Disclaimer
A number of points are relevant to consider when looking at these forecasts. Our forecasting method is proprietary, uses alternative data as inputs, and has not been vetted by academic study or peer review. So, you should use it at your own risk. Having said that our track record is enviable, but past success does not guarantee future results.
Additionally, the current environment should be considered. This election is so different in so many ways that correctly forecasting its outcome is considerably more difficult. Take this forecast, as well as all others, with a grain of salt.
Consensus
The consensus clearly states that Biden will easily win, with a probability of around 90%. In addition, forecasts based on traditional data and models fall into a tight range, implying that expectations have clustered. Consensus seems to be saying, with high probability and a tight range of outcomes, that there is little risk involved in this election.
Another point of interest is that polls have been more consistent in this election cycle than in 2016. In other words, they have shown Biden having a sizable lead throughout the race – which also implies lower risk.
Our Data
We look at what people are doing on-line. We believe that, in an on-line heavy world, trends and underlying conditions are more easily identified when looking at on-line data than relying on traditional methods, most of which were conceived of prior to the advent of the Internet.
Social media metrics, and our proprietary Social Media Influence (SMI), have formed the base of our analysis. For the 2020 election, we expanded to incorporate cohort analysis as well, such that we rate +1,000 politically relevant accounts, in order to better understand underlying conditions instead of focusing on just the candidates.
We have also increased our coverage of other alternative data sources for this election. Some included sources are search data (Google and Bing), page views (Wikipedia), site rank (Alexa), and on-line shopping (Amazon).
Differences Between Alt and Traditional Approaches
Under normal circumstances, our systems tend to identify trends prior to traditional ones. In this sense, our systems have an inherent advantage in the early identification of trends.
Another main area where an alternative data approach would have an advantage over traditional ones is when there is considerable contention around a topic, event, or election. The contention would need to be so great that poll respondents may alter their answers. We could currently be in such an environment but can only confirm post-vote.
Big Picture View
Polls and traditional analysis basically say that the country has moved left. Polls clearly state that Democrats have gained traction since 2016, and more specifically during most of 2020. Real world events seemingly support this view. The fact that polls forecast a Republican incumbent losing the election and the Democrats likely taking control of the Senate/House as well makes this shift exceptionally clear. Such a drastic change in government without the entire country shifting left would be highly unlikely.
Our data, however, tell us that the country has moved to the right. This goes beyond the topic of Trump or the election. We note that conservative-leaning topics, causes, and news-sources have generally outperformed.
The Fox News versus CNN rivalry is a microcosm of this battle. 20 years ago, these news stations were more or less middle-of-the-round. Now, they so lean in one political direction that they not only attract certain people but repulse those with opposing views. The fact that Fox News has gained traction in front of CNN supports the concept that the country has shifted right. This same pattern has repeatedly shown itself in other areas.
Competing Versions of Reality
The large divergence between traditional data (Biden will easily win, Democrats will take both House and Senate proving the leftward shift of the country) and our alternative data (Trump will win with the country shifting slowly right) portrays two completely different versions of reality.
Which Set of Forecasts or ‘Version of Reality’ is Correct?
We will only know after Election Day.
However, we believe after sifting through mountains of data and also reading through traditional polls and commentary that the key factor is in information flow.
Every election seems to have a key to it. Something that is often ignored during the election but then becomes common knowledge afterwards.
In 2020, we believe this key could be the performance of information providers, for lack of a better term.
They are politically biased providers of news, opinions, and general information. The lines have blurred to the point of making a major media organization similar to an anonymous (or at least not verified) social media account, in the eyes of those wanting to consume a political viewpoint.
Someone looking for left-leaning commentary might tune into CNN or MSNBC, read a blog post by the Palmer Report, follow a social media account of Bernie Sanders or AOC, or watch a Biden campaign video on YouTube. To them, these events would be more or less equivalent.
The same can be said for conservatives, except swap out CNN for Fox News, read Breitbart instead of …
The sense that lines have blurred between these vehicles (TV, social media, newspapers, on-line video, blogs, etc.) makes the echo chamber larger than previously thought.
Additionally, if people increasingly see these as interchangeable we are no longer talking about silos but about the same phenomenon.
What we have found, through various cohort analyses, is that the universe of right-leaning information providers have outperformed their left-leaning equivalents.
This implies, though does not prove, that the right-leaning echo chamber is growing faster than its peer.
If we are correct in our forecast and Trump ends up winning, we believe others will point to the strength of the right-leaning echo chamber of information providers (or however they describe it) as being the key element in deciding the election.
One of the main points that leads us to believe that our forecast will end up being correct (or at least closer to correct than the consensus) is that we highly doubt Biden could win in a landslide if the right-leaning echo chamber of information providers has better growth.