US Presidential Elections: Déjà Vu 2016. It's not over till it's over!!

This tweet was gleefully retweeted by Trump. Nate Silver and his website 538 used to command a status of a rockstar within the polling industry after he predicted all 50 states correct during the 2012 US Presidential Elections. However, in 2016, he gave Trump less than 25% probability to win and ended up an international laughing stock. His iconic status almost washed away overnight!

According to Silver's recent admission, enthusiasm for Trump only got that much stronger in 2020! Two nights ago post his illness, Trump conducted a campaign rally in Johnstown, Pennsylvania and before that in Sanford, Florida where a sea of reds braved the COVID pandemic to greet him, many without masks and showing an utter disdain for social distancing norms. In Johnston, thousands of his supporters additionally defied rains to participate in his rally. Seeing the crowd numbers and enthusiasm, Trump appearing in very high spirits, broke out into a dance whose video went viral! Since US Labour Day 7th September, over an estimated 250,000 have attended Trump's rallies as compared to just a couple of hundreds for Joe Biden-Kamala Harris.

Meanwhile Biden and Kamala Harris rallies are marked with extremely poor crowds! Arizona a state where the RealClearPolitics (RCP) average is 2.7% in favour of Biden and having 11 Electoral College Votes serve to illustrate the pathetic degree Joe Biden and Kamala Harris are reduced to, virtually struggling to turnaround public apathy. A couple of days ago, they visited the state for a joint rally but found to their sheer embarrassment not a single person turned up except for media personnel. The Democratic leaning media swung into damage control by spinning the headlines to "nobody turned up because the public was not invited". But the moot question is - how did the media turn up if there was no invitation? So did Biden-Harris jointly fly to Arizona just to address a press conference?

At the time of writing, with less than 3 weeks go to election, the RCP average indicate that nationally, Biden leads Trump by a whopping 9.8% margin. The latest ABC News/Washington Post poll suggests Biden leads Trump by a 55% to 43% margin among likely voters. In the 21 previous presidential elections since 1936, there have only been only five challengers who led at this time. Of those five, only one (Bill Clinton in 1992) was ahead by more than 5 points. None of those five were registering more than 48% of the vote in the polls. In other words, if polls are to be believed, Biden is the first challenger to be above 50% at this late juncture in the campaign. With that kind of spread, election maths suggest Biden should win a landslide, a claim Democratic surrogate media like CNN is making, constantly repeating like a parrot.

However, it's not merely the contrasting size of rally crowds which respectively dot the Trump and Biden rallies alone that jars the perception of the election being already done and dusted. Consider this. If Joe Biden is actually whipping up a Tsunami Wave, it's totally bewildering why the Gallop Poll found 56% of voters feel Trump will easily romp home to just 40% who felt Biden would do so.  Leading bookmaker Coral is offering 7-4 outsider odds for Trumps but Biden remains odds on favourite at 4-9. However 75% of bets placed with Coral is on Trump! 

The same poll found that 56% of voters feel that they are much better off than in 2016 even amidst a pandemic and job losses! The comparative numbers for George Bush was only 47% (2004) and  just 45% for Barak Obama (2012). If that's not construed a massive approval rating, what is?

US elections are usually won or lost on the economy and on the economic front  the DowJones is on fire! The return for the S&P 500 index SPX over the past three-month period thus far has risen by about 7% since Aug. 3. That equity-market performance measure has been the best predictor of U.S. presidential elections since 1984, proving 87% accurate since 1928, LPL chief financial market strategist Ryan Detrick notes.

What's even more confounding is Nancy Pelosi, Speaker of the House of Representatives, calling a rushed press conference to announce that the Democrats were immediately initiating yet another attempt to impeach Trump. If Democrats were indeed happily cruising to a landslide win in less than 20 days as the likes of CNN their media surrogate wants us to believe, what was the need for impeachment, particularly when they lack the requisite two third majority in both houses to succeed in this task?

If at all the Democrats had succeeded in weaponising polling and media coverage with the intent of creating a public perception of a Biden landslide, anecdotal evidences instead appear to point to the miserable failure of this objective. As of now, the expected "bandwagon" or "contagion" effect as an outcome of such tactics is no where readily detectable!

Why this dichotomy? 

The question of polling accuracy came to a head last week after CNN published a national survey showing Biden with a whopping 14% lead - 55% to 41%

As in 2016, the Trump Campaign's typical refrain is that the polls are underestimating Trump's appeal. The Trump Campaign retweeted John Mclaughlin (one of their campaign pollsters) opinion that the survey was "skewed" against the President! The Trump Campaign followed this up serving a legal notice to CNN to withdraw this survey.

In the meanwhile, a Washington Post analysis found that Trump had not led a national poll in the RCP database since February this year and that polls after this period showed the momentum for Biden and against Trump.

