Survey Article: Unfounded Pessimism about Electoral Accountability

I. EMPIRICAL FACT PATTERNS

Evidence that voters are ignorant of public officials’ choices or government policy, drawn from a substantial literature in political science, may seem particularly puzzling when juxtaposed with empirical results about accountability that have been broadly ignored in normative democratic theory.⁸

There is now a large body of work, based on explicitly experimental or quasi-experimental research designs, that provides powerful causal evidence that weakening mechanisms of electoral accountability exacerbates problems of corruption, while strengthening these mechanisms benefits voters. Consider the following examples:

• A study on Brazil explores the effects of introducing an electronic voting system that made it much easier for illiterate citizens to cast valid ballots. It finds evidence that this reform significantly increased redistributive programs, including spending on healthcare for the poor, substantially reducing low-weight births.⁹

• Evidence from a study on the Voting Rights Act in the United States suggests that the elimination of literacy tests led to higher public spending in counties with larger black populations, with the effect of substantially improving educational enrollment and services for black citizens.¹⁰

• A study on the expansion of voting rights to women in the United States shows that, within a year of suffrage extension, public health spending increased by 35%. This extra spending was linked to a significant drop in child mortality.¹¹

• A study on term limits in Brazil shows that the incumbent mayors who could not run again because of term limits misappropriated 30% more funds than their (otherwise similar) counterparts who were eligible for re-election. The implication is that the absence of a reelection incentive for term-limited mayors led to the misappropriation of a substantial portion of the Brazilian budget for social spending.¹²

• A similar study in the United States finds that, controlling for other differences, term-limited governors in the U.S. preside over lower economic growth and higher taxes, spending, and borrowing costs than their similar, but non-term-limited, counterparts.¹³

Further findings from related literatures generate similar puzzles:

• Why, if elections are largely random events, is there extensive evidence of a strong feedback loop from policy change to public opinion and back to policy change?¹⁴

• Why, if voters are so poorly informed about the candidates they vote for, do ideological extremists who win a coin-toss primary against a more moderate candidate go on to have a substantially lower chance of winning the general election than when such a primary is won by the more moderate candidate?¹⁵

• Why, if elections are largely arbitrary events, are incumbents apparently so obsessed with service to voters, legislative votes, etc?¹⁶

These results present a puzzle for democratic theorists who view the claims about elections being largely meaningless or ineffective as settled science. It is clear that, at least sometimes, it (a) matters that voters can hold elected officials accountable (because they are not term-limited, for example), (b) matters who (by being enfranchised) has the ability to hold them to account, and (c) matters whether there are obstacles (like literacy tests) to exercising this ability. These implications do not seem to sit well with the idea that elections are “largely random events,” such that elected officials are essentially free to do as they please without fear of being held accountable.

We are not claiming that the empirical evidence uniformly supports the hypothesis of meaningful electoral accountability. As on virtually any empirical question in the social sciences, the evidence is mixed, and there is room for debating which interpretation is best. That is a part of our point – democratic theorists’ treatment of Achen and Bartels’ work as settled science is simply not correct. A more subtle point is that variation in empirical evidence on accountability suggests that there are important differences in contexts – in particular, in institutional detail within the large set of “institutions of representative democracy” – that may be driving this variation.¹⁷ That is another reason to doubt sweeping generalizations about the limits of electoral accountability.

The skeptical position thus appears weaker once one broadens one’s survey of the relevant empirical literature. In the rest of the paper, we explain why the empirical and theoretical foundations internal to the skeptical position are also shaky.

II. A CONCEPTION OF ACCOUNTABILITY

In this section, we explain how we will understand the idea of accountability for the purposes of this paper. We will say that a mechanism for holding an agent accountable for x is an arrangement that gives the principal the ability to penalize or reward the agent for the observed value of x, where x can be an action (e.g., consistently voting for abortion rights) or an outcome (e.g., low inflation).¹⁸

The institution of periodic elections, within a broader democratic system that includes protections for basic political rights, freedom of the press, and so on, is an example of such a mechanism. It provides citizens (the principal) with opportunities to observe government action or measures of government performance and to remove incumbent officeholders from power on the basis of their observations. Citizens are thereby able to penalize (remove) or reward (retain) their agent (the incumbent officeholder) on the basis of the observed values of these measures.

