Last Minute Moneyball: How To Maximize Your Impact Before Election Day

Eric Bolton
10 min readOct 29, 2020

--

We’re now less than a week from the most important election of our lifetimes, and the culmination of an undeniably emotionally harrowing year. There is so much at stake — healthcare, climate, racial equality, reproductive rights, the Supreme Court — that it’s become virtually impossible to sit on the sidelines without going insane.

And while it may be that voters’ minds are already pretty set, and 66 million ballots have already been cast, there’s still a lot that can be done to have an impact this election, even if that just means turning out a few extra voters. In fact, that can make all the difference.

But where will turning out voters make the most difference? Where should one make phone calls, send texts, and donate?

The short answer is:

  • Arizona, Wisconsin, Iowa, and North Carolina for the Electoral College
  • Iowa, Montana, Arizona, and Maine for the Senate
  • NY11, NJ2, NM2, TX10, FL26, CA25, IA1, UT4, TX24, and TX21 for the House

Your time volunteering in these races, if effective, could be worth up to $80,000 per hour. Read on to find out why.

In 2002, statisticians Andrew Gelman and fellow colleagues set out to answer the tricky question of where American voters held the most sway. Boiled down to its essence, they found this was equivalent to figuring out how likely it was for an election to be decided by just one vote.

It makes sense if you think about it: if an election is decided by exactly one vote, then every single vote was decisive. Had any individual voter on the winning side stayed home, the result would have been entirely different. So if Gelman and his colleagues could measure that tiny probability —the likelihood of an election being decided by a margin of precisely one vote — then they could measure how likely it is for voters to have a decisive impact.

Now, obviously, Gelman and his colleagues were examining an idealized scenario that almost never happens. In a country with hundreds of millions of voters, you’re never going to see an election decided by only one vote, right?

Well, it‘s not as far-fetched as you’d think. And that’s in part because the fate of America is not actually in the hands of hundreds of millions of voters.

U.S. elections more often come down to very few voters in a few states. Trump’s victory in 2016 was handed to him by 78,000 voters in Pennsylvania, Wisconsin, and Michigan. In 2000 in Florida, which decided Bush’s election, the number at which the Supreme Court decided it was time to stop the recount was 537 votes.

What if that number had been 536? Or 250? Or 10? Would things have gone differently? In such situations — effectively coin tosses — it’s easy to see how a single voter could completely change the fate of the nation.

So the “decisive single vote scenario” is not even that unlikely. Which means a single vote could determine whether our multi-trillion dollar federal budget ends up being spent on healthcare… or on tax cuts. In the case of very close tipping point races, a vote is essentially a very valuable lottery ticket. All it takes to claim it is heading to the polls.

So where can one find the most valuable lottery ticket in the United States? To answer that, there are three questions we need to answer in turn:

  1. For any given race (District, Senate Seat, or Presidential), what is the chance that a voter will be decisive?
  2. For each given chamber (House of Representatives, Senate, or White House), what is the likelihood that the given race will decide the chamber?
    (e.g. the Senate is exactly 50–50, and the race ends up being the closest Senate race)
  3. How much money would a change in government end up affecting?

In the case of number 1, we can use polling data from RealClearPolitics to estimate the likelihood of individual races being settled by a single vote. Here, some statistics come into play, which you can read about in Gelman’s paper.

For number 2, we can borrow FiveThirtyEight’s notion of tipping point states to understand which states are likely to change the outcome of the presidential race. As for the Senate and House, we can use voting record data from VoteView, as well as party affiliation where it exists, to determine which individual candidates are more likely to vote for conservative causes — or liberal ones. Here’s a chart showing the recent Amy Coney Barrett vote, for example:

That pink Republican “No” was Susan Collins of Maine

(I used this data to determine the partisan lean of Independent candidates)

Finally, for number 3, let’s keep things simple and just plug in the United States’ federal budget for 2019: $4.45 trillion.

Combining all of these answers, we can calculate the top 10 most valuable races for each chamber of government— White House, House, or Senate:

Left is Electoral College, Center is House, Right is Senate

Looking closer, these rankings make sense — they correspond to races that are not only close, but also have fewer voters, which means they are more likely to be decided by very few votes. (Alaska ranks highly here because typically only about 300,000 people vote each year there)

Now of course, with the Senate, House, and White House split between two parties as they are now, major changes in government funding are hard to make happen. Even in the midst of the COVID crisis, Senate Republicans were recently unable to agree on an urgent stimulus bill with House Democrats.

So perhaps we should say a vote is truly “decisive” only if it means all chambers of Government end up controlled by the same party. If we count only those scenarios, here are the top 10 rankings per-district (all chambers together), and per-state (Senate and Electoral College together):

So we have our winner: the election “lottery tickets” in Iowa’s second district are worth a whopping $64,575. Here’s what that looks like on a map:

(By the way, the idea of multiplying the likelihood of a decisive vote by the budget it affects is not a new one, and you can read all about it in this excellent blog post by Robert Wiblin.)

In those darkest colored districts, any single voter could completely change the federal budget for the next two years. So phone banking, which can reach up to 5 voters per hour, suddenly makes a lot of sense. Some strategies have even more reach, such as polling place vote tripling, which can reach up to 45 voters per hour (but involves physically going to a polling location).

Now if we assume that 28 votes are added for each 1,000 voters who pick up the phone, and that Iowa District 2 votes are indeed worth $64,575, then that means that your time phone banking in IA2 would be worth over $9,000/hour for your candidates of choice in Iowa.

