Unfortunately the VP debate appears to have resulted in a similar mixed outcome as the first presidential debate:
For the first presidential debate, see our snap poll analysis here.
A principal theme has been apparent throughout the 2020 race, namely that traditional metrics point to Biden/Democrats being the favorites while alternative data mostly points to Trump/Republicans.
This same theme has continued into the VP debate.
Perhaps most interesting are the snap poll results from the news providers based in swing states, namely Pennsylvania and New Mexico. In both cases, Pence ‘won the debate’ according to respondents.
Most, though not all, of the snap polls mentioned in this post were organized by local news sites / TV channels. From this perspective, the results provide reasonably high quality and insightful data which seems illogical to ignore.
Snap Polls
Flash polls, snap polls, or on-line surveys are taken by an organizer, normally a news site but also at times by a social media account, and open to the public. These are mostly anonymous, though some ask basic demographic questions. The advantages of such snap polls are that they attract many more respondents than normal polls and are anonymous. While traditional polls might have somewhere between 500 and 1,000 respondents for a post-debate poll, these snap polls can easily go to over 100,000 responses. The downside is that we are not sure if the makeup of the participants matches the electorate. There is also a potential problem with voting multiple times.
In the current environment snap polls should be watched carefully. A highly contentious political environment, which the US clearly has at present, will normally produce distortions in traditional polls because they are not anonymous. This was the issue in 2016 with the ‘hidden Trump voters’ and we believe it has continued into 2020.
So, even with their weaknesses, snap polls should be taken seriously.
One other interesting factor is that many of the snap polls this election cycle have been conducted by relatively small and/or regional organizers. In previous elections, we saw major national outlets conducting these snap polls. The Drudge Report had conducted various snap polls after debates in previous cycles.
For whatever reason, the snap poll phenomenon has shifted in 2020 to smaller and/or regional organizers. We believe this lends more credibility to the outcome as those looking to game the outcome would frankly need to work a lot harder to find such snap polls.
Snap Poll Results
This snap poll was conducted presumably through multiple local news sites, which may be owned by a common parent company. At least three sites carried the same snap poll, nj.com, oregonlive.com, and al.com. These three sites appear to serve New Jersey, Oregon, and Alabama. But, it is unclear how many other sites fed into this snap poll. The number of responses was 39,271. Results:
Interestingly, this same snap poll was conducting after the first presidential debate and had 300,467 responses the morning after the debate. A comparison between the responses from each of the debates shows far fewer responses for the VP debate, which makes sense.
Next, we have a snap poll from tvline.com. They do not provide number of respondents. They conducted a snap poll for the presidential debate as well. Results:
A third snap poll was conducted by wevv.com on its twitter account, @my44news. This same organizer conducted a snap poll for the presidential debate on its site receiving approximately 9,000 responses. The VP snap poll on twitter only received 59 votes. Results:
The number of responses is unusually low. Normally this would not be worth mentioning but it shows that the likelihood of manipulation appears low as the results are similar to other larger snap polls.
A fourth snap poll was organized by krge.com. This site likewise conducted a snap poll for the first presidential debate. There were 18,340 responses. Results:
The next snap poll, we did not cover in our previous post on the first presidential debate. It was conducted on twitter by @6abc, which according to its twitter profile: “Philadelphia area news, traffic, weather, politics and sports from WPVI, including breaking news & AccuWeather forecast.” Seemingly, this snap poll was organized by the Philadelphia area ABC news affiliate. It had 79,452 votes at the time of writing. Results:
Another snap poll was organized by actionnewsnow.com which is local news organization focused on Northern California. There were 2,655 responses. Results:
The largest snap poll that we know of was from WPRI, which is a local TV news provider in Rhode Island and Massachusetts. It received 478,229 responses. Results:
Why You Should Take Note of these Snap Polls
In general, we see these snap polls as being more than just entertainment.
The organizers are legitimate, mostly highly focused local news providers. Further, it seems as though respondents would be mostly from their local coverage areas, making these snap polls regional. Before coming to further conclusions, let’s take a quick look at some of the organizers of the snap polls.
The snap poll organized by the series of presumably linked locally focused news sites appears extremely neutral. It includes left leaning states (New Jersey and Oregon) as well as right leaning states (Alabama). We do not see an obvious problem or bias here.
The wevv.com snap poll also seems reasonable. According to Wikipedia it is a “CBS/Fox/MyNetworkTV-affiliated television station licensed to Evansville, Indiana, United States, serving the Tri-State area of southwestern Indiana, northwestern Kentucky and southeastern Illinois.” Basically, it is a regional news provider. This region leans right and could likely have a conservative bias.
The KRQE snap poll also looks reasonable. According to Wikipedia: “KRQE, virtual and VHF digital channel 13, is a dual CBS/Fox-affiliated television station licensed to Albuquerque, New Mexico, United States and also serving the state capital of Santa Fe.” The organizer of this snap poll appears to be a regional TV news provider. New Mexico voted for Clinton in 2016 and most forecast it will vote Biden in 2020, though we see it more as a general swing state.
