Polls, Bets, and Stakes for the 2024 Presidential Election

A Thinking About Thinking Perspective on the state of polling, and the stakes of the race

Hi Folks,

I’m going to be changing up the column mid-season, similar to what I (mostly) did last year. The “Sunday Drive” half of the HHG is quite useful early-season when usage changes are frequent and even small changes may portend larger movements later on: think back to the early-season notes on the Jaguars, Bengals, Buccaneers, or Cowboys backfields, which all had strong predictive value (sorry about the Bears). However, as the season goes along, fewer teams are making major, non-injury related changes to their allocation of touches, which means I don’t think you’re getting the best value out of two RB-focused columns per week (and I really don’t have the capacity with my legal career to write three in most weeks).

So the entirety of the Hitchhiker’s Guide is going to move to later in the week, typically Friday Evening or Saturday Morning. I’ve done this already once or twice this year on an as-needed basis, but effectively it will be the regular Streamer Stock Watch column, with extra notes on any non-streaming backfields that are worthy of additional discussion.

The early-week column however will now be rather flexible. Late season I will be mostly focused on playoff best ball and then FFPC playoff challenge content. I will likely do a mailbag, and I’m going to do some dynasty-focused stuff as well in the near term.

But today I want to discuss the thing that’s likely on a lot of your minds at this particular moment — the 2024 U.S. Election.

I’m not going to lecture at you about who to vote for — and in fact won’t touch upon my normative perspective of the race at all until the last section — though I suspect for anyone who follows me on twitter you can likely guess where I stand.

As you know, I’m Canadian — so while I have a neighbourly stake in what direction you folks choose, at the end of the day it’s not my Country. What I do think is worth discussing however, and what fits in with the scope of this substack, are the probabilistic considerations at play in this election, and how to think more efficiently about elections.

First, there has been a lot of discussiuon about polling in this election and how that intertwines with newly created presidential betting markets. (Full disclosure I’ve made one bet on this election with a Canadian Sportsbook, and it was Kamala Harris at +200 just prior to Joe Biden dropping from the race)

I want to discuss my perspective on some of the mistakes people make in handi-capping this election, the human motivations behind some errors I suspect we will see in the polls this cycle, and the potentially damaging real-life consequences of polling errors and poll denialism.

Lastly, I will provide a probability-based line of reasoning I hope swing voters reading this column consider, in terms of the stakes of this election and how to choose if you don’t like either option.

If this is the last thing you want to read about from your fantasy football substacker, no worries — I will see you later this week for football-exclusive content I promise.

The State of Play

The presidential election will effectively come down to seven states: Arizona, Georgia, North Carolina, Nevada, Pennsylvania, Wisconsin, Michigan.

If you set each of those to “undecided” and assume the favoured candidate wins every other state, your starting ground is 226 electoral votes for Harris and 219 for Trump. You need 270 to win.

Pennsylvania = 19

Georgia = 16

North Carolina = 16

Michigan = 15

Arizona = 11

Wisconsin = 10

Nevada = 6

Here’s your current polling average in each of those swing states (and a couple more) in the Silver Bulletin model. If you find Nate Silver personally annoying and therefor prefer other models such as FiveThirtyEight, SplitTicket, or any other reasonably well-calibrated forecast, you’ll get a similar spread.

There are several theoretical combinations of those 7 states that can bring either candidate to or above 270. But not all combinations are equally likely.

Furhter, states such as Texas, Iowa, Alaska, Minnesota, New Mexico and New Hampshire could conceivably swing to the underdog, but are very unlikely to determine the presidency, since in the event that Harris or Trump flips one of these states they likely already secured enough of the swing states to win the election.

Generally speaking, the electoral map as a whole is correlated, and the more similar two states are in terms of demographics and geography, the more correlated they are. For example, if we knew Kamala Harris was going to win Wisconsin, we should be slightly more optimistic she wins Arizona, and a lot more optimistic she wins Michigan.

If you’re in Harris’s position, she has two clear paths to 270+

1. Sweep the “blue wall” of MI-WI-PA

2. Win MI-WI + NV and one of NC or GA

There are other paths of course, largely ones that include winning Pennsylvania but not both the other blue wall states (each of which she polls better in) or that include winning both Georgia and North Carolina. But those two are the simplest path to the minimum delegates required to win.

