There are just 22 days to go until Americans head to the polls (some are already doing so by mail and early voting). It’s a familiar story by now — just like Remain and Hillary Clinton were on course to win in 2016, so too is Joe Biden. According to ‘conventional’ measures, such as FiveThirtyEight (538) and Real Clear Politics (RCP), Biden is on course to defeat the incumbent by a substantial margin. In the past month, we have been greeted with headlines declaring, ‘The coming Biden landslide’; 538 has given Biden an 86% chance of winning the election. In most battleground states, the RCP average gives Biden a comfortable lead over Trump. If one were to only focus on the averages offered by 538 and RCP, you would be forgiven for believing that Trump’s chances of re-election are very slim.
However, there are some polling firms which indicate that the 2020 election is a lot closer than most polls are showing. Some of these, are, I believe, ‘outside the mainstream’ or outliers. I say this not to disparage them; nor does it imply that they should not be taken seriously. Just the opposite — they deserve to be taken seriously because they have an impressive record of predicting election outcomes.
So, let’s begin with identifying some of these polling groups. The polling groups which show that Trump is winning in the key battleground states are the Democracy Institute and the Trafalgar Group. Both firms can boast of an impressive track record when it comes to polling elections in the past. The Democracy Institute, for instance, correctly predicted that Leave would win the ‘Brexit referendum’ (its final poll put Leave ahead of Remain 50–47, with 3% undecided); they also correctly forecasted that Trump would win the 2016 election over Hillary Clinton.
So, too, did the Trafalgar Group. Unlike most polling firms, the Trafalgar Group recognised that some portions of the 2016 electorate were different to the ones in the past. For example, the Trafalgar Group believed that several states could break for Trump on account of a ‘shy Trump voter’ — those voters who supported or were likely to vote for Trump but were reluctant to express their views in public. In an interview shortly after the 2016 election, Robert Cahaly, chief pollster/senior strategist at the Trafalgar Group, believed that the primary reason why most polling organisations were wrong was because they were not polling ‘a segment of voters’ that had (re)entered into the electoral fold; in the words of Cahaly, ‘a lapsed voter’. Specifically, the Trafalgar Group discovered that there was a ‘Trump surge’ voter comprising of people who had not voted since 2006 (or even before this year) but were going to vote in 2016. As such, most pollsters were not surveying these ‘lapsed voters’ to obtain an accurate ‘picture’ of the electorate.
Their final projection (see below) mirrored the final electoral college result on election night — with the only ‘incorrect calls’ being Nevada, Wisconsin, and New Hampshire (in New Hampshire, Trump only lost by 2,800 voters). In fact, Cahaly suggested that Wisconsin might end up breaking for Trump due to the number of shy voters.
Perhaps even more remarkably, Cahaly enjoyed further success in 2018 when, contrary to almost every polling firm, he projected that Ron DeSantis would win the gubernatorial race in Florida during the 2018 midterms; he also predicted that Rick Scott would prevail in the Senate race in the same state. The final margins were tight, but like 2016, Cahaly predicted the correct winner(s) when most polls were displaying the opposite.
So, what are the polls released by the Democracy Institute and Trafalgar Group saying as the race stands? In its most recent poll, released in early October, the Democracy Institute has found that Trump is not only winning in key battleground states, but that he is ahead of Biden in the national vote 46%-45% (a true outlier).
In battleground states, which the Democracy Institute have defined as: Iowa, Wisconsin, Michigan, Florida, Minnesota, and Pennsylvania, Trump is leading with an average of 47% to Biden’s 43%. The Democracy Institute also polled three states individually: Florida, Minnesota, and New Hampshire. Once again, the Democracy Institute has Trump leading in all three: 48%-44%; 46%-44%; and 45%-43% respectively. If all this holds, Trump would win the electoral college by 320–218. Adding to the value of the data from the Democracy Institute is the fact that they have been tracking the race for a few months. Consequently, they have been able to track the shifting electoral ground and the preferences of voters. As an example of this, before their previously released poll in late August, the Democracy Institute was showing that Biden was maintaining a consistent lead in Wisconsin (the Kenosha riots have changed this); in the same poll, issues of law and order/safety had now overtaken the economy to become the first priority for likely voters. The economy follows closely behind as the second most important issue for voters — both issues, according to the Democracy Institute, heavily favour Trump electorally.
