We need to talk about voting behavior. Working class support for both Donald Trump and Brexit is striking, but what seems to be happening is more general. Why do so many people now vote against both their apparent economic self-interest and the political traditions within which they were brought up?

Mainstream academic approaches to analyzing why and how people vote date back to the 1950s and have little to say about this.

One approach is empirical, inductive and essentially sociological. It built on the post-war revolution in high quality academic survey research which began with the American National Election Study (ANES), was emulated in many other countries, and evolved into what became known as the ‘Michigan School’. The core argument is that people are socialized, by various mechanisms, to have underlying propensities to support certain parties – to have certain levels of ‘party identification’. In aggregate, party identification in the electorate as a whole provides the long-term baseline upon which the short-term ebb and flow of competitive party politics plays out. Especially in countries, like the United States and to a lesser extent Britain, where electoral turnout is relatively low, winning elections is on this account more about ‘mobilizing the base’ – of core party identifiers – than about seducing supporters of rival parties.

The other mainstream approach is theoretical, deductive and essentially economic. A political reworking of simple microeconomic theory, this models political interaction as ‘rational choice’. People are assumed to have self-interested, intrinsic preferences and to act “rationally” in ways that help satisfy these. When deciding what to do at election time, rational voters behave instrumentally. They evaluate each party’s policy offer and cast their ballots in ways designed to contribute to expected policy outcomes as close as possible to those they most prefer – to their ‘ideal points’. On this interpretation, electoral politics is precisely about parties trying to seduce the supporters of their rivals, by strategically manipulating their policy offers. In US presidential elections, for example, successful candidates are expected to make one, more extreme, policy offer to their party’s primary voters, moving towards the centre ground in the general election to appeal to the electorate at large.

Each approach to analyzing voting and party competition has generated mountains of literature over the decades. If we are honest, however, neither has much to say about why, in the US for example, lifelong working class Democrats switched to supporting the wealthy property developer and TV celebrity Donald Trump, or why, in Britain, many working class voters fervently supported a Brexit which is, objectively, more likely to harm than to help them economically.

Alex Schuessler’s beautifully written book A Logic of Expressive Choice (NB not THE Logic of Expressive Choice) was published in 2000 and now seems ahead of its time. Starting his training in anthropology but graduating in political economy, what Schuessler offers us with refreshing modesty are “contours of a logic of expressively motivated choice” (p. 5). In a nutshell, on this account Trump or Brexit voters are not voting instrumentally to try and influence an election result, but expressively, to become part of a movement with which they want to associate. They are not trying to produce something when they vote, they are consuming something.

While this might make the notion of expressive choice seem to blend old-school party identification with old-school instrumental voting, there is a powerful and crucial added ingredient. The socialized preferences and/or party identifications of adult voters are not assumed to be fixed for all time, but to evolve endogenously as a result of social interaction with others. As we will see, this helps us to understand political bandwagons and ‘momentum’ – phenomena which have largely confounded mainstream approaches to making sense of voting behaviour.

Schuessler motivates his argument with a simple, arresting, and memorable example, set in an Italian-American café in Boston’s North End. This has a jukebox with a repertoire of classics such as Frank Sinatra’s New York, New York and Dean Martin’s Volare. Like most jukeboxes, it plays each song once, no matter how many customers have paid 50 cents to select it. People effectively make their selections in private, though can of course go back to their table and tell their friends what they selected.

This jukebox has the unusual feature of displaying the day’s most popular song. New York, New York was typically the most popular, though occasionally it was Volare. The big question concerns why anyone pays 50 cents (in twentieth century money) to play a song they know – almost for certain because the jukebox actually tells them so – will shortly be played anyway, at someone else’s expense? In economic terms, if you want to hear New York, New York, why don’t you simply free ride, enjoying the song but letting other customers pay for it? Schuessler’s answer is that you don’t feed the machine 50 cents because, instrumentally, you want to produce the voice of Frank Sinatra (since putting nothing into the machine will have the same effect). Your 50 cents lets you consume an association with Frank Sinatra – “I chose that song” – allowing you to express your “Frank Sinatra-ness”. The very act of paying 50 cents to select New York, New York is itself part of the pleasure arising from your visit to the jukebox, as much as listening to the song.

And so it is, according to Schuessler, with voting. Voters don’t turn out to vote because, instrumentally, they think they might influence the outcome of the election. They vote because, expressively, it makes them feel part of something.

