We are finally here. Election Day.
What does big data and Aunty AI have to say about the US Presidential Election 2020? Will Trump win a second term? For the precursor to this story to date, here is my previous prediction which sets out our methodology.
So, are the polls right in saying that “Biden is in pole position” (according to the Economist, the global publication, the other economist)?
Pole position off poll data is like sticking women on boards and ticking off diversity: doesn’t mean a thing.
So let’s use some real data and real machinery to analyse this sudoku.
Overnight, our prediction team across the world have collaborated to pull together this cutting edge analysis. We mustered every bit of technology computing firepower (including an army of servers) and human brainpower to analyse petabytes of the world’s open-sourced online data to measure human behaviour and engagement of the US market.
I know you want to know the answer right now. Here we go.
Revisit: Affect over time
In the last report, we saw a late surge of positive affect (emotion) towards Biden in mid-October. Until that point, he was just ‘the other guy’.
So did Biden hold in the last 2 weeks?
The Affect Iceberg shows Biden maintained positive affect for the first week (I suspect in line with Obama coming out in full force on the campaign trail while Biden was largely sidelined. Did you know that Obama paid more taxes as a teenager scooping ice cream at Baskin Robbins than Trump in his entire tax-non-paying career? ).
You can also see the affect tide quickly shift back to Trump around the time of the 3rd presidential debate. Trump knows how to work TV and drama like a reality TV star. Wait, he is a reality TV star.
So the chart shows that in the last few days, we see affect swing back to Biden. Is this making you dizzy?
Notice the height of positive affect and depth of negative affect increased for both candidates. Things are hotting up. The market is getting invested in this. The election is happening.
But, the Affect Iceberg view is no longer useful or predictive.
Let’s revisit our emotion wheels.
Revisit: Trump vs Biden (English speaking market)
A quick measure of emotions of the market the day before Election Day…
Wow! Overall, the English speaking market collectively does not like Trump anymore!
Trump’s disapproval has intensified in the last 30 days (refer to my previous article for prior charts). This emotion wheel sends a clear signal that overall, the US English speaking market intensely dislikes Trump.
Meanwhile, Biden gained ground and demonstratese overall positive emotions (flipped from overall negative a month ago). The strength of the green shows an intensifying positive.
At this point, one could call Biden as winner – at least for the popular vote.
But remember – this is not how this game works. Dig deeper. Another lens, please.
Revisit: Trump vs Biden (Spanish speaking market)
In 2016, Clinton surprisingly lost the election despite all major polls predicting Clinton to win with the support of Hispanic and Latino voters who traditionally voted Democrat. Everyone got it wrong because they were using the wrong data and the wrong method.
Don’t measure the sky with a ruler. There are better ways to read the stars and see the future.
Using the unbiased method of AI, we measured the Spanish market view in 2016. These charts are too good to not share again. You can see, Clinton saw red and did not get the Latino vote. She lost, contrary to the world view based on polls.
So for 2020, it was prudent to consider the US Spanish speaking market view again. We can’t listen to polls anymore or predict the future based on opinions and historical assumptions.
The world has changed and what people deeply care about has also changed.
Latinos are the largest minority in the US with an estimated 60 million people, of which 32 million are eligible to vote – that’s 13% of all eligible voters.
Polls say “68% of Latinos are likely to vote Trump”. But are they telling the truth? Surveys, polls and focus groups are biased opinions. Voting is very personal, sometimes a secret shame for Trump voters. I suspect this will happen again in 2020.
Again, polls don’t mean a thing.
So let’s do this the right way using market data and AI. What is the Spanish view of Trump and Biden 24 hours before Election Day?
Now, look at that! This tells a very different story to the English speaking view, doesn’t it? Latinos dislike Biden more and are likely to vote Trump.
My key takeaways:
- Trump has stronger expectation (blue) – good for Trump
- Biden has stronger negative emotions (red) – fear, dislike and anger – good for Trump
- Both are not intensely liked or disliked
Are you confused?
Not to worry, there’s a logic behind the way we analysed this: in layers, like peeling the onion. Let’s hope the result won’t make you cry.
This read on Latinos will be important to note later. Latinos favour Trump.
So now enter the main arena where the real game of prediction begins.
Maven 2020 Methodology: World first way to predict the US Election by Battleground States
Unbiased, predictive, without asking a single person a question.
Thousands of content pieces (petabytes of data) were analysed with natural language processing to measure market engagement and affect. Over 400 emotions are analysed, mapped and charted. Over 140 narratives were analysed over 10 wo-man hours.
We measured Trump vs Biden along multiple dimensions:
- Engagement – what did people really care about?
- Affect – was there movement or momentum in a positive or negative direction?
- Emotion – how did people feel? And how intensely?
These dimensions were analysed for Trump and Biden against the Top 6 issues discussed in this election – consider these as the attributes of choice in voter’s minds:
- Law & Order
This was a multi-dimensional, multi-lens, very comprehensive view of decrypting a very complex soduku. We looked for common trends, themes and relativities. Where results were ‘too close to call’, we went deeper, and then made a call.
