It will be a while until we have a good understanding of public opinion in the aftermath of the death of Supreme Court Justice Ruth Bader Ginsburg. Most high-quality polls are conducted over three or more days, so weâll have to wait to see many full poll results. Even then, those results will reflect only the state of play at the outset of whatâs sure to be a long fight.
The early results are decent for Biden. We do have our first results from a few quick online surveys. YouGov found that voters didnât want President Trump to appoint a new Supreme Court justice before the 2021 presidential inauguration, by a margin of 51 percent to 42 percent, and that if President Trump did appoint a new justice, they did not want the Senate to seat him or her.
Ipsos found that a much larger majority of Americans â 62 percent â think that the winner of the presidential election should replace Justice Ginsburg, while just 23 percent disagreed.
Why the big difference? Question wording. The results of issue questions are often highly sensitive to the wording of the question. In this case, the huge difference between the Ipsos and YouGov results might be explained by a common bias in questionnaire design: acquiescence bias.
In general, people tend to be nice and agreeable, so questions that simply ask voters to agree or disagree with a statement will generally find higher support for âagreementâ than a similar question with a forced choice between opposing ideas.
In this case, if the pollster simply asks whether the respondent agrees that the winner of the election should replace Justice Ginsburg, most voters might say âyes.â But if you give the same respondents a second option â whether President Trump should choose the nominee â you can bet that plenty of Trump supporters would take the option, even though they might have agreed, if asked the other question, that the winner should decide. Thatâs probably a better reflection of the politics of the issue.
More of the same in the horse race. There werenât too many polls this weekend, but most of them were very consistent with what we thought we already knew: a modest but still significant advantage for Joe Biden.
Maybe the most noteworthy one was an NBC/Wall Street Journal poll, the first live-interview, education-weighted national survey in a while. It found Mr. Biden up by eight points among registered voters, with a net one-point shift in the presidentâs direction since August. On balance, thatâs better for Mr. Biden than both the trend line and the average.
Status of the polls at the end of the day: Waiting to see how Supreme Court politics will reshape the race.
Weâve been lucky to have a lot of high-quality state polling this month. It has given us a really good look at several states where we received almost no high-quality polling in 2016, like Minnesota and Wisconsin. At times, these polls have suggested that Joe Biden is faring extremely well.
But the wave of state polling has come at a cost: fewer national polls. On Friday, we got some. And they werenât as great for Mr. Biden as the state polls, raising some questions.
Cloudy national picture The national polls contained a mix of good and bad news for both sides.
The good news for Mr. Biden: Ipsos, Data For Progress and Global Strategy Group/GBAO/Navigator Research showed him with a comfortable lead and holding his ground â or gaining â compared with prior surveys.
A four-point lead for Mr. Biden from AP/NORC, a particularly bad number for him when you consider the poll is of all Americans, not likely or registered voters, and its prior survey showed him up by 12 in June.
A seven-point lead from U.S.C./Dornsife, which remains about four points smaller than it was a week ago.
In the middle is an eight-point lead for Mr. Biden from NPR/Marist. The trend line here is decent enough for Mr. Biden, but the poll is not weighted for education, which usually means the poll underestimates the president.
Why the national polls matter The national popular vote wonât decide the election, as Hillary Clinton can attest. So who cares about the national polls?
They have one really important use: We can track change over time. Most firms have conducted several national polls this year, so we can say how things have changed over recent months. Polls by the same firm in the same state, on the other hand, tend to happen less frequently. That makes it hard to say whether things have been changing over the last few weeks.
Taken in isolation, the national polls we do have over the last week suggest that the race is stable or even moving a bit toward the president, who now trails by just six points in our average.
A state-national gap? In contrast, the state polls in the last week most certainly do not suggest that the race is moving toward the president, or that thereâs a six-point race nationally.
Todayâs polls were no exception. Hereâs one simple way to look at it: If you take a straight average, Mr. Biden ran about 8.7 points ahead of Mrs. Clinton across all of the state polls Friday. Add that to her 2.1-point victory in the popular vote, and maybe youâd guess Mr. Biden was up 10.8 points nationwide. You can see that for yourself on the right-hand column of our daily poll tables, if you havenât noticed:
What explains the difference? Noise is always a possibility, and it could also go in the other direction â we could get a Biden +16 result from a high-quality pollster tomorrow, for instance. But I do think that would be a bit of a surprise, given how many mid-single-digit results weâve seen this week.
Another possibility is that Mr. Biden has actually lost a bit of ground nationally while holding his own or is even gaining in the battleground states. Thatâs certainly plausible, especially since Mr. Biden has tended to make his largest gains among white voters, who are disproportionately overrepresented in the battleground states, while struggling to match Mrs. Clintonâs numbers among nonwhite voters. Where is Mr. Bidenâs best state poll result today, compared with the 2016 result? Itâs Maine, the state where non-Hispanic whites represent the largest share of the population in the country.