If you’re in Trump’s position, the path is effectively the opposite. Hold the south and win at least one blue wall state.

To be clear, the forecasts all take this into account, but for reasons I’ll explain in a bit, my position is that the polling averages in each state have a wider margin of error than reported. As a result, I’d slightly prefer to be in Harris’s position because she has more pure ‘outs.’

Trump is effectively dead-on-arrival if he can’t pick up one of three highly-correlated states in the blue wall — each of which has gone the same way in each of the last two very close elections. Harris is likely toast if she gets swept as well, but that’s less likely to happen given the polling averages in each state. The most likely breach is Pennsylvania, and while I would definitely bet on Trump to win contingent on him winning Pennsylvania, Harris does have other plausible outs in other regions that might be somewhat favourable for her.

Polls are not Refs — They’re not Keeping Score

Twitter has become the worst place possible to garner a legitimate expectation of the election results.

I understand polarization of politics, and it makes sense that personalities and voters on either side of the political spectrum would fall into very strong opinions about who should win. The natural cycle of social media in its current algorithms is that if you are a person who cares a lot about politics and engages with political posts — especially of a particular persuasion — you are going to get fed more and more of that and it will create an echo-chamber. How much of that echo-chamber is high-quality information is to some extent up to you, and to some extent up to the algorithm.

But what’s surprised me this cycle is how much each side is invested in projections of who will win. The amount of whining and conspiracy-raising from political hobbyists on twitter regarding polls and forecasts is quite something. This is not a football game, no team is actually in the lead. Assuming all else is equal, a forecaster making methodological changes that show Trump or Harris with a 70% chance to win vs. the current 50-50 makes no actual impact to who will win the race. And yet, people are constantly whining that their side is being under-estimated as though the refs are calling questionable holding penalties against them.

After the latest round of NYT/Sienna College polls, right-wing accounts flooded the internet to claim it was a voter supression effort. This is obsiously un-true. Even if you have zero faith in institutions to be honest, you should trust in incentives. Nate Cohn, who oversees the polling process for NYT/Sienna — gains far more financially and reputationally from being correct that anything else and has zero incentive to purposely put out polls he believes to be wrong. Further, in what universe would these polls suppress anyone’s vote?

If we were to assume the polls are well constructed, it shows a race squarely within the margin of error in every swing state, and near deadlocks in Gerogia, Pennsylvania, and Michigan. While yes, if you built an election forecast based exclusively off this set of polls, Harris would be favoured, the path for Trump to win the election based on these results inolves winning two states he’s tied in, and one of three states that he trails in by 1 point or 2. That’s not a particularly unlikely parlay. I’m sure we all watched dozens of third downs that were converted despite lower odds this Sunday.

{Also — voters who are intimately familiar with NYT/Sienna polls of swing states and their effect on the electoral count are all going to vote in every single election]

It is undeniable that the right is the more conspiratorial segment in society — and not just in America. A massive part of the Trump appeal is to present as a human molotov cocktail against the “establishment,” which many folks disdain, and a large number of his supporters are low propensity voters who were engaged by Trump explicitly on the promise of massive disruption. It’s not a shock that folks craving disruption of the system are also the most mistrustful of institutions, media, and — in some cases — data. It is equally unsurprising that Trump has found new supporters such as Tulsi Gabbard and Robert F. Kennedy Jr. — who have a long record of anti-establishment, conspiratorial positions — and that some of these positions have been taken up by silicon valley actors who are heavily invested in crypto-currency, the market for which depends on institutional distrust.

I’m not saying that all conspiratorial thinking is bad, or that all institutions are trustworthy. I have a certain libertarian streak to my politics as well, and while that’s overriden to a large extent in my voting behaviour by other considerations, most of the times I find myself agreeing with a Trump voter, it’s likely due to a mutual dislike and distrust of authority. It’s just that this particular category of conspiracy makes no sense because there is no incentive to deliberately lie in one direction about the state of polling. (There is incentive to manipulate your data to avoid publishing outliers)

Naturally, consrpiratorial thinking is not exclusive to the right. Below are a couple samplings of particularly inane comments that represent what I call the “Blue MAGA” wing of democrat-twitter, which also embraces poll denialism and media loathing.

