Last October (FAIR.org, 10/3/22), I warned about “The Persistently Faulty Record of Generic Ballot Polling.” The message was that it’s dicey to predict House election outcomes based on the national polling.
Still, many in the media relied on the generic ballot—which asks voters across the country which party’s candidate they prefer for the House of Representatives—to shape their prognostications about the 2022 House elections. And for an understandable reason: Nate Silver’s 538 (6/5/17) has argued that the generic ballot is “the best tool we have for understanding how the midterms are shaping up.”
In late October, 538’s generic ballot average began moving in favor of Republicans. According to the RealClearPolitics average, the shift began a month earlier.
And journalists noticed.
“Two weeks out from the midterms, evidence points to a re-emerging red wave that could sweep in GOP control of both chambers,” Axios (10/23/22) reported, noting that “the latest public polling shows Republicans pulling ahead on the generic ballot.”
Josh Kraushaar of Axios (10/23/22) referenced the trend in an article headlined “Red Tsunami Watch,” which suggested that “it’s now very possible House Republicans win back the majority on November 8 with more than 20 House seats.”
A New York Post (11/4/22) article also noted the trend and suggested it was “yet another sign of the GOP’s momentum advantage with less than a week to go until Election Day.”
CNN (11/2/22) reported its own poll on the generic ballot, showing Republicans leading Democrats by 4 points. This was significant, because even “closely divided generic ballot numbers have often translated into Republican gains in the House.”
That last statement reflects the fact that the generic ballot averages, as compiled by both 538 and RCP, had historically underestimated Republican strength, and needed to be adjusted to take past errors into account.
What the faulty record showed
To be clear, the generic ballot can only measure the national popular vote. The national vote, in turn, usually overstates the number of House seats Democrats actually win.
The New York Post (11/4/22) cited the generic ballot poll as “yet another sign of the GOP’s momentum advantage with less than a week to go until Election Day.”
In fact, as I indicated in the October article, for the previous ten elections (2002–20), the national vote has, on average, overstated Democratic strength by an average of 3 percentage points—primarily reflecting how gerrymandering makes congressional elections a non-level playing field. In practice, that means the Democrats—on average—would have to win the national vote by 3 points just to break even in the number of House seats they won.
But there’s another complication: how well the polls predict the national vote itself.
As I showed in the October article, the generic ballot averages compiled by both 538 and RCP over the previous ten elections have overpredicted Democratic strength in the national vote—by an average of 2.4 and 1.1 percentage points, respectively.
It’s important to note that journalists typically do not treat the generic ballot results as indicative only of the national vote. Instead, they present the results as though they predict the actual percentage of seats each party will win.
Thus, for purposes of this article, we can examine the overall accuracy of the generic ballot in predicting House seats, recognizing that the overall number reflects both 1) the accuracy of the generic ballot in predicting the national vote, and 2) the accuracy of the national vote in reflecting the distribution of House seats.
2022 House seat predictions
In the October article, I showed that 538’s generic ballot overstated Democratic strength in winning House seats (or understated Republican strength) for the past ten elections (2002–20) by an average of 5.5 percentage points—which translates into miscalling the House results by an average of 24 Republican seats (that is, 5.5% of the 435 total House seats).
538‘s final forecast (11/8/22)—combining the generic ballot with other polls, fundraising, voting patterns and expert ratings—gave Democrats less than a 1 in 3 chance of doing as well as they did.
This election, the final 538 generic ballot compilation showed Republicans leading by 1.2 percentage points, which—if correct—would give them a five-seat margin in the House.
Anyone familiar with the skew of 538’s generic ballot average in favor of Democrats, however, would want to take into account that average 24-seat underprediction of GOP seats. Add that number to the predicted five-seat margin for 2022, and the results would suggest Republicans winning the House by 29 seats.
Similarly, the RCP final generic ballot average for 2022 showed Republicans winning by 2.5 percentage points, or 11 House seats. But RCP also has a history of underpredicting Republican seats over the previous ten elections, by an average of 4.2 percentage points, or 18 House seats. Add those two House seat numbers, and RCP, like 538, would seem to predict a Republican win by 29 seats.
Given that both aggregators seemed to indicate a similar election outcome, it is understandable why many journalists might have been expecting a decisive GOP victory in the House—maybe not a “red wave,” but certainly more than a red ripple.
2022 House election results
With the final votes in all congressional districts recently completed, Republicans have actually won 222 seats to the Democrats 213, giving the GOP a nine-seat margin—compared to the 29-seat margin one might have expected, given the adjusted final generic ballot averages of both 538 and RealClearPolitics.
The unadjusted 2022 generic ballot results of these two sites were actually quite close to the final outcome. As it has mostly done, the 538 average underpredicted GOP strength, but by only four seats; RCP overpredicted GOP strength by just two seats.
Is there a lesson to be learned?
Prior to the election, Walter Shapiro of Roll Call (9/27/22) suggested 2022 was an unusual election year that did not fit the pattern of election years past—or as he put it, “You just can’t account for the weirdness of 2022.”
He happened to be right: 2022 did not fit the general pattern of most elections.
As the chart shows, 538’s generic ballot average suggested Democrats would lose by 1.2 percentage points, and they actually lost by 2 points—an overstatement of just 0.8 points.
Only one other time in the past 22 years has 538’s generic ballot average come within one percentage point of the actual distribution of House seats—in 2018. In all other years, the error was substantial—from a minimum of 4.6 percentage points (20 House seats) to 12.1 points (53 seats).
RCP also shows 2022 to be an unusual election year. Its generic ballot showed Democrats losing by 2.5 points. They lost by 2 points. That half-point error is the closest margin for the past 11 elections.
Two other years were close (within 2 percentage points), but the other eight years had substantial errors—from 3.8 to 11.4 percentage points, representing 17 to 50 seats.
So, the lesson is—as I warned before the election—the generic ballot is a precarious tool for predicting election outcomes.
If the generic ballot is, in fact, “the best tool we have for understanding how the midterms are shaping up” (as 538 suggests), perhaps media prognosticators ought to be more constrained than they were last month.
Or, better yet, perhaps the media should actually follow Walter Shapiro’s advice: “For those tempted to predict the midterms—don’t.”
The post In 2022 Midterms, Media Were Again Misled by Generic Ballot appeared first on FAIR.
This content originally appeared on FAIR and was authored by David W. Moore.
David W. Moore | Radio Free (2022-12-09T20:45:43+00:00) In 2022 Midterms, Media Were Again Misled by Generic Ballot. Retrieved from https://www.radiofree.org/2022/12/09/in-2022-midterms-media-were-again-misled-by-generic-ballot/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.