Is Democracy Bunk?

Attention conservation notice: Christopher Achen and Larry Bartels have written a very good book that deserves a wide audience. Their critique of behavioralist and liberal (Manchester not FOX) verities raise real doubts about how political scientists and others study and justify democracy. Their points also matter for prominent IR theories of foreign policy behavior. Their book suffers from a lack of comparative perspective, an overreliance on cross-sectional observational data, and some presentist biases in their history–all of which argue for more, not less, research in their program.


High school civics teachers across the United States preach a happy catechism of the virtues of American democracy. The people form a body of free citizens. Endowed with the power to vote, these citizens choose representatives to advance their interest in lawmaking and enforcing the laws. Those representatives act according to the will of the people, and should they disobey, they will be replaced through the peaceful revolution of the ballot box by a new representative who will serve the people’s bidding. Democracy thereby constitutes a self-correcting machine for the translation of the wishes of the people into the best possible policy.

In Democracy for Realists (2016, Princeton University Press), Christopher Achen and Larry Bartels refute that litany. In place of the happy vision’s individual citizens calculating whether the government of the day has succeeded, they describe a world in which individuals invent facts to justify the positions their group identifications has supplied them with. In place of a citizenry rationally deciding that the government of the day is competent despite setbacks beyond its control, they show that even an exemplary president is apt to lose if some event beyond his or her control causes a spike in prices or joblessness. And instead of a democratic system correcting its errors and improving the policies it produces, they depict instead a myopic Leviathan randomly lurching from policy to policy, reversing itself on a whim, responding only to the tyranny of popular opinion.

Achen and Bartels are not the first political scientists to bemoan the structure of American democracy, or even the first within the past several years. Jacob S. Hacker and Paul Pierson claimed that rising economic inequality had generated a new class war in which the rich were soundly thumping the poor in Winner-Take All Politics: How Washington Made the Rich Richer–And Turned Its Back on the Middle Class. Nicholas Carnes documented how the professional-class origins of most government officials led the government to systematically underrate the policies important to the poorer in White-Collar Government: The Hidden Role of Class in Economic Policy MakingMartin Gilens mapped out similar arguments in Affluence and Influence: Economic Inequality and Political Power in America. Bartels himself earlier wrote Unequal Democracy: The Political Economy of the New Gilded Age, which argued that Republican presidents are entirely happy to manipulate the economy in their interests while hapless Democrats are left behind to clean up their mess.

Like those earlier works, Achen & Bartels are soundest when they go after the least considered, but most widely held, beliefs about how American democracy functions. They reserve their strongest broadsides for what they call the “folk theory” of democracy. (Despite its name, they make clear that they think that a great many experts are reluctant to abandon this core faith of American democracy.)

Democracy, they admit, has some advantages. Its chief advantage relative to dictatorship is in generating legitimacy for rulers (p. 317). The inevitable defeat of the ruling party in a subsequent election also guarantees turnover in office (p. 317), while the legitimacy of elections itself generates “loyal oppositions” who can resist the incumbent government without resorting to treason. Finally, the democratic process both assists in the cultivation of pro-social behavior–all the tenets of “good citizenship” those high school civics courses meant to teach us–and deters politicians from overtly misusing their office.

Yet Achen & Bartels take pains to show that these justifications for democracy–legitimacy, turnover, loyal oppositions, and good citizenship–are not the same as those that the folk version of democracy sanctifies. It is, in fact, probably impossible for any democracy above the scale of a small group to really produce a voting system that would really reflect anything called the will of “the people.” As Achen & Bartels note (p. 320), “One need not spend much time in Washington, D.C., or any other capital city of a democracy to learn that, for the great majority of issues that the government decides, no ‘ordinary citizens’ are actually involved in the policy-making process.” Limiting their critique to Washington actually overstates the degree of public involvement that most citizens can expect to have. The same observation would apply with practically the same force to Sacramento, the New York City Council, or even the Advisory Neighborhood Commissions that handle zoning and alcoholic-beverage permits for DC’s many neighborhoods. Even the activists and others who do show up to monitor their elected and unelected officials’ behavior hardly constitute a representative class. As Edward Carmines and James Stimson observed, “Political activists–those Norman Rockwell good citizens who give money, ring doorbells, and wear buttons with candidate, not designer, names–are abnormal. Their interest is atypical. Their conspicuously political behavior, in a culture of mass noninvolvement, is deviant.”

