So where are we at this point in the race?
Before I start, lets discuss polls.
During the 20th century, American political polling was driven by the Gallop Organisation (Gallop) who were considered the gold standard of pollsters.
However, once mobiles and the internet went mainstream, from the 90’s onwards, polling got more complicated and the voters harder to find. The days when most voters were contactable by a landline were over. As landline polling became less reliable with younger folks, the polling, even for experienced professionals like Gallop, got harder.
In 2012, in their final forecast, Gallop projected that the Republican Mitt Romney would win the 2012 election. That was wrong. Obama won comfortably. Shortly after that, Gallop gave up on election forecasting polling.
Since 2014, polling has become increasingly useless as a guide to predicting who will win elections.
We all remember how the polling overwhelmingly indicated that Hilary Clinton would win the presidency. All those state polling which showed her miles ahead of Trump in the Rust Belt days and weeks before the election melted upon impact with electoral reality. Things barely improved in 2018, with polling totally wrong with the Florida election, to give just one example, where the Republican candidate Ron DeSantis won. Ignoring the outlier Trafalgar Group, the average polling lead for his opponent was 5%. DeSantis won the election by 0.4%.
There are many reasons for the decline in polling over the last decade, but a key reason is its extremely difficult to reach voters these days. Many people refuse to answer pollsters, especially Republicans and those living in the countryside. Also, folks who work during the week simply don’t have time to answer pollsters. So those pollsters that don’t do their work through the weekend, will not get a proportionate sample and capture those voters (often Republican leaning types) who are too busy to waste time answering polling questions (which can run to 72!).
Getting polling right is expensive, takes time (longer than most pollsters allow) and should use multiple ways to get to those hard-to-find voters who are key to an accurate polling sample. Most pollsters simply fail to do that these days which is why they have been unreliable during the previous election cycles. Even those pollsters, Big Data/People’s Pundit (PP) and Trafalgar Group (TG), who do a better job than most, aren’t perfect.
So, when trying to work out who is going to win an election, I don’t rely on public polling. Registration data is a very useful metric. It proved to be more reliable in 2016 then the polling, which overwhelmingly suggested that Clinton would win the election. Only certain states provide party-based registration data, but of the data available its looking good for the Republicans, in particular the battlefield states of Pennsylvania, Florida and North Carolina.
Primary election data/models are also useful because they are a powerful proxy for turnout and enthusiasm for each side. A few different political analysts use primary data. One is Professor Helmut Norpoth, whose Primary model has forecast the majority of US elections. His forecast, based entirely on the primary election data, is a Trump landslide in November. There are those who critique his model but despite that he has a better track record then most pollsters so I take their criticisms with a pinch of salt.
On a common-sense level, his model makes intuitive sense. If millions of people are prepared to vote for Donald Trump in a primary election with no real opposition (unlike 2016) in historic numbers, its likely that those same Republican voters will come out in big numbers on election day.
A variation of that primary based model is the Washington State primary results.
Henry Colson of the Washington Post reviewed the data and wrote “…the party that led in the primary also won the seat in the general every time. That track record is why the state’s recent primary results are so important. With more than 99 percent of the votes counted, Republicans have a higher share of the total vote than they did in 2018 in eight of the state’s 10 congressional districts.” The conclusions from the Washington State data are;
- Republicans are doing 5% better in rural and small metro places compared to their 2018 mid-term performance.
- On a national scale Trump will be much more competitive in swing states such as Wisconsin and North Carolina than the polls currently show.
- Trump and GOP House and Senate candidates across the country could out-perform expectations once again. Indeed, nine of the House GOP’s top Democratic target seats have at least substantial portions of voters who live in rural and small-town areas. Key Senate targets in Maine, Montana, North Carolina and Georgia have similar profiles.
- The Republicans are struggling in metropolitan suburbia and haven’t improved since the 2018 mid-terms results.
The above data matches what I am reading in terms of registration (GOP are doing much better than the Dems in registering new voters, particularly voters who have never voted before) and the more reliable pollsters out there. Other non-polling metrics also indicate an enthusiasm gap, whether they are Trump/Pence signs which vastly outnumber Biden’s or the huge numbers turning up for Trump rallies across America in comparison to the Democrats.
