…Joe Biden. Barring Donald Trump succeeding in his pending legal challenges, the former VP will become the 46th president of the United States. Mr. Trump has not yet conceded to becoming only the second incumbent to be dispatched (George H. W. Bush was defeated by Bill Clinton in 1993) in the last 40 years. And one cannot ignore the irony of President Trump grasping at technical legal straws in an attempt to remain Commander in Chief. Nonetheless, we have gleaned some extremely pertinent information about this election – far beyond Trump vs Biden - and what it means to our nation moving forward.
Despite his defeat, Donald Trump performed materially better than most polls predicted; the same thing happened in 2016. This merits the question, is the reason most polls failed to have predictive value because of faulty surveying methods, that certain pollsters were biased, some combination thereof, or something entirely different?
To that end, let us begin with a top-down approach and examine the popular vote. The final polls conducted before the election projected the following: USA Today had Biden ahead 52% to Trump 44%. The Morning Consult predicted the same result, 52% to 44% for Biden. Quinnipiac University had the former VP in the lead by double digits, 50% to 39% for Trump. You Gov expected Biden would beat Trump 53% to 43%. The Wall Street Journal/NBC News also foresaw Biden winning by 10 points.
FiveThirtyEight, a statistical aggregator of everything from sports to politics, compiled a weighted average of an array of polls they ranked from “A” to “D” based on several proprietary factors. Despite their more robust process, even FiveThirtyEight’s model showed Biden prevailing by 8.4 percentage points. At this juncture, almost every ballot has been tallied. Biden (presumably) won, but only by ~3%.
FiveThirtyEight repeated its aggregative process at the state level. In some instances, its model was satisfactory, in others, it suffered shortcomings. In AZ (Biden 48.7% - Trump 46.1%), GA (Biden 48.5% - Trump 47.4%), NC (Biden 48.9% - Trump 47.1%) & OH (Trump 47.5% - Biden 46.8%) FiveThirtyEight was accurate. However, there were some big misses. In MI, the outlet predicted Biden would win the state 51.2% to 43.2% for Trump (Biden appears to have won the state by ~1%). In WI, Biden was expected to win 52.1% to 43.7% for Trump (Biden won by ~1%.) In NV, it was predicted that Biden would win the state by 5 points (Biden won by ~1%.) In FL, FiveThirtyEight modeled that Biden would secure the state by 2.5 points (Michael Bloomberg spent 100 million dollars in Florida to help the former VP secure a victory. Trump won by ~3%). In TX, FiveThirtyEight foresaw Trump winning the state by 1 point; he won 6.
One poll that did prove to be accurate, especially in contested states, was Trafalgar Group, which predicted that Trump would win FL 50% to 47%; lose WI by 1 point and lose in MI by 2 points. Accuracy does not appear to be an anomaly for Trafalgar Group. In fact, it was one of few polling outfits to correctly call the outcome of the 2016 presidential election, predicting a Trump victory. Their prize? Hate mail and a “C-“ grade by FiveThirtyEight. We wonder then, what grade FiveThirtyEight should bestow upon itself.
Regarding the legislative branch of government: in the upper chamber, a plethora of polling outfits predicted that Democrats would flip the Senate, leaving them in control of all branches of government. Instead, despite having to defend ~twice as many seats, the GOP appears (there will be two runoff contests in Georgia on January 5th) to have held onto their slim majority. Specifically, pollsters were convinced Dems would unseat Susan Collins in ME (she won by 9 points) and that Jamie Harrison had a credible opportunity to oust Lindsey Graham in SC. Taking their cue from the polls, donors funneled almost 60 million dollars to Mr. Harrison to get him over the hump. It was the costliest senate race in American History; Lindsey Graham won by 14 percentage points.
In the lower chamber, most polls predicted that Democrats would add to their majority to the tune of between 10 and 15 seats and solidify their hold on the House of Representatives. The result: the GOP appears to have gained at least 5 seats. This week Democrats hosted a contentious call with party leaders to express their shock and displeasure.
At TQC, we find it peculiar that while almost all the polls were inaccurate, the skew was consistent: almost all polling outfits predicted Biden and Dems in both chambers of congress, would outperform. Conversely, almost none predicted that Trump and other members of the GOP would do better. Like weather forecasters, pollsters are proficient at continuously tweaking their models to achieve more predictive value. However, the same polls that overestimated Mrs. Clinton in 2016, did so again with Mr. Biden in this election. Given all the (supposed) resources allocated towards fixing what went wrong in ’16, only to arrive at the same place four years later, is disheartening, to say the least.
According to FiveThirtyEight, there is a small – but statistically insignificant - bias towards democratic candidates. On its website the organization argues, “Over the long term, there is no meaningful partisan statistical bias in polling. All the polls in our data set combine for a weighted average statistical bias of 0.3 points toward Democrats.”
From the information derived from the quantitative data we examined coupled with a holistic overview, our conclusion is that survey methods used by pollsters were deficient (more on that later in this post) and a major contributor to the polling mishaps, some pollsters possibly allowed (at least unconscious) bias to seep into what should have been an objective exercise; and that groupthink was a key factor in the polling snafus.
