Charlie Cook Breaks Down What The Polling Really Says About The Midterm Elections

The Cook Political Report

Charlie Cook, one of the country’s foremost experts on election forecasting, political polling and the state-of-play in American politics at any given time, spoke with Mediaite for an in-depth look at the challenges pollsters are facing and how both media and the public should read the barrage of polling ahead of the November midterm elections.

Cook, who is the founder of the Cook Political Report as well as a political analyst for both the National Journal and NBC News, began by arguing that polling was not nearly as off in 2016 and 2020 as most people have been led to believe.

While offering a handful of tips for how best to interpret polls, Cook emphasized that most people criticizing polling today “need to realize that the popular and electoral vote is not the same thing.”

He also offered a detailed look at how the current election cycle has shifted toward the Democrats, but concluded “this election is far from over.”

Read the full Q&A with Charlie Cook, which has been edited for length and clarity.

Mediaite: We saw in 2020 final polling averages over predicted Biden’s vote total, and in some states like Wisconsin and Pennsylvania by quite a large percentage. What do you see as the key challenges for pollsters this election cycle?

There is no question that polling is more challenging today than it was 30 years ago. Then people had landline telephones, few had or used Caller-ID and voice mail thus didn’t screen incoming calls that way, and robo-marketing calls had not yet reached an infuriating level. Simply put, it was easier to get a representative sample, now they have to work hard to work around each of these challenges and the cost of quality survey research has gone through the roof.

Having said that, much of the debate over polling dates back to the surprise 2016 presidential election outcome and Donald Trump’s election over Hillary Clinton. Ironically, the polls were not quite as “wrong” that year as many seem to think they were. Many believe that “the polls called Hillary Clinton the winner in the presidential race, she lost, therefore, the polls were wrong.”

The problem with that thinking is that national polls only look at the national vote, do not break out individual states, when of course the election outcome is based on state-by-state results and the Electoral College.

The RealClearPolitics average of national polls, going into the 2016 Election Day, showed Clinton ahead of Trump by 3.2 percentage points, 46.8 to 43.6 percentage points while 9.6 percent were undecided or going to a third party or independent candidates. In the final national popular vote, Clinton won by 2.9 percentage points, 48.2 to 46.1 percent, so the national polls called the national popular vote winner correctly, the margin was off by three-tenths of a percentage point. So, the national polls were off, but not by much at all.

In most individual states, the polls did fairly well too. The off was in three states, Michigan, Pennsylvania, and Wisconsin, all had Clinton ahead by a few points, Trump ended up winning those three states by fewer than 78,000 votes out of just under 14 million cast in those states, 137 million nationally. There was some under-sampling of Trump voters, pollsters have studied very closely since, revising their sampling and weighting of data for various groups.

But it should also be pointed out that those three states had among the most restrictive early voting policies in the country, thus virtually all the votes were cast on Election Day, unlike most other states where early and mail voting has become quite prevalent, making “Election Day” for many days or weeks before the official date of the election.

One mistake is that many focus on the spread between two candidates and ignore the presence of undecided voters, many of whom eventually decide, but perhaps days or weeks after they were interviewed for a poll. It was clear, particularly given what was going on with FBI Director James Comey’s press conference on Hillary Clinton’s emails, undecideds broke heavily towards Trump at the end. Did that make polls taken while those people were still undecided, wrong?

The national polls were off by a small amount and got the popular vote call correctly, it was in three states where the real off was, people need to realize that the popular and electoral vote is not the same thing and that splits can happen, and that is not the fault of the pollsters.

Yes, there are problems with polls, but there is also misinterpretation by journalists and their readers, listeners, and viewers as well.

Is there a concern the pro-Trump vote may be undercounted again as these voters may be less willing to participate in polling? Are there any changes to methodology that pollsters can/are taking to address this?

Pollsters and political scientists have and continue to study this closely, it is likely to be several small items that combined to be somewhat larger, the Shy-Trump voter phenomenon may exist but there appear to be a couple of other things as well. It is interesting that the polling in both the 2014 and 2018 midterm elections was very accurate, only the 2016 and 2020 presidential elections were off, and again, less than popularly thought.

Clearly, there is a group of people that refuse to be interviewed, low-trust voters, and more of them were supportive of Trump than of Clinton or Biden. Pollsters are looking at how their samples are constructed, which types of voters are least likely to participate and what efforts can be made to either reach them or to weight up those of that kind that they do manage to interview. It is a work in progress, but as I said before, polling is having challenges but was not nearly as wrong as some would like to believe, and some of this was the confusion with what national polls might show and what ends up happening in the Electoral College.

How do you believe the media should cover polling in general? Do you have any specific advice or words of caution for reporters? Any common mistakes you see in coverage?

First, I would not obsess over the spread, the gap between the leading and lagging candidates. I would pay more attention to the vote each candidate is getting, vis-à-vis 50 percent, and know that undecided voters often decide, rarely breaking evenly between the two candidates. Normally, few end up going with incumbents, to the point that some say for incumbents, “what you see is what you get.” Also, keep in mind the margin of error of a few points that exist in any poll. Be dismissive of any poll that reports results past a decimal point, that implies a level of precision that does not exist in polling, and I think is misleading.