Whether citizens exercise this ability—whether they actually hold elected officials to account—is a separate question. Whether they exercise it well—holding officials to account in a manner that advances their interests—is another matter still. For example, voters might hold a politician accountable for keeping their campaign promises in a strict fashion, thereby giving the politician incentives to keep campaign promises even in circumstances in which it would be better for the voters if the politician deviated from the promises. Or, as Achen and Bartels claim, they might penalize incumbent officeholders for outcomes beyond their control, like shark attacks or natural disasters. The mere fact that voters hold officials accountable is thus not by definition good for voters.

To delimit the scope of our inquiry, we will focus on two mechanisms by which elections might deliver accountability: through the provision of incentives to the elected official and through the selection of superior types of officials. By exercising their ability to hold elected officials to account for an array of important outcomes (e.g., the rate of unemployment), voters may provide officeholders with incentives to promote the outcomes that serve voters’ interests (low unemployment). That is, the knowledge that voters will hold them accountable in the future may affect officeholders’ behavior in a manner that promotes voters’ expected welfare. With respect to the selection mechanism: by exercising their ability to hold elected officials accountable for an outcome like the rate of unemployment, voters select for the type of officeholder who might be better able, or more willing, to take actions that reduce unemployment. In that case, elections could generate accountability even if officeholders were indifferent to the prospect of winning re-election.¹⁹

Of course, elections may have other benefits for voters, independent of accountability (e.g., the salutary effects of alternation in power on social peace, undermining unhealthy ossification of officeholders by bringing in new people with fresh ideas, etc.).²⁰ Because we want to focus the discussion on disputes surrounding the capacity of electoral systems to generate effective accountability, we set such other benefits aside, without denying their importance.

We will say that elections generate minimally effective accountability if the incentive or selection effects of holding officials to account make voters better off than they would be under the counterfactual scenario in which they voted at random and the behavior of prospective candidates and elected officials was based on the anticipation of random voting. Viewing elections as “largely random events” is tantamount to denying that electoral accountability clears even this rather low bar. We will say accountability is more or less effective according to how far it improves voter welfare beyond this baseline via the incentives or selection channels.

With these definitions in place, we can now distinguish between several skeptical claims about accountability. The first and strongest is that elections do not generate any accountability: voters do not exercise their ability to hold elected officeholders accountable for any actions or outcomes, much less exercise this ability effectively. That claim is prima facie untenable. The universe of actions and outcomes for which the electorate might, in principle, hold elected officials accountable is vast: if voters can observe an action or outcome, they can condition their voting behavior and the conferral of rewards on it. In principle, voters could hold officeholders accountable for the state of the economy, the outbreak of war, the occurrence of famines, or the number of deaths during a pandemic; the enactment of tax reform, individual roll call votes on proposed legislation, or other policies; but, also, for the frequency of hurricanes, wildfires, and natural disasters; or even for shark attacks, solar flares, and the outcomes of sporting events. Moreover, there are theoretical and empirical grounds for believing that voters do hold officeholders accountable for extremely and obviously poor governance. How could it be, for instance, that – as Amartya Sen famously documents – there has never been a famine in a democratic country if elections do not generate any accountability?²¹ Indeed, Achen and Bartels themselves allow that elections can prevent politicians from pursuing actions that a sufficiently broad coalition of voters regard as unacceptable: “No president will strangle a kitten on the White House lawn in view of the television cameras.”²² Moreover, according to them, there are events that voters do hold officials accountable for, such as shark attacks, droughts, and sporting events. As Achen and Bartels put their contention about “blind” retrospective voting, in “most cases, incumbents will pay at the polls for bad times, whether or not objective observers can find a rational basis for blame.”²³ So the claim that elections do not result in any accountability at all is not only implausible, but also inaccurate as a statement of a skeptical thesis that anyone in the empirical literature appears to actually advance.