Jukebox to ballot box might seem a huge and improbable jump. Not really, however, when you think about how much of the ‘real’ economics of capitalist societies, in which voters spend their everyday lives, is characterized by expressive appeals to consumers. Generations of Coca Cola ads, to take an example leveraged by Schuessler, are about the sorts of people who drink Coke, people with whom you can associate if you buy a Coke. These ads long ceased to be about the relative merits of the various fizzy drinks on the market. The same can be said for fashion and sportswear branding. Consumers pay a premium to proudly display brand names with which they want to express an association, despite the fact they can easily get functionally equivalent goods for less money. The entire advertising industry at the heart of most capitalist economies is fundamentally about creating and manipulating expressive desires among consumers. Why should we expect things to be different when those very same consumers are also voters?

Schuessler provides strong anecdotal support for this argument with a potted history of the emergence advertising in the “real” economy – sketching the Coke-Pepsi wars – and its political analogue. Pepsi nearly went bankrupt with a classic instrumental appeal – ‘twice as much for half the price’ – switching very successfully in the early 1960s to an expressive appeal that had nothing to do with facts – ‘come alive! You’re in the Pepsi Generation!’. Following suit, Coca Cola switched from the instrumental ‘cold crisp taste of Coke’ in 1958 to the expressive ‘I’d like to buy the world a Coke’ by 1971.

Advertising in US presidential elections followed a similar path, culminating in the seminally expressive 1984 ‘morning in America’ campaign by Ronald Reagan. This is no accident because ‘[n]ot only do the marketing of soft drinks and marketing of presidents share a common origin, but to the present day, they share many of the same practitioners’ (p. 69). In each case, slick marketing of fundamentally expressive appeals has proved very effective. According to Schuessler this underlies the failure of both ‘advertising blind economists”’ (p. 69) and ‘political economists modeling electoral behavior’ (p. 70) to understand how their respective markets really work.

Circling back briefly to Trump and Brexit, Schuessler’s argument implies that neither Trump nor Brexit voters are behaving instrumentally with the intention of furthering their underlying economic interests. Rather, they are buying into the expressive appeal that many people ‘like them’ are supporting Trump and Brexit. By jumping on this bandwagon and associating themselves with Trump (to “Make America Great Again”), or Brexit (to “Take Back Control”), they become part of something that makes them feel better.

Schuessler does not make explicit predictions about how the manipulation of expressive appeals to voters might develop in real world politics, but his “speculation” about role of the emerging media market in this is prescient. It is worth quoting in full:

… with the proliferation of customized information channels via cable television and Internet distribution, the potential for providing demographically and psychographically targeted narrowcast information is tremendous. At the same time, such customized information delivery loses its crucially important common-knowledge dimension: recipients can no longer believe that they are witnessing a public event (that is, implicitly witnessing others’ witnessing, or being witnessed by that witnessing) as they receive such information. Consequently, the most potent feature for mass advertising and campaigning appears to involve the use of a two-tier, simultaneous narrow- and broadcasting strategy. Such would provide specific product or candidate information via customized channels, while at the same time allowing this tailored information to resonate with the ambiguous, symbol-intensive portrayal of the product or candidate in the explicitly public realm. On informal observation, this appears to be a strategy gradually emerging in the product market, and there is enormous potential for it in electoral campaigning (p. 88).

Schuessler could not possibly have foreseen the sophistication and precision with which today’s electorate is segmented and then micro-targeted by very widely-used information channels on social media controlled by vast multinational corporations – though this adds force to his argument. But he did foresee that campaigns based on narrowcasting such as this must also be coordinated with broadcast campaigns stressing the more general expressive appeal of the political “product” being marketed. He didn’t predict Trump or Brexit, of course. But he did describe a change in the political and media climate that led to marketing campaigns asking voters in effect to “join the Trump generation” or telling them that “things go better with Brexit”.

This is all very well at an informal level – the jukebox analogy really sticks in the mind – but where does it take us? Can we turn these insights into a more systematic and up-to-date approach to analyzing and predicting voting behavior in mass electorates? If the argument is no more than that people turn out to vote for certain candidates or options because they’ve been manipulated by advertisers to feel good about doing this, in the same way they feel good about buying a Coke, then Schuessler is well aware that ‘… we essentially risk reducing the status of our approach from explanation to mere description’ (p. 47).

The second half of the book, therefore, sets out to sketch the ‘contours’, as Schuessler puts it, of a more formal model of expressive voting. This model builds on instrumental voting models by adding an expressive component to each voter’s cost-benefit calculus, and then deploying the types of argument currently used to explain purely instrumental voting. This is a reasonable and sensible approach, but very much a work in progress, in part for reasons not normally discussed in a book review. Alexander Schuessler departed the academy not long after publishing A Logic of Expressive Choice. He left to found what became a highly successful private sector start-up, grounded at least partly in the “momentum” and “bandwagon” effects he describes in his only book. A regular academic career would have taken the model contours set out in the latter part of this book and systematically filled these in. The author would have written and published technical papers and articles refining and strengthening the core ideas, and promoted these at faculty seminars and conferences around the world. This is how a new model gains traction in the profession, but none of this happened. The potential of the model sketched in A Logic of Expressive Choice is therefore yet to be realized.