There are other possible differences, like the effect of campaign ad spending. The polling firms could be relevant, too: The state polling firms have generally been of higher quality. But state and national polls have generally been in alignment this cycle. If this recent difference persists, weâll really dive in.
State of the race at the end of the day Mr. Biden ends the week with a clear advantage in the battlegrounds, but still some question marks about whether or how things are shifting nationally.
The good part for Mr. Biden is simple: On Thursday, he led in every poll of North Carolina, Florida and Arizona â five polls in all. President Trump carried all of these states four years ago. Together, winning those states would probably put Mr. Biden well over 300 electoral votes.
The good news for the president is that he didnât trail in any of those states by very much, and he stayed close in a state that has seemed in danger of slipping away: Arizona.
Trump still hanging on in Arizona? On Wednesday, I noted that if you penciled in Wisconsin and Michigan for Mr. Biden â given his recent run of strong leads there from high-quality pollsters â he would essentially be one battleground state away from crossing 270 electoral votes. The most plausible options are probably Pennsylvania or Arizona, assuming he also carries Nebraskaâs Second District.
The polls have been pretty split in Arizona, with some showing a close race and others showing a commanding Biden lead. With Mr. Biden eyeing just one more state with a comfortable edge, a commanding Biden lead in Arizona would have significant consequences for the electoral map.
He led a Monmouth University poll by a slim margin, while a Kaiser Family Foundation poll showed him up by five. That latter result is probably closer to two or three points because itâs a poll of registered voters, rather than of likely voters. Arizona is a state where a poll of registered voters will overstate Democratic support quite a bit because it has a large and heavily Democratic, but relatively low-turnout, Latino vote.
Arizona voted for Mr. Trump by 3.5 points in 2016, so itâs hard to consider the results bad news for Mr. Biden. But if heâs up by only three points in the state, as todayâs polls might suggest, maybe Pennsylvania â where there has been less polling recently â might be the better option for Mr. Biden to find his 270th electoral vote. And the polls havenât been all been great for Mr. Biden there, either.
Think twice about that Biden lead in North Carolina Today there was a Suffolk University poll showing Mr. Biden up by four points in North Carolina, which would be a strong showing in a state where our averages suggest a dead heat.
Unfortunately, this poll has a major problem: The share of respondents who are college graduates is too big. In polling lingo, this means they didnât âweight by education,â and itâs critical for understanding modern polling and what went wrong in 2016.
The problem is simple. College graduates are likelier to take political surveys, but they are also far likelier to oppose Mr. Trump. So if pollsters donât do anything to make sure they properly represent people without a college degree, theyâll have results that are too good for Democrats. The most common way to address this is to âweightâ by education, which simply means you give more weight to your respondents who arenât college graduates. Many pollsters didnât weight by education in 2016, and itâs one of the major reasons the state polls were biased toward Hillary Clinton. Many state pollsters have addressed this since then. Suffolk doesnât appear to be one of them.
If you look at the poll release, Suffolk reports that about 49 percent of its respondents have a college degree or higher. This is way too high, and particularly a problem in North Carolina. The state has a huge educational divide among white voters: Mr. Biden routinely wins white college graduates in North Carolina polls in liberal bastions like Raleigh, Durham and Asheville, but Mr. Trump can lead white voters without a degree by 50 points in rural evangelical areas.
Pew recently published a handy table of estimates for the share of the electorate with a college degree. As it notes, thereâs no such thing as the ârightâ number, since youâre trying to estimate the composition of an electorate in an election that hasnât happened yet. But the truth is probably closer to Pewâs estimate for North Carolina, 38 percent. (The Times/Siena polls have that number at 37.) This poll would almost certainly have been much better for the president if it had been weighted by education. In a state like North Carolina, it could have erased all of Mr. Bidenâs lead in the survey.
Best for last: On Friday, New NYT/Siena College polls in Maine, Arizona and North Carolina My colleagues Alexander Burns and Matt Stevens will have the story, with some key Senate results, posting first thing in the morning. Keep an eye out for it.
State of the race at the end of the day A clear Biden advantage, but still searching for a comfortable lead in the 270th electoral vote.
Another good day for Biden in Wisconsin. On Tuesday, I said that weâd have an unusually clear picture of the race in Wisconsin if the ABC News/Washington Post poll Wednesday found Joe Biden ahead by at least five points. Thatâs exactly what we got: The poll showed Mr. Biden up by six points among likely voters in Wisconsin.