Also, watch for any signs of late-breaking trends. In 2020 for example, after President Trump performed poorly in the September 29 debate, the bottom fell out both for him and for GOP candidates, as Republican-tilting voters became disillusioned, Democrats more motivated and undecideds began moving toward Democrats. Then in the closing weeks of the campaign, as expectations of a big Democratic win built and talk of the Blue Wave, that Biden would win the presidential race easily, Democrats would expand the size of their House majority and win control of the Senate with several seats to spare, padding that new majority.

At that point some undecided voters grew concerned about whatever democratic socialism was, earlier talk of defunding the police, packing the Supreme Court. Would Democrats talking about “Medicare for All” take away their private health insurance and exactly what is a Green New Deal and would it kill jobs?

Also, some Republicans who had grown disheartened after the debate re-engaged. Without warning, the Blue Wave turned into the Dead Sea. The Biden margin in quite a few states was less than expected, he won an Electoral College majority by fewer than 126,000 votes total, in Arizona, Georgia, Pennsylvania, and Wisconsin. Instead of gaining House seats, Democrats almost lost control, holding on by fewer than 32,000 votes spread across five Congressional districts, won only 48 seats in the November election, only capturing control of the Senate in two January 5 runoff races in Georgia, after Trump’s tantrum after the November balloting, attacking both the Republican Governor and Secretary of State in Georgia.

Democrats won their 49th seat by fewer than 94,000 votes and the 50th with less than 55,000.

Much has been made about the Democrats rebounding on the generic ballot in recent weeks, what do you make of this trend and should voters/media put much stock into it? Should House races be viewed as local or national elections in this hyper-polarized environment?

Despite the Senate being evenly divided and Republicans only a handful of seats away from a majority in the House, the old axiom that the campaign doesn’t really begin until Labor Day has been ignored.

Earlier this year it looked like a classic midterm election with the party of the sitting President likely again to suffer terrible losses. President Biden’s very low approval ratings, inflation at a 39-year high, gasoline prices at record levels, concerns about inflation rampant, and fear that a recession was in our future created a terrible political environment for a governing party trying to maintain control. Some believed that Biden and Democrats had not delivered on what they promised, others thinking that Democrats had over-reached, tried to go too far given how narrowly divided the country is and how close the election had been.

But came summer things changed. After the Supreme Court’s decision in June overturning the Roe v. Wade decision that gave women the right to an abortion, exacerbated by recognition that “trigger-laws” of previously passed abortion restrictions in many states meant that bans and restrictions automatically went into effect, a number of governors and state legislatures began moving to enact even greater restrictions or bans and reports of judges stopping women and in some cases children from getting abortions and doctors afraid to do anything that could be themselves afoul of the law, the motivation of Democrats and pro-choice voters increased dramatically and there was movement toward Democrats in quite a few races.

Mid-summer brought some better economic news with inflation slowing, gasoline prices though still high dropped by a dollar a gallon and a flurry of legislative successes in Congress tapped down some disenchantment with Democrats.

Worries of Republicans over-reaching grew, not just triggered by the abortion fight but also revelations from the investigation of the January 6 attack on the Capitol Building, efforts to overturn the election, increased speculation that Trump would run again followed by the FBI search of his home in Mar-a-Lago effectively broadened that spotlight of attention to become more of a floodlight, encompassing not only the issues that had plagued Democrats but also concerns about Republicans, the election shifting from a referendum on Biden and Democrats to a choice, the alternative being a Trump-led Republican Party.

But the election is just under 60 days from now, and just as things are different than 60 days ago, they may be different two months from now. There will be plenty of events between now and then, not the least of which are several inflation reports, the direction of gasoline and good prices, interest rates, and who knows what else. This election is far from over.

What key races do you believe the media should most focus on? Which polls or forecasters would you recommend or warn against?

I would suggest that those really interested in individual races focus on the “Toss Up” races identified by the people who watch races closely for a living, study which candidates win and which lose, the voting behavior of individual states and districts, and importantly have no partisan ax to grind.

“The Cook Political Report with Amy Walter,” “Inside Elections with Nathan Gonzales,” and “Sabato’s Crystal Ball,” each have cracker jack staffs sifting and winnowing the races to point to which ones really matter. Those who are serious about it also keep a close eye on Nate Silver’s FiveThirtyEight and the polling compilations on RealClearPolitics. The Cook Political Report’s Partisan Voting Index (PVI) provides a starting point in assessing how any one state or district votes compared with the nation as a whole, using presidential voting for the last two elections, jurisdictions that do not lean more than three or four points from the nation as a whole are the places where it is most likely for competitive races to occur, the PVI is instructive in what natural tilt a state or district might have. C-SPAN carries many of the debates in key races, which are worth watching and give texture to assessing races.

The last piece of advice is to try to look at these races as clinically and dispassionately as possible, push out of your mind who is more qualified, who “should” win or might do the better job, start off with the voting patterns in that state or district and how that candidate does or does not match up with how the voters there tend to vote.

It has long been said that in real estate, the three most important factors are “location, location, and location,” in politics it is “timing, timing, and timing.” A candidate in one year and political environment may have a much greater or lesser ability to win in a different year and political climate, there are partisan headwinds and tailwinds that can help or hurt a candidate a great deal, with the direction of any last minute gust of wind making a real difference for those who have not already voted and yet are undecided. This is an art based on a science and should be viewed that way.