I am rotating the voter in my mind
Thinking about voters as masses and individuals
(note: this blog was originally at cauldron.llc, and is now also a Substack, because it seems easier for folks to follow and get updates here. I will continue to maintain the blog page as well.)
When I was in college and spending a lot of time trying to understand political science papers, my friends would joke that these papers should start with “assume a spherical voter”. This is very funny if you are simultaneously taking classical mechanics and spending a lot of time making diagrams of frictionless planes with perfectly spherical objects in vacuums. We briefly considered making t-shirts.
Sphericalness aside, doing any consideration of elections larger than about 20 people requires you to simplify the voter down to an easily managed concept. While each individual may have quirks and unusual traits and unpredictable responses, The Voter as concept reacts in (somewhat) predictable ways to (somewhat) predictable forces. I have been reliably assured that while this is not true of bacteria it is annoyingly true of mice, although they lack the thumbs to write you mean letters about it.
How you imagine The Voter has impact on your approach to strategy. My experience is almost entirely in American politics, so the way I imagine The Voter places them in a strict two party first past the post voting system, with large, periodic, and high-profile national elections. We know from surveys and studies that most people vote for the same party over time. This fact is a major part of what makes modeling and targeting for political campaigns possible- if you can figure out that someone’s a Democrat (by either their registration, their past votes in primaries, donation behavior, or direct surveying), odds are good they’re going to keep being a Democrat. Whether or not they’ll vote is a whole other problem, but it’s still extremely related to how much they’ve voted in the past. People are blessedly predictable.
None of this observational stuff tells you *why* people behave like this. You can imagine a Voter who’s essentially drawing from a bag of multicolored marbles each time they go to vote, but has such a skewed number of marbles in the bag that they’re a Democrat every time. You can also imagine a Voter who’s doing careful, rational calculations each time, and similar ends up a consistent Democrat. For the practitioner, these models would have wildly different implications on how you tried to appeal to voters. Luckily for us, political science has been agonizing about this for ages and we can grab some decent models from them. Within studies of political behavior, the two models that most influenced my thinking are the Michigan Model (and many, many responses and re-responses, which we’ll get to later) and the Spatial model.
Note: This is not a great survey of political science literature, nor I am wildly up to date on current arguments.I would strongly recommend you go back to the original reading here, if you’ve got the time. For the Michigan model, see The American Voter (1960), followed by The Responsible Electorate (1966). For Spatial voting, see An Economic Theory of Democracy (1957). Also, consider consulting a reading list for an American Political Behavior course.
Spatial
Spatial voting is interesting because it is impractical when applied to an individual voter, but macro-level useful when applied to The Voter en masse. The theory argues that individuals have ideal preference points along an axis (or several axes), and will vote for whichever candidate comes closest to that preference point. This sort of single-peaked preference model requires voters to understand the available spectrum of options over which to have a preference (i.e. no taxes to 100% taxes) and to also understand where each politician lands on that spectrum (15% taxes, 60%, etc). As a model of a live political campaign, the spatial theory quickly traps your Voter in a nest of different axes for different issues, where the actual position of the politician may be uncertain or just not well advertised enough for the voter to know. (I.E. What exactly *is* Trump’s stance on abortion? If you listen to him, he’s not going to ban it, but you’d get a radically different idea if you listened to people around him, or to Democrats). As anyone who’s done polling can tell you, nailing down specifically what a voter’s ideal policy is (even on an issue they care about) is a complicated nightmare. Most people don’t have a perfect ideal point that can be mapped on an axis.
For me, the thing that’s useful about spatial voting theory is that it implies that policies matter. Michigan model theory, which we’ll get to in a second, could lead you to imagining a voting public that has no idea what policies a party or leader would enact, and doesn’t much care. This is clearly wrong. Voters do seem to have preferences, not necessarily in the political science sense of rigid thought-out ideal points, but in the sense of preferring one policy over another. Sometimes this preference is just a vague sense of “ew” that falls apart on closer examination (see opinions on “late term abortion” here), and sometimes this preference is a desire to avoid dealing with a messy issue and simply have their lives improve (immigration, homelessness, etc).