The WPVI-TV or the @6ABC snap poll also seems reasonable to use. It provides news to the largest city in Pennsylvania, clearly one of the most watched swing states for the 2020 election.
actionnewsnow.com is a very local news provider. According to its site: “Action News Now is the shared news brand for television stations: KHSL, KNVN, the CW, and Telemundo in the Chico-Redding DMA. Action News Now is dedicated to delivering “Coverage You Can Count On” to the nearly 175,000 Northern California households who watch Action News Now every month, with an additional 2 million+ monthly mobile & online pageviews. Action News Now is uniquely positioned to serve the needs of viewers and advertisers, alike, with live, breaking, local news & weather, plus national news, live sports & programming from major networks: CBS, NBC, The CW, and Telemundo.” Interestingly, Chico voted for Clinton in 2016 while Redding voted for Trump in 2016. The fact that this station also provides “The region’s ONLY local Spanish newscast on Telemundo” should also be noted.
Lastly, we should look at the outlier WPRI which had the most pro-Harris results as well as the largest number of responses by far. According to Wikipedia: “WPRI-TV, virtual channel 12 (VHF digital channel 7), is a dual CBS/MyNetworkTV-affiliated television station licensed to Providence, Rhode Island, United States and also serving New Bedford, Massachusetts.” Presumably, this local news provider would tend to attract its local base which according to previous elections has overwhelmingly voted Democrat. In 2016 the margin of victory for Clinton in Massachusetts was 27% and in Rhode Island was 16%, some of the states with the highest margins of victory for Clinton.
For the most part, we see these snap polls as offering legitimate insights into the 2020 election.
It might be reasonable to throw out the snap polls conducted by news organizations based in the Indiana/Kentucky and the Rhode Island/Massachusetts regions as they could have a local bias. In fact, these results went more or less in the direction that you might expect.
Perhaps most interesting are the snap polls from organizations based in New Mexico and Pennsylvania, which traditionally have had much more closely divided elections. The fact that they show Pence winning should be taken note.
CNN Poll
CNN conducted a poll post-debate which included 609 respondents. Results:
According to CNN: “The CNN post-debate poll was conducted by SSRS by telephone and includes interviews with 609 registered voters who watched the October 7 vice presidential debate. Results among debate-watchers have a margin of sampling error of plus or minus 5.3 percentage points. Respondents were originally interviewed September 30 to October 4 either by telephone or online, and indicated they planned to watch the debate and would be willing to be re-interviewed when it was over. Respondents initially reached online are members of the SSRS Opinion Panel, a nationally representative probability-based panel.”
New Information from the VP Debate
Very little.
The CNN Poll from the first presidential debate put Biden at 60% while its VP debate poll put Harris at 59%. These results look remarkably stable.
Snap polls from the first presidential debate showed a slight advantage for Trump while those from the VP debate also show an advantage for Pence.
(This conclusion could be argued as the largest snap poll did point to Harris having won however it was also from the region which produced a lopsided 2016 election outcome which implies that these results reflect the local respondents. The snap polls from swing states showed a clear Pence victory, which we believe is more relevant.)
Conclusion
The nation is extremely divided as are the data! Measuring using:
Other alternative data, which for the most part is more anonymous, also continues to point to Trump and the Republicans as having the advantage.
Similarities to 2016
As an afterthought, we should mention that a similar situation occurred in the 2016 election.
Prior to Election Day, ZettaCap had repeatedly highlighted the divergence between forecasts based on traditional and alternative data analysis methods.
We also warned of relying too heavily on non-anonymous sources of data in a contentious political environment. Specifically, we forecasted that Trump would win Michigan, Pennsylvania and Ohio partially due to these distortions.
Currently, we see similar distortions in the 2020 election. We also see the mainstream putting their full faith into data that is not sourced anonymously.
On this site we have repeatedly highlighted data based insights that should cause forecasters to pause.
Biden-themed products are not selling well on Amazon in comparison to Trump. In some categories, his products barely sell at all.
Liberal-leaning sites are not receiving the rush of traffic that you would expect in the case of an upcoming landslide electoral victory. In fact, conservative biased sites have seen their rankings on average improve while those of liberal biased sites have fallen over the previous months.
Our Social Media Influence (SMI) cohort analysis shows conservatives doing much better than liberal accounts. This includes both ‘activist’ type accounts as well as traditional news/opinion based accounts. We also see conservative accounts run by African Americans and Hispanics as doing extremely well. The fact that these demographics have historically voted Democrat is another red flag.
There is enough evidence in our alternative data universe to conclude that this election is not the slam dunk that most forecasters believe. The +90% probability forecasts of a Biden victory, we believe given the aforementioned issues, should be ignored. This election is much much tighter than the consensus believes.