Patrick Ruffini
Subject: Are the presidential polls flawed?
Bio: Pollster with Echelon Insights
Reading: Party of the People: Inside the Multiracial Populist Coalition Remaking the GOP is here
Transcript:
Larry Bernstein:
Welcome to What Happens Next. My name is Larry Bernstein. What Happens Next is a podcast which covers economics and politics.
Today’s topic is Are the Presidential Polls Flawed?
Our speaker is Patrick Ruffini, who is a pollster with Echelon Insights and he is also the author of the book Party of the People: Inside the Multiracial Populist Coalition Remaking the GOP.
The election is just days away, and voters from both parties are looking at the polls to predict what will happen on election day.
I want to learn from Patrick whether polls are trustworthy and whether by their very construction that they are defective. I want to hear from Patrick how to interpret polling data and where the misses will likely come from.
I recently held a conference in Washington DC with a bunch of my friends where we chatted with Patrick. This podcast will be different than normal because some of the questions will be asked by my friends.
This is a very opportune time to discuss where we are in this election process. I would like to hear about the quality of polls and whether the Latino and black populations are really leaning more towards the Republicans than usual. And of course, the state of the election. Good luck, Patrick.
Patrick Ruffini:
I'll need it. I want to take a step back and talk about where the Republican Party was before Donald Trump descended the golden escalator. They were reeling from a defeat in 2012 in which they had expected to win but lost by four points. And as they dissected the exit polls from that election they found that they did even worse among non-white voters than they did in 2008, which was the historic Obama election.
They went backwards when they were expecting to move forward, particularly after a thumping victory in the 2010 midterms. But lo and behold, a record low share of the Hispanic vote. Black voters turnout rates exceed white turnout rates in that election, which is stunning in the context of American history.
In the wake of that election, you had huge recriminations about Mitt Romney taking a too hard line stance on immigration. You had Mitt Romney policy of self-deportation, which was maniacal, and he lost all the Latino voters.
Donald Trump is singing a different tune in the early stages of his campaign and everyone at the RNC at the time is beside themselves as he rockets up in the polls. As many people were very worried that Donald Trump was squandering this opportunity against the hated Hillary Clinton.
Well obviously, things turned out very differently in that election because the message that ignited the Republican base for Trump seems to be appealing to swing voters in places like Michigan, Pennsylvania, and Wisconsin, which are the three deciding states in that election.
Fast forward to 2020 and Trump is in trouble, but for different reasons. We had a global pandemic, which he did not handle very well and in the polls, it was suggested he was going to lose by as much as eight points, which was the polling average heading into election day.
That narrow win for Biden materializes and Trump does a lot better than the polls had suggested. In fact, the polling error in 2020 was even greater than it was in 2016. When you look at the difference in the national popular vote, Trump ends up overperforming the polls in the national popular vote by about four points. He's still four points behind. But remember, we don't have a popular vote. We have an electoral college and the so-called electoral college bias, the difference between the 270th electoral vote and the popular vote grows even wider to about 3.7 points to where the tipping point state is Wisconsin. It is within 0.7 percentage points in 2020 and the polls pre-election had Biden winning in Wisconsin by eight points.
What is up with these upper Midwestern battleground states? It is a place where you have a lot of blue-collar white voters. You have outright majorities in these states that are white voters who do not have a college degree, who are precisely the voters who gravitated by about 15 to 20 points towards Donald Trump, which was not at all factored into the pre-election models in 2016. Now, in 2020 pollsters thought they had fixed it, but hadn't. But on top of that, you had a surge for Trump in Latino areas in South Florida and in South Texas, and among Asian voters in places like Little Saigon in Southern California.
Remember going back to that 2012 autopsy report where Republicans say the way to win back Hispanic voters is to do pass a comprehensive Immigration Reform Bill? Fast forward 12 years where the Republican candidate for president, Trump now the third time around is running on a platform of mass deportation and looks set to reach a modern high percentage of the Latino vote.
In 2020 Trump does slightly better with African American voters while at the same time turnout drops from its Obama-era highs in cities like Detroit, Philadelphia, Milwaukee, and Atlanta.
The Democrats have to make up for it by appealing to people in the suburbs and college educated people who are moving away from Trump for the same reasons that non-college and working class voters are moving towards him.
The story of 2020 the Republican Party becomes more of a multiracial party, but not in the sense that the establishment 2012 autopsy report envisioned of we're going to be more welcoming to these different communities.