A more modest skeptical claim would be that elections do not produce the right kind of accountability. More specifically, the claim might be that elections do not produce minimally effective accountability, as defined above, or that they produce minimally effective accountability only in limited circumstances. The skeptic might concede that voters hold officials accountable for natural disasters beyond the elected official’s control, but not for the official’s actual conduct, with the implication that elections might as well be random events. Or the skeptic might concede that voters will hold officials accountable for extremely and obviously bad outcomes, like famines or catastrophically poor governance, but not for more mundane and subtle varieties of rent-seeking and incompetence. Such moderate skepticism is our primary focus. We will explain why neither theoretical considerations nor available empirical evidence provide strong support for it.

There is an even weaker skeptical claim that our argument does not directly address. One might concede that elections generate effective accountability but deny that they succeed at inducing elected officials to govern as well as or better than their counterparts in alternative regime types like lottocracy or epistocracy. For example, a skeptic might assert that voters hold officials accountable for certain important outcomes (e.g., economic growth, unemployment rates, etc.), but that, due to putative pathologies, like recency bias or unreflective partisanship, the benefits from such accountability are not sufficient to justify electoral democracy, at least under the existing institutional forms, when compared to the benefits from having a more diverse chamber generated by lottocracy. Evaluating that kind of claim would require a full specification of the relevant alternative regime types and a broad comparative analysis, including quantifying and comparing measures of the relevant effects under electoral accountability and under the alternative regime type. It is not clear how this could be done empirically since many of the alternative proposed regime types do not exist. Furthermore, it’s not clear how to evaluate empirically the accountability effect associated with representative democracy, given the heterogeneity of forms of representative democracy. The implication is that, if critics allow that representative democracy can generate some accountability of the right kind, they would need – as a result – to make significantly more precise their constructive claims, in order for the overall comparative argument to go through.

III. ACCOUNTABILITY & “IRRELEVANT” EVENTS

Having explained our conception of accountability for the purposes of the paper, we turn our attention to exploring the compatibility of representative democracy and effective accountability. We will assume, for the sake of argument, that critics of representative democracy are right that citizens are generally ignorant about policy issues as well as their representatives’ conduct. This makes the key question whether it is right to infer from this that elections do not create effective accountability. This section argues that empirical evidence that democratic theorists commonly accept as undermining the possibility of retrospective accountability is much weaker than often thought. The following section argues that the relationship between voter knowledge and effective accountability is much more complex than typically appreciated in the democratic theory literature.

Many political scientists have doubted the significance of voters’ policy ignorance, supposing that citizens could induce officials to deliver good policy just via retrospective voting—that is, by reelecting incumbents when times are good and voting them out of office when times are bad.²⁴ Retrospective voting would not require citizens to have policy preferences or to be aware of the specific policy decisions of incumbents. Achen and Bartels reject this theory, arguing that “Voters consistently and systematically punish incumbents for conditions beyond their control.”²⁵ The thought is that if incumbents know that voters are just responding to events beyond their control, then there is little reason for officeholders to expend costly effort to improve the lives of their constituents.

The first piece of evidence that Achen and Bartels present in defense of their position relates to a spate of shark attacks that took place along the New Jersey coast in the summer of 1916 and decimated the tourism industry there. No government could have anticipated or prevented such attacks, but Achen and Bartels nevertheless provide evidence that – relative to the 1912 election – Woodrow Wilson fared particularly poorly in the coastal communities where they occurred.²⁶ They interpret this as evidence that voters punished Wilson for the shark attacks, despite his evident lack of responsibility for them. Their second piece of evidence is that, over twenty-six presidential elections and across all states, incumbents fared particularly poorly in states hit by droughts or floods.²⁷ It is this evidence of a systematic relationship between election outcomes and events that are beyond the control of politicians that underlies their claim that election outcomes are “in an important sense, random” and “from the standpoint of democratic accountability, essentially arbitrary.”²⁸

