–>
October 24, 2022
Yes, folks, the red wave is real and coming to your town soon.
‘); googletag.cmd.push(function () { googletag.display(‘div-gpt-ad-1609268089992-0’); }); }
I have discussed the types of bias that can occur in political polling in several previous articles, one of which you can read here. Unfortunately, that bias is still out there. More on that later.
I have also discussed the hidden conservative vote, now officially recognized by pollsters and referred to as partisan non-response bias. This means Democrats want to talk to the pollsters, and Republicans don’t.
In 2020, polls overestimated Joe Biden’s support by roughly 4% nationwide and by an average of 4.3 points at the state level. The only factor the American Association for Public Research could point to for the polling miss was partisan non-response bias.
‘); googletag.cmd.push(function () { googletag.display(‘div-gpt-ad-1609270365559-0’); }); }
What is partisan non-response bias?
Imagine you are a conservative working at a woke corporation or a leftist college. Or imagine you are a union member or are friends with a group of liberals. If you admit you are a conservative, you could lose your job, ruin your career, or be shunned or ostracized.
It does not stop there. Conservatives are banned on social media despite our First Amendment rights. The FBI is raiding the homes of prominent conservatives, including former president Trump. And President Biden considers MAGA Republicans to be a threat to democracy. Under these circumstances, would you admit you favor conservatives in a poll? Would you even answer a poll?
Since the level of partisan non-response bias can be assessed only after an election is over, how do we make an estimate now? After all, you don’t know what you don’t know.
The simplest thing to do is use the 2020 Biden nationwide polling miss of 4%. It might be higher in red states like Ohio, which has more registered Republicans, and lower in blue states like Colorado. This is a reasonable assumption until it is clear that pollsters have learned how to account for it. Right now, I consider 4% low.
There are structural forms of bias in the polls themselves. Sometimes it is easy to spot.
‘); googletag.cmd.push(function () { googletag.display(‘div-gpt-ad-1609268078422-0’); }); } if (publir_show_ads) { document.write(“
For example, a poll of likely voters is usually more accurate than a poll of registered voters. In 2014, pollster Nate Silver, whom I have quoted before, claimed that midterm polls of registered voters are biased toward Democrats by over 2.5%.
Another place I like to look is the Gallup Party Affiliation Poll. Gallup has been surveying political party affiliation monthly since 2004. Suppose we average the ten months from December 2021 to September 2022 and add the independent leaners. In that case, 45.8% of voters are or lean Republican, and 47.5% are or lean Democrat. This gives Democrats a 1.7% edge.
But keep in mind that partisan non-response bias affects this poll just like the others. If we account for non-response bias, Republicans should have an advantage. However, for the sake of argument, let’s assume they are tied. Let’s also assume that roughly 1,000 people were polled, though I cannot confirm that.
The sampling error in a poll of 1,000 people is +/- 3%. This means that the reported numbers for Democrats and Republicans could be off by 3% one way or the other. If we add 3% to the Democrat number and subtract 3% from the Republican number, Democrats have a 6% advantage. If we reverse the process, Republicans have a 6% advantage.
Why do we care? This means any national political poll cannot have more than a 6-point split on party affiliation with either a Democrat or Republican advantage and be considered within the margin of error. In other words, a poll outside the margin of error could be biased.
The most recent Politico/Morning Consult survey of less accurate registered voters has Republicans with a slight advantage at 37-36.5, without counting leaners. Surprisingly, independents are 16 points below the Gallup estimate. Democrat women, at 21%, are the largest segment, while independent women are the smallest, at 12%.
The Economist/YouGov poll of likely voters gave the Democrats a 9-point advantage.
The CNN poll of 800 registered voters is hard to figure because there are no numerical splits, just percentages. For example, 47% of their respondents voted for Biden vs. 44% for Trump. Then they claim that the poll contains 24% liberals vs. 34% conservatives.
Now let’s look at the Generic Congressional Vote polls. Historically, for Democrats and Republicans to have a 50-50 chance of winning the House of Representatives, Democrats would need to have somewhere between a 3.6- and a 5-point lead in the Generic poll average. Let’s call it 4 points to make calculating easier. That means that if Democrats and Republicans are tied in the Generic Ballot, Republicans are actually up by 4 points.
The current RCP Generic average gives Republicans a 3.1% edge. However, if we remove the Economist/YouGov poll from the Generic Ballot average because it has a Democrat bias, the Republican average rises to almost 3.5 points. If we add this to the 4 points Republicans already have, they should have a reasonably solid 7- to 8-point voter advantage.
Now let’s take a quick look at a few other factors. First, the midterm turnout bias for the party out of power should favor Republicans by 5%. While this is a valid statistic of a historical trend, it’s not something you can count on to estimate Republican strength.
Voter enthusiasm is sometimes hard to gauge because it is often tied to current events. After the Roe decision, Democrat enthusiasm went up. After the Mar-a-Lago raid, Republican enthusiasm went up. According to Rasmussen, Republicans have an 8-point advantage. According to the Morning Consult Election Tracker, Democrats have a 5-point advantage. As the Joker said, “Hubba, hubba, hubba, money, money, money, who do you trust?”
The last item I want to discuss is the top concerns of voters. According to the Harvard/Harris poll, inflation, crime, and immigration were the top three issues. Abortion was fourth.
According to Statista.com, inflation and the economy were the top issues. Abortion was in eleventh place. And according to a CBS News poll, inflation, the economy, and crime were the top three issues. Abortion was sixth.
My point is, this election is about eggs, butter, gasoline, crime, and fear of the future, not abortion. With partisan non-response bias at 4%, the generic ballot favoring Republicans by 7–8%, and a favorable advantage in enthusiasm filling our sails, a 12-point red wave is not out of the question.
I’ll leave you with this.
Just after the 2020 presidential election, pollster Richard Barris claimed that other polling organizations were responsible for the poor poll results due to their own political bias, saying, “This industry is dominated by left-wingers. And a big, big problem is they’re trying to profile the voting behavior of people they don’t understand and may even despise.”
Next, the governors. Then the Senate.
Image via Pixabay.
<!– if(page_width_onload <= 479) { document.write("
“); googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-1345489840937-4’); }); } –> If you experience technical problems, please write to [email protected]
FOLLOW US ON
<!–
–>
<!– _qoptions={ qacct:”p-9bKF-NgTuSFM6″ }; –> <!—-> <!– var addthis_share = { email_template: “new_template” } –>