Trump is speaking in very blunt terms about challenges with immigration and with crime, and not only does it not hurt him with communities of colors, but you also have those communities flocking and supporting him. So this time around, polls are obviously a lot closer.
Larry Bernstein:
You mentioned that the Wisconsin poll was off by eight points in 2020. I think polling is good. When you look at a single state, it's hard. There's going to be a large standard error. And so maybe we shouldn't give too much credence to any particular state poll. What I would do is this for the battleground states add up all the battleground states into one poll and then take an average.
How should we think about error in polling and the quality of polls and how much credence should we give it?
Patrick Ruffini:
The polling error in 2016 and 2020 was unevenly distributed. I wish we were more relaxed and could take the average, but that doesn't seem to be the order of the day when any poll show something different than other polls are showing. I am not here to tell you that something different is going to happen than what the polls are saying. I don't think anyone really knows that and anyone can say that with real confidence.
The polls are saying there is a wide error band around because the margin of error is very misleading. We could be measuring perhaps the wrong group of people. If you have a specific type of person who isn't answering polls, then that's going to be a challenge. The poll survey is not going to be representative.
Larry Bernstein:
The best example of that is in 1936, the biggest presidential poll was by Literary Digest. That magazine polled 2 million people and Landon was predicted to win but he lost. The poll was wrong by nearly 40 points. What that poll’s flaw was that its magazine subscribers were overwhelmingly Republican, and since the pool was not representative of the US population that even though the number of people surveyed was enormous, they polled the wrong group and got an erroneous answer.
Patrick Ruffini:
And that's what's happening in these upper Midwestern battleground states that there's something off specifically in 2020 and 2016 about the white non-college voter. And it is not all white non-college voters are created the same. You have somebody in Georgia who has been voting 90% Republican over the last five election cycles where realignment has already happened. They are fixed in their partisan beliefs.
But when you have a large group in transition such as you had in the Midwest in 2016, the polls are trying to measure things that they're not particularly suited to do is when you have a group that was formerly Democratic, maybe still registered as Democrats but is moving over to the Republicans. That seems to be a uniquely challenging task for pollsters to accurately measure what that population is.
We want polls to measure change, but to some extent the electorate we are modeling is the past electorate.
Larry Bernstein:
In the past few weeks, the polling data suggests that Harris’s lead has shrunk from plus 3 to plus 1. You have mentioned that there might be an error in the representative sample that calls into question what her real lead is. But do you believe that there have been real changes or a trend and if so by how much?
Patrick Ruffini:
Changes are probably happening, but the rate of the actual level of change in the electorate is overstated because what you see in these periods of high enthusiasm, particularly after the debate or after Harris becomes the candidate because you have certain populations who are willing to take polls more.
By the way, I would have said the same thing for Republicans right after the first Trump assassination attempt, the response rates were sky-high, and I don't care what weighting you use that is the ghost in the machine that's going to show up in your final result.
I believe that those trends are describing something real but probably the real vote share is moving by 0.3 for every point movement in the polls just because differences in response rates and polls exaggerate what is going on underneath the surface.
Larry Bernstein:
One big surprise in the polling data is that Black males are supporting Trump much more than previously. Do you believe it?
Patrick Ruffini:
We're not going to know until after the election happens, and I see the county level and precinct level data. I think you are going to see a gradual decline in the Democratic share of the Black vote as generational turnover happens. You have younger black voters who are less partisan than their parents and grandparents who grew up in the civil rights era when this Democratic voting norm was established.
There is a strong sociological basis for why you're going to see racial depolarization. What Trump specifically has done for Republicans is catalyzed things that were probably already going to happen in 12 years, but it's making them happen today.
If you just look at the African Americans who identify as conservatives, Hispanics who identify as conservatives, there was something like a 35 to 40-point shift in those groups in 2020.
What you're seeing is a macro trend. If it's happening in Miami, in South Texas, in the Latino precincts in Milwaukee, we have to discount localized explanations. Racial block voting loyalty to a single political party is increasingly becoming a thing of the past. You are increasingly seeing Latinos moving to the suburbs as the population grows more prosperous. It's increasingly becoming enmeshed within mainstream of American society.
The scholarship on this is that it is not about policy preferences; it is about a social voting norm. Once a certain threshold is reached where if it is Republicans hitting 15% that social norm becomes unenforceable and then you have a cascade effect from there.
