The Generic Ballot Rollercoaster

In the run-up to the 2022 midterms, various organizations issue a new generic congressional ballot poll seemingly every day. One day, Democrats are up by several points and look to have the momentum. The next day, Republicans are up by several points and they appear ready to take back the House and perhaps the Senate, as well. It’s simply not logical for the country to swing like a pendulum one way on a Monday and the other way on a Tuesday, so what gives?

by David Burrell, CEO at Wick

5 min read

The Light Version

In the run-up to the 2022 midterms, various organizations issue a new generic congressional ballot poll seemingly every day. One day, Democrats are up by several points and they appear poised to hold the House and Senate. The next day, Republicans are up by several points and they appear ready to take back the House and perhaps the Senate, as well. It’s simply not logical for the country to swing like a pendulum one way on a Monday and the other way on a Tuesday, so what explains the wild swing in the generic ballot polls?

The short answer is: while pollsters are setting quotas for physical geography values, like northeast, south, midwest, or west coast, they are not correctly setting quotas based on political geography values such as republican-led, democrat-led, swing or competitive congressional districts. As a result of this shortcoming, especially in popular non-probability sampling techniques, there can be a significant imbalance in the polling leading to huge swings in results.

But how big is the impact?

Well as a starting point: Looking at Wick Insight’s recent non-probability poll when we collected 1,500 complete nationwide using panels, in Republican-represented districts the spread was +10% in favor of the Republican candidate. For Democratic represented districts, the spread was +23 in favor of the Democratic candidate. So you can probably start to see issues with a pollster not setting political party as a strata.

Fortunately, when our team does non-probability sampling we use a really cool feature of our software platform, Wick, that finds the respondent in our custom database of hundreds of millions of respondents and, while we are still in the field, it can append new variables to the survey data set. The political geography variables are just a few of the hundred available that enrich a respondent’s survey record and empower new insights and stories from the research.

Read the entire article to get the full story on the generic congressional ballot rollercoaster.

The Full Story

In the midterm elections, the president’s party historically has lost seats in the House, especially in a president’s first term in office or when the president’s approval rating is particularly low. In midterm elections since 1906, the party that occupied the White House lost seats in the House 27 out of 30 times, or 90 percent of the time, with losses averaging 31 seats. The exceptions were 1934, 1998 and 2002, when the incumbent party was able to pick up seats in the House. In the Senate, the incumbent party lost seats 21 out of 30 times, or 70 percent of the time, with losses averaging three seats in the Senate.

In the 2022 midterms, one could expect history to repeat itself. Indeed, many headlines have been rife with Armageddon predictions for the Democrats suggesting potentially historic losses this year. Rising inflation, increasing gas prices, and an ongoing border crisis among other issues seem to have taken root to the detriment of Democrats. On the other hand, the Supreme Court’s Dobbs decision seems to have fired up the Democrat base. As we approach the middle of October, the latest generic congressional ballot polls project gains for Republicans. Even Jen Psaki, former press secretary to President Biden, agrees that if the midterms are a referendum on her former boss, the Democrats can expect to lose.

As you know, the generic ballot is a poll question that asks voters who they will likely support for the House of Representatives in the November election—a Republican or Democrat. There are no specific candidate names mentioned, only the major party that voters will choose.

At the end of August and despite the historical gains of the opposition party in midterms, Democrats were leading Economist-YouGov’s generic ballot 46 percent to 38 percent. By the end of September, that same Economist-YouGov generic ballot poll showed that eight-point lead for Democrats had totally evaporated with the race for Congress tied at 44% for each party. 

Even though major polling organizations like 538 and Pew consider the generic ballot as an important tool for accurately predicting the partisan distribution of the national vote, we must ask several questions given the wide swing in just a month:

Generic Ballot Polls Do Not Focus on Competitive Districts 

Because it is, in fact, “generic,” the generic-ballot question asks people across the country whether they are voting for Republicans or Democrats.  While this information might be interesting, it’s not necessarily indicative of whether the opposition party will actually pick up seats or not. Why? We don’t know whether those responding to the polls live in key swing districts that are actually competitive or whether they are from districts that are either heavily Republican or heavily Democrat. 

Unfortunately, not all non-probability polling set geographic quotas for district partisanship despite the fact that as shown in the chart to the right, there is a strong, direct correlation between respondents’ choice for generic congressional ballot and the respondent’s party.

As a result, as shown in the following two charts, there can be a significant imbalance in the polling leading to huge swings in results. In the first chart, when 7% too many respondents live in Republican congressional districts, the polls will skew by 1.7 points in favor of Republicans. In the second chart, when 7% too many respondents live in Democrat congressional districts, the polls will skew by 5.4 points in favor of the Democrats.

Results of Generic Congressional Ballot When 7% Too Many Respondents Live in Republican Congressional Districts

Results of Generic Congressional Ballot When 7% Too Many Respondents Live in Democrat Congressional Districts 

In order to predict which party is going to gain or lose seats in the House, it’s vitally important to focus on the swing districts and districts that lean just a little slightly to one party or the other.

SUMMARY

When we adjust for partisanship, the polls point to the Republicans winning big on election day. ABC News Washington Post concluded:

Among those living in congressional districts that are rated as at least somewhat competitive by ABC’s FiveThirtyEight (neither solid Republican nor solid Democratic), registered voters favor Republican candidates by a wide 55-34 percent — nearly as big as the Republican lead in solid GOP districts (+24 points). Democrats lead by 35 points in solid Democratic districts, pointing to a potential overvote where they’re most prevalent.

According to this same report, “In midterm elections since 1946, when a president has had more than 50 percent job approval, his party has lost an average of 14 seats. When the president’s approval has been less than 50 percent – as Biden’s is by a considerable margin now (42.9% average according to RealClearPolitics) – his party has lost an average of 37 seats.”