'If we've got it wrong, so does everyone else'
A battleground pollster with a unique approach to the 2020 election talks about what she's seeing in the data.
Welcome to Margin of Error, a newsletter from me about the polls and the way they are covered.
If you’re new here, thanks for signing up! If you find it interesting, I’d be grateful if you would encourage others to sign up.
If someone sent this your way or you found this post through Twitter or other channels, make sure to subscribe below.
Here’s where we stand with six (6!) days to Election Day:
National poll averages: Biden +8.5 (via 538) (-2.2 from last week)
Florida: Biden +1.8 (-2)
Pennsylvania: Biden +5.2 (-1.3)
Michigan: Biden +8 (+0.1)
Wisconsin: Biden +8.9 (+1.6)
Arizona: Biden +2.8 (-1.1)
Georgia: Biden +1.8 (+1)
North Carolina: Biden +1.9 (-1.3)
Texas: Trump +1.5 (+0.3)
Ohio: Trump +1.6 (+1.5)
A programming note: More analysis will return next week, ahead of and/or on Election Day. But now, a bit more about the Great Lakes battlegrounds.
We know that a lot of pollsters in the battleground states of Michigan, Wisconsin, and Pennsylvania got it wrong in 2016 (or, at least, ran out of time).
The Great Lakes Poll, which is run by Baldwin Wallace University in Ohio in partnership with Oakland University and Ohio Northern University, is an interesting case study of polling in the 2020 race. The poll was launched in early 2020 to study the three states that decided the 2016 Electoral College margin (along with Ohio).
This week, I talked with Dr. Lauren Copeland, associate director of Baldwin Wallace’s Community Research Institute, about having a clean slate with the poll, the key metrics of the 2020 race, and whether there are really “shy Trump” voters in the Midwest.
The conversation has been lightly edited for length and clarity.
First, a bit of background: How did you get into the field of polling?
I was in the right place at the right time with the right skill set. (Or part luck, part preparation.) In 2015, a position opened at Baldwin Wallace University. They were looking to grow the Community Research Institute, as well as hire someone to teach courses on American politics and research methods/data analysis. I applied, and I got the job.
What are your go-to resources for polling- and data-related information?
FiveThiryEight is my go-to resource for the latest updates on the presidential election. I especially enjoy their Politics Podcast. I am also a member of the American Association for Public Opinion Research, as well as the Association of Academic Survey Research Organizations. These organizations help me stay up-to-date on best practices for survey research.
Can you give me a bit of the backstory on how the Great Lakes poll started? What made you believe there was a need for a comprehensive poll of the Great Lakes states?
For 2020, Baldwin Wallace University’s Community Research Institute wanted to focus on state-level polls since the Electoral College determines which candidate wins the presidency, not the national popular vote. We honed in on Michigan, Pennsylvania, and Wisconsin because they struck us as the most important states in the 2020 election; the Democrats could retake the presidency in 2020 if they held on to the 20 states Hillary Clinton won in 2016 and flipped Michigan, Pennsylvania, and Wisconsin. We were also interested in whether Ohio was still a battleground and bellwether state; Trump’s eight-point win in 2016 and a strong showing by Republican candidates in 2018 led many to challenge Ohio’s status as a battleground state. Some even argued Ohio was off the table for 2020. (Of course, now we know that Ohio remains competitive even if the contest is much closer here than in the other Great Lakes states.)
You are in the unique position of starting a poll the cycle after polls in those four states (especially Michigan, Wisconsin, and Pennsylvania) missed the mark last time. As you prepared to start your own survey, what overall mistakes from 2016 did you learn from? What did you try to build on? And what makes you confident in your methodology?
Since the 2016 presidential election, the CRI has made several changes to its methodology. First, like most polling organizations, we weight the data by education. Second, we put caps in place for central metro areas and fringe metro areas in each state so that we don’t overrepresent the urban vote nor underrepresent the rural vote. Third, we put caps in place for age so that younger people, who are among the least likely to vote, would not be overrepresented in the data. Finally, all the Great Lakes polls are weighted by gender, race/ethnicity, annual household income, and of course, education.
We feel confident in our methodology because our results are very much in line with current polling averages on FiveThirtyEight. In our latest poll, which was released October 9, Biden led Trump in Michigan by 7 points, in Wisconsin by 6 points, and in Pennsylvania by 5 points. In Ohio, Trump was up 1 point. To me, this says that if we’ve got it wrong, so does everyone else. This seems unlikely.
One adjustment many pollsters have made, which you also do, is weight for education. Why do you view that as necessary?
The short answer: When you don’t weight by education, you underestimate support for Trump.
The longer answer: Many polls did not weight for education in 2016, leading them to underestimate Trump’s strength in key swing states. People with lower levels of educational attainment may be hard to reach in public opinion polls, but they still turn out to vote, and they are much more likely than people with a four-year college degree to vote Republican. We see this in the topline report from the latest Great Lakes Poll. Although the results suggest that Biden has cut into Trump’s lead among white voters without a four-year college degree, Trump still has an edge.
There are a lot of people still obsessed with the notion that there could be "shy Trump voters" lurking outside of the polls. Do you give that any credence?
We should always be prepared for some degree of polling error, but I think shy Trump voters are largely a myth. If there were shy Trump voters, we’d see larger differences between polls that are conducted online versus by phone, with lower levels of support for Trump in polls conducted by phone.
A lot of people believed that Michigan, Wisconsin, and Pennsylvania would be battleground states again. Has it surprised you that Ohio has been close, and actually (in your poll and in the averages) the closest of all four states?