Larry Bernstein:
Early voting has started, what are we seeing in the swing states?
Patrick Ruffini:
You're seeing real warning signs for Democrats right now in Georgia and North Carolina, early voting where black voting is down by about 10 or 12 points.
Larry Bernstein:
Why did Trump go to McDonald's?
Patrick Ruffini:
The immediate direct thing is this Kamala's claim that she worked at McDonald's, which she hasn't really provided any substantiation for, and he completely turned that against her. You have to just look at the contrast between the two candidates right now and the willingness to mix it up and take risks.
You'd never seen Mitt Romney, John McCain or even seen George W. Bush doing that and he was pretty folksy. Trump put himself in situations that relate to the average American far greater than Republicans in the past.
Larry Bernstein:
What did you make of Kamala's decision to be interviewed by Fox?
Patrick Ruffini:
She is doing outreach to the Haley voters. She is campaigning in Pennsylvania with Liz Cheney. If they went to peel off Republicans, that was not a good reason to go on Fox. The main benefit she got from doing that was to appear strong. You are going into the Lion's den, and you are taking the questions. What I will say is I do not think she has done well in friendlier softball interview scenarios, but she does better in these adversarial situations like the debate and the Fox interview.
That said, there are a lot of questions she did not answer in that Fox interview that are going to be fodder for Trump specifically around the border. What did you know about Biden's cognitive abilities and how would you be any different from Biden on policy? She is repeatedly refused to answer those questions, and those were all issues that were flagged in the Fox interview. Nonetheless, it was received well among her supporters because it is going into the lion's dead and coming out of it in one piece.
Larry Bernstein:
Does campaigning in purple swing states matter?
Patrick Ruffini:
There's this great academic book called Battleground by Daron Shaw who does the Fox poll and is a professor at UT Austin. He has worked in Republican campaigns trying to measure the impact of campaign activity. One of the conclusions is that battleground state travel does matter on the margin, probably a half a point, but when a Republican or Democratic candidate has the advantage in visits to a state that can move voters within that state.
Larry Bernstein:
Do you agree with the idea that Hillary would have won Wisconsin in 2016 if she had visited the state in the last few days before the election?
Patrick Ruffini: I was skeptical that it might have done anything. She would have had to travel there significantly more. If you buy the political science, she would have had to travel there significantly more than Donald Trump did. And that it goes to why didn't they do it? They fell in love with this idea that we are going to go and realign a bunch of Sunbelt states. We are going to go into expand the map into Arizona, maybe Texas, but potentially Georgia, North Carolina, those types of states. Those have been the backbone of the Republican coalition for decades. So, if you can swipe some of those states out from under the Republicans, particularly with Latino voters. But if you are banking big on Latino voters in a state like Wisconsin, that is 90% white but seems to be in the bag for you in the polls.
Larry Bernstein:
What is the Harris Campaign trying to accomplish during this final week before the election?
Patrick Ruffini:
The Harris campaign when they got into the race did follow what the focus groups told them to do. But now they pivoted strongly towards emphasizing this democracy January 6th appeal. In the final stretch, campaigns sometimes make decisions for emotional not necessarily logical, or poll driven reasons.
The decision about what the closing argument of the campaign should be is often a very personal one. As a candidate, how do I want to be remembered? What do I want this victory to be about? In September or August, the Harris campaign was very logical and cool headed about its strategic decisions. Now it's reverting.
Same thing with the Biden campaign both in 2020 and 2024 he wanted to make this a high-minded crusade for democracy when this was something that doesn't necessarily land with non-college educated working class voters.
Larry Bernstein:
With a very close election, how important is Wisconsin?
Patrick Ruffini:
Wisconsin surprises us because if it does end up being the tipping point state again that if you had a situation on election night where Trump sweeps all the Sunbelt states as expected, and Harris manages to win Pennsylvania, and everyone's saying it's lights out for Trump at that point, Wisconsin or Michigan could save him. And that's because Wisconsin in particular is a much more rural state. It's got a lot more white working-class voters who still vote Democratic, particularly in the rural areas. They're fundamentally up for grabs. I worry that the polls may be missing something.
Larry Bernstein:
What do you mean by the polls are missing something?
Patrick Ruffini:
What are the biases? Not necessarily of any individual poll, but the polling industry as a whole. There's an enormous incentive for pollsters to herd to release results in line with the average. I was wrong with everyone else, I don't lose any clients ultimately because of it. This picture being painted right now of all these states are within one or two points, and I'm never going to release a poll that shows, God forbid anyone up four that could be a misleading. But don't discount Wisconsin as a deciding state.