This will not be an easy task. The key question concerns whether we can construct a systematic argument about the creation of expressive value for voters. (Schuessler does not really do this.) If we are forced to explain this in terms of the dark arts of marketing gurus, then we may not have achieved very much. This is a little unfair since the deep question ‘where do our preferences come from?’ is also problematic for more traditional approaches to analyzing voting behavior.

Instrumental accounts of ‘rational’ voting, for example, typically take voters’ fundamental preferences as given, ignoring how these are generated. They do not model the endogenous evolution of preferences, and for this reason find it difficult to account for bandwagons. Taking fundamental preferences as given, however, mainstream “spatial” models of instrumental voting have built on early work by Hotelling and Downs to develop a comprehensive “common space” technology for analyzing voting and party competition (Hotelling 1929, Downs 1957).

In the real world, of course, voters’ preferences are high-dimensional – they concern a wide variety of very different matters. In the world of theoretical modellers, however, drastically reducing the dimensionality of this high-dimensional preference space has proved a very fruitful way forward. This works because voters’ preferences of bundles of different issues tend in practice to be correlated. If you know Voter X opposes abortion and gay marriage and supports capital punishment, you can probably predict X’s position on open carry gun laws. Of course you won’t always be right, but you will typically be right, which is what matters. These correlations allow us, in this case, to use a “latent” liberal-conservative dimension of social policy to summarize voters’ preferences. We can similarly summarize voters’ preferences on another important bundle of issues using a latent left-right dimension of economic policy. As it happens, we can summarize lots of important information about voters’ preferences using just these two latent dimensions. We can however add other dimensions, for example, referring to environmental or foreign policy, if these usefully add to our description of the political world.

This low-dimensional representation of voters’ preferences, taken as a given without going into where these preferences come from, underlies several generations of “spatial” modeling of instrumental voting and party competition. Parties and candidates compete with each other by staking out positions on particular policy dimensions. Naïve voters support the closest party. More sophisticated voters take account of how the electoral system transforms votes cast into election winners. In light of this they strategically support the party or candidate they judge most likely to deliver policy outcomes close to their ideal points.

Crucially, this theoretical approach can be applied to systematic empirical research, given widely accepted methods for estimating voters’ preferences (opinion surveys) and party policy positions (expert surveys or systematic analyses of party manifestos). The entire approach would have had far less traction, would in effect have been a Sudoku, if there had been no agreed way of measuring voter preferences and party positions and in this way applying spatial models to the real world. All of this is by now a mature technology, subject to incremental refinement and improvement, and has generated an immense literature about voting and party competition. Schuessler now adds an expressive component to voter utility, over and above issue preferences, so is there an analogous systematic way to describe and analyze this?

To see the what is a stake, let’s return to the problem with which Schuessler motivates his book, the problem of why people vote, a.k.a. ‘the paradox that ate rational choice theory’ (Fiorina 1990). If the net return to voting for some voter is R, the benefit of affecting the result of the election is B, the probability of effecting the result is p, and the cost of voting is C, then R = p∙BC a rational instrumental voter will turn out if p∙BC > 0. Since p is infinitesimally small in large electorates, turning out is not instrumentally rational for almost any small cost of voting. Riker and Ordeshook ‘solved’ this problem in 1968 by adding the positive benefits of complying with what they describe as ‘civic’ duty (D) to the turnout calculation (Riker and Ordeshook 1968). On this account, turnout is instrumentally rational if p∙BC + D > 0. Assuming that voters are instrumentally rational and observing that many of them do in fact turn out to vote it must be that D > C – the civic duty benefits of voting must outweigh the costs.

Schuessler echoes Ferejohn and Fiorina in observing that this solution has no analytical bite (Ferejohn and Fiorina. 1974). Adding civic duty to the calculus ‘… even if the term substantively is a correct one– reduces its purpose to that of a sponge … ‘(p. 34, emphasis in original). Absent any reasonable way to systematically measure the fundamentally unobservable value, D, of civic duty, this becomes a purely descriptive equation filler which can be used to classify any voter’s turnout decision as instrumentally rational. The lack of bite arises, not in the logic of the argument but because the value of the D term is fundamentally unobservable, and might indeed be anything which explains the behavior under investigation. The result has been that attempts to explain turnout as instrumentally rational have largely been abandoned. Notwithstanding this, models of instrumentally rational behavior by voters, once those voters have turned out for reasons which are not explained, have continued to be an important part of the mainstream.

