This is an excerpt from Neel's series "5 Keys to Campaign Strategy." Read the remaining 4 parts on Medium.
When you’re working on a political campaign — either seeking office yourself or working for someone you believe in — you’ll always be operating in a tough environment. There are always too many votes to get, too many babies to kiss and hands to shake, but never enough money, time, or volunteers to do it all.
Your team has to learn the strategy for optimizing your scarce resources to win the race. Through BlueBonnet, I'm working as a data scientist on a Florida congressional race where our guy is seeking an upset win in a district tilted 13 points against him, and I’ve learned a thing or two about using math, data, and stats to squeeze every possible vote out of a district.
This post is part one of a five-part series where I’ll be sharing strategic frameworks that’ll help you win the race or, if you’re just a voter, understand how political campaigns operate and some common pitfalls that politicians, parties, and even election forecasters make. We’ll draw on some concepts from statistics, economics, game theory, math, and optimization to boot. (You can read the remaining four parts on Medium.)
The first step in campaigning is figuring out how you’re going to get to 51% of the vote (or, at least, a plurality). At the start of your campaign, the local party will usually give you a huge list of all the registered voters in your district, plus demographic and party information on each — the so-called “voter list.” Your job is to turn this list into votes.
Simply asking everyone on the list to “please vote for me” isn’t a very efficient strategy, though. If you’re a Democrat, telling a diehard Republican to vote for you probably won’t work; reminding them to head to the polls will only work against you. Meanwhile, sending campaign mail to your staffers is probably a waste of paper, since they’re going to vote for you anyway.
Four ways to interact with voters
The more efficient way to campaign is to narrow your aim to a few key groups and adopt a slightly different approach for each. To do this, you grade voters by how likely they are to support you and how likely they are to vote. This forms a handy 3x3 matrix:
The persuasion-GOTV matrix. It divides potential voters into 9 segments and shows you the right strategy for engaging with each. GOTV means “getting out the vote.”
If someone is definitely not going to vote and is definitely not going to support you, there’s no point wasting your time on them. Ignore them. The people you need to persuade are undecided or ambivalent voters — these are the people you’ll want to talk to at town halls, woo with policy positions, or otherwise engage with in the typical “politician” way. This is simple enough.
Many campaigns make the mistake of ignoring the people who support them but aren’t habitual voters. These people like you, but without encouragement, they might not turn out. This can be a huge problem: one of the big reasons for Hillary Clinton’s 2016 loss was that 4.4 million Obama voters stayed home. Clinton’s woes with the white working class are a bit overrated — if she’d turned out just a few more hesitant, Democratic-leaning voters in WI, MI, and PA, she’d be president.
“The critical ‘swing voter’ is not the Biden-Trump switch voter. It’s the individual who is weighing if it’s worth voting at all.” –Brittany Shepherd
This is why it’s so essential to get out the vote, or GOTV. In today’s polarized age, it’s nearly impossible to get someone to switch sides. It’s much easier to convince someone who likes you to show up and vote. Phonebanking, handing out sample ballots door-to-door, reminding people when and where to vote, driving people to the polls, and all the other unglamorous volunteer operations are key to this mission.
Finally, the people who love you and are going to vote for you come hell or high water don’t need to be reminded how awesome you are or how to vote (though it helps). Instead, these are the folks you’ll want to ask to donate to you or volunteer for your campaign — in other words, activate. If someone’s fired up about you, don’t restrict them to just voting for you; harness their extra energy!
Ignore, Persuade, GOTV, Activate — these are the four buckets you’ll want to lump your voters into.
But there’s an additional “Defense Against The Dark Arts” component here: unethical opponents might add a sleazy fifth bucket.
In 2016, the Trump campaign split voters into five groups: swing voters to persuade, supporters to GOTV, Trump-lovers to activate, Trump-haters to ignore, and, crucially, hesitant Hillary voters to suppress. Trump’s campaign literally had a list of Black voters — the canonical group of strongly-Democratic voters who don’t always show up to vote — that they labeled “Deterrence.” Campaign manager Brad Parscale planned to run anti-Hillary ads on Black radio stations with the intention of disillusioning these voters and getting them not to switch to Trump but rather to stay home.