Ray Iwanowski:
Some of Trump's choices for rally locations in the last month, Coachella, Long Island and none of those are winnable states or locations. What's up with that?
Patrick Ruffini:
Giving way to emotion in the final stages of the campaign. Trump wants to campaign in his own state, or I'm going to send a message by going into deep blue California. There could be some value in that you get a news cycle where you have got adoring crowds in California. Maybe we want to win the popular vote, which would be an enormous boost for Republican morale if they not only were able to eke out an electoral college majority but also win the popular vote majority.
It is more an emotional than a rational decision, but the reality of scheduling, what else is he doing in New York City? He's doing a fundraiser that day too, so why not? He's not going to get to Arizona, Pennsylvania in time, so why not potentially do a rally too and also then get some media coverage out of that. It is probably 70% emotional and 30% strategy.
Think that his choices of rally locations within the battleground states have been very deliberate to go directly into Democratic base areas, going into Detroit to tap into the black male vote. He was up there with a rapper in Detroit with the Muslim mayor. That is potentially going to shift substantially away from Harris and the Democrats and as well as going to rural areas in Wisconsin where people have been voting Democratic up until recently to solidify some of those gains.
Larry Bernstein:
The next question comes from Jay Greene.
Jay Greene:
You brought up the error that occurs outside of the standard error provided with the model. So that's the sampling error.
But then there's also error about the weighting of the sample to ensure representativeness, which are a set of assumptions, and the error in those assumptions is not quantified or reported.
Larry Bernstein:
You mean by that if you sample 100 people, and you ask them if they voted for Biden or Trump in 2020 and the ratio of Biden to Trump in the sample is 75/25 but the state voted for Biden 60/40, then the pollster can do an adjustment for the polling result.
Jay Greene:
And then there’s a third error, which is a likely voter. You must estimate your sample, but then you also have to estimate who's going to show up. So, you have three types of error in polling at least and only one of which is quantified and reported. And the compounding of those errors. Maybe campaigns are not wrong to not pay very close attention in the stretch to polling because maybe it's not that informative given how really all it's telling you is, well, gee, it's close, and that's about it.
And combining all the polls as Nate Silver and others do in a Monte Carlo simulation makes this problem worse because the choice is about how to ensure representativeness of sample or to estimate likely voter. These are, as you say, correlated across pollsters, which means that the error is correlated. They're not independent observations, they're correlated observations, and therefore the probability is not correctly calculated by Monte Carlo simulation. So, what would you say about to convince a politician that they should be paying attention to data towards the end? Why should they?
Patrick Ruffini:
Well, yeah, I mean, you've correctly stated the garbage in garbage out problem that a lot of polls have. I think particularly if those issues exist, pollsters like myself what we view our job primarily is to provide advice as to what groups in the electorate are going to matter most or what messages are most likely to move people. It is as much art as it is science. It's very important for campaigns to show that they're competitive or ahead for the purposes of fundraising. The most important use for polling is what issues resonate with what voting constituencies that are up for grabs for me.
Colin Teichholtz:
It seems like for both candidates, there is more willingness to jump on each other's campaign points and just adopt them. Whether it's not taxing tipping, or it's a competition to see which one will give more money to families that have children or whatever it is, is it giving these campaigns too much credit that they're listening to polling feedback?
Patrick Ruffini:
Trump particularly intuited this, that elections are not won or lost on policy. I'm going to abandon decades of Republican orthodoxy on economic issues if that's what I need to do to win. That was something that previous Republican nominees would have avoided for ideological reasons, or they thought they were going to get a massive blowback within their constituency groups that they weren't going to be able to withstand it.
Trump is the first post-policy Republican nominee. I am strong enough to overcome whatever internal blowback within the party that I will get from saying, I might vote for the Florida pro-choice abortion measure because I delivered the three Supreme Court justices so you can trust me on this. In a campaign they're absolutely meaningless.
Larry Bernstein:
Polls used to call people on landlines. Now they text voters to see if they want to participate with an online survey. What kind of people are willing to do these online polls? Does this process undermine the pollster’s requirement of using a random sample?
Patrick Ruffini:
I heard from my phone vendor today that it is taking between 600 and a 1000 texts to get a single response in battleground states.