Trick Or Treat
Groupthink behavior in polling is analogous to a teenager’s behavior on Halloween. Despite the “street cred” a single teen might get from putting on a costume and “egging” somebody’s house all by himself, he (or she) is unlikely to do so; the consequences of being caught are too daunting. However, as a member of a group, that same teen is very likely to “egg” a house. He would not receive as much “street cred” for a successful operation but if he were caught, the consequences would be diffused among multiple participants.
In polling, even if certain pollsters trust their data, groupthink can permeate. For example: Let us assume that polling outfits A, B, C, D, and E run their respective models and predict that Joe Biden will win the state of Wisconsin by 10%, 11%, 12%, 1%, and 9% respectively. As we can observe, polling outfits A, B, C & E’s forecasts are harmonized with each other. Polling outfit D’s forecast is materially < the mean. The quant geeks behind polling outfit D are secure in their process; they have Ph.D.’s in multiple disciplines from top universities, computers at their disposal that can process a trillion calculations per second, and have tweaked their models from ’16 to adjust for the “shy Trump” voter phenomenon. Further, if their prediction is proven correct, they will receive “street cred” for their accurate, contrarian view.
Polling outfits A, B, C & E would be wrong, but the criticism lobbed at them would be diffused among a group of pollsters. However, the quant geeks behind polling outfit D are also human beings. The risk of an egg on their face and the utter embarrassment from being wrong – and alone – is high. To avoid that possibility, polling outfit D capitulates and changes their inputs so their prediction closely resembles that of the other polling groups. We think given the highly contentious nature of this election (and last election), groupthink played a major role in the polls being so off base.
In 1984, trustbusters forced AT&T to break itself up into eight separate companies. Before that seminal moment, the oligopolistic structure of the U.S. telecommunications business rendered the cost of making (and receiving a collect) phone call very expensive. Indeed, until the 1990’s, people had to “think” and ask themselves, is Jane or John Doe was “worth” calling?
Being on the receiving end of a phone call – so long as it was not from your cheap cousin calling collect – was a treat. After all, if somebody was going to pay a handsome sum to get in touch, they must have something important to share! It is no surprise then that in the mid 1990’s, ~33% of randomly selected people would pick up their land lines and responded to pollsters’ questions, a healthy participation rate that encompassed a vast cross-section of voters.
By the turn of the century, the cost of placing a phone call began to drop precipitously. In fact, using the phone has been one of few services in America that has seen both its absolute, and inflation adjusted cost, collapse.
Naturally, as the cost of making – and receiving – phone calls (including cell phone calls) dropped, an increasing number of them were placed. Soon, instead of being excited by the phone ringing, more and more people became annoyed by the constant occurrence. It is no surprise then that currently, less than 1% of randomly selected individuals pick up their phones and respond to pollsters’ questions, an unhealthy participation rate that is probably not indicative of a larger subset of the voting public.
In the digital age, analog polling methods are simply not as effective.
Despite Joe Biden’s (presumed) victory, generally Republicans outperformed in this election. Democrats did not live up to expectations. The “blue wave” many were expecting turned out to be a blue ripple.
The consensus envisions the GOP underpinned by old white men lacking college degrees. This view is no longer completely accurate. To the dismay and disbelief of many Democrats, the Republican party has been successfully courting Hispanic voters, including some first-generation Americans. Hispanics are the largest (~18%) minority America. More specifically, ~30% of them voted Republican in this year’s election. They were a big reason Trump prevailed in Florida and Texas. As The Economist so eloquently argued, “a more welcoming policy towards immigrants does not automatically translate into more votes from immigrants.”
Republicans also made headway among black men (less so among black women). Blacks make up ~13% of the electorate. ~15% of black men voted Republican this year. More significant than these static data points are the trends: both Hispanics and Blacks are slowly warming to the GOP. Given the changing demographics of America, this bodes well for the Republican party over the long term.
To that end, Republicans should cease wasting resources in a futile attempt to lower voter turnout via arcane technicalities in voting laws. Instead, they should channel those energies towards continuing to broaden their own base. Finally, to the depraved minority of Republicans who practice white-identity politics that advocate a stealth separation of the races via a general fear of being "left behind" and other terms purposefully so general that anybody who wants to, can identify with, stop! These sophomoric tactics are morally bankrupt and are doing a disservice to the GOP: the very people these fear mongers are targeting are some of the same people that will make up a large proportion of the GOP’s base in the future.
Democrats continued to outperform with Americans armed with college degrees. As was the case in previous elections, a disproportionate number of them voted Democrat. Democrats have done a good job courting this component of the electorate. As a result, college-educated Americans continue to desert the GOP. Keep in mind however, only 35% of American’s are college graduates.
In our view, Democrats are in the process of making a potentially cataclysmic mistake in too often assuming they have Hispanic voters in their back pocket. Indeed, it is unfathomable to us why some Dems expect a Latino, especially one who emigrated from a country where socialism proved to be ruinous, to fall in line with a party who’s most vocal and well-known members declare themselves to be (democratic) socialists. This is stunning and borders on an odious blend of elitism and arrogance.
Democrats must do a better job earning the black vote as well. Almost every major city in America afflicted for decades with corruption and violence has been run almost exclusively by Democrats – ones with answers for everything, solutions to very little, who promise their constituents bread, leave them crumbs and take most of the loaf themselves. Irrespective of past and/or current policies, these politicians have (for the most part) failed and continue to disenfranchise many black citizens they are entrusted to serve. If this continues, over time, more blacks will abandon the Democratic party.