Of course, polls could have been wrong all along. However, that seems unlikely, given the consistency of the results in a number of polls, administered by many different polling organizations. For the Lake-Hobbs contest, the results displayed by FiveThirtyEight came from 13 different polling organizations, as shown in Figure 2. Only one (Marist) had Hobbs winning:
Figure 2: Lake-Hobbs polling organizations and results
Did 12 out of 13 companies make the same error, and in the same direction?
We see the same pattern with the senatorial contests. Figure 3 displays the final polling averages, as produced by RealClearPolitics, for seven key races. In each case the GOP candidate was leading just before the election, even though these races were in different states and conducted by different pollsters, and for different candidates. This pattern does not suggest polling error. The actual election results (courtesy of Reuters) are also shown.
Three of the seven contests were won by Republicans, but look deeper. In every case, the Democrat candidate’s election results were sharply higher than the final polling percentage. For the most part, that was not the case for Republicans, as we can see in the right-side columns in Figure 3.
Figure 3: Senatorial polling versus results
Figure 4, below, presents the same information in graphic form. Keep your eye on the “zero” line. It shows the expected results for both Democrats and Republicans. Democrats are entirely above the “expected” line, signifying that they beat the predictions in every case. On the other hand, Republicans are much lower, and mostly below the expected line.
Figure 4: Senatorial polling results, shown graphically
Before we address ballot-harvesting, let’s summarize: Perhaps I missed it, but it seems there was no last minute erosion of support for GOP candidates. And polling error is unlikely, given that the GOP lead was shown by many different pollsters, in different states, and for different candidates. Yet, the Democrat candidates consistently beat expectations by quite a lot. Why was that?
Ballot-harvesting
It is just a theory, but I believe that aggressive ballot-harvesting is the most likely explanation for the disparity between pollster predictions and actual results in the 2022 midterm elections. In the movie, 2,000 Mules, True the Vote (TTV), the organization run by Catharine Engelbrecht and Gregg Phillips, described the prevalence of ballot-harvesting in the 2020 election. Although its estimates are only rough approximations (based on cell phone analysis), TTV believes there were up to 200,000 and 275,000 harvested ballots in Arizona and Pennsylvania, respectively. Nevada was not included in the TTV analysis, but since the end of 2020, Nevada has taken all restrictions off of harvesting. It is now legal to harvest ballots in Nevada, and it can be done anonymously in any quantity with ballots from unregistered homeless people, drug addicts— anyone. [Editor's note: See also phantom voting, previously published at American Thinker by contributor Jay Valentine, here and here.]
When it comes to the legality of ballot-harvesting, there is confusion. As noted, it is now completely legal in the state of Nevada. In most other swing states, including Pennsylvania and Arizona, it is (technically) allowed only within tight limitations (e.g., for family members). That said, large-scale harvesting operations have been performed in many states, and I believe they often lead to serious election law violations. There are three reasons for this:
- In some states, administrators ignore the practice, and may even see it as “empowering” for voters in “marginalized communities.”
- It is very easy to cheat when ballots are returned via mail or drop boxes without a hard identification requirement.
- Harvesters are usually paid “per piece,” and have incentives to find or steal as many ballots as possible.