Larry Bernstein:
It's one in a thousand agree to take the survey poll and you need a thousand people to be in your sample, that would require that you send out 1 million texts!
Patrick Ruffini:
Yeah. With all the spam and fundraising texts that channel is quickly getting ruined.
Larry Bernstein:
I think this incredibly damning for the polls because the only people who answer the surveys are freaks. And the critical assumption is that the individuals that are polled are representative of the population, but they are not. They are peculiar and then the statistical assumptions are flawed.
Patrick Ruffini:
Only one in 500 people are answering that text, right? So, the idea is that you're getting a particular type of person that is different from the rest of the population in unknown and unquantifiable ways, but perhaps related somehow to attention paid to politics. And probably somebody who is the opposite of the people in this room. Those are the people who we are not measuring. Now, I talked about the ways where this hasn't really been too big of a problem in the past because those people were equally likely to be Republicans as they were Democrats, and typically it works out.
But there is a scenario where we see this polling error when Trump specifically is on the ballot. But is it a certain type of low propensity, low information unaffiliated voter who just thinks Donald Trump is better than Republicans in general, right? It is somebody that it's not necessarily earlier that he is somehow over-indexes among this unmeasurable group of voters. And that would be my concern. If we have a polling error in favor of Trump, if we have a polling error that in the end Harris wins, then I'll go with the scenario that a pollster tried something to really lock the Trump vote in place, and there was just too much Trump fatigue in the electorate. But it's a good chance. It's one of those two things.
Larry Bernstein:
Why are the Republican senate candidates underperforming Trump in the polls?
Patrick Ruffini:
Most of the Democrats running the cycle are incumbents. The incumbency advantage in running in Congress hasn't vanished, right, it's maybe gone down from 10 points to one or two points. And when you're talking about states that may be within one or two points, that could be enough to make the difference. And it just so happened in 2016, and even in 2020, a lot of those Republicans who were running with Trump on the ballot were themselves incumbents. So, you had Pat Toomey in Pennsylvania, Rubio in Florida, Ron Johnson in Wisconsin, and there is absolutely some level of incumbent advantage that you're going to see manifest. Now, once that incumbency flips to align with whatever the dominant partisan leans in the state. Maybe, Sherrod Brown gets reelected. But once that seat flips, it is unlikely we are going to see a Democrat win in Ohio for a long time to come.
Larry Bernstein:
Do you think it is likely that we will know who won on election night?
Patrick Ruffini:
Nate Silver’s forecast where he looks at the modal outcome. I'm not talking about the median number of electoral votes in the Nate Silver forecast that each side gets, but I'm talking the modal outcome. What's the most likely combination of states? And the most likely combination right now is either a Republican sweep or a Democratic sweep of all the seven battleground states, and there's about a 40% chance, one of those two things that we're going to wake up on election morning. And one of those two things is going to have to happen because high uncertainty and correlated error.
Larry Bernstein:
Who's going to win the presidential election, and what's the fair bet? I want a probability.
Patrick Ruffini:
I've been 55-45 probability on Trump. I would have said that even when Biden was running. I just don't know that she has done enough to change what the fundamentals of this race are about. And you look at what's going on around the world and incumbent parties losing three times as many elections as they've won post-COVID. The widespread discontent with the current administration that Harris is currently a part of. The fundamentals favor Trump. Do they favor Trump as much as they would favor a generic Republican or Nikki Haley? Probably not. And that's making it very close. That's making it a high uncertainty, but it does tilt ever so slightly his way. Thank you.
Larry Bernstein:
Thanks to Patrick for joining us today.
If you missed our previous podcast the topic was comparing White vs. Black Working Class Voters. Our first speaker was Batya Ungar-Sarayon who is the author of the new book entitled Second Class: How the Elites Betrayed America’s Working Men and Women. Batya interviewed over 100 Working Class Americans to explain their support of Trump’s policies.
Our second speaker was Corey Fields who is a Professor of Sociology at Georgetown and the author of a book entitled Black Elephants in the Room: The Unexpected Politics of African American Republicans. Corey discussed why the Republicans have been historically unsuccessful attracting Black people and why he thinks it will be no different in the upcoming presidential election.
You can find our previous episodes and transcripts on our website whathappensnextin6minutes.com. Please follow us on Apple Podcasts or Spotify. Thank you for joining us today, goodbye.
Check out our previous episode, White vs. Black Working Class Voters, here.
Share this post