“[The Trump campaign’s] aim? Not to win their votes, it seems, but to get them to stay at home on Election Day.” –Krishnan Guru-Murthy
In other words, Trump shaded in another cell in our matrix:
The Trump 2016 campaign added a special square to the usual persuasion-GOTV matrix: deterring hesitant supporters of their opponent.
(It’s unclear if Trump’s Deterrence gambit worked, but Black turnout was definitely down from 2012, and that definitely hurt Hillary.)
Deterring your opponent’s supporters is unethical and, arguably, undemocratic, so don’t do it. Just know that your opponents might, and work extra hard to GOTV among your hesitant supporters.
Expected margin impact
Trump realized that turning away a Hillary supporter was just as good for his final margins as swinging an undecided voter toward him was. He had, wittingly or not, discovered the mathematical concept of expected margin impact.
Think about it like this: how much is a given voter expected to change your margin of victory? If someone votes for you, they’re adding 1 vote to your margin of victory, but if someone votes against you, they’re subtracting 1 vote. If someone likes you but may not vote, there’s a coin flip between them voting for you (+1 vote) and them staying home (0 votes), so their expected margin impact is +0.5. A passionate voter who is torn between you and your opponent has a roughly equal chance of voting for you (+1) or against you (-1), so their impact averages out to 0.
It’s easiest to understand in this matrix:
A chartfor computing the expected margin impact of a voter.
Now you can start comparing the effectiveness of various strategies. GOTV will turn a hesitant supporter (+0.5) into a definite voter for you (+1), which swings the expected margin by +0.5. Trump’s Deterrence plan sought to turn hesitant Hillary voters (-0.5) into non-voters (0), which also swung his margin by +0.5. As I said, Trump realized that suppressing a Hillary voter was just as good as turning out his own voter.
You’ll notice that the single most effective thing you can do is to get someone who’s definitely voting for your opponent (-1) to vote for you (+1), for a margin impact of +2. That’s four times the impact of turning out one of your own supporters, but I’d argue that it’s also four times harder. The exact math will differ from campaign to campaign, but suffice it to say that you should consult this matrix as you’re crafting your strategies and figuring out how much effort to put into persuasion versus GOTV.
This matrix also explains the adage that “a vote for a third party is like a vote for your opponent.” It’s not technically correct — it adds 0 to your candidate’s margin, while voting for the opponent subtracts 1 — but it is still problematic because you’re missing the opportunity to add 1 to the better candidate’s margin.
The takeaway from all this is that you can scientifically break down your giant voting list and figure out the optimal strategy for talking to each group of voters. Tools like NGP VAN even include “scores” that give each voter a score from 0 to 1 of how likely they are to support you and, sometimes, how likely they are to vote. So you, too, can construct a matrix like the ones we made above.
Once you’ve made the matrix, you can start mathematically comparing various strategies to maximize the bang for your buck. Is it better to have your volunteers GOTV in a particular neighborhood or phonebank with swing voters? Well, if 20 of your 100 volunteers will get someone to vote, that’s a margin impact of 20 * 0.5 = 10, and if just 5 of your volunteers will get an undecided voter to pull the lever for you, that’s a margin impact of 5 * 1 = 5. With this math, you can optimize your use of limited time and money (and what campaign doesn’t have limited time and money?)
I’ll leave you with a pair of personal anecdotes. When I was a kid, my parents and I would work with the local party to drive elderly voters to the polls. My mom would tell me that these were dozens of literally free votes — the only thing stopping these folks from voting for our side was physically getting to the polls. Meanwhile, in 2016 I spent a whole day knocking on the doors of undecided voters in New Hampshire and think I might have swayed a grand total of one or two voters.
I'll be sharing the remaining four entries in this "Keys to Campaign Strategy" series on Medium, so feel free to head there and given them a read. If you'd like to put some of these strategic insights to use, make sure you apply to join BlueBonnet!
ABOUT THE AUTHOR: Neel Mehta is a BlueBonnet Data Fellow working to help flip Florida's 17th Congressional District. He's a product manager by day and a writer, civic techie, and data nerd by night.