Achen & Bartels go beyond earlier writers. The problem as they see it is not only that American democracy now fails to provide the results that we want from it; instead, they argue, it is that the entire apparatus by which we justify and study that democracy–a melange of liberal and Enlightenment nostrums–doesn’t correspond to how people think or act. In this, they go far beyond the normative path of a book they do not cite, Bryan Caplan’s The Myth of the Rational Voter: Why Democracies Choose Bad Policies (2008), since these are not deficiencies that more education (and less government) can remedy. Rather, they argue, we need to reboot the study of democracies and democratic governments by considering how group attachments and identities supply otherwise uninformed voters with beliefs that trump facts (pun very much intended).

Why Democracy Doesn’t Work As It Should

Two observations make up the core of their critique of existing arguments for democracy. Both the folk theory and more sophisticated defenses of democracy require that voters be able to accurately assess how governments have performed in office. One mechanism through which voters should accomplish this is called “retrospective voting”–casting ballots about the competence of politicians (a trait that will influence their future performance) by inferring that competence from their prior performance. If, in other words, voters want the economy to grow, then they should seek to ascertain whether the incumbent government’s policies and administration helped that growth or not. This task is made harder by the fact that judging a government’s competence requires separating the government’s performance from the random events that also affect economic performance.

In an ideal world, voters would only reward politicians for good economic growth if those politicians had brought it about, and similarly would only punish them for poor performance if they had caused it. Factors beyond the governments’ control (a drought or the discovery of oil, say) would not affect voters’ evaluations. Yet that is precisely what the data do not show. Voters apparently hold governments responsible for everything that happens under their watch, good and bad. This “myopic retrospection” means that politicians will often have an incentive to pursue policies that will lead voters to reward them in the short term even if it harms them in the long term. (For a review of this literature, see Andrew Healy and Neil Malhotra’s work, starting with a 2013 essay in Annual Review of Political Science.)

Achen & Bartels’ best-known vignette along these lines concerns voting patterns in New Jersey during the 1916 elections, in which, they meticulously argue, voters who lived nearer the Jersey shore punished incumbent president Woodrow Wilson for a series of highly-publicized shark attacks. Even Wilson’s greatest critics would hardly argue that he should be faulted for the actions of sea monsters. But voters did blame him. Such failures of voters to assess what is good and what is bad in government performance wrecks the theory on which democratic responsiveness is based. “The result is that, from the viewpoint of governmental representativeness and accountability, election outcomes are essentially random choices among the available parties–musical chairs….[that] simply put a different elite coalition in charge.” (p. 312)

The problem goes even farther. Not only do voters fail to properly attribute responsibility, they also fail to apprehend reality. A precondition of most justifications for democracy, and an assumption shared by most models by which economists and social scientists study political life, is that some facts are shared by everyone. Yet a long series of studies show that when we think about politics–even about objective questions, like whether inflation has increased or decreased over a president’s term–people do not answer objectively but with partisan-tinted glasses. (An example not from Achen & Bartels: Even though 63 percent of Americans supported invading Iraq in 2003, only 38 percent recall supporting it today. Democrats are likeliest to have forgotten their earlier support, in line with their party’s new position on the war.)

The consequences of these partisan filters prove profound for the folk theory. It’s unsurprising that, when a new issue arises, voters will sort themselves into a party that agrees with them on that new issue. It’s much more surprising that voters are even more likely to take their party’s opinion on that new issue as their own. And it’s devastating for many theories that voters will believe, against all evidence, that their party already shares their opinion even when it doesn’t. Achen & Bartels diagnose this second cluster of behaviors as ways in which “it feels like we’re thinking”–that is, people receive their senses of identities from party and other group identifiers, and then act in accordance with beliefs that accord with those identifications, instead of actually working out positions on their own.