Nate Silver is a common target for Blue MAGA, in part because he has idiosyncratic, libertarian instincts on several issues and is happy to share them, in part because he was correct about the need for Joe Biden to leave the race long before most democrats agreed, in part because he’s a little annoying, and in part because some people take unfavourable polling averages as a personal attack on their preferred outcome.

For what it’s worth, Ann Selzer is Silver’s highest-rated pollster, and he’s stated on numerous occasions he’s voting for Kamala Harris.

The frustrating thing about all this — as someone who earnestly believes that most election forecasts are trying to give us the most accurate picture of the election possible — is that this election can’t end 50-50, and a bunch of people who have zero understanding of probability on one side or the other are going to feel affirmed in their overconfident tribalism.

Now does any of this matter? Yes and no.

No rational person should let what the polls say at this point of the election affect their actions if they are not actively making strategic decisions for (or betting on) one of the campaigns. You should vote for your preferred canddiate and hope for the best either way. Alexandria Ocasio-Cortez has characterized the race as within the “margin of effort” which is the perfect terminology if you’re trying to use the polls in a helpful fashion to motivate voters and volunteers on your side of the aisle.

Polls serve a purpose largely to give information to campaigns in terms of where to spend their money, what voters to target, and whether major strategic changes are necessary. There is not much value for the general public to be gleaned from them.

That being said, people’s inability to grapple with probabilistic outcomes has extreme risk-factors for what happens after the election.

At least the vast majority of poll denialists on the left seem to me rather harmless: they think the polls are under-estimating Harris, and soon they will either be proven right or wrong about that, but I have seen zero compelling evidence that a material number of democrats would doubt the actual result of the election should Trump be victorious.

But poll denialism is now becomming inseperable on the right from skepticism about actual election results, all of which is completely without merit.

Despite this election being roughly 50-50 — give or take your subjective analysis — it is extremely concerning how many influential folks on the right such as Elon Musk, Shaun Maguire, of course Donald Trump himself, and others alternate between propping up debunked concerns of election fraud, with overconfident prognostications of a Trump blowout. If Harris does win, rather than having already accepted that it has half a chance of happening, a far too large minority of Trump voters may find the result inconceivable, and act on it — in similar ways to what we saw on January 6th, 2021.

Unfortunately, the heightened scrutiny experienced by pollsters in recent cycles has led them to make some — in my opinion — flawed methodological choices for 2024, that ironically leaves us more open to a dramatic polling error than usual.

What Does it Mean for the Polls to be Right or Wrong?

We always hear the phrase "the polls were wrong” or “the polls were right” after each election. Typically this is mis-used.

Strictly speaking it’s impossible to discern whether a poll is “right” given we only have one outcome.

In statistics, any time you make a projection in an exercise such as polling, there is going to be a margin of error, and a confidence interval associated with that prediction.

If your poll shows Harris 48, Trump 46, the margin of error is 4, and the confidence interval is 95%, what that means is:

Assuming the poll is methodologically sound, 95% of the time the actual result is going to be somewhere between Harris 52 — Tump 42, and Trump 50 — Harris 44. The Harris 48—Trump 46 number only tells us the median of that range, and that there is a slightly larger portion of the total range of outcomes in which Harris wins than Trump does.

Technically speaking, this poll is “right” if in 95% of the simulations of this election, the result falls within the associated margin of error.

Obviously that’s not how people actually percieve of whether the polls are “right” — in large part because we don’t get to see hundreds of outcomes through which we can determine how well calibrated the polls actually were.

If you have a model for instance that projects the result of all 1230 NBA games and offers a percentage chance for each team to win, you can test the calibration of that model over time, to determine whether it’s accurate. Ideally what you’d want to see is that when the model says a team has a 70% chance to win, 70% of such teams win that game over time. If 90% of 70% favourites win, the model is not sufficiently favouring these teams, if only 55% of those 70% favourites win, the model is over-favouring such teams.

With only one presidential election every four years however, it’s very difficult to ascertain whether an unexpected result is part of the natural variance the model accounts for, or whether the model was poorly calibrated.