However, the evidence that voters are systematically responsive to outcomes beyond politicians’ control is controversial in the broader political science literature.²⁹ In an influential paper studying the electoral consequences of the broader universe of shark attacks from 1872 to 2012, Anthony Fowler and Andrew Hall conclude that “there is virtually no compelling evidence that shark attacks influence elections, and any such effect–if it exists–must be substantively tiny.”³⁰ A different paper finds that tornadoes, on average, have a negative effect on incumbent reelection rates, but the effects appear to depend on how incumbents responded to the disaster: while they suffer no penalty for tornado-related deaths (which seem more likely to be beyond an incumbent’s control), there is a significant penalty for failing to deliver relief aid.³¹

For the sake of argument, set these empirical doubts aside and suppose that voters allow events beyond the incumbent official’s control to influence their voting behavior. Is this grounds for inferring that, “from the standpoint of democratic accountability,” elections are “essentially arbitrary”?

Let us first dispense with a mistaken thought about effective accountability that might otherwise muddy our thinking. It is in general neither irrational nor counterproductive for a principal to hold an agent accountable for an outcome that the agent does not control—however “unfair” it might feel to the agent. A principal might hold an agent accountable for outcomes that indicate the agent’s “type” – attributes of the agent that are outside the agent’s control but that matter to the principal, such as the agent’s deeply ingrained dispositions and character traits, values and preferences, intelligence and other natural talents, including those that affect how difficult it is for the agent to undertake costly actions (effort) on the principal’s behalf. If the principal can observe an outcome that reliably indicates the agent’s type, then the principal will rationally hold the agent accountable for the outcome, and profit from doing so, even though the agent has no control over the outcome.

So if natural disasters, shark attacks, sporting events and the like are “irrelevant” events, it must be true both that (1) the official has no control over them and (2) voters cannot infer anything about the official’s type from them. So defined, the determination of whether an event is irrelevant can be somewhat subtle. That a region experiences a drought is an irrelevant event, but that the drought brings a certain amount of suffering or economic hardship in its wake is not—at least, not if the incumbent official’s choices and type mediate the effects of the natural event, and so convey information about the incumbent’s competence and other qualities.

This observation points to a more general and abstract distinction between two possible claims that one might entertain on the basis of the evidence Achen and Bartels and others have produced. The first claim is that irrelevant events have effects on voting behavior. The second is that voters hold officials accountable for irrelevant events, where this describes voters’ observing that a certain event has occurred and treating its occurrence as a reason to vote for or against the incumbent. One cannot infer the second claim from the first. In fact, irrelevant events may influence voting behavior precisely because voters hold officials accountable for relevant events. (In the context of the drought example: if voters hold the official accountable for the costs of a drought, then the drought itself, an irrelevant event, should in expectation have nonzero causal effect on voting behavior.³²) Thus, a relationship between irrelevant events and voting outcomes could emerge because voters are rationally holding officials accountable for outcomes they do partly control, and this accountability may effectively promote the voters’ interests.

One might naturally wonder whether this can explain why incumbent officials would incur on average a penalty after events like droughts. Wouldn’t a drought also provide an opportunity for rational voters to reward the competent incumbents if it reveals how well they prepared or responded? Not necessarily. If elections are effective as selection devices such that incumbents are, on average, more competent than challengers, then voters would be more likely to re-elect incumbent officials than not in the absence of new information. In such a setting, the revelation that a competent incumbent has prepared well or responded well can have only a limited positive effect on the re-election rate. In the extreme case where voters always re-elect incumbents in the absence of the information about his competence that the drought reveals, the drought creates an opportunity for voters to learn information that can only hurt the incumbent’s reelection rate: either he is competent (as expected), in which case the outcome is the same as it would be in the absence of the drought (he wins), or he is incompetent, in which case the impact of the drought provides evidence of his incompetence and his probability of winning is less than it would have been in the absence of the drought.³³