Over at The Trafalgar Group, the polls they have released show Trump is ahead in Arizona (47.8%-43.8%), Florida (48.7%-45.6%), Ohio (47.6%-43.9%), Michigan (their most recent poll shows the race is close at 46.7%-46.0%) and North Carolina (47.8%-46.1%). With Trump ahead in Florida by about three points, this tracks closely with the Democracy Institute’s poll on the state. Biden, however, is ahead in Wisconsin (47.7%-44.9%) in their most recent poll on the state; he is also leading in Pennsylvania (47.3%-44.9%). They have Minnesota as a tossup. It is also worth noting that the Trafalgar Group have Trump comfortably ahead in both Texas and Georgia, contrary to much media chatter that Biden is competitive in these states. If these projections on North Carolina, Florida, Ohio, and Arizona hold, Trump will only need to win one rust belt state to ensure his re-election.
It is not just the Democracy Institute and Trafalgar Group which project that the result will be close. John Zogby Strategies’ polls show that the race has been narrowing since June; in their most recent poll, Biden is leading Trump by only two points (49%-47%). Harvard-Harris’ early-October poll shows that among a two-person race, Biden holds a four point lead over Trump (52%-48%); Emerson College has Biden leading by four points as well. The point being is that many polling firms do not expect this election to be a Biden walkover; as such, one’s perspective on the race depends entirely on which polls are consulted (and promoted by 538 and the media).
Polls, even those few which show Trump in the lead, do not provide a full picture of the election, however. They are simply a ‘snapshot’ in time which ignore election fundamentals and other data points. In fact, people like Allan Lichtman and Helmut Norpoth, who both have impressive records of predicting Presidential election results, do not consider polls in their ‘election models’ at all. Lichtman uses ‘thirteen keys’ to predict the outcome of an election, while Norpoth utilises data from both parties’ primaries to forecast an election result. Incidentally, both are predicting different winners this year (in contrast to 2016, where they both predicted a Trump victory). One data point (besides polls) which indicates that a number of battleground states are trending Republican is the net gain in registered voters for the Republicans since 2016.
North Carolina is one such state where Republicans have increased voter registrations since 2016, a state Trump won by 173,000 votes (or 3.66%) in 2016. As it stands, RCP has Biden ahead by only 1.4%. However, voter registration numbers since 2016 show that the state is trending more Republican. In early-October 2016, the number of registered Democrats numbered 2,689,045, while there were 2,044,281 registered Republican voters in the state. In the last four years (up to early-October 2020), the Democrats now have 2,574,523 registered voters (a loss of 114,522 registered voters), while the Republicans have gained 127,217 voters (meaning the total number of registered Republican voters in the state stands at 2,171,498).
Worst still for the Democrats, there has been no surge among African American voter registration, a key demographic in any Presidential election (in fact, there has been a slight drop). While there has been an increase in the number of registered Hispanic voters, this may not be a boon for Biden at all. According to some exit polls in 2016 (and these should be treated with caution), Trump may have won as much as 40% of the Hispanic vote in North Carolina (the real margin was likely lower). According to a Civitas Institute poll, 42% of likely ‘other race or ethnicity’ voters (which include Hispanics) are ‘definitely’ or ‘probably’ voting for Donald Trump; only 30% say the same for Biden, with just under 24% undecided. In fact, the ‘problem’ for Biden among Hispanics has not escaped the attention of the mainstream media. CNN has written that, ‘Joe Biden’s Hispanic voter problem is real’; the Washington Post has issued a similar story. The results from North Carolina mirror the uptick of Hispanic support for Trump seen across several states; none of this increased support is as apparent in Florida. As an aside, both the Democracy Institute and Trafalgar Group have seen an increase in Hispanic and African American support for Trump this year.