Typically when people say the polls are wrong they don’t mean it to say the result is outside the margin of error. What they mean is the person who was projected to lose won, and the person who was projected to win lost. But in close races, the trailing candidate will still be in position to win a race a material percentage of the time, and when they win by a small margin, it was likely within a poll’s margin of error. Meanwhile, when polling identifes the “correct” winner, but is far off the correct margin, fewer people decry the polling despite it actually having been far less accurate.

We’ve seen both those scenarios occur in the last two presidential elections. In 2016, Hillary Clinton was a favourite on election day (though not an overwhelming one) and lost. In 2020, Biden was a substantial favourite on election day and won. Additionally, he won roughly the number of states he was expected to win as well, but by much smaller margins in the ‘blue wall’ than was anticipated.

We Should Predict a Polling Error — We Shouldn’t Predict its Direction

The conventional wisdom has become that polling does not accurately reflect Trump’s popularity, and since his polling is in a much stronger position today than at this time 4 or 8 years ago, he’s likely to win. He remains the betting favourite despite being in a statistical tie according to most models largely due to bettors anticipating a third consecutive over-performance.

One such example of this is Justin MacMahan, whom I quote-tweeted earlier today:

There are a few flaws in this line of reasoning:

1. The polls were fairly accurate in 2018, and actually underrated Democrats in 2022 — though not by as wide a margin as Trump was underrated in 2020. So Republican voters as a whole are not systemically unable to be reached by polls. The argument depends on Trump alone being under-polled.

2. Trump doesn’t have a history of out-performing the polls in his 2016 or 2024 primaries. So again, republicans aren’t systemically under-polled vs. democrats, and Trump isn’t under-polled vs. Republicans. You have to bet on this being both a systemic phenomenon and one unique to one specific matchup: Trump v. Democrat nomineee for President.

3. You have to assume that the Trump polling overperformances reflect an underlying systemic cause rather than simply being a thing that happened twice. As I discussed above, polls have a margin of error, and a percentage chance of falling outside that margin of error, so it’s not that inconceivable polls could miss the same way twice completely by chance without any predictive value.

4. Pollsters are weighting their results. A common misconception is that pollsters are merely calling up 1000 people and recording the answers. That’s where the idea that voters who are “hard to reach” are less likely to be captured by polls. Every pollster has a different process, but most are weighting their polls in some manner — either by party ID, education, age, likelihood of voting, or some combination of factors in hopes of avoding non-response bias. A particularly common (and in my opinion, flawed) screen this year is for pollsters to weight by “recalled vote” a.k.a. which candidate the voter supported in 2020. In theory, this helps correct polls from “mising” Trump voters — but it also means that polls are serving a very narrow purposes. In essence, they’re building their polls based on the premise that the 2024 electorate is much the same as the 2020 electorate, and the race will be decided by a few flippers within largely consistent demographic groupings.

Please don’t hear what I’m not saying. I don’t think pollsters are necessarily going to underrate Harris in this election, nor am I confident they won’t underrate Trump again. I just don’t think the fact that they underrated Trump in two previous elections is very informative to our expectations of what will happen this time — and to the extent that it would be, it’s equally likely that pollsters over-correct for the previous error than under-adjust.

Given that polling generally shows a 50-50 race, I think it’s about equally likely to be Trump or Harris that wins. However, while I think the race is highly uncertain, I’m not particularly convinced that it’s going to be that close.

The reason for this is poll herding. Poll herding means publishing results that are intentionally closer to the average of all polls rather than simply publishing your findings in a vacuum.

I’ve already raised some of the concerns I have that polls are potentially going overboard with the screens and weights leading to over-fitted results. But even setting that aisde, it’s simply implausible bordering on impossible that if pollsters are not putting their thumbs on the scale (before or after the fact) that we would be seeing so many polls published within such a tight range in each swing state.

This article by Nate Silver does an excellent job of summarizing the issue, and I’ve excerpted a section from it demonstrating how unlikely it is that the state of polling in this race is happening organically.

And this brings me to the difference between a poll, a polling average, and a forecast.

