Source: Harry Enten on Twitter.
As elections analyst Ryan Matsumoto put it yesterday: “A lead of 15 points in the polling average is quite frankly a very different situation than ‘the polls said Hillary would win.’”1
Tomorrow, September 14th, is the final day to cast a vote in the recall election of California Gov. Gavin Newsom (D). Since a midsummer scare for Newsom, the polling has moved in his favor – and especially so since voting began on August 16th. As you can see from the chart below, the FiveThirtyEight polling average has keeping Newsom (i.e., “No” on the recall question) ahead by 16.6 points (57.5% to 40.8%). No poll has had Newsom losing since SurveyUSA’s August 4th release showing removing the governor winning by double digits. In fact, only three polls have ever shown the “Yes” option winning in the 32 polls tracked by FiveThirtyEight since January; one was in February and the other in early June – both had “Yes” up by just one point.
Source: FiveThirtyEight
In the middle of the summer, there were several polls showing the recall question to be close. However, only four of them had Newsom winning by five points or less (and a fifth was the August 4th SurveyUSA poll referenced above). All of these polls took place between July 19th and August 12th. That created a media narrative that Newsom was very much in danger of being removed from office. I assume Newsom’s political team was seeing something similar in their own polling considering how seriously they took the possibility that voters would oust the governor. Aside from this three week midsummer period, Newsom has been in good shape in the polls. And with voting ending tomorrow, he has opened up a nearly 20-point advantage in the polling. Polling errors of this magnitude are so rare one might even say they never happen (but there is always a first time for everything).
There is an interesting debate taking place among polling analysts about the polling in the recall election. Here are the options:
Campaigns matter and Newsom’s strategy is responsible for widening the gap we are seeing in the polling.
The three polls in July that showed the race extremely close (UC Berkeley and Emerson) or Newsom (SurveyUSA) losing were actually outliers and the polling has actually been stable all along in favor of Newsom.2
Pollsters’ likely voter screens were incorrect and pollsters have adjusted them in response to the high Democratic turnout we see in the early ballot returns.
On all three points, we will have a much clearer idea whether any of the theories is accurate once the election is over and the results are certified. But they cannot all be true at the same time. For instance, if we take the late July and early August polls as outliers we cannot also say that Newsom’s campaign made the difference in widening the gap in his favor by late August and early September (although it could still make a difference in the margin of victory if the most recent polling is accurate). The third point is difficult to ascertain because pollsters tend to hold their turnout models close to the vest, considering them proprietary or at least something approaching that. The third option can work with option one, but not for option two. For instance, Newsom’s campaign might have increased Democratic voter turnout in such a way that it confounded some pollsters’ turnout models. However, if the turnout models for the potentially outlier polls were incorrect then those polls might, in fact, not be outliers (although it is hard to see how SurveyUSA’s first poll in the race – in which Newsom was losing by 11 points – is not an outlier).
The table below shows all of the available public polling in the race since voting began on August 16th:
Source: FiveThirtyEight3
A month ago, Emerson (A) released a poll showing the recall essentially tied with keeping Newsom at 48% compared to 46% favoring removal. This morning’s poll shows significant movement to Newsom since that early August poll with 60% of respondents saying they will vote “No” on the recall question.
Since early September we have seen five polls from pollsters who surveyed the race earlier. While the data bounce around a bit in earlier months, all five pollsters show a double-digit advantage for Newsom heading into the final day of voting. Note that none of these pollsters released survey results in early summer. “Mid-Summer” means mid-July to early-August (in which survey field time bridged July to August). “Late Summer” is early-to-late August. Any poll released in September – even if in the field partly in August – is listed under “September.”
Source: FiveThirtyEight
Data for Progress and Suffolk University have released only one survey each in this race. Only two other pollsters that have released multiple polls in the race but have not released one yet in September: Change Research and Public Policy Institute of California. I expect - but do not know for certain - that Change Research will have one more poll either today or tomorrow. PPIC does regular polling so the schedule for the September poll might not be consistent with the election calendar.
The polling should not be surprising since Newsom’s approval ratings are much higher than those for Gov. Gray Davis (D) on the eve of his recall in 2003.
According to 2003 exit polls, Mr. Davis’s approval rating was at just 26 percent, with 73 percent disapproving, on Election Day. Voters were deeply dissatisfied, blaming him for the skyrocketing bills and rolling blackouts that emerged from the electricity crisis.
Recent polls show California voters generally think Mr. Newsom is doing a good job. A YouGov poll found 53 percent of likely voters approved of his performance, with 38 percent disapproving. Another poll from the Public Policy Institute of California found similar results, with likely voters at 53 percent approve and 43 percent disapprove.
Ryan Matsumoto, The Upshot, September 13, 2021
The turnout thus far supports the polling in the race. Democratic turnout is up. So far, slightly more Democrats have returned ballots than they did at this point in last year’s presidential election. Republican turnout is not nearly as strong as it needs to be to overtake the Democratic turnout. Could that change on Election Day? Well, Republicans could show up strongly at the polls tomorrow but there are just not enough of them at this point. The “Yes” campaign doesn’t need just a strong right-wing turnout; it needs a depressed Democratic turnout. And it does not appear that we are seeing such a depressed turnout.
I am sure you have already heard some folks saying, “but a lot of Democrats want to recall Newsom.” My response to that is “bullshit.” This is just a right-wing talking point. There is no evidence that any significant number of Democrats want to recall Newsom.4 Perhaps they have their problems with the governor, but those concerns are largely coming from the left and enough seem to understand that recalling Newsom means putting in a fascist talk show host as governor. Once we have the unofficial results from the election we’ll revisit the three questions above and see if we can make more sense of what the polling told us over the course of the campaign.5
Remember, the final polling in the 2016 race had Clinton winning by about three points nationally. She won the popular vote by just under that. The polling over this past weekend represents the “final polling” in the recall election - and there is no Electoral College to upend the popular will this time.
Although if the polling in the last two weeks is accurate, Newsom’s numbers were stable until September when they did in fact increase.
FiveThirtyEight adjusts polls based on things like a pollster’s track record of success when calculating its polling averages. That is what the “Adjusted Leader” column shows. The polling average in this race is calculated using the adjusted figures. When I calculate a polling average on this page I typically use a simple method based on the reported poll results, but usually cutting off the historical time period at some point. As a result, my calculated polling average can differ from FiveThirtyEight’s even when we are using the same polls. https://fivethirtyeight.com/features/polls-policy-and-faqs/
In fact, the polling is showing that more Republicans are voting No (~15-20%) than Democrats are voting Yes (~5-7%).
If Newsom loses, the damage that will have on the polling community will be enormous. This would be the biggest and most consistent miss across pollsters that I can remember. By a lot.