RCP still has Biden ahead by an average of 3.7 points in Florida. Yet, like North Carolina, the weight of data which suggests that Trump is doing far better in Florida is overwhelming. Starting with voter registrations — in 2016, like in North Carolina, the Democrats had an advantage in voter registration of 330,428. At the end of August 2020, this gap had narrowed to 183,596. In the previous four years, the Republicans have gained 444,922 registered voters, with the Democrats adding 298,090 voters (a net gain of 146,832 for the Republicans since the end of 2016). Breaking down registration’s county-by-country, some of the largest gains for the Republicans have come in counties that have heavily voted for the Democrats in the past — such as Miami-Dade.
Apart from net gains in voter registrations for Republicans in Florida, another concern that Democrats are increasingly voicing is the Biden campaign’s failure to reach out to Hispanic voters. In fairness to Biden, the groups of Hispanic voters (such as Cuban Americans, and a growing Venezuelan population) in Florida often vote more Republican than the national average. Despite this, Clinton still carried the Hispanic vote in Florida 62%-35%. In a sign that Biden is struggling to generate the enthusiasm among Hispanics he needs to win Florida, some polls are now indicating that Biden has a wafer thin lead among Hispanics. In Miami-Dade, the Latino vote is split 50–50. A majority of Cuban Americans had already voted for Trump in 2016 (54%-41%); in this election, some polls believe that Cuban Americans will vote for Trump by an even larger margin than 2016. In the 2016 election, Clinton won the Miami-Dade County by nearly 30 points (and could still not carry the state); Joe Biden is ‘only’ winning Miami-Dade by 17–21 points. If Clinton could not carry the state despite winning by such an awesome margin in Miami-Dade, the odds of Biden winning in Florida are, indeed, very long. To be sure, Biden could do better among some demographics/groups than Clinton did. But, the shift of voters toward Trump in a heavily Democratic area such as Miami-Dade suggests that Biden will not make the required inroads to offset these losses.
And what’s true of voter registration favouring Republicans in North Carolina and Florida is also true in Pennsylvania (where Republicans have added far more registered voters since 2016). If people do plan on voting for Biden in record numbers, then this is certainly not reflected in the voter registration patterns. Fundamentally, it has been the Republican ground game which has helped delivered these registration gains. The ground game can also help convince undecided or independent voters to vote for Donald Trump on election day. With one month to go until the election, the ground game is being utterly dominated by the Trump campaign– not just in Florida but in other key battleground states such as Michigan. Writing in mid-September, Time Magazine has reported that, ‘Biden’s campaign is all but invisible to the naked eye’ in this key swing state. Michael Moore, who correctly predicted that Trump would win in 2016, has labelled Biden’s ground game in the state as, ‘worse than Hillary’s.’ It is a similar story in Florida. With some polls demonstrating that the election is a close contest in battleground states, the effectiveness of Trump’s ground game could give the incumbent an edge. Incidentally, in early October, the Biden campaign announced that they would be upscaling their ground game after all — a recognition that the race is tighter than most polls let on. When the Trump campaign have already been canvasing for months however, it is unknown what impact a one-month ground campaign will have.
Up till the end of September, rather than relying on people knocking door-to-door, or even getting people to show enthusiasm for their candidate, Biden’s campaign was attempting to connect to people virtually/online. On the occasion that Biden did visit a state, the events were small and intimate. Granted, the efficaciousness of this strategy will only be known after the campaign. However, the parallels between this approach and that of the Trump campaign (which is relying on traditional methods such as door-knocking and holding campaign rallies) is stark. In no small part, the success of the Trump campaign’s ground game in comparison to Biden’s is due to the enthusiasm of people voting for him.
