This will be our last predictions market update piece of 2020. Throughout this race, many in both mainstream and alternative media have called into question the validity of polling given the systemic errors (however overstated they may be) in 2016. By the same token, betting markets have come into focus as a promising alternative to polling and modeling. Prior to the 2016 election, the New York Times’ UpShot model, which was neither a poll nor a betting market gave, Hillary Clinton a 99% chance of winning the race, and we all know how that turned out. Around that same time, PredictIt (a betting market in which investors wager on political outcomes) was giving Trump as high as a 36% chance of winning. Whether that was the correct assessment at the time or not, it ended up being far closer to the truth than mainstream consensus, which saw almost no doubt in Clinton’s victory.
Today, we see a not too dissimilar situation: Trump is trailing by significant margins in national polling and trailing in most battleground states, and most analysts are confident in a Biden victory. Granted, these are different races – there are about half as many undecided voters as there were in 2016, which ended up being critical to Trump’s victory, and Biden’s lead has been larger and more stable than Clinton’s ever was. This is not to mention the unprecedented proportion of mail-in ballots, and all the potential problems that incurs; independently of any concerns about voter fraud, a larger proportion of mail-in ballots are thrown out than in-person ballots.
Trump’s likeliest path to victory this time around diverges significantly from 2016. Trump’s 2016 win depended on late-breaking voters in the Rust Belt going in his favor. Trump’s 2020 win, if it materializes, will likely run through the Sun Belt. Trump needs to win Arizona, Texas, Florida, Georgia, South Carolina, and likely Pennsylvania. Let’s take a look at what the markets say about those states. Keep in mind that these numbers add up to over 100% given the rake taken by the bookie, and that these numbers do not represent polls but the implied probability of victory as assessed by bettors who have a financial interest in getting the outcome right. These are the states that are most likely to be the deciding factors, so I’ve left out some other states that appear uncompetitive, such as Michigan and Wisconsin. Any scenario that leaves us with a narrow race in Michigan and Wisconsin necessarily entails a catastrophic polling error, far more severe than what we saw in 2016.
Despite Biden’s very narrow lead of 0.5% in the RealClearPolitics polling average, PredictIt investors are still giving Trump a slight lead over Biden. This is probably based on a momentum factor, the trendline in Arizona polling being in Trump’s favor over the past few weeks. PredictIt gave Biden a 70% chance of winning Arizona a little under a month ago, so this market has tightened substantially.
This is optimistic but not outrageously so. Trump has a polling lead there of a little over 1%. Democrats have been salivating at the thought of Texas turning blue cycle after cycle, but that is assuming the voting demographics remain the same. Contrary to popular conception, Trump is doing better with Hispanics in 2020 than he was in 2016. Nonetheless, while most would assume a Trump victory in Texas is assured, Biden is within striking distance.
This is one of the more volatile state-level markets on PredictIt. A month ago, Biden was favored to win with an implied probability of 55%. Around October 19th, the market switched in Trump’s favor. This is, again, likely due to the momentum factor, as it was around this time that polling in the state started to move in Trump’s favor. Today, the RealClearPolitics polling average gives Biden a lead of 1.7%, though it is worth noting that the two most recent polls both have Trump up by 1%. This is the most critical state of all the critical states – there is no plausible scenario that I can see where Trump loses Florida but wins the election.
This market has been remarkably stable, at roughly the same level it was a month ago. Georgia is tragically lacking in high quality polling, but what little there is indicates momentum in Trump’s favor.
Trump has a very narrow lead, of half a percent, in the RealClearPolitics polling average. The last two weeks have seen a huge turn in his favor, practically eliminating a Biden lead of 3.5%. The latest Emerson poll there has it literally in a dead heat, with both candidates at 47%.
This is where the bullish Trump prediction runs into trouble. Trump’s most plausible path to victory runs through PA, where he is neither favored to win by prediction markets, polls, or modeling. But Biden’s lead is narrow, and polling is trending in his favor: Biden currently leads in the RealClearPolitics polling average by 2.5%, and in the middle of October, that lead was over 7%. A win in Pennsylvania is close to necessary for Trump, but not for Biden. Biden still has plausible paths to victory if there is a polling error in Trump’s favor in PA – Trump does not. It is not impossible, and I am certainly not making a prediction, just assessing probabilities. A 45% chance of an event happening does not mean that event will not happen.
All things considered, this is where the investors – and it is crucial to stress that these people have a financial self-interest in getting the results of this election right, not pushing a political agenda – see the race. This race has been remarkably stable, with Trump always trading in the range of 48% to 37%. If you are a hardcore Trump supporter, you might view this as overly optimistic for Biden, but a 43% chance is close to a toss-up. PredictIt is saying that Trump has almost a coinflip’s chance of winning. Which means they are counting on Trump winning every single toss-up that I just listed. We will know within a few days whether that will end up being a reasonable estimate. What happens after that may end up being even more consequential – this is an election that will be contested no matter the outcome. God-willing, that contest will be peaceful and resolved quickly. But I would not bet on that outcome. Given the mood of the country over the past few months, it is far from impossible that it will be contested in the streets, and we should all be praying against that outcome.