In a perfect world, each poll would exist in its own vacuum and would be completely agnostic about its results. Even if every poll in each state made the same methodological choices, they would naturally reach different people and you would see variant results between them. Rather than picking and choosing the “best” poll, we could then simply average all polls to reduce our margin of error by expanding the sample, and then build a model of the mostl likely winner based on a set of polling averages nationally and in each state. In that world, polls showing a major outlier from the median are equally valuable to those which match the median exactly.

Think about it like you’re using a DFS simulator — perhaps one of the thousands simulations predicted your score exactly, but the best measure of the quality of your lineup was the collective of all simulations.

Unfortunately, we do not live in that world. We only get to see one possible outcome of each election, and those outocmes only come once every four years (at least presidentially). Therefore, Polling firms are not mindless auto-generated simulations content to fulfill their purpose as a miniature segment of a range of outcomes. Instead, they are run by humans who despite having a better understanding of statistical principles than 99% of the population, would rather herd their data to bring it in line with the median and/or their expectations of the race, than to simply press publish and attempt to explain what polling does and does not tell you.

One exception to the herd is J. Ann Selzer, the legendary Iowa pollster who recently put out a result of Harris +3 in Iowa. Perhaps that’s because she is “mother,” or perhaps it’s because she has one of the simplest polling processes among highly-rated firms — a process that has yielded both a strong track record of results historically, and one that often separates from the conventional wisdom. If you’re wondering, Selzer’s poll was directionally accurate (i.e. more favourable for Trump) compared to the polling average in both 2016 and 2020. What if the way to find Trump voters (or any voters) was not to over-fit yourself into a straight jacket but to simply ask human beings if they planned to vote and then believing them?

Again, I don’t want you to hear what I’m not saying: I like Selzer’s methodology. It fits with my notions of what the purposes of polling should be, and I think you have a far better chance of correctly diagnosing changes in the electorate over four-year cycles if you don’t go into your poll with a pre-determined sense of what the electorate will be.

However, I am definitely not saying that her one single poll is likely to be of more value than the entire polling industry at large.

Instead, I wish that all polling firms operated without herding so that we could have a better feel for the race overall. I’m quite sure the actual average we have is not as robust as it should be since pollsters appear to be thumbing the scale. When that’s the case, there is a reasonable chance of a major polling error in either direction, which would be bad news for trust in the democratic process. (But is potentially actionable for bettors)

Who will win the Election?

I DO NOT KNOW … and have very little confidence in the answer to this question. But with that caveat out of the way…

I think it’s slightly more likely than not, and materially more likely than betting markets imply, that Kamala Harris will win the election. Given Trump is currently favoured in most betting markets I think you can get value even without my gut instinct, but here’s my reasoning noentheless:

The polls which are traditionally the highest-quality, non-partisan, and have generally been most independent of the herd — such as Selzer and the NYT/Sienna — have shown a couple interesting trends that I think favour Harris.

It was commonly held wisdom for much of this race that Harris was a favourite in the popular vote, but would need to put up a margin of 2-3 points or more in the popular vote in order to win the electoral college. The fact she’s currently projected to win by about 1 point in the popular vote would of course speak poorly of her odds to win if Trump’s electoral college advantage is as strong as it was in 2016 and 2020.

But based on the least-likely-to-be-herded polling, I don’t think it will be. Trump’s win in 2016 was largely on the back on massive margins among non-college educated white voters, in particularly non-college educated white men. This was a very helpful demographic in picking up key swing states along the “rust belt” — and we saw him sweep Michigan, Wisconsin, Ohio, and Pennsylvania — relatively white states with a long history of voting democratic (or in Ohio’s case, being a swing-state) on the back of ‘blue-collar voters.’

What Trump has lost in exchange for those gains over the elections since is the support of college educated white voters, largely in suburbs. This group is the most likely subset to have flipped from Bush/McCain/Romney Republicans to modern-day democrats or ‘never-trumpers.’

What appears to be happening in 2024 is that Trump has more or less maxed out the dials on non-colleged educated white men. But he’s lost ground with non-college educated white women, and has continued to suffer losses with college-educated white folks. However, he’s off-setting these losses nationally by making gains with non-college educated black and hispanic men.