But with increasing questions on the accuracy of polls being asked, the Democrats while conceding statistical errors are often unavoidable for surveys, argue that a double digit plus national lead is large enough spread for Trump to be guaranteed a denial of second term. Sampling errors like 2016 won't be enough to overcome it!

Is the US polling biased?

The study commissioned by the American Association for Public Opinion Research (AAPOR) to look at the 2016 polling debacle noted that while the polling trend line since 2000 show no political partisanship, they indeed exhibited a marked bias against Trump in 2016.

"For a given election, whether the polls tend to miss in the Republican direction or the Democratic direction appears random. While U.S. pollsters may be guilty of pointing to the wrong winner on occasion, as a group their work does not reveal any partisan leanings.....

In 2016, the average signed error in state polls was 3 points, showing an over-estimation of support for the Democratic nominee..... The polls in non-battleground states under-estimated Trump’s vote margin against Clinton by 3.3 points on average (signed error); the under-estimation of Trump’s standing was 2.3 points in battleground states.....

Wisconsin polls exhibited the largest average signed error (6.5 points), with polls there showing Clinton ahead by between 2 and 12 points in the final two weeks before she narrowly lost the state (47.2 percent to 46.5 percent).

Ohio polls also under-estimated Trump’s margin by a substantial 5.2 points on average, indicating he had a small lead, though he went on to win the state by eight points.

Polls in Minnesota, Pennsylvania and North Carolina each under-estimated Trump’s margin against Clinton by an average of four to five percentage points, while polls in Michigan and New Hampshire under-estimated his standing by 3.5 percentage points on average. Under-estimation of support for Trump was smaller in Florida, Arizona and Georgia."

As in 2016, there now exists relatively a visible divergence between national and state polling which are particularly evident in the battleground states. What did in Hillary Clinton was such a divergence obscured her under performance in battleground states for wich she tactically paid a heavy price.

Yet, in the closing days of the 2016 election cycle, much of the elite media created and promoted the illusion that then-candidate Donald Trump had little hope of winning the presidency, and would likely lose this election badly. Investor Business Daily observed:

"NewYorkTimes columnist Jim Rutenberg seemed to acknowledge the unprecedented opposition directed towards Mr. Trump from members of the press: "If you view a Trump presidency as something that's potentially dangerous, then your reporting is going to reflect that. You would move closer than you've ever been to being oppositional."

Accordingly, with the news media being so antagonistic in their coverage of Mr. Trump, it should not be surprising that their own polls would be more favorable to his opponent. In order to create a feeling of malaise among Trump supporters, and to discourage them from voting, the media used their polls to promote the narrative that Mr. Trump could not and would not win the presidency."

Two major factors that led to polling missing their mark as identified by the AAPOR study had been oversampling of Democrats and the Shy Trump or lying factor. While the degree of oversampling of Democrats have reduced, its practice is still widely prevalent in 2020. Further a study by Cloud Research found 11.7% of Republicans and 10.5% independents said they would not give their true opinion, vs. 5.4% of Democrats who said they were unwilling to disclose who they are actually voting for. Because the media is constantly demonizing and caricaturing Trump and his supporters, and with Hillary Clinton describing them as "basket of deplorables," still ringing in their ears, Trump supporters would be reluctant to admit their support to strangers. Polling agency Trafalgar found that voters responding to "robocalls," consistently registered support for Trump 4.5 percent higher than when they were talking to a live pollster. By adjusting this weightage to their data, Trafalgar catapulted as the most accurate pollster in 2016!

The third factor perhaps the most significant that could have a perverse impact on the results is turnout. In 2016, the Black voters turnout relatively declined as compared to 2012 while those of rural white, less educated, blue collared voters, particularly in the Rust Belt states relatively increased. This tipped the scales for Trump to the shock of Clinton.

More than 5.6 million people have already voted early in the presidential election, vastly exceeding the pace of 2016. Democrats todate have outvoted Republicans 2.4 times. 

Florida has had the most early voters (948,000), followed by Virginia (770,000) and the battleground states Michigan (637,000) and Wisconsin (577,000). South Dakota has seen the greatest percentage of increase in voters in relation to 2016 overall turnout. Twenty-three percent of South Dakota's 2016 turnout has voted early. That's followed by Virginia (19% of its 2016 turnout), Wisconsin (19%) and Wyoming (15%).

This however does not necessarily give Democrats a big advantage. While 70% of Democrats prefer early voting, a similar percentage of Republicans prefer voting physically on election day. The disadvantage of mail in ballots is that they are more vulnerable to be rejected or getting lost or missing the deadline due to postal delays than physical voting. Additionally, with Biden-Harris preferring to remain in their bunkers,  Trump set himself a punishing campaign schedule visiting swing states, Trump accordingly possess the advantage in the homestretch of being able to swing fence sitters and also enthuse his base to vote, even to prod them to take to early voting.