Some of this recapitulates what tribes of political scientists have been arguing for generations. An early classic, the 1960 book The American Voter, for instance, described the ways in which partisan identities and loyalties shaped vote choice in similar terms. Michael X. Delli Carpini and Scott Keeter similarly documented just how little American voters knew about politics in a 1997 volume (What Americans Know About Politics And Why It Matters). But many earlier writers did, in Achen & Bartels’ reading and in my own, nevertheless shy away from the more pessimistic implications of their arguments, assuming instead that democracy will somehow all work out right in the end. “Defenses of the conventional faith, conceding a few difficulties but affirming the fundamental verities, generally predominate in both popular and scholarly conversation,” they write (p. 298).

Achen & Bartels do not embrace any such illusions. They insist that the evidence must be followed to its conclusion, and that the folk theory and its zombie progeny must be destroyed: “The folk theory is like the ether theory of electromagnetic and gravitational forces: it is based on 19th-century intellectual foundations, and the empirical evidence has passed it by” (p. 99). Some popular candidates to replace it, like participatory or deliberative democracy, they dismiss as unworkable, both because such theories also assume an Enlightenment-derived rational individual citizen as a locus of reason and because they neglect the fact that not everyone is equally prepared to participate in deliberations, leading such outcomes to be at least as de facto unrepresentative as the current theory. Instead, they argue, reforms to create a more “effective democracy…should not simply [try] to maximize popular influence in the political process, but to facilitate more effective popular influence” (p. 303). Parties, donations, and campaigns do not thwart democratic responsiveness; they instead remain fundamental to it.


Achen & Bartels aim their work at scholars of American government and, secondarily, democratic theorists. They sketch a research program that puts social groups back at the core of individuals’ identity and, therefore, models of both individual voter behavior and of social choice and institutions.

Although they do not mention it, their writing matters a great deal for scholars of foreign policy, who often rely on models that make precisely the same assumptions about reason and reward as the models of domestic politics A&B shred. Any model of IPE in which politicians are assumed to choose policies over trade to minimize post-hoc voters’ judgment should, at a minimum, be viewed more skeptically; the fact that trade agreements face substantial popular opposition when such models assume that they should be overwhelmingly popular should only reinforce such skepticism. Similarly, if scholars believe that war is a random outcome, then the case for risk-seeking democratic leaders engaging in bellicose behavior (say, a presidential candidate behind in the polls) might be much greater under Achen & Bartels’s terms then (I think) current models assume. (We’re limited here by the fact that most of our relevant observations derive from a period of relative international stability.) More crucially, models that imply that voters evaluate competence of leaders based on war performance fall prey to exactly the same critiques Achen & Bartels level here.


There are, as ever, some caveats to be drawn here. I wish Achen & Bartels had either done more to justify their focus on the American case; although they nod to comparative work (p. 49), they also state that “there is little evidence to suggest that changes in electoral institutions will be sufficient to ensure popular control of public policy through electoral competition.” Maybe, although that’s setting the bar awfully high. Certainly, there’s plenty of evidence for a weaker claim–that different institutions induce variations in the ability of voters to monitor and sanction government performance (e.g. Powell & Whitten Palmer & Whitten, Whitten & Palmer). Achen & Bartels take the U.S. as representative, when it is instead an extreme outlier. As Gerber & Green write, “The United States has the busiest election calendar on earth. Thanks to the many layers of federal, state, and local government, Americans have more opportunities to vote each decade than Britons, Germans, or Japanese have in their lifetimes. (p. 1)” It may be that many of Americans’ problems derive from the combination of the volume and the complexity of their system. And, to be fair, it has traditionally been rather more difficult to demonstrate that policy failings derive from the action or inaction of one branch or one party.

With all that said, I think that the Achen & Bartels case on retrospective voting might be stronger than it seems in some regards than it seems. Grigore Pop-Eleches’s work on protest votes in post-Communist parties seems in fact to suggest that myopic retrospection is alive and well in Europe, and contributing to extremism and destablization there. Similarly, King et al’s study of “Ordinary Economic Voting Behavior in the Extraordinary Election of Adolf Hitler” (which A&B don’t cite) accords with Achen & Bartels’s claim of the general anti-incumbency swing of the early years of the Great Depression.