This trade however is not a favourable one for him in the electoral college. While he’s likely to win back Arizona, and may win back Georgia, his result in Michigan, Wisconsin, Pennsylvania, and yes, Iowa, is likely to be worse relative to the popular vote than in 2016 or 2020. That’s not to say he won’t win those states, but while in previous elections he performed better in those states than he did nationally, my predicition is that his share of the vote in PA, WI, and MI will be more or less equal to his national vote share.

If polls under-estimate Trump nationally by a strong margin, then he will likely win the race. But if they don’t, I suspect Harris can win the electoral college with a thin popular vote victory.

Further, if I had to guess which way the polls are off, based on how they are trying to overfit the results to make sense in the context of the 2020 electorate, my guess is that Harris numbers in swing states are potentially more likely to be skewed downward in order to keep the results in swing states more consistent with national polls. I’m not necessarily saying that’s a result of post-facto herding, but mostly the result of an over-reliance on screening and weighting.

I also think it’s a bit more likely that in a third consecutive race of Trump v. Democrat that this favourable exchange in support in terms of the electoral college pushes Harris over the top, than that Trump’s total popular vote ceiling has massively increased.

However, that is very much a guess. If you told me Trump was was +150 to win tomorrow, I wouldn’t bet it because I would never bet on Trump. But I would not bet Harris at the corresponding number and I would be calling Trump a +EV bet in this column. (The actual odds are — depending on where you look — +140 for Harris and -160 for Trump)

What I have more conviction on is that the process behind the polling in this election is poor, and could give rise to a significant polling error. As a result, if I were betting on the election today, my best bets would be (in order): any favourable [relative to other sites] line you can find to play an OVER or an alt-number for the total electoral votes by either winning candidate, Kamala Harris to win states she’s not favoured in (my favourite would be North Carolina), followed by a straight Harris to win bet.

If you care (which you shouldn’t) my official guess is this:

Kamala Harris — 292 (including Wisconsin, Michigan, Pennsylvania, Nevada, North Carolina)

Donald Trump — 246 (including Arizona and Georgia)

The Stakes of the Election

I’ve spent all of this post up until now leaning into the nature of this substack — one that is focused on probabilies and ways of interpreting and reconciling unexpected or unknowable events. But I’m going to shift gears a touch here at the end and discuss another probabilistic aspect of this election:

What is going to happen if Harris wins? And what is going to happen if Trump wins?

I suspect the vast majority of you have already made up your mind to vote for either or neither candidate. But for anyone that hasn’t or anyone talking to anyone who hasn’t; consider this perspective:

It is highly uncertain which party will win the House of Representatives — and that result is highly correlated to which party wins the presidency. It is however far from uncertain which party will control the Senate. According to 538, the Republicans have a 92 percent chance. (Conditional on a Harris presidential win, the odds likely shift closer to 85%, and into the high 90s conditional on a Trump win)

I happen to like most of Kamala Harris and the Democrats’ policy agenda, and to the extent that I don’t — it’s largely that it does not go further. I would like to see the democrats legitimately embrace universal health care, broad reforms to the Supreme Court, and — most critically to me at least — enact a substantial departure from their current foreign policy agenda. (Though I assure you the Republicans would be just as hawkish — and liekly even more hawkish — in their unconditional support of the Netanyahu government in Israel if you happen to share my largest grievance with the Democrats)

But if you don’t like Kamala Harris’s policy agenda I have great news for you: there is a 80-90% chance she won’t be able to pass a single item of policy over the next two years that the likes of Lisa Murkowski or Susan Collins — moderate Republican Senators — won’t vote for.

A wealth tax? Not going to happen.

Major increases to capital gains taxes? No shot.

Green New Deal? In your dreams.

Frankly, even with a Democrat-controlled senate, Harris will struggle to pass any legislation that does not have broad support among centrists politicans in her own party. And if Harris’s campagin is any indication, she’s willing to adopt positions to the right of Joe Biden’s. If you think her actual positions were the ones she adopted in the 2019 primary, I think you’re incorrect. But if you are correct, it won’t matter, because those policies cannot be passed.

Under Harris, the most likely scenario is a maintenance of status quo, with the potential for incrimental changes in health care and on the edges of the economy, and — depending on the makeup of the senate and her willingness to be rather cavalier with executive orders — increased protections for reprodutive rights.