We are still three weeks away from election day but overall, I am still reasonably confident (with a 60% probability) that President Trump will get re-elected.
So, what about the electoral college map? Well, everything I am reading and watching seems to suggest that despite what the mainstream polls are suggesting (remember all those media articles suggesting Texas was going blue in 2016!) The GOP are looking pretty good in Florida, Ohio, Utah, Georgia, North Carolina and Arizona (which despite a small shift away in registration numbers should remain red).
Trump’s polling among African-Americans and Hispanics are better than 2016 and he is doing particularly well among Cuban and Venezuelans in Florida. In 2016, according to exit polls, Trump got 8% of the African-American vote. Even the mainstream polls suggest that he will do better than that this time around on the back of rising support among working class Black men (between 11 to 15%).
Among Hispanics, Trump polling is now consistently in the 30’s, a significant improvement on his 28% share of the Latino vote received in 2016. Rich Baris, has commented that 2020 will be a class/income-based election, with blue-collar voters of different backgrounds and races starting to coalesce around the Republicans whilst the Democrats consolidate among the upper-middle class. This is very much John Greer’s thinking – who accurately forecast the 2016 election – that class is the major factor driving an electoral realignment within American politics. The Democrats are increasingly the party of the high-income elites and the upper middle classes whilst the Republican Party under Trump is increasingly a blue-collar and middle-income mass party across the flyover states.
Of course, alignments are a decades long process and don’t happen in one election cycle. But the trends you saw in 2016, with working class non-college educated whites shifting to the Republicans whilst college educated white females increasingly moving to the Democrats is likely to accelerate. What is more interesting, and far less picked up by the mainstream media, is the growing support for the GOP among blue-collar minorities, in particular men, from Hispanic and African-American backgrounds. Greer has written for a while that it is this shift that will prove key to Trump’s re-election in 2020 which he thinks will be a landslide with approximately 350 electoral college votes.
The other key consideration is turnout. Students are usually a powerful card for the Democrats in getting their vote out. This year, thanks to Covid, colleges are locked down and student registration and turnout will be significantly reduced from 2016. The data from Washington State suggests – and this tallies with what other analysts are saying – that pollsters are under-estimating the turnout among rural non-college educated Republican voters who will come out in droves for Trump.
My forecast, and this is currently interim and not final, is the following:
- The Republicans will keep their majority in the Senate.
- The Republicans will do better than expected in the House but I’m not in a position, right now, to predict who will definitely win a majority in the House.
- Trump is likely to win the southern states of Arizona, North Carolina and Florida. I’m also confident he will win Utah and Ohio again.
- Among the Rust Belt states, the election looks very tight at the moment, with both TG and PP in their latest polls showing Trump 2 to 3% behind Biden in the key Rust Belt state of Pennsylvania. The Rust Belt looks competitive but Trump needs to win over the independent and remaining undecided voters to get those key states over the line.
- Nevada, with its large Hispanic population, could also be in play. A recent Politico article warned that the Democrats could lose the state and recent chatter online suggests that the Trump campaign are increasingly bullish about flipping the state. I’m going to tentatively predict a narrow Trump win in Nevada.
- Outlier states, primarily Colorado and New Mexico, both with large Hispanic populations, may prove closer than the mainstream polls suggest.
- New Hampshire is likely to remain Democrat.
So, to summarise, I think the Trump team have around 250 to 260 EC votes that are reasonably secure and a cluster of battlefield states, primarily in the Rust Belt, in play over the next few weeks. If, and it is a big if, the Democrats manage to win the Rust Belt – Minnesota, Wisconsin, Pennsylvania and Michigan, Joe Biden will win the election.
However, if my hard predictions prove correct, and in addition Trump wins over say 2 out of the 4 Rust Belt states, he wins the election. And that excludes any further wins in the south-west, like Nevada. Despite what the mainstream media and polls suggest, the pathway to a Trump win is wider than they think but this election is not over yet and the winner will likely be decided in the Rust Belt.
I will proceed to make a final forecast shortly before the election.