I also think that the debate over the role of sanctioning error might be a little overblown. Michael K. Miller argues that myopic retrospection is bad, but “the main contribution of random, nonpolitical factors such as sports outcomes, weather, and shark attacks is noise.” And some of the more famous non-Achen & Bartels claims for myopic retrospection have encountered some sophisticated objections. It may be, instead, that the chief effect of myopic retrospection is long-term suboptimal investment in prevention relative to flashy mitigation, as Healy & Malhotra argue in their study of U.S. natural disaster policy. Yet that would still redound to Achen & Bartels’s side, given global problems of climate and environmental mitigation.

As a minor point, I think there are some ways in which Achen & Bartels cherry-pick (lemon-pick?) U.S. history. Their book is monumentally well-informed historically, and it seems churlish to fault them on the issue, but when an author identifies party loyalties as critical tribes in determining voting history, I want to understand the processes by which such loyalties change. Parties took generations to evolve as mass institutions, and some parties (like the Whig Party) collapsed, while others (like the Democratic Party) survived despite massive failures. Moreover, we know that one polity in the United States evolved and lived (thankfully only for four years) without political parties (see Jenkins 1999 and Jenkins 2000). It’s also hard to reconcile the authors’ pro-James Madison viewpoint with the inconvenient fact that Madison’s political realism couldn’t accommodate himself to the “mischiefs of faction”. Nevertheless, here again the opportunities for future research seem great: why did major social movements, like the Women’s Christian Temperance Movement, flouish and then vanish? Did such movements exert similar claims on citizens’ loyalties? What about people who “irrationally” opt-in to third parties like the Greens and the Libertarians? Studying how third-party filters affect voters’ choices would be difficult but theoretically rewarding.


Finally, by rethinking the roots of representation and citizens’ behavior, Achen & Bartels have perhaps inadvertently opened up some big questions again. On the methodological and empirical side, do the techniques of “normal” social science (“garbage-can regressions“, as Achen called them) even begin to approximate the sorts of questions we want to study? After all, such data presume that individuals can be studied as monads, but Achen & Bartels point toward dismantling such methodological individualism. Can we take self-reported labels for group identification as measures of group pressure? Say what you will about party ID, but at least we can cross-check those sorts of labels with other data sources; there’s no registry (thankfully) we can check to see if someone is really black, or a southerner, or a Red Sox fan. (Scholars of religion and politics have long faced similar questions.) What would it mean to place a complex notion of identity–which is largely not manipulable–at the center of political-science theorizing? True, social psychologists claim to manipulate such factors through priming and other framing effects, but I’m not sure we want to borrow another field’s replication crisis just now.

Many scholars would answer that we already have answers to these questions, but that they’re not explicit and probably not very good. But Achen & Bartels should probably talk more about what the implications of their ideas mean for working social scientists (at a minimum, we should probably become as familiar with social-network analysis as we are with event history analysis).

Let’s roll the dice on democracy

It is, however, ultimately the normative questions that matter more. Achen & Bartels argue that group identities structure cognition–that instead of the individual rational judge serving as voters, social identities teach people how to think about what. This, however, sets up a trap. After all, the reason that the high-school civics faith matters is because we believe it matters.Without Enlightenment myths to serve as positive arguments for democracy, instead of Achen & Bartels’s (and others’) arguments against dictatorship, then why should anyone ever act as a democrat? And if nobody acts identifies or acts as a democrat, then how can democratic institutions generate the legitimacy and security that Achen & Bartels advance as the advantages of democratic systems over others?

Indeed, without normative myths behind them, one might conclude that a system of choosing rulers purely at random would suffice to guarantee rotation in office. That method, known as sortition, was a feature of many pre-modern democracies. Given that Achen & Bartels already view modern elections as quasi-random, it would seem almost Pareto-improving to go all-in on randomness; at least that way we would spare ourselves the expense and tedium of presidential contests. It is interesting to think through what sortition would mean for the United States; at a minimum, we would have more minority members of the legislature, and many fewer millionaires in office. Any transition to such a system would pose real questions for our new Solons–should government chosen at random be limited or expansive in its powers? And it would be interesting to see if such a system would lead over time to greater or lesser social provision or bellicosity (I suspect both). At the very least, thought experiments enshrining randomness rather than hiding the randomness already inherent in the system would force us to reconsider once again whether the argument for democracy is normative or instrumental.