Right now the American economy is strong. Inflation has had significant detrimental effects on common people across the world, but has largely stabilized and the United States is not an outlier in that respect.

However, the United States is a postive outlier in terms of growth, consumption and investment compared to other highly-developed countires in the post-Covid economic environment.

This is not to say that the economy is excellent in all ways for all people. It’s not. And I’d be happy to discuss with anyone ways in which the United States, Canada, or several other countries could do a better job of ensuring that that the spoils of the economy are more fairly distributed than either the Republicans or the Democrats have proven willing to do.

But that’s not for this column.

In this column, I’m simply preaching that if you try to consider the state of the country through a data-driven lens, things range from fine to good, but are a far cry from being a disaster.

So you have a passble (if not better) economy, and a political environment in which one candidate has a very minimal chance of enacting any policy which median voters may find troubling. What about the other candidate?

First of all, conditional on Trump winning the Presidency, his party is quite likely to control all three chambers of government. This means that if there is anything you’re worried about regarding his agenda, it’s a lot more likely to actually happen than anything you don’t like with Harris.

One of those items is that Trump plans to impose broad, sweeping tariffs on imported goods. He can commission up to a 15% tariff for at least 150 days without congressional approval, so the only chance it doesn’t occur is if he’s lying or eventually changes his mind. That’s extremely troubling, since tariffs — which apply a tax on the cost of importing goods from a particular Nation — are fundamentally guaranteed to raise prices on any goods subject to said tariff. That policy will be extremely inflationary.

What else can Trump do by executive order? Well he can enact at least some aspects of his mass deporation scheme. And the mass deportation of un-documented immigrants will create a labour shortage, which is also inflationary.

And again, he’s very likely to have the support of his party in the House and Senate — so he can get quite a lot done, potentially including federally regulated resirictions on reproductive rights.

If you love Trump, that probably sounds like great news but I doubt I could sway your vote anyhow. But from the perspective of a “double-hater,” it’s just objectively true that Trump — based on the nature of his policies and the odds of who controls the Senate — has a lot more potential to do harm than Harris.

And all this is fundamentally elementary compared to the most pressing concern in my opinion — which is that Donald Trump refuses to accept the results of the 2020 election, conspired to have it over-turned, and roused a mass of supporters at the capital building, before standing by as supporters holding his flag breached the capital in an attempt to disrupt the certification of the election.

What are the odds that Trump attempts to retain power after the expiry of his 2nd term? Or that he attempts other measures to undermine free and fair elections?

It is indisputable that he has authoritarian instincts and desires. How strong of a possibility it is that he withstands them, or that Senate and House Republicans rebuke him, is unknowable. But those Senate Republicans had an opportunity to ensure he could not run again by convicting him after his second impeachment trial, and chose not to.

If you were to ask me, I’d say that conditional on a Trump presidency there is only a 5-10 percent chance that the 2028 presidential election is either non-existent, conducted contrary to the constitution, or materially compromised. But I would have said there was a 0% chance of that conditional upon any other Democrat or Republican nominee in my lifetime, certainly including Kamala Harris, and certainly including defeated Republican primary challengers such as Nikki Haley or Ron DeSantis.

Maybe you disagree and you think there is only a 1% chance, or even 0.1% chance of this disasterous scenario. But ask yourself please if you think it’s higher than zero.

Are you 100% certain that Donald Trump — with 4 years of power and at least two years of complete control of the federal government — will allow for a free and fair election in 2028 that he will not participate in, nor unlawfully influence, and that he will accept the result of regardless of who wins?

This is the real-life application of assymetric outcomes we discuss in this column every week. Even if you think that Trump will be better on the margins for the economy or any other issue you care about, how much better do you think the median outcome of a Trump presidency is vs. a Harris one to take on even the slightest risk of such a damaging worst-case scenario?

If you’re in a swing state and don’t feel inclined to reward either candidate with your vote, how much utility is your pride in denying these candidates your vote worth to you, relative to the chance that it indirectly supports a Trump presidency, and the chance that presidency fulfills your worst fears about him?

Probabalistically speaking, I think the choice is clear.

Wishing you all the best, in an exceptionally stressful time. Remember to love and respect all members of your commuinity, no matter what.