What about irrelevant events where the officeholders’ actions do not plausibly mitigate their effects on voter welfare (say, in the proverbial example of shark attacks)? Even here, evidence that such events influence voting behavior is consistent with effective accountability. For the sake of argument, imagine, with Achen and Bartels, that voters take some satisfaction in punishing incumbents for irrelevant events, as a form of catharsis. If holding office is sufficiently valuable to the incumbent, the voter need not “promise” to retain the incumbent with certainty as a reward for taking good actions in office. Posit, notionally, that retaining with probability .7 would suffice to induce such actions retrospectively. This creates “slack” in retention that voters may, without undermining accountability, “spend” on expressing their subjective sense of how well their own lives are going in light of natural events even when officeholders cannot affect the marginal impact of such events. Thus, in the wake of an event that improves their subjective sense of happiness, voters might re-elect the incumbent with the probability of, say, .9, but following an unhappy event, might do so with probability of, say, .75. Assuming, just for the sake of argument, that there was persuasive evidence of incumbents being held responsible for shark attacks, this is an explanation that is perfectly consistent with effective accountability.

The preceding discussion makes clear that inferences about accountability in the context of seemingly irrelevant events are far more nuanced than critics of electoral democracy have appreciated. Evidence of a relationship between such events and voting outcomes is compatible with effective accountability. Still, even if one concedes this point, and grants that these studies are not evidence that voters are responding to what they ought to ignore, one might have the intuition that the real threat to effective accountability stems from their failure to acquire and respond to more straightforwardly relevant information. As we explain in the next section, in many pertinent contexts, this intuition turns out to be wrong: it is a mistake to assume that more informed voters will get more effective accountability.

IV. ACCOUNTABILITY & VOTER INFORMATION

Work in contemporary democratic theory often assumes without argument that lower levels of information are bad, even disabling, from the perspective of facilitating effective accountability. For instance, in an influential article, Guerrero writes, “If I do not know what you are doing and have done, I cannot hold you accountable for it. And if I only know that you have done A (rather, perhaps, than B or C), but I have no idea what A amounts to (I know nothing about the issue for which A is a candidate proposal), or how it differs from B or C, I might as well not even know that you have done A—my ability to hold you accountable is equally impoverished.”³⁴ The intuition that one cannot hold representatives accountable without substantial information about their actions, and that improving voters’ information of those actions is beneficial, is straightforward.

Drawing on important work in political economy, we explain why this intuition about electoral accountability is not in general accurate.³⁵ Less information can, in an important range of realistic settings, lead to more effective accountability. The intuitive story implicitly takes for granted a number of restrictive conditions, including (a) voters’ having precise knowledge of what policy/action is best, (b) separability of incumbent’s actions, (c) absence of adverse selection in electoral choice, and (d) very special properties of random events’ effects on social welfare. Individually, these conditions are neither necessary nor sufficient to make the intuitive story work. But understanding their effects helps isolate mechanisms underlying the intuitive story. When these conditions fail, the argument that voters benefit from having more information breaks down.

We begin by introducing a useful distinction. Recall that in section 2, we said that a mechanism for holding an agent accountable for x is an arrangement that gives the principal the ability to penalize or reward the agent for an observed value of x, where x can be an action or an outcome. The ‘outcome’ could be a state of affairs that includes within it a range of important considerations — a level of unemployment, a level of inflation, a level of growth, relative peace or prosperity, and so forth. An outcome can be a noisy measure of inputs (e.g., particular legislative decisions, legislators’ competence, etc.) because, among other things, there may be forms of luck (such as exogenous events) that, along with those inputs, help determine the outcome. When employing outcome-accountability, the principal conditions the agent’s reward (re-election) directly on the outcome, and not on the inputs. In contrast, action-accountability depends upon the principal observing the inputs and conditioning the agent’s rewards directly on them.

We will next explain, with respect to each of the four conditions noted above, how less information and outcome-accountability may be more beneficial to the principal than more information and action-accountability.³⁶

V. RESPONSIVENESS TO THE WEALTHY

We have explained why it is a mistake to conclude from the research on voter ignorance that elections cannot generate effective accountability. There is a related skeptical thesis about electoral democracy that, like the claim about accountability, has become a piece of received wisdom among democratic theorists. This is the oligarchy thesis, according to which policymakers are responsive mainly to a small segment of (wealthy) citizens. There is an extensive empirical literature on the thesis, and we cannot do it justice, given space constraints. With this in mind, we focus our discussion on considerations, parallel to those raised in the discussion of voter ignorance and accountability, that suggest that the thesis deserves more skepticism than democratic theorists typically appreciate. Like the problematic inferences from voter ignorance discussed above, the uncritical acceptance of the oligarchy thesis reflects a failure to think through the possible mechanisms of accountability, as well as an incomplete and one-sided reading of the empirical literature.

Two studies from the literature on biased policy responsiveness, in particular, have influenced democratic theorists. First, in Affluence & Influence, Martin Gilens examines correlations between survey respondents’ expressed policy preferences and enacted policies. He concludes that:

The American government does respond to the public’s preferences, but that responsiveness is strongly tilted toward the most affluent citizens. Indeed, under most circumstances, the preferences of the vast majority of Americans appear to have essentially no impact on which policies the government does or doesn’t adopt.⁴⁶

Gilens thinks a campaign finance mechanism may explain the finding since “donating money is the only component of political participation that mirrors the patterns of representational inequality revealed” by the book.⁴⁷

Second, in Unequal Democracy, Larry Bartels, using roll call votes of U.S. Senators in the late 1980s and the early 1990s, finds that Senators “were vastly more responsive to affluent constituents than constituents of modest means.”⁴⁸ Indeed, like Gilens, he (a) finds no evidence of responsiveness to the policy views of poorer citizens and (b) observes that the documented pattern of responsiveness is broadly consistent with the shape of campaign contributions.⁴⁹ Largely on the basis of these two studies, democratic theorists often treat it as an uncontroversial fact about modern democracies that their policymakers, by and large, respond only to the policy views of their affluent citizens.⁵⁰

The science is far from settled, however.⁵¹ The extent of disproportionate responsiveness to high-income citizens is far more contested and uncertain than democratic theorists typically acknowledge. For instance, one recent study of the U.S. finds that for 90% of issues, a majority of middle income citizens agree with a majority of high-income citizens and when they do not, the majority opinion of middle-income citizens prevails 47% of the time.⁵² This is perhaps not as often as perfect political equality would require, but certainly more than what Gilens’s claim of “essentially no impact” predicts. More generally, an excellent recent review reports that although the literature “collectively points to a positive income gradient in political responsiveness, there is considerable divergence in results across models and studies,” and key early studies “overestimate inequalities in political responsiveness.”⁵³ It’s also noteworthy that the extreme version of the oligarchy thesis, according to which policy is simply driven by the preferences of the most advantaged, sits uncomfortably with the strong persistence, across advanced democracies, of middle-class-oriented fiscal policies, including programs like progressive tax codes, public pension, and healthcare systems.⁵⁴

Even setting aside the preceding empirical concerns for the sake of argument, there are reasons for skepticism about the proffered causal interpretation of Gilens’ and Bartels’ results. Overall, (a) the evidence does not seem to support the claim that disproportionate responsiveness to high-income citizens is more pronounced in more inegalitarian countries; and (b) at least in the United States, there is no evidence that the results are driven by responsiveness to the rich in those policy domains that particularly affect the distribution of economic resources.⁵⁵ Finally, it has been notoriously difficult for political scientists to empirically pin down a causal relationship between campaign contributions and elected officials’ voting behavior.⁵⁶

This causal identification issue is important for democratic theory for at least two reasons. First, if there is a causal relationship, not understanding the underlying mechanism puts in doubt the kinds of comparative institutional claims that the empirical work has been called upon to support. For example, if the correlation reflects the fact that the affluent are more likely to take cues from elected leaders, who are themselves the captured servants of special interests that might also capture lottocratic representatives, then the correlation might persist even if representatives were selected by lottery rather than elections. If the explanation for the correlation is instead that the wealthy extract quid-pro-quos through campaign contributions, then the implications for the comparison between elections and lotteries are different. The bottom line is that without knowledge of the underlying causal mechanism, it is difficult to know how the empirical evidence provided in support of the selective responsiveness thesis bears on central debates in normative democratic theory.

