Exit & Opinion Polls India

 

Just a week remains for election day. The question is how pollsters would fare on result day?

As for Clinton, a general perception has been created that Biden this time is going to win because his national polling lead is just really big. In 2016, while the overwhelming majority of these polls fell by the wayside, only a handful of polling agencies, all dismissed as outliers, made the accuracy grade - among them IBD-TIPP, Trafalgar and Rasmussen. IBD-TIPP in fact had been the most accurate poll right through 2000 to 2016 and taken most seriously as a pollster. Rasmussen though who did well in 2016, a time series analysis of their performance shows that they tend to be erratic in terms of accuracy. Trafalgar made waves in 2016 hitting the bullseye in key swing states like Pennsylvania and Michigan where Trump flipped from blue to red  by a wafer thin 0.7% lead. They also distinguished themselves very well during the 2018 mid polls.

The latest IBD-TIPP survey finds Trump trailing Biden 7.2 points, 51.6%-44.4%, in a four-way presidential poll of likely voters. Their tracking poll highlighted that some key features remain the same over time - "Biden's lead in the suburbs has remained solid; Trump has lost a significant share of his 2016 voters; and Biden holds a clear advantage among those who voted for a third party or didn't vote in 2016."

Unlike 2016 wherein IBD-TIPP had been an outlier which predicted a possible Trump victory, their numbers this year are practically more mainstream - within the range of the RealClearPolitics (a poll aggregating site) average. Trump needs to be at least within 3 points in the national popular vote total to have a shot at victory. The IBD-TIPP survey indicates Trump lacks national momentum to close the gap viz it just looks incredibly unlikely that anything dramatic is going to happen over the next week that could radically change the outcome. If this is so Biden could be expected to register a landslide victory.

However IBD-TIPP also offer the following caveats:

"President Trump has a clear advantage when it comes to voters' intensity of support. Among those who intend to vote for Trump, 78% back him strongly, vs. 65% among intended Biden voters.

The IBD/TIPP presidential poll finds that an equal 8% of Trump and Biden voters say they're at least somewhat likely to change their mind before Election Day.

Meanwhile, a matching 38% of voters expect each candidate to win. Yet 48% think most of their neighbors will vote for Trump, and just 35% think their neighbors will mostly back Biden. The big difference comes from rural areas, where 71% say their neighbors back Trump."

It's the last paragraph that rings an extremely jarring discordant note. If 2020 is indeed a wave election for Biden then why are majority of likely voters inclined to believe that their neighbors are most likely to vote Trump rather than Biden? The Democrats besides cannot hold a candle to Trump's ability to enthuse his base. His rallies are attended by tens of thousands in key swing states while Biden-Harris virtually struggle to get even a hundred voters for their rallies. In key swing states, Republicans have outnumbered Democrats in voter registrations. After weeks of Democrats outvoting them by mail, Republican voters stormed early voting precincts in person this week, taking large bites out of their opponents’ historic lead in pre-Election Day ballots. This has practically unnerved the Democratic Party who initially was betting the record early turnout gave them a winning edge. This edge looks now highly suspect.

 

Are there something amiss in the polls? Can Trump repeat an upset win a la 2016 again?

The false optimism of the 2016 Clinton win could be traced to a flawed way of thinking about how state-level races relate to one another, and a misperception about the state of the Electoral College -  factors still prevalent. State Polls are a better predicator of Electoral Votes (EV) than popular votes. So let's look at the latest outlier polls in the key swing states. This table is what we can find:

In 2016, surveys that showed Trump winning were ignored by most media outlets or summarily dismissed as partisan polls.

If we go by outlier polls as shown in the above table, particularly those polls which excelled in accuracy in 2016, what we find is that Trump is very much in play with the exception of perhaps Minnesota. So what are the significant methodological differences between these outlier polls and mainstream? These are two broad radical differences.

1. The mainstream polls tend to oversample Democrats

2. The outliers mostly factor in the "Shy Trump Voter". Trafalgar for example adjusts their polls for a "social desirability bias" effect, the hypothesized tendency of some voters to calibrate their responses to polls towards what they believe the survey taker would like to hear. Trafalgar uses a proprietary model of calculating a polling sample that takes into account the social pressure that leads many people to hide their support of a controversial, polarizing candidate like Trump, even from anonymous pollsters.

As in 2016, these two factors remain pertinent which basically explains contrasting predictions for Trump by polls. What if we go by outlier polls? The table below illustrates the results:

We see a neck-to-neck race, Trump with around 244 EV and Biden around 222 EV, both short of the magic number 270 EV to clinch the Presidency.

There are 5 tossup states vis Arizona, Michigan, North Carolina, Pennsylvania and Wisconsin accounting for 72 EV. Trump is relatively better positioned in Arizona (11 EV) and New Carolina (15 EV). If he manages to take these two states - he is through with 270 EV. However, to hedge any unexpected reverses in the red states list, he needs a buffer. Pennsylvania with 20 EV accordingly becomes a must win for him since it offers him much better chances to win than either Michigan or Wisconsin.

Biden, on the other hand needs to must win not only Pennsylvania with 20 EV and Michigan with 16 EV, but also 3 out of 4 remaining tossup states. This looks an uphill task for him based on projections of outlier polls.

 

 

The criticality of Pennsylvania can be brought home by both Trump and Biden repeatedly visiting the state in the home stretch. Besides the state post elections are most likely to be a flashpoint that could lead to nasty and prolonged legal fights over who won the presidency for the following reasons:

1. The ongoing legal chaos around how to vote leading to confusing voters.

2. Prevention of the mass rejection of ‘naked ballots’ has become a focus of voter advocates

3. Voters won’t be guaranteed an opportunity to fix other discrepancies with their mail-in ballots

4. An expected GOP push to stop the counting when results take longer than usual to report?

5. An expected increase in provisional ballots could provide fuel for baseless fear-mongering.

The huge Electoral College bias is why coverage of the 2020 race can end up giving you whiplash — either the polling looks like a boring landslide for Biden or else it’s an exciting toss-up, with nothing in between. Added to this, post poll uncertainties also rules high. So grab your can of popcorns and stay glued to your television sets from November 4th. We are likely to witness drama as never before scripted by US Democracy!