Schuessler is more consistent in insisting that we should not assume the preferences which cause voters to make the journey to the polling booth are quite different from the preferences which guide their behaviour once they have got there. But in effect he does this by replacing Riker and Ordeshook’s D term with his own E term, explaining voter turnout in terms of the expressive value of associating with one of the particular options on offer. Does the same problem not arise, therefore – once more even if the term substantively is a correct one? Crudely speaking, even if the logic of the arguments in the second half of A Logic of Expressive Choice is impeccable, where do these arguments get us if we don’t have a reasonable method for systematically measuring the value of the E term, the expressive value to a given voter of attaching to some particular outcome? Surely expressive benefits are as fundamentally unobservable as ‘civic duty’? If so, is this another rhetorically compelling argument – and it is indeed rhetorically compelling – that ultimately lacks analytical bite?

Deeper questions concern the sources and exploitation of expressive value. While spatial voting models say nothing about the sources of intrinsic voter preferences, they do explicitly analyze what party strategists can do to exploit these. If I am a Trump or a Brexit strategist, however, what does Schuessler tell us about how to both enhance and exploit the expressive appeal of my client among key sections of the electorate? What, in short, can we say systematically about the mysteries of political tradecraft in a world of expressive appeals? I don’t know the answer to this question, but it is the sort of thing Schuessler would have worked on had he stayed longer in the academy.

Part of the solution will be to find a way to systematically describe expressive benefits in a manner analogous to spatial descriptions of intrinsic voter preferences. Are there, for example, dimensions or types of expressive value that Trump voters derive from supporting Donald Trump? Presumably the marketing gurus who promoted Trump and Brexit talked and argued with each other in systematic terms, and did not just come into the office to recount their latest dream or brainwave. These were successful marketing campaigns, so there must be something systematic to be learnt here. We need to understand much more about the production and exploitation of expressive value.

There are two types of answer to the problem of the unobservable value of the E term. The first is that, for all the equations and figures in the second half of the book, the fact that the E term is fundamentally unobservable will stand in the way of building a comprehensive technology analogous to the spatial voting model. How would we apply the approach systematically to analyse, for example, the German general election of 2021? We can certainly use the language of expressive voting, with the benefit of hindsight, to tell a convincing story about this election, but can we use it systematically to predict the result ex ante?

It’s quite possible this problem will be solved by a hack. The sophisticated and mature survey methodology used by the Michigan School, for example, deploys measures of “party identification”, a quantity which is both synthetic and unobservable. These measures have the virtue of being widely accepted in the profession, even if it is difficult or impossible to demonstrate that they validly capture the underlying theoretical construct. In the same vein, it is clearly possible to imagine questionnaire designers coming up with a battery of questions that the profession would accept as validly measuring the expressive value of particular campaign stimuli to individual survey respondents, thereby adding crucial legs to the argument.

The second type of answer is that a model can have value even if it’s not possible to measure all of the key quantities involved. The acid test for a good new model is whether it tells us something we didn’t already know. The arguments about political momentum and bandwagons in the second half of the book are convincing demonstrations that this approach does have something new and important to tell us. While expert political commentators often describe political campaigns in terms of bandwagons and momentum, which clearly featured in the Trump and Brexit campaigns, mainstream academic approaches have little or nothing to say about these.

Schuessler’s arguments about bandwagons and momentum depend on an assumption and/or observation that the expressive value of some stimulus to a voter depends in part on how many others have chosen to associate with it. Fashion brands and Coca Cola provide expressive benefits to you because you know that other people you esteem are consuming them. The job of marketers is to make sure that you know this. It’s easy to see, therefore, that bandwagons and momentum can be generated if the expressive value of associating with a particular candidate increases as more people come on board, and some tipping point has been reached. Precisely what that tipping point might be in any given case will be hard to say, given the difficulty of measuring expressive benefits, but highlighting the possibility of momentum and bandwagons, when no mainstream model does this, is an important achievement for Schuessler’s approach. It shows he is onto something.

In conclusion, A Logic of Expressive Choice makes a powerful rhetorical case that many people turn out and vote because this allows them to attach themselves to particular candidates or causes which provide substantial expressive benefits. Coming before the era of Trump and Brexit, both explained much more easily in terms of expressive than of instrumental behavior, and before the evolution of relentless and sophisticated micro-targeting of expressive appeals to voters by giant multinational social media platforms, the book seems remarkably prescient. The fundamental unobservability of the expressive benefits of some stimulus to individual voters means that this approach is hard to operationalize. But this does not mean that it is wrong, and progress might be made in proxy measurements. Ultimately, the explanation of bandwagons and momentum is both intriguing and impressive.

Good books often provoke an argument more than they resolve one; this book is an excellent example of this. If Schuessler had stayed in the academy, his argument would likely have grown more legs. As things stand, it sits there waiting for some bright spark to pick it up and run with it.