A pure spatial-voting Voter would be someone almost clinically evaluating the policy positions of various candidates, and hyper sensitive to swings on issues they care about. In practice, unless they held issue positions neatly balanced between the two parties, they’d probably end up a consistent partisan. They’d probably find our modern political environment confusing, and be one of those people who frequently cites written policy platforms on Twitter. Actually, the population of twitter seems to behave the most like spatial voters, which you can see as some pundits begin to embrace Trump as he seems to align with their narrow policy preferences. This behavior can seem cynical and craven to more partisan-aligned actors, and just sort of silly to people paying attention to Trump’s overall level of follow through (low as hell).
Michigan Model
The Michigan Model posits that party ID is the most important force in determining The Voter’s vote, that party ID is a heritable, long term, relatively stable attachment, and that issues have comparatively little impact. The American Voter is the foundational text here, almost immediately criticized for its overly dim view of voters by V.O. Key in The Responsible Electorate. At a macro level, this is pretty clearly Just Correct. Especially as politics has become increasingly nationalized, ticket splitting has declined, and even local politicians have been utterly subject to national partisan forces. Most voters will vote for whatever party they selected last time, even if the candidate or situation has changed. I always note that the best way to guess what will happen in an election is to look at what happened last time. Despite how chaotic politics can feel, there’s a lot of underlying stability.
A Voter under the Michigan Model is for the most part a partisan political animal, driven by warm feelings and habitual affiliation with their party. You could pick far worse models for explaining my personal political trajectory (raised a Democrat, vote as a Democrat, work as a professional Democrat). These Voters are likely to build their policy positions (if they have many) on top of their political identity, and pay most attention to what their party espouses as it’s core beliefs. Under this framework, ideological consistency with their party is not because voters have been sorted into parties but because they inherited that ideology from their party ID.
The American Voter is somewhat dismissive of the idea that voters might react to policy or partisan shifts in the political parties. It notes that these voters (paying enough attention and actually shifting party) are a small part of the electorate. Importantly, these high-attention switchers are not all possible switchers. Swing voters in this theory are disengaged in the extreme, not tied emotionally to a party, but simultaneously not up to date on policy issues. They are doing very nearly bag-of-marbles stuff, in which their vote might change based on what they saw that day, or who looked good on TV. (I tend to think of this as tossing additional marbles into the bag, but I don’t think this is a good model unless you personally love to think about micro adjustments to probability. And very tiny marbles.)
Both At Once
My preferred way of fitting these theories together is to say they’re both right. Most voters are highly highly partisan (this might be because they inherited that partisanship, or because they’re taking cues about what party they ought to belong to, or because they’re made a studied evaluation of the facts and chosen a party closest to them ideologically, it really doesn’t much matter in a very polarized system). They have some level of awareness of policy positions, and can be moved, but it’s a high lift to counteract partisanship. These policy positions are not necessarily coherent or well-reasoned, because most people are busy and have lives.
If you lean too hard on the spatial model, you might convince yourself that politicians have complete flexibility in which voters they appeal to. This is not practically true. Politicians exist in an environment where their party is a defining part of their identity. Being A Democrat comes with a whole host of signals about your values, your behavior, and your beliefs. Voters very reasonably interpret messages from individual politicians in the context of all the other messaging from their party and affiliated figures. If a party figure in good standing without any other “maverick” branding suddenly espouses a new policy position adopted for campaign purposes, the reasonable voter might be skeptical. You can’t escape your party brand without continued aggressive positioning, or by Not Being A Democrat (the Osborne strategy).
But elections are not entirely about Michigan Model partisanship. If they were, we’d see flat ratios of R to D over time, affected only by birth and death rates. While most people may not be movable, elections are about margins and ratios, and in the scheme of things there are a lot of people who might potentially switch their vote. And we know this matters, because we do see variation in election results over time, and we can watch certain geographies trend R or D.
Hyperfocusing on the swing voter
Okay, so there are a small % but large number of people who are likely to change their vote in response to *something*. At the same time, there is an overwhelming amount of information floating around. A standard Voter might be getting messages from 2 presidential campaigns, 2 senate campaigns, house, governor, state legislative, and 10000 outside groups. I don’t know if you’ve read your political mail recently, but there’s a ton of information contained in it. This guy wants to lower taxes, but someone else says he wants to defund schools, but someone else says he’s a threat to traditional families, but someone else says he actually just wants to preserve tax cuts for the wealthy. Which of these things are true? Which do you have to care about?