In fact, given Biden’s initial reluctance to campaign on the road, it is something of a small miracle that he is visiting states in the first place. Shortly after the conclusion of the Democratic National Convention, where Joe Biden and Kamala Harris were confirmed as the Democratic ticket, Biden, perhaps unsurprisingly, indicated that the election was winnable from his home — stating that the campaign will continue to operate from there. Yet, only a few days later, Biden’s campaign team, contra this previous statement (and amongst much nervousness in some Democrat quarters), stated that their man would, after all, be hitting the campaign trail sometime after Labour Day (7th September). The itinerary was vague, but the decision itself signalled a significant reversal in strategy. As it turned out, Biden held a (short) speech in Pennsylvania on August 31st and went out to campaign in other battleground states — all before the declared start-date of sometime after Labour Day. Despite the plethora of polls indicating that Biden is comfortably on course to win the race for the White House, the actions of the Biden campaign, as detailed above, show almost the opposite.
Evidently, from posturing that the election could be won all from the comfort of their home in Delaware to hitting the campaign trail (albeit in a lacklustre manner), something must have ‘spooked’ Biden’s team into changing course. Going by the content of his speech in Pennsylvania, which focused heavily on issues of riots and law and order, there is some indication of what this push was.
By way of illustrating that the ‘law and order’ issue forced Biden’s hand to resume campaigning, on August 25th, a few days before Team Biden reversed course, over at CNN, Don Lemon was a delivering a sobering analysis to fellow CNN anchor Chris Cuomo. Lemon, in an about-face from his earlier positions, now believed that the riots in Kenosha were harming the Democrats; they (the riots) needed to stop and Biden had to address this issue before it was too late. Most damningly, Lemon admitted that the issue of rioting was, “showing up in the polling. It is showing up in focus groups. It is the only thing — it is the only thing right now that is sticking.”
The problem with this is that there were no polls from the preceding week for him to reference how the issue of riots in Kenosha were harming the Democrats’ fortunes in November. While earlier polls did ask about crime and protests, the results could hardly be described as a catastrophe for Biden. When the first polls filtered through post-Kenosha which did feature questions about handling crime and attitudes toward protestors, the results did not reflect Lemon’s concern. Some polls actually showed that voters trusted Biden more on issues of crime; others even demonstrated that a majority of voters still sided with protestors. Yet, given Lemon’s urgency, you’d believe that the polls he was referencing painted a bleaker picture for Biden.
How does one square this off? Either Biden’s numbers have genuinely improved since late August, or Lemon and Cuomo had access to polls that the public largely do not have access to — internal polls. If this is the case, then it is surely no accident that Biden is spending most of his (limited campaign) in the key battleground states of Wisconsin, Minnesota (where, at the start of early voting, he went to the Democratic stronghold of Duluth, a further signal that the election is tight in Minnesota), and Pennsylvania (all three states had, incidentally, some of the worst riots through the summer.) It is also no accident that since the riots in Kenosha, Democrats have adopted a more critical stance against riots. As an aside, if Biden is spending most of his time in the Midwest and rustbelt, he is not campaigning in other battleground states like Florida or North Carolina. In fact, given both campaigns’ actions so far, one could venture to say that both think Florida is either: not as important as it was in 2016, or they recognise that Florida is not as competitive as it was four years ago. By way of comparison, in 2016, Florida was one of Clinton’s (and Trump’s) most travelled to states. So far, Biden has only travelled there twice. Curiously, given the importance of Florida to his re-election, Trump has also not spent a great deal of time there (even before his Covid-19 diagnosis); and, with his focus still on the Midwest, this suggests that the Trump campaign is confident of a win in Florida.
Another data point which suggests that the race is a lot closer than most polls suggest is the ‘enthusiasm gap’, a factor which overwhelmingly favours Trump in almost every poll. This is especially true on the question of whether voters are ‘very enthusiastic’ to cast their vote for either of the candidates. In a YouGov poll of registered voters in late August, 78% of Republicans responding were ‘very enthusiastic’ to vote, compared to only 69% of Democrats. In the same poll, in a further signal that Biden is not generating enthusiasm among voters, only 38% of people voting for Biden are doing so because they like him, while 42% are doing so mainly to oppose Trump. By contrast, 71% of respondents voting for Trump are doing so because they like him (an astonishing gap of 33%!).