It puts him in a strong position in the state, hovering at or above 50 percent with a consistent lead in an unusual number of recent high-quality surveys.
Wisconsin was the âtipping pointâ state in 2016 â the one that pushed President Trump over the Electoral College threshold. It remains crucial to Mr. Trumpâs re-election hopes, and many thought Wisconsin would be a real challenge for Mr. Biden. The state has a large share of white working-class voters, who swung hard to Mr. Trump in 2016. On top of that, Mr. Biden seemed to be weaker in the suburbs of Milwaukee than he was in suburbs elsewhere in the country. And finally, the unrest in Kenosha last month seemed to give Mr. Trump another opening.
But unrest in Kenosha didnât lead to a change in the state of the race in Wisconsin, and that might say a lot about the national political environment. If the president canât break through in Wisconsin now â in the immediate aftermath of the Republican convention and unrest in Kenosha â when could he?
If you pencil in Mr. Bidenâs leads in Wisconsin and neighboring Michigan, heâs essentially one flipped battleground state away from crossing 270 electoral votes.
An eye-popper in Minnesota ABC News/Washington Post also released a poll of neighboring Minnesota, showing Mr. Biden ahead by 16 points in a state Hillary Clinton won by only 1.5 in 2016.
To be clear: Itâs pretty unlikely that Mr. Biden is actually ahead by 16 points in Minnesota. Itâs probably a noisy, outlying result. But the poll joins a host of other high-quality surveys â including from CBS News/YouGov and The New York Times/Siena College â showing Mr. Biden ahead by at least nine points in the state. His lead might not be 16 points, but itâs large, and itâs real.
An even bigger eye-popper in Maine Quinnipiac released a poll of Maine with incredibly strong results for Democrats: a 21-point lead for Mr. Biden in a state Mrs. Clinton won by only three points in 2016, and a 12-point lead for the Democrat Sara Gideon in a hotly contested race against the incumbent Republican senator, Susan Collins.
A word of caution: Quinnipiac has leaned quite a bit to the left in this cycle; I recommend nudging these to the right a few points in your head before taking them to the bank. And thatâs even before considering the possibility that itâs an outlier, like the ABC/Washington Post poll in Minnesota. That said, you could cut Mr. Bidenâs lead in half and it would still be a strong result for Democrats.
Most of all, know that Maine is a very underpolled state. Itâs hard to say itâs an outlier without any other polls for comparison. Tread carefully. Fortunately, weâll have a second opinion in a few days from a Times/Siena College poll.
A split between state and national polls? If you look over the state polls so far this week, you see a run of strong results for Mr. Biden â some of his best of the cycle. But if you look at the national releases, you see stability, or even a bit of tightening.
One possibility is that itâs just noise: The national polling is pretty sparse and often of fairly questionable quality. And some of the state polling for Mr. Biden â like the Quinnipiac poll â comes from firms with a record of showing him doing particularly well.
But itâs also possible that it reflects a real split, perhaps driven by demographics: Most of the great results for Mr. Biden in recent state polls have come in overwhelmingly white states, and there are plenty of national (and state) poll results suggesting that Mr. Biden is running ahead of Mrs. Clinton among white voters but faring worse among nonwhite voters. If so, it might lead to seemingly surprising results for Mr. Biden in overwhelmingly white states like Minnesota and Maine without corresponding national leads.
Of course, itâs not necessarily a bad trade for Mr. Biden. After all, Mrs. Clinton probably would have traded a few points nationwide for greater support in the Midwest.
Good news for Democrats in the Senate. Along with the wide lead for Democrats in Maine, Quinnipiac found a tied race in the South Carolina Senate race. And OH Predictive Insights found the Democratic challenger Mark Kelly up by 10 points in his Arizona Senate race. Theyâre not the first polls to show an unexpectedly tight race in South Carolina and a wide lead for Democrats in Arizona. Weâll get another poll in Arizona on Thursday from Monmouth University.
This is around the time when convention bounces start to diminish. Itâs still too soon to say whether President Trumpâs bounce will fade or endure, but Tuesday was arguably Joe Bidenâs best day of state polls since the Republican National Convention.
The best news for Biden in a while in Florida. A poll from Monmouth University showed Mr. Biden up four percentage points among likely voters on average, his best result from a nonpartisan, live interview pollster there in several weeks. He held a wide lead in Florida over the summer, but it has gradually slipped â in part because of a somewhat surprising weakness among Latino voters. The Monmouth poll shows no signs of that weakness today, with Mr. Biden leading by 26 points among Hispanic voters, comparable to Hillary Clintonâs performance four years ago. If Mr. Biden can match Mrs. Clinton among Hispanic voters, heâll be in a strong position: Polls consistently show Mr. Biden running ahead of Mrs. Clinton among white voters.