Of course, many people deal with this by throwing the damn mailers away and going on with their day. Which like, fair, it’s a lot. But we know that some information leaks in around the edges, we know that political communication to some degree “works” because people persist in running RCTs about it to double check (thank goodness). How are people parsing all of this, and how are they coming out of that with an adjusted opinion of who to vote for? Some of them, who I would lovingly call weirdos, are reading all of it and doing extensive googling. Most of them aren’t doing that. Here, we can dip back into political behavior for some potential explanations for how this synthesis and consumption might work.
Converse (Nature of Belief Systems In Mass Publics) offers the idea that the electorate is split up into groups with different “levels of conceptualization”, approximately level of attention to politics/how they form their beliefs. Level 1 is Ideologies, weirdos on twitter who compare candidates to abstract ideals of Liberalism or Conservatism, and level 5 is folks with a party ID but no concept of policy at all. I don’t entirely buy that these levels exist in complete isolation from one another, or that you can sort voters into them neatly, but it certainly serves as an excellent summary of ways voters might evaluate the options.
For my purposes, I’m most interested in levels 3 and 4, which are “group interest” and “nature of the times”, respectively. Group interest relies on assessing if a politician is broadly for or against “your type of people”. This is a useful heuristic for a voter who wants to figure out if a politician is going to be good for their interests but doesn’t have the capacity to read up on the details of what the politician plans to do. If a candidate is positioned as pro-worker, or pro-business, this gives you information depending on what category you’d sort yourself into. In modern politics, this might read as “identity politics”, but to my mind it’s really about trying to evaluate how much a candidate has to materially offer you by an easier method than reading about their policy ideas. For Democrats, this type of voter is increasingly difficult to reach, as the perception that Democrats are “for” college educated extremely liberal voters grows. Level 4 matches up with ideas about thermostatic response. The fact that prices were high due to inflation was clearly a huge factor in 2024 vote. If you assume that the party in power has some control over how things are for you, voting based on if things are good or bad makes sense. This may not be fair- politicians don’t have magical economy levers, or control over world events- but if you primarily care about politics in terms of “does this improve my life”, it’s certainly sensible enough.
Our prototypical not-very-engaged swing voter might use one or both of these strategies to figure out which candidate to vote for. They’re both heuristics for evaluating complex information without spending a ton of time and mental energy on it, relying on messaging about groups or practical effects on your life.
Practical Voter-Imagining
The most important part of constructing a mental model of the standard Voter is remembering that that person does not look like you. If you’re interested enough in politics to go through this exercise, you are probably an extreme outlier in terms of political knowledge and attention. This is not a normal set of behaviors, and you shouldn’t assume that’s how most people are. Most folks are habitual partisans, deriving their ideological positions from partisanship first, and having a middling to low amount of knowledge of policy. Most people will vote for whatever party they voted for in the past.
There’s a small percentage of people who are likely to switch their vote between elections. These folks are not ultra-high information and somehow torn between the two parties, but are disengaged and not paying a ton of attention. They’re also more likely to have incoherent and heterodox policy views.
When you’re trying to communicate about your candidate’s ideas and brand, the vast majority of what voters will hear is your party affiliation. Deviating from the national brand is difficult. When you do manage to communicate about your specific policy ideas, that messaging has to filter through all the heuristics voters use to evaluate candidates. If you’re proposing increasing the minimum wage, but the general vibe is that you’re focused on rich people, that policy idea is going to get filtered out and ignored. Voters are looking around at their friends, family, and people they generally consider to be “like them” to figure out if you’re “good for them” in a way that probably won’t tie back to any single specific idea.
Policy can have real impacts, but so can general world happenings that may have zero to do with you and your campaign. The literature on shark attacks and votes for president may be a little sketchy, but a bad economy, high inflation, or a pandemic, are all going to take a toll on incumbents.
So, before you get incredibly bogged down in observing a single voter someone found in a diner, it’s worth taking a step back and asking yourself if this person is usefully close to the average voter, or idiosyncratic in a way that makes them a terrible representative. If they’re incredibly invested in trade policy or something, they’re probably a bit odd. If they’re operating on pure vibes plus a generalized sense of self interest, that might be truly average. Having a conception of how voters en masse operate can help you avoid some rabbit holes on individual people’s beliefs. That’s not to say those individuals aren’t important (they are!), but when you’re only able to hear from a tiny tiny slice of the electorate, theory can help fill in for that unheard from mass.