This trend has maintained into September. In another YouGov poll (20–22 September), 75% of Trump voters are casting their ballots for him (rather than to oppose Joe Biden); only 43% of Biden’s voters say the same (an eye-watering 57% of Biden’s voters are voting for him simply to oppose Trump). The Democracy Institute’s recent poll has also found this enthusiasm gap: 83% of Trump’s voters are ‘strongly or very enthusiastic’ about their candidate; only 49% of Biden’s voters can say the same.
The evidence of this enthusiasm gap is there for all to see. The crowds-upon-crowds of support that President Trump continues to generate is something that Democrats cannot hope to compete with. Evidently, voters are enthusiastic about his message. Seriously, whenever Trump has a campaign event, keep an eye on videos posted on Twitter to see the number of people that have turned out to see him. Even the roads which lead up to the event are thronged with people to simply greet Trumps’ convoy. Then there are the lines of people waiting to get into the event itself. As for Biden? Well, the best they can do is awkwardly host socially distanced conversations on a dying lawn with a misplaced table.
This trend is important for one simple reason: its impact on turnout; more specifically, the turnout of voters casting their ballots for Donald Trump in comparison to voters for Biden. In the rustbelt alone, Richard Baris (another pollster with an impressive record of tracking elections) has found that on account of greater enthusiasm among Trump voters, the race in that region is a dead-heat at 48.7%-48.6% in Biden’s favour. Needless to say, the enthusiasm gap will favour Trump in other battleground states such as Florida and North Carolina.
Ultimately, as Matthew Goodwin has written in UnHerd, the Democrats believe that an overwhelmingly anti-Trump message sprinkled with a few hints of what they will do once in power will be enough to drag them over the finishing line. Whether this is a cause or result of so many people voting for Biden simply because he is not Donald Trump is irrelevant — the assumption that Democrats can ‘Trump bash’ their way to victory is highly dubious at best. As evidence of this, despite the constant bombardment of Trump, he still leads on important metrics — such as the economy (even polls which show Trump losing substantially show that voters believe Trump will handle the economy better), immigration, and dealing with China (important to note here is how much of the American public’s attitudes have soured towards it). In Gallup’s most recent poll, 56% of registered voters stated that they are better off now than they were four years ago; this number is far higher than it was in previous election years.
None of the data/factors that have been mentioned here necessarily point to a decisive victory for Trump. At present, the United States is too polarised to discuss ‘landslide wins’ (certainly in the electoral college). But what this data does indicate is that Trump is, in fact, leading by some important metrics which could well decide the election; more than anything, the data demonstrates that the race is a lot closer than most polls indicate. And, in the one place where Biden is definitively leading, the polls, even these should be treated with caution.
Joseph Cotto has written an article which explains the general shortcomings of most media/university polls. After pollsters understated the support which Trump enjoyed in 2016, many would like to assure us that the issues which plagued pollsters have since been accounted for. However, as Sean Trende has warned at Real Clear Politics, many pollsters still got the final results wrong in gubernatorial and senate races in the Midwest and Florida (in most cases, Republicans ‘overperformed the polls’) during the 2018 midterm elections, the first real test to see if polls had improved since 2016. As such, according to Trende, it is not immediately obvious that pollsters have, ‘really fixed the problem at all.’ There is the additional issue that many pollsters are failing to accurately poll substantial groups of people all together. Sometimes, this manifests as a shy Trump voter (in many cases, a voter like this will lie to pollsters) — a phenomenon backed by research (this concept will be discussed below).
In other cases, pollsters may be missing large groups of people altogether. For example, since 2016, in the states of Wisconsin, Pennsylvania, and Michigan, ‘registration among whites without college degrees is up by 46%’ according to the New York Times, a key voting demographic for Trump. This is not a result of population growth, either — the population of whites with no college degrees has only increased by 1% in those states since 2016. Moreover, as per the same article, this demographic of newly registered voters (older white men with no college education) are underrepresented in polls even today. This is corroborated by Pollsmart MR, who have found a, ‘substantial block of older, white, male voters without a college degree’ who plan to vote for Trump in 2020. Because this block does not usually vote, pollsters are ‘missing’ them on account of tight likely voter screens. To further corroborate this, in his polling for the Democracy Institute, Patrick Basham also believes that men are significantly more enthusiastic to vote this year — especially for Donald Trump. This is a contrast to previous election years, whereby women typically constitute at least 52% of the electorate; this year, Basham reckons that the gender split could be much closer on account of this enthusiasm.