Let’s start the engines. Grab your coffee.
PREDICTION STRATEGY: Trump vs Biden by the Battleground States
There has never been an election as volatile and unpredictable as 2020, as we can see with the shifts in emotions and engagement over such a short time.
All elections are decided on the Battleground States. We outline how we arrived at our overall prediction, piecing together a state by state view.
The result is going to be very clear.
Here is a summary of the Battleground States in order of importance. This is how this game of monopoly, in my view, will be played out:
For those who hate tables, I’ll walk you through the highlights of our thinking.
1 FLORIDA is KEY
Florida is one of two Battleground States that has a high Latino population (who we assessed to favour Trump). It counts for the second largest number of votes after Texas, but most importantly, it will be one of the fastest states to be counted. If Trump wins Florida, he will pull all legal avenues to call the election early and stop counting. He will tip the monopoly board upside down. Florida once again may play a key role in history ala 2000 Bush vs Gore.
So let’s examine Florida closer.
The top 2 issues in Florida are Jobs and Covid. Trump has the strongest engagement overall in Florida and is favoured in both attributes.
Check out the emotions for each attribute side by side…It’s pretty clear, isn’t it? Biden does not engage Florida on the key issues that matter to this state.
- Jobs are very important to Florida, including the Latino population, who favour Trump
- Trump is very strong on Jobs – high positive emotions and expectation (this is his home state so I would expect this to be strong)
- Despite Florida recording 4111 new cases per day, up 22% from last week, it appears that Trump’s swift Covid recovery reflects favourably. We see high expectation (blue) as the dominant emotion (this is a very strong signal).
Trump is also very strong on Law & Order, while Biden is intensely negative.
Maven Prediction: Trump to win Florida. Note: Polls tip Biden to win by a narrow margin.
2 ARIZONA is CLEAR
Arizona mirrors Florida with Jobs and Covid as the top issues deciding this election. They also have a high Latino population who favour Trump.
Biden wins on Jobs.
Trump has generally mishandled his response to Covid and in Arizona, this matters a lot. The emotion wheel says it all for Trump on Covid with high negativity as Arizona records 1330 new cases per day, up 33% from last week. We assessed that Biden on Covid was irrelevant in Arizona, so overall Biden wins on Covid by default.
Maven Prediction: Biden to win Arizona.
3 TEXAS is WEIRD
Texas is the largest Battleground state with 38 electoral votes, historically Republican. This week we heard disturbing news that Trump tweeted “I LOVE TEXAS!” with a video of a Biden campaign bus being swarmed by Trump supporters, trying to ram it off the freeway.
This is a key Battleground State where we measured Law & Order as a top issue for the election. On Law & Order, we see Biden in favour.
In Texas, Jobs and Leadership are also top issues for this election. On Jobs, it is too close to call. On Leadership, Trump is the clear leader in favour. I’m calling out strong ‘indifference’ and ‘lack of concern’ in Texan voters towards Trump towards concerning his Leadership.
Interesting side notes: Texas recorded the highest daily average of new Covid cases in the nation at 6401 per day, up 18% from last week. Yet, Covid was not an election issue. Is there a correlation here? In this case, ignorance is not bliss.
Maven Prediction: Trump will win Texas.
4 ALL BATTLEGROUND STATES
To prevent death by emotion charts, going through state by state, I have summarised the rest of the analysis below – you are most welcome!
We consistently measured Jobs and Covid as top issues in this election.
Based on this analysis, we conclude that Trump will win most Battleground States, and hence we conclude that he will win the US Election 2020.
We assess that Trump is favoured on all issues except for Values (surprise surprise). It is clear from looking at the details that Biden -‘being the other guy’- is not enough. Biden has not sufficiently engaged the market on the key issues that matter. Biden hasn’t done enough, or at least, enough that mattered.
Final word: This onion made me cry
What the weeds uncovered…
Wading through the content that formulated these aggregate measures of engagement, emotion and affect, it was very clear that this election is not focussed on the future but on the now. It is disconcerting that not much content reflects a debate about either side’s plan for the future – there appears to be no plan for the future, or at least the market is not finding this content most engaging right now.
This campaign has been riddled with distractions and drama, which took everyone’s attention away from the reason the highest office exists in the first place – to serve the people of the nation with policies and plans that are relevant and matter to changing people’s lives.
Understandably, the US is in a crisis with Covid out of control. Jobs and Covid are clear election issues for 2020. There is no time for idealism, it’s just survival.
Voting is very personal. For many, exposing the truth about who you will vote for can be scary (particularly for secret Trump supporters). Voting is a deeply personal question – one that we should respect. We are not in their shoes.
It’s easier to analyse a situation from the other side of the world than to live in it every day. It’s evident that the data bears disappointing news. Trump has a strong chance of winning this election, contrary to all major polls and the wishes and hopes of millions of people in the US, and the world at large.
Trump will win in 2020.
I hope I am wrong, but I back the data, our methodology and our work.