Trump's poll slide is over. He is now on the roll

Biden leads the national polls massively but state-level data shows his victory is by no means guaranteed. There are three approaches to predict the outcome of November 3rd elections. The first is through statistical forecasting models, the second by polling data or surveys and thirdly through using an Artificial Intelligence model. All approaches suggest that though the windows for Trump to win the elections is very much narrower than Biden, they are still wide open. And all these approaches further suggest either Trump will romp home easily or an error margin away from the Presidency

FORECAST MODEL

Political scientist Prof Helmut Norpoth who teaches at Stony Brook University, in New York State evolved the Primary Model, a statistical representation of U.S. presidential races based on data going back more than a century..

The Primary Model has correctly predicted five of the past six presidential elections, and when applied to previous elections, correctly predicts an impressive 25 of the last 27, missing only the 2000 election in which George W. Bush defeated Al Gore and the 1960 election in which John F. Kennedy defeated Richard Nixon — two extremely close and contested votes marred by allegations of voting inaccuracies.

The model utilizes data from presidential primaries to help predict who will win. Accordingly, Norpoth predicts Trump would win a landslide this November with almost 100% certainty!!

POLLING METHOD

One of the reasons why the RCP averages are skewed in favour of Biden is because till last month mostly Democratic leaning polls were being published. It's only from September the Republican leaning and more independent leaning polls started kicking in.

Anti-Trump bias has not disappeared but appeared to have only grown even stronger in 2020. In 2016 it was the 'outlier' polls from agencies like Trafalgar, Rasmussen, IBD-TIPP etc who were invariably lower ranked pollsters and considered mostly leaning Republican that topped for poll accuracy. In particular Trafalgar made waves for their methodology though ranked only C grade and to a lesser extent IBD-TIPP, also ranked as a C grade pollster. Invariably, the ranked pollsters fell by the roadside and was subjected to public ridicule. However, they are back in 2020 to figure prominently in the RCP basket of averages, both nationally and the state. However, Trafalgar ranked as the best pollster by RCP during the 2016 elections do not find a place in the RCP basket of national average though considered for state polls.

Found below is a table that provides the latest RCP polling averages state wise, compared to latest polls

These 14 states account for 212 Electoral Votes, out of which polls are indicating that Biden leads in 13 states accounting for a whopping 174 Electoral College Votes. This leaves Trump leading only in a single state with 38 Electoral votes.

But what happens if we factor in the latest state polls?

a). Except in North Carolina where Biden increases his RCP average, the overall momentum is against him and for Trump. The increase in North Carolina besides is not statistically significant.

b). Factoring the latest state polls, Biden now leads in only 7 states with 78 Electoral College Votes. Trump also leads in 7 states but could corner as much as 134 Electoral College Votes in these swing states.

c). Further even in the 7 swing states leaning Biden, Democrats are strongly leading in only 3 states viz Virginia, Minnesota and Hampshire which together adds up to only 27 Electoral College Votes. Trump requires less than 3.6% to swing all these 7 states. This is a very much a doable challenge particularly factoring in the bias towards Democrats inbuilt into these polls. If he does succeed, he wins by a landslide!

ARTIFICIAL INTELLIGENCE MODEL

Expert.ai's political forecasting model was successful in predicting the Brexit vote where pollsters failed.

"Trump was the focus of far more social media activity than Biden, accounting for almost 60% of all the posts . Expert.ai analyzed, compared to slightly less than 17% for Biden.

But Biden ranked higher in terms of positive emotions such as "success" and "hope," while Trump scored higher on negative emotions such as "fear" and "hatred."

The only positive emotion on which Trump scores better than Biden, according to a statement from Expert.ai, is "action."

It's not over till it's over

It's unwise to put faith in the Biden-Harris ticket just because of the double digit leads the national polls are indicating for them. Such massive leads have disappeared in 1948, 1968. 1976. 1988, 2016. Even otherwise, Joe Biden's leads in the key swing states are very slim. Even Democratic Party heavyweights like New York Governor Andrew Cuomo doesn't trust these leads in his recent media interview. So did the Obama and Biden Campaign advertisements just released.

To win Trump would only need to beat Biden in three of these six states: Michigan, Pennsylvania, Wisconsin, North Carolina, Florida, and Arizona. That looks very much doable!  Like in 2016, Trump remains only an error margin away to victory. In 2004, 2012 and 2016, we have seen late swings over 5%. That's a Democratic nightmare if it happens again.

In short, though these maybe all contrarian projections, and whether you like or dislike Trump, do not write off him just yet!