Meanwhile, your time spent on polling place vote tripling would be worth over $80,000/hour. Maybe worth booking a flight?

Now, does this all mean it’s rational to simply pay an Iowan $60,000 to vote for your favorite candidate?

The short answer is no. First of all, it’s illegal to do so. And second of all, that’s about as rational as dropping $60,000 on lottery tickets, knowing that whatever you won you’d have to dedicate to government spending (I suppose that’s more or less what happens anyways).

The more nuanced answer is that while that may not be the most rational decision for an individual, it undoubtedly enters the calculus of big corporations and billionaires who can sink that kind of money into politics.

In fact, they usually don’t even need to spend that much to find massive returns on investment: an analysis back in 2009 showed that multinational corporations got $220 back for every $1 they spent on lobbying for tax cuts. So there are more efficient ways to spend money in politics, if you’ve got a lot of it.

But let’s take a look now at how our estimated value of votes actually matches up to campaign spending. According to OpenSecrets, these are the most expensive races by campaign spending this year, sorted by a “per vote value” calculated using 2018 turnouts:

One thing is clear: voters in the New York 14th race are getting way too much attention (given that Ms. Ocasio-Cortez is nowhere close to being unseated). Also, the money spent on the Kentucky and South Carolina senate races (Mitch McConnell and Lindsey Graham), while reasonable, could be better spent elsewhere. In the analysis we just conducted, Kentucky Senate votes are worth less than $50 apiece. And South Carolina votes, while valuable, pale in comparison to votes in Iowa, Montana, Arizona, and Maine. This is because even though South Carolina is now a toss-up like any of them, these races are far more likely to be the ones that end up flipping the Senate.

In 2002, the year Gelman’s paper on voter power was published, something else was etched into the halls of statistical history — the now famous story behind Moneyball. After a heart-breaking end to the 2001 season, the Oakland Athletic’s general manager Billy Beane decided to focus solely on statistics in making purchasing and trading decisions. The Oakland A’s went on to find such a ludicrous degree of success in doing so that at one point they won 20 games in a row — an MLB record. And they did so with one of the smallest budgets in the league.

Now, don’t bother comparing the American voter to the Oakland Athletics. In the face of a hugely unfair and unrepresentative system — characterized by the outdated Electoral College, the imbalanced Senate, the gerrymandered House, and the pervasiveness of lobbyists and big money — we’re more like a little league baseball team facing a team of Hall of Famers.

Those who live in populous, reliably red or blue states like California and Texas (I live in New York) are painfully aware that their state’s tens of millions of votes often make no difference in presidential elections, due to how the Electoral Map is drawn. Case in point: not one mention was made of the California wildfires in the vice presidential debate, compared to 10 of fracking.

Those who have power often have it because they knew how to game an unfair system. It’s been happening a lot more lately: the Electoral College overturned the popular vote twice in the past 20 years. The Senate majority represents a smaller swath of Americans than the Senate minority. And despite some promising efforts to fight gerrymandering, we still have horribly snaky districts like Maryland’s 3rd — yet more evidence politicians often pick their voters, rather than the other way around.

Maryland’s 3rd district

The Oakland A’s never won a World Series under Billy Beane, but they completely changed how baseball was played, eventually inspiring the Red Sox’s 2004 World Series win. Their methodology arguably created a fairer game, one that valued players for their talent and ability over their flashy looks.

Likewise, if American voters want to change politics for the better, they need a better understanding of how politics can be made to work for them, not against them. Hopefully, this analysis gets us at least some of the way there, by demonstrating the massive discrepancies in the value of people’s votes that our unfair rules have created.

Now, with that knowledge in hand, go forth, make donations, canvass, and place some damn phone calls — not to Kentucky or Queens— but Iowa, please!

In general, if you want to learn more about how to fix the U.S. electoral system, head on over to Sam Wang’s excellent Princeton Election Consortium. Almost everything there is dedicated to keeping voters informed about the unfortunate quirks of America’s political landscape, and how to repair them.

You can find all the code for my analysis (originally developed for the 2018 midterms) here.

And here’s a list of links for getting involved in the most crucial races cited in this analysis, in rough order of how important they are:

Links for Democrats

Iowa (Senate, House — IA1, IA2, and Electoral College)

Arizona (Senate and Electoral College)

Montana (Senate)

Maine (Senate)

Kansas (Senate)

New York (House — NY11)

Wisconsin (Electoral College)

New Jersey (House — NJ2)

New Mexico (House — NM2)

Texas (House — TX10, TX24, TX21)

Florida (House — FL26 and Electoral College)

California (House — CA25)

Utah (House — UT4),

Alaska (Senate)

North Carolina (Senate and Electoral College)

South Carolina (Senate)

Minnesota (Senate and Electoral College)

Again, for context: Left is Electoral College, Center is House, Right is Senate

Links for Republicans

Iowa (Senate, House — IA1, IA2, and Electoral College)

Arizona (Senate and Electoral College)

Montana (Senate)

Maine (Senate)

Kansas (Senate)

New York (House — NY11)

Wisconsin (Electoral College)

New Jersey (House — NJ2)

New Mexico (House — NM2)

Texas (House — TX10, TX24, TX21)

Florida (House — FL26 and Electoral College)

California (House — CA25)

Utah (House — UT4),

Alaska (Senate)

North Carolina (Senate and Electoral College)

South Carolina (Senate)

Minnesota (Senate and Electoral College)

--

--

Eric Bolton
Eric Bolton

Written by Eric Bolton

Senior Software Engineer @ Applied XL. Previously: ML Engineer at Columbia University, ML Scientist at WSJ