The potential for Ohio to be a competitive state remains true, if voters on both sides have strong turnout. The lopsided eight-point win by Trump in 2016 was largely due to very strong support from Republicans combined with many voters who either switched from voting Democrat to voting for Trump (particularly among the labor vote), chose third-party candidates (4% of the vote in Ohio), or stayed home (voter turnout in the eight urban counties dropped by 133,000 from 2012 to 2016).
The same factors driving voter turnout in Ohio are at work in the other three states, which is why Wisconsin and Michigan appear to be back in the Democratic column, and Pennsylvania leaning in that direction. It should also be noted that in 2018, Michigan and Pennsylvania re-elected Democrats as governor and to the Senate, while Wisconsin turned out conservative Republican governor Scott Walker by a one-percent margin. In contrast, Ohio voted for Republicans for governor and all statewide seats except for re-electing Democratic Senator Sherrod Brown and electing two Democrats to the Ohio Supreme Court.
Where do you see the future of the Great Lakes states electorally? Some debate has started over whether its working-class population will be less friendly to a non-Biden Democratic nominee in the future and that Democrats will shift focus to the Sun Belt and spots like Arizona. Others see it swinging back to Democrats in a non-Trump Republican world.
The Great Lakes states will continue to be important in presidential contests. The six midwestern states that border the Great Lakes total 84 electoral votes; PA and NY add 49 electoral votes. While the Sun Belt states continue to grow and diversify, issues such as water scarcity and natural disasters affecting the South and the West Coast may result in longer term population shifts to the Great Lakes states, which would stabilize and possibly grow their Electoral College vote count.
In the presidential race, what do you see as the key differences between 2020 and 2016?
We sometimes forget that 2016 was an open-seat election without a sitting Vice President as a potential candidate. Lingering concerns about the economic recovery, as well as the strengthening of the left via Bernie Sanders run for the nomination resulted in lower turnout for Hillary Clinton. Working class voters turned away from Clinton and voted for Trump, while many Black voters, disenchanted with both candidates, stayed home.
The 2020 election is unique in having so many variables at play, such as the effects of the coronavirus on voter turnout, the increase in use of the mailed ballot and early voting (due in part to the coronavirus), negative reactions to Trump’s style, and the sudden drop from economic growth to unemployment levels not seen since the Great Depression. The net result may be a historic high in voter turnout, but also the possibility of close contests, recounts, and court challenges.
You noted some of this, but what do you see differently in how voters view Trump and Biden vs. their attitude toward Clinton? Your latest poll shows that a plurality of voters in Michigan, Wisconsin, and Pennsylvania view Biden favorably and Trump pretty unfavorably, while in 2016 both candidates were viewed very negatively.
The consistent net unfavorable ratings for Trump are part of what is driving turnout in favor of Biden, who is seen as a ‘safe’ alternative. This is also largely why Biden succeeded in the primaries, despite having little financial support. Biden voters are not necessarily enthusiastic about him, but they also do not see him negatively. Clinton had strong support from voters who wanted to see a woman as president but had unfavorable ratings due to perceptions about her style and concerns about being part of the ‘establishment.’
What bit of data (in your polls or even elsewhere) has surprised you most so far in 2020?
The majority support for Black Lives Matter in combination with majorities agreeing that the concerns and needs of rural communities are not understood
The widening margins in favor of Biden in Wisconsin and Michigan, while Pennsylvania remains close
Strong support in Ohio for election reforms that expand accessibility but also increase accountability
The high percentages of people directly affected by the coronavirus, particularly related to employment status
Probably the biggest bit of silver lining for the president, both in your polls and elsewhere, is how voters view his handling of the economy. What is your read on why the majority of voters still approve of his handling there, even as the country continues to grapple with the economic fallout from the pandemic?
It’s true that a majority of likely voters in our polls approve of Trump’s handling of the economy. But if you look at approval levels by party ID, big differences emerge. In Michigan, for instance, 95% of Republicans either “strongly” or “somewhat’ approve” of “the way President Trump is handling the economy,” compared to 45% of Independents and 11% of Democrats. Party ID definitely shapes people’s perceptions of the economy, and this isn’t a new phenomenon either.
What do you do for fun to escape the world of polling and elections?
It’s harder to escape the world of polling and elections with the coronavirus outbreak, but I enjoy running; hiking; exploring; getting lost in a good book, especially historical fiction; hanging out with my family; and playing monkey-in-the-middle with our corgi. {Ed. note: Photo of Trudy below, thanks to Lauren!] I have two young kids; they keep me grounded and focused on what matters most.
Anything else I didn't ask that you wanted to mention?
I am fortunate to be part of an incredible team, which includes Dr. Tom Sutton, Dr. Aaron Montgomery, Shawn Salamone, and Dr. Julie Newcamp. For the Great Lakes Poll, we also partnered with Dr. Rob Alexander (Ohio Northern University) and Dr. Terri Towner (Oakland University in Michigan), and we owe at least some of our success to their participation.
Finally, the BW-CRI serves as a laboratory for teaching BW students research methods, providing students with hands-on experience in applied research projects. Since Fall 2019, more than 20 students have had a behind-the-scenes look at public opinion polling, including the nuances of survey research design, data analysis, weighting, as well as how to communicate our results to the public. Their ideas have also inspired questions we’ve included in our polls, and I’m grateful I’ve had the opportunity to work with them.
Thanks for reading Margin of Error. If you have any tips, comments, or insights about polling, email me at bplogiurato@gmail.com, or find me on Twitter @BrettLoGiurato.
If you liked what you read, share it with a friend and make sure to subscribe.