US Election: Mallard Forecast 12.09.2020
The Mallard Forecast this week is unchanged with both candidates having a 49% chance of winning the US Election.
Note: Forecast above is to win the Electoral College (270 seats). Joe Biden has a 75-85% of winning the popular vote. Donald Trump has the advantage with the Electoral College, meaning he needs fewer net votes to win than Joe Biden. We see the tipping point at which Biden becomes the favourite is +4 national vote (see last week’s forecast for information). Odds by national vote modelling are conducted by a Bayesian Regressive Model using parameters: 2016 key state margins; polling information from Biden vs Obama, which indicates that Biden polls as high as Obama, in some cases higher, in Democrat areas.
Plus ça change, plus c’est la même chose; The more things change, the more they stay the same.
A super-forecasting friend commented on the last week’s edition of the US Election Forecast, “This makes sense to me. You have the polls on one side, and almost every piece of non-poll related information on the other. The case for Biden is the polls and nothing else. Maybe that is enough, but I doubt it”.
This is exactly our rationale at the Mallard. The polls point to a Biden win, that is true. However, the supplementary, non-election polls continue to point to Donald Trump: Reduced Support for Black Lives Matter; Economic Recovery; Economic Trust; Public key policy indicators favouring Trump; Electoral College Advantage; Betting Markets; Focus Groups and Anecdotal Evidence.
This does not mean that the polls should be discounted. Far from it. The polls are vital in our model. The difference between our model, and other forecasting by pure data modellers, is that we also look at what the polls do not show. We also factor in Bookmakers odds, and anecdotal evidence from focus groups and credible article data.
This week I have collaborated with Sarah Stook to explain why we think Bookmakers odds are important, and college educated suburban voters could be key come election day.
Polling vs Betting – Ewell Gregoor
If you step out of the Twittersphere for a few moments and went on a trip to your local betting shop, you may be surprised. Whilst Twitter continues to tell us that the Election is Joe Biden`s to lose. The bookmakers tell a different story. It is clear from the Table below that the Bookmakers and Leading Forecasters see the race in a somewhat different way.
|Bookmakers Implied Odds||50%||48%|
What does this mean? Well, firstly, as I said in a Tweet last week, anyone who thinks Biden is has over a 70% chance of winning should run to their local betting shop, as there is some serious money to be won. On a serious note, it does beg the question, why is there such a difference between Bookmakers and Pollsters?
Betting odds are slightly different to poll forecasters as they have to consider betting shifts. As Bookmakers naturally want to protect their profits, this makes sense. GQ Magazine published an analysis of August betting markets this week. In a turnaround from July, Donald Trump has overtaken Joe Biden in terms of betting volume, with 8.7 million dollars wagered on Trump, whilst Biden is trailing on 7.9 million dollars. This resulted in a post-convention odds shift of 7% for the incumbent. So, while the polls indicate that Trump had no real convention boost, which the data-forecasters continue to tell us. The odds say differently. Not only did Trump get an odds boost, his swing of 7% represented the biggest post-convention odds shift in betting history, beating fellow Republican George Bush who had a 6% shift in the 2004 Election, where he went on to beat Democrat Candidate John Kerry. The omens are good for Trump, both Obama and Bush in their elections (2000 – 2012), had an odds shift in their favour following their conventions. And of course, they went on to win their respective elections.
The fact that so much money is being wagered on Donald Trump is surprising, considering the Bookmakers are offering odds as much as 34% better value compared to what the experts are suggesting is likely.
So who should we believe?
Bookmakers have been taking bets on US Politics long before the creation of scientific polling research. Robert S. Erikson and Christopher Wlezien from the University of Texas analysed Betting Markets vs Polls throughout history. They found that Polls are not conducive to Bookmaker performance, and that the Bookmakers predicted Elections better before polling was introduced in 1932. Indicating that polling isn’t always everything and can often lead to discounting other important information that is key in calculating odds. The authors also concluded that Bookmakers odds pre poll introduction are at least as good of a predictor of events as Scientific Data Research is. Whilst not significant, there is a trend that pre-poll Bookmaker Odds outperform Scientific Data Research when predicting American Elections.
It is worth remembering that the gambling industry is worth almost 500 billion dollars, implying they usually get the odds right.
Suburban and College Educated – Sarah Stook
There’s always an assumption in US politics. If you tick a certain box, you probably vote a certain way. A middle-aged black woman in Illinois is expected to vote Democrat. An older, white male labourer in Mississippi is expected to vote Republican.
In terms of white voters, Trump and the Republicans tend to have the biggest trouble with the college-educated. To them, it should be a concern. The demographics that vote against them- ethnic minorities and the educated- are expanding to the point where states like Arizona are likely to flip from relatively safe red to soft blue. Things, however, aren’t always as they seem.
The polling site Five-Thirty-Eight (named for the number of Electoral College votes), has an interesting piece of data. In four states- Texas, Florida, Georgia and North Caroline- white college graduates remain safely Republican. This bucks the national trend of their peers preferring Democrats, especially in this election. It seems a surprise that this demographic in certain areas remains conservative, but one must remember the factors that decide a person’s political views aren’t always just their colour and level of education.
For Trump, this is a piece of good news. Whilst Florida is a perennial swing state, the other three are usual reds who have started to swing in recent times. The Senate election between Democrat Beto O’Rourke and incumbent Republican Ted Cruz led to the most expensive race in US senate history, with the media focused on it like nothing else. In Georgia, a close Governor race between Republican Brian Kemp and Democrat Stacey Abrams led to accusations of voter suppression and electoral fraud when Kemp won. Ms. Abrams refused to concede and still has not nearly two years later. Despite the closeness of these races, there’s steadfast federal support for Republicans.
If you look at the suburban, college-educated white voter, there’s a chance they could swing Republican this time around. It’s a simple reason- the riots. As cities burn and politicians deftly call them ‘peaceful protests,’ there’s a growing concern that a Democrat victory would lead to police being defunded. Those in the suburbs and nearby towns look on with concern. As liberal whites either join in or condone the riots, others are concerned.
Should they either stay at home or hold their nose and vote Trump, it’s a gift for the incumbent. Red states could stay red. Swing states like Wisconsin, Michigan and Pennsylvania might stay in his column. Some believe that Minnesota, the only state not to vote Reagan in 1984 (albeit only by three thousand votes), could turn red- it’s been close before and its city of Minneapolis was the start of the current unrest.
This is the demographic to watch.