As proof that pollsters are failing to gather representative samples of likely voters, take a look at the maps below. This was posted by Nate Cohn on Twitter to show where samples were collected from in Ohio (left) and Pennsylvania (right). As you can see, large portions of both states remain blank — likely voters in those regions have simply not been contacted or have most likely refused to respond to pollsters in the first place. Essentially, the failure to adequately obtain representative regional samples also inflates Biden’s lead as rural voters are not sampled appropriately. By contrast, those in urban areas may be too enthusiastic to respond to pollsters, thereby further exaggerating Biden’s lead. For an extended rebuttal to Cohn, see this thread here.
Then, there is the problem of shy Trump voters, a concept that Nate Silver or 538 still does not accept. Anecdotally, I am sure that many of you possibly know someone in your family, your work, or social group that will be voting for Trump in November but is reluctant to admit it. Some polling firms, however, such as the Trafalgar Group, not only believed that people had incentives to mislead pollsters but devised some methods to work around this problem. Going back to Cahaly’s interview shortly after the election — one question that the Trafalgar Group used which could potentially weed out shy Trump voters was asking voters who they thought their neighbours were likely to support. A direct quote from the interview follows:
“In the 11 battle ground states, and 3 non-battleground, there was a significant drop-off between the ballot test question [which candidate you support] and the neighbours’ question [which candidate you believe most of your neighbours support]. The neighbours question result showed a similar result in each state: Hillary dropped [relative to the ballot test question] and Trump comes up across every demographic, every geography. Hillary’s drop was between 3 and 11 percent while Trump’s increase was between 3 and 7 percent.
According to a recent Harvard-Harris Poll released in early October, when voters were asked the question, ‘Who do you think your neighbours are mostly voting for?’ — 43% said Trump, 36% said Biden, and 21% were not sure. In no way is this suggestive of the final vote share, but it does indicate that the number of shy Trump voters have not diminished (it also means that pollsters have likely missed a large portion of voters who plan on voting for Trump in November). In fact, this could also be the reason why Don Lemon was so anxious about the impact of the Kenosha riots. In a recent interview detailing some of his findings in Wisconsin, Cahaly has stated that the Kenosha riots have resulted in a substantial portion of ‘angry people’ dismayed at how the riots were handled in the state; he predicts that the ‘hidden’ vote for Trump could be ‘north of 6%’ in Wisconsin. In the same aforementioned interview, Cahaly has opted against disclosing the methodology utilised by the Trafalgar Group to ‘discover’ shy Trump voters; what he does share, however, is that that ‘voting bloc’ has not dissipated.
The current state of the race is remarkable in many ways. What stands out, however, is the lop-sidedness of data points which support either candidate. Essentially, the main data point which shows Biden winning is his performance in the polls. Yet, the data points discussed here — trends in voter registration, the enthusiasm gap, trust on the economy — all indicate that the race is much closer than these polls let on. As a result of this, many Democrat and Republican voters are living in two completely different realities (one of which will be violently upended come election day), with even fewer thinking that the race is a dead heat. If I were to make a prediction today, I believe that Trump is on track to win at least 259–260 electoral college votes. If this is true, then Trump only needs to pluck off Minnesota, Wisconsin, Michigan, or Pennsylvania to win the election; Biden, by contrast would have to win all four states, a far more unlikely event. Alternatively, and much less likely to occur, Trump could win Nevada and New Hampshire (thereby nullifying the need to win one of the four aforementioned states); with Biden recently travelling to Nevada to shore up support there, this possibility should not be dismissed just yet. As the election quickly approaches in a deeply polarised country, we can only hope that civil unrest in the wake of the results will be kept to a minimum, and that both parties will honour the tradition of accepting the results of a democratic election.