Now, gauging the support of Hispanic voters in Florida is not easy. About a third of the stateâs Hispanic voters are Cuban, and they are overwhelmingly concentrated in the Miami area â the toughest area of the state to reach in a survey. As a group, those voters lean Republican. But the other two-thirds are heavily Democratic and live across the state. On top of that, Hispanic voters are harder to reach in general. Theyâre younger and concentrated in urban areas, and many speak Spanish as a first language, which adds further difficulties â and costs â for pollsters.
All that to say: In Florida a lot will hinge on how pollsters can measure a relatively small group of hard-to-reach voters. So interpret any single result among Latino voters with caution, especially in Florida.
Another poll showing Trump trailing badly in Wisconsin. One place where the polls have offered consistently bad news for the president is Wisconsin, where Mr. Biden has held a steady lead. Today, a CNN/SSRS poll added to the consensus by showing Mr. Biden up by 10 points, one of his largest leads there this cycle. The firm also gave Mr. Biden a three-point lead in North Carolina, another result consistent with a clear national advantage for the former vice president. One note of caution: CNN/SSRS polls have tended to tilt to the left compared with the average of polls so far this cycle, as well as in 2018.
Tomorrow, we expect another poll of Wisconsin from ABC News/Washington Post. If it joins the club of high-quality pollsters showing at least a five- or six-point lead for Mr. Biden, that would yield about as clear of a picture as youâre going to get in a battleground state so far from an election.
A stable day nationwide. There werenât many national polls today, but the handful we did get were largely consistent with their prior results and with a fairly stable race.
Odds and ends Morning Consult had a relatively weak result for Mr. Biden in Minnesota, though thereâs plenty of other recent polling there showing Mr. Biden with a wider lead. Florida Atlantic University showed a tied race in Florida, though the firm doesnât have much of a track record and its methodology is a mixed bag. Virginia Commonwealth University gave Mr. Biden a double-digit lead in Virginia.
Simple as it is, âWhat do the polls say?â is by far the question Iâm asked most â by my colleagues, my editors, my friends and family, and by hundreds of you on Twitter. Even if youâre not looking for polls, it can sometimes feel as if the polls find you. More often than not, confusion follows.
Thatâs why we built this page. Ideally, itâs a place where you can get the state of the race at a glance and the polling news of the day, just about every day.
Most days, Iâll summarize the most recent polling. Sometimes, that summary will be ânothing much happened,â and you can go about your day. Other days, Iâll highlight major polls or longer-term trends, or Iâll go on a bit of a methodological digression that might help you make sense of the sometimes arcane world of polling. Or maybe an individual poll result will make enough of a splash to merit a post of its own.
Youâll find polling averages of the most competitive states. The goal is to summarize the state of the polls. Itâs not a forecast or a prediction. Just because weâre showing you the average of public polls doesnât mean weâre telling you that theyâll be right. The polls have been wrong before, and theyâll be wrong again.But despite their flaws, polls remain the best way to measure attitudes across a huge and diverse country. We might not even be aware that an election is close if we talk only to our like-minded friends and relatives. (For those who want to know more about our polling averages, we published a detailed account of our methodology here.)
Youâll also see a table showing what the results might look like if the polls are off by about as much as they were in the last two presidential races. Of course, the polls could be more accurate, less accurate or exactly the same as they were the last two times. And this particular measure â the average error over the final three weeks â is a little unfair to those pollsters, who tended to produce more accurate results over the final week. But it serves as a simple reminder that thereâs still a wide range of possible results on Nov. 3.
Youâll find information about our polls, conducted in partnership with Siena College. Weâll be polling nearly nonstop between now and the election. Weâll have more to say about our specific plans soon.
And finally, a big thanks to our friends at FiveThirtyEight, who compile polling data, including the set used here, and make it available to the public. And, of course, to the pollsters themselves, who do the hard (and expensive) work of conducting these surveys in the first place.
Our poll averages include all polls collected by FiveThirtyEight. The estimates adjust for a variety of factors, including whether a poll represents likely voters, whether other polls have shifted since a poll was conducted, and whether a pollster has leaned toward one candidate in a state or nationwide. Polls are weighted by recency, sample size, and by whether they’re conducted by a firm with a track record of success. More details here. Source for polls: FiveThirtyEight polling database. * In Maine and Nebraska, two electoral votes are apportioned to the winner of the state popular vote, and the rest of the votes are given to the winner of the popular vote in each congressional district. (Maine has two congressional districts, and Nebraska has three.) â Poll error in 2016 is calculated using averages of state polls conducted within three weeks of Election Day.
News – The race is more stable than we thought.