The second reason that the underlying causal story is important is that the normative implications of the observed correlation depend on whether it reflects some incidental feature of recent American politics (e.g., its rules of campaign finance and lobbying, its high level of economic inequality, or its particular electoral rules) or, instead, some relatively unavoidable feature of representative democracy (e.g., the fact that the affluent are more likely to run for elected office). While critics of representative democracy often seize upon this literature to make arguments for quite dramatic institutional reforms, without a better understanding of the underlying causal mechanisms, the implications are unclear.

Even if one sets aside questions about the robustness of the findings, as well as the mechanisms driving them, such findings would still not support the implications that democratic theorists often derive from them. In light of the well-documented ignorance of citizens on policy questions, it need be neither surprising nor unfortunate that correlations between citizens’ immediate policy opinions and actual policy outcomes are weak. That is insufficient to rule out the possibility that low-income voters are, in fact, getting effective accountability — just via outcome-accountability, rather than action-accountability. This could be the case if voters cast ballots primarily based on their sense of well-being (rather than holding incumbents accountable for enacting particular policies). Indeed, as we have seen, it is possible that they could be better off, as a result, than they would be with action-accountability.⁵⁷

A related possibility is that, for the electorally pivotal bulk of voters, there is, in fact, considerable outcome-accountability in place, but it is realized alongside accountability failure, from the perspective of the same voters, on some number of issues on which representatives also feel pressure to deliver action-accountability to their wealthy constituents (perhaps, as Gilens and Bartels suggest, for reasons related to the need to raise campaign funds). This interpretation would generate the reported empirical pattern, and in light of it, it would be correct to say that high-income citizens were getting a measure of accountability not available to their poorer counterparts. But it would not follow that representatives were unaccountable to low-income citizens, nor that some other regime type could do better for them.

In sum, while many democratic theorists have concluded that representative democracies are really only responsive to high-income citizens, the empirical evidence for this claim is highly controversial, the claimed underlying causal mechanism is not well-supported, and the evidence is ultimately compatible with effective accountability to poorer citizens.

VI. CONCLUSION

It is common in democratic theory to assert that empirical political science supports an extremely skeptical view about elections’ capacity to generate effective accountability. Theorists cite influential studies that highlight voters’ relative ignorance of policy issues, their apparent tendency to punish politicians for events that are beyond the control of those politicians, and the disproportionate correlation between the policy preferences of wealthy citizens and policy outcomes. This skeptical view about the potential of electoral accountability often underpins relatively radical reform proposals, including lottocracy and epistocracy.

But the skeptical consensus in the democratic theory literature ignores a large body of important empirical work on electoral accountability suggesting that competitive electoral environments reduce corruption and deliver important benefits to voters. It also ignores findings from formal work on electoral accountability, which reveal – contra the assumption in much of the democratic theory literature – that citizens can often do better when they have less information. Meanwhile, the evidence of voters’ punishing politicians for events beyond their control as well as the disproportionate correlation between the preferences of high-income citizens and eventual policy changes are open to a wide variety of interpretations, including ones that are consistent with effective accountability.

The paper’s key implication is that the interpretation of the empirical picture underlying some prominent arguments for radical reform in the democratic theory literature is not well-supported. The case for reforming electoral systems, rather than doing away with them, may be stronger than is often thought.

ACKNOWLEDGMENTS

We thank Bob Goodin and three anonymous reviewers for helpful comments.

COMPETING INTERESTS

The authors declare that they have no competing interests.