ELECTORAL ACCOUNTABILITY: FOUNDATIONAL QUESTIONS

The subtitle of The Democratic Dilemma—Can Citizens Learn What They Need to Know?—incisively cuts to the core issue from more than 60 years of research on electoral accountability. To understand, diagnose, and improve the limitations of voter behavior and electoral institutions, this is the primary question we must answer.

The predominant strain of empirical literature in political science has always been at least pessimistic about this question. The American Voter (Campbell et al. Reference Campbell, Converse, Miller and Stokes1960), the ur-text for survey-based analysis of political behavior, revealed a shocking lack of engagement, awareness, and discipline in the views that members of the mass electorate hold on politics and political issues. Generations of scholars in American voter behavior have reinforced, refined, and elaborated this point. This literature has evolved to an unqualified negative assessment: Democracy for Realists (Achen and Bartels Reference Achen and Bartels2016) flatly declares that “elections do not produce responsive government” before one even opens the book.^{Footnote 1}

To accountability pessimists, the accepted diagnosis for this unfortunate state of affairs turns ultimately on the limitations of individual voters. Of course, most pessimists acknowledge that the pressing demands of everyday life—combined with the collective nature of political activity—create a strong incentive for rational ignorance (Achen and Bartels Reference Achen and Bartels2016). Part of the problem, then, is that a collective-action problem among rational voters induces them to “free ride” on others to do the trench work of holding officials accountable, thus leaving too few of them to do it effectively. However, to many pessimists, the problem goes much deeper: as ordinary people acting in a political capacity, voters have an unsophisticated understanding of how issues fit together, a defective architecture for assessing political uncertainties, and deep biases in seeking and parsing information about political performance. In other words, citizens as voters are victim to many of the same cognitive limitations and defects that they face in all facets of life. Solving the collective-action problem that underlies rational ignorance cannot change these defects; ultimately, there are psychological limits of accountability, not only strategic limits.

Disillusionment stems from the original, implicit question of this literature: If voters were able to rationally hold elected officials accountable, what would their attitudes and their voting behavior look like? Their issue positions would be stable and organized in meaningful correspondence with the positions of elites, enabling two-way communication. (At the very least, they would know the key issues—those taken up by elites in institutions—and have some position.) In their electoral choices, they would respond only to some events—those deemed “relevant” by a scholar. Voters particularly affected by a policy would know more about it and cast their votes based on it. Repeatedly and spectacularly, voters have failed to conform to that standard. The result may be the only academic literature in which everyone understands “Homeric” to connote an oafish, shortsighted, ignorant cartoon character, held as a fitting metaphor for our objects of study (Bartels Reference Bartels2005).

THE DIALOGUE IN ELECTORAL-ACCOUNTABILITY RESEARCH

However, for much of this literature’s history, an important minority position existed. Downs (Reference Downs1957), of course, in empirical innocence, simply assumed the rationality of voters, and the coherence of their issue positions collapsed to a single dimension. More subversively, because in direct dialogue with the Michigan School, Key et al. (Reference Key1966) laid out “the perverse and unorthodox argument...that voters are not fools.” Fiorina (1981); Sniderman, Brody, and Tetlock (Reference Sniderman, Brody and Tetlock1991); Page and Shapiro (Reference Page and Shapiro1992); Popkin (Reference Popkin1994); Lau and Redlawsk (Reference Lau and Redlawsk1997); and Alvarez (Reference Alvarez1998) all provided significant, empirically grounded statements consistent with this proposition.

Of course, the rationality of individual voters is neither necessary nor sufficient for an affirmative answer to The Democratic Dilemma’s subtitle question. More explicitly than their strictest rational-choice predecessors, Lupia and McCubbins (Reference Lupia and McCubbins1998) squarely acknowledged cognitive limitations of voters in their theoretical arguments. Still, they largely affirmed that voters can learn what they need to know. Elections present voters with a relatively simple set of choices and, cognitive limitations notwithstanding, Lupia and McCubbins showed that in experimental settings capturing the strategic kernel of real elections, voters can distill their information adequately to make a decision that promotes their interests.

The Democratic Dilemma is important because it showed this clearly, with several features that are unusual in this field. First, it explicitly considered the strategic interaction of elites and voters. A core question of this book is not only whether voters can parse information but also whether they can parse biased information provided by self-serving elites against the background of goal conflict. This is crucial to understanding the prospects for real-world accountability, but no prior work in the literature gave any cause for optimism in such a context. Second, its research design is based primarily on the rarefied but controlled setting of a laboratory experiment. This allows for unparalleled control over the decision context so that alternative explanations other than the efficacy of voters can be ruled out. Third, The Democratic Dilemma emphasized the institutional foundations of political accountability, taking the psychological capabilities of voters largely as given. This is important because voters’ cognitive architecture seems to be more difficult to change than formal institutions. Thus, a reasonable approach to improving accountability is to make those institutions work as well as possible given that architecture (a point amplified by Krupnikov’s article in this symposium). Fourth, the cognitive architecture assumed in The Democratic Dilemma is decidedly short of the standard of full rationality.

However, the importance of The Democratic Dilemma has been amplified by the past 20 years of political science research on electoral accountability. It stands now as one of the last major empirical statements of accountability optimism in the political science literature on voter behavior. For 40 years after publication of The American Voter (Campbell et al. Reference Campbell, Converse, Miller and Stokes1960), the accountability pessimists and optimists were engaged in a dialogue. The pessimists, as the far-larger contingent, were in the driver’s seat. Nevertheless, the literature was punctuated by important counterpoints from optimists. The optimists largely accepted the empirical findings of the literature but offered their own with a sharply contrasting interpretation of the prospects for accountability. Working ultimately in a comparable empirical idiom, these strands pushed the others to make clearer theoretical statements and devise more discriminating empirical tests. This may or may not be “progress” in the sense of philosophy of science—and the existence of two competing strands of the literature implies a lack of consolidation on a single dominant paradigm—but the engagement was good for both sides. In addition to providing clarity to the theoretical claims and empirical tests, the engagement across this divide supported circumspection by each camp about proposed institutional reforms. It is easier to hedge our bets about the value of grand institutional reforms when we engage with colleagues in a mutually accepted format in which neither side has obviously yet prevailed. The debate between optimists and pessimists now seems to be largely over in political science—or at least on hiatus—and the pessimists have won (cf. the symposium on Democracy for Realists in Critical Review, vol. 30, 2018).

ACCOUNTABILITY OPTIMISM RESURGENT

In fact, it is premature to declare a winner in the debate—more so to declare it irrelevant. A literature has continued to flourish that highlights the efficacy of voters in holding politicians accountable. An empirical strand has taken root, much of it in economics. Numerous papers in this strand have shown that when a group of people is empowered to vote, public-policy outcomes shift in the interest of that group—and these shifts are mediated by increases in the affected group’s turnout.^{Footnote 2} For example, Cascio and Washington (Reference Cascio and Washington2014) showed that the Voting Rights Act (VRA) of 1965, which removed literacy tests for voter registration, increased blacks’ share of public spending. Focusing on the US county level, they found that the VRA’s removal of literacy tests increased voter turnout and increased state transfers that benefit black voters, compared to similar counties in states that had no literacy test. Similarly, women’s suffrage was found to increase social spending by state governments—public-health and child-welfare spending in particular—as well as the ideological slant of elected officials’ voting records (Lott and Kenny 1999; Miller Reference Miller2008). These effects are consistent with long-standing evidence that women are more actuated than men by social dimensions of public policy. Focusing on Brazil, Fujiwara (2015) showed that electronic-voting technology effectively enfranchised poor and less-educated citizens, which in turn shifted public spending in the interest of these groups. In particular, government spending on health care increased, with corresponding increases in utilization and beneficial outcomes for those less educated.

In political science, a smaller literature on accountability optimism has begun to develop. These papers take as given the empirical evidence marshaled by pessimists^{Footnote 3} and show that it is actually consistent with electoral accountability operating about as well as could be expected in light of postulated information asymmetries. The pessimistic literature from opinion surveys implicitly asks, “Given our theory of rational voter behavior, how well do actual voters conform to it?”; the newer literature asks, “Given the actual behavior of voters, in what sense might it be rational?”

For example, Ashworth, Bueno de Mesquita, and Friedenberg (2018) considered the broad finding that voters respond to events outside the control of politicians—most famously, droughts, shark attacks, and the like—and thus seem to blame politicians for events they cannot possibly control. This “blind retrospection” (Achen and Bartels Reference Achen and Bartels2016) seems not only clearly irrational but also damaging for electoral accountability. Ashworth, Bueno de Mesquita, and Friedenberg (2018) noted that although the event may be outside of a politician’s control, the effects of it are not: they depend on the quality of preparedness, mitigation, and relief—all in turn dependent on policy and organizational skill of political leaders. Therefore, exogenous events provide information about those skills. When the information is bad, a rational voter responds by punishing incumbent politicians who might be reelected in the absence of the exogenous event. This creates exactly the correlation between natural disasters and incumbent punishment documented by pessimists but with a dramatically more optimistic interpretation about prospects for effective electoral accountability.^{Footnote 4}

Gailmard and Patty (Reference Gailmard and Patty2018) similarly took as given the empirical evidence of supposed voter pathologies—in this case, from Healy and Malhotra (Reference Healy and Malhotra2009) showing that voters reward incumbents for disaster-relief efforts but not for disaster prevention, which has been interpreted as a sign of voter myopia. We argue that voters probably are less informed than politicians about the need for prevention efforts but also may believe that some politicians are “corrupt” in that they benefit from prevention spending even when it is not publicly useful. In view of these beliefs, a rational voter would take prevention spending as bad news about incumbent corruption; strategic incumbents, in turn, underprovide it. Conversely, because voters directly experience harm when disaster damage occurs, they know exactly when relief efforts are beneficial to them and have no trouble holding incumbents accountable for providing disaster relief when necessary. The result corresponds to empirical evidence that relief efforts are plentiful and electorally rewarded, whereas prevention efforts are neither. However, there is nothing Panglossian about these results, and the “optimism” they deliver must be qualified. Consistent with empirical evidence, we assume that prevention spending is more effective than relief; thus, its underprovision reflects a real loss of welfare. Although we show that empirical evidence is consistent with electoral accountability operating as well as it possibly can, given the information asymmetries between voters and politicians, this does not imply that it produces objectively good policy. It does imply that no other institution, short of one that magically eliminates incentive conflicts and information asymmetries, could better deliver on voters’ interests.

Finally, Fowler (2018) considered a broad array of empirical evidence that was offered for what he calls the “partisan intoxication” thesis: the idea that voters are not well informed or much concerned about policy effects; instead, elections are simply “roll calls of intoxicated partisans.” It is difficult to imagine a more profound failure of electoral accountability as that implied by the partisan-intoxication hypothesis. However, Fowler showed how most of its major empirical bulwarks—that is, predictive accuracy of party ID for vote choice, stability of party ID over time and across generations, and cuing from copartisan elites on specific issues—can be readily reconciled with effective accountability by voters concerned about a coherent set of values and interests. In other words, in many tests, a reasonable interpretation of voting based on values and interests is observationally equivalent to partisan intoxication.

In summary, there is ample reason for at least qualified optimism about the ability of voters to operate the machinery of elections, thereby pursuing their interests through public action. It is difficult to square the empirical results in this literature with a strong form of pessimism; elections appear in these cases to make government responsive to newly enfranchised voters. The recent political science literature takes the pessimists’ evidence as given and shows the marked flexibility of interpretations consistent with it. This work helps us to understand how and to what extent elections produce government that responds to voters’ interests. Somewhat ironically, the political science literature’s recent insistence that elections do no such thing seems to reflect theoretical commitments that the evidence alone does not imply.

CONCLUSION: REVIVING THE DIALOGUE

These arguments provide ample reason to continue the dialogue about electoral accountability. Unfortunately, the optimists and pessimists now labor in mutual ignorance to a greater extent than in the past.^{Footnote 5} This is undesirable for the electoral-accountability literature as a whole. The empirical literature in economics is unconnected to the political-behavior literature’s canonical themes of mass civic engagement and ideological consistency. For a reduced-form estimate of the effect of a turnout change on policy outcomes, we can argue that these themes are superfluous. Yet, the voting booth cannot be left theoretically as a “black box.” We must understand how these voters understand their choices; how they incorporate information about policy responses; whether there are other countervailing policy effects not picked up by voters (or the research); whether anticipation of voter behavior by politicians drives the results; and so on. In political science, the pessimists must understand the plasticity of interpretations consistent with their data, which in turn imply that their normative arguments about democratic theory and institutional reform—which rest on one specific interpretation—are less compelling than they first may appear. Moreover, we formal theorists seeking to advance a qualified optimism must understand that taking existing empirical findings as stylized facts to be reproduced in models will be less convincing to our behaviorally inclined colleagues than deriving novel implications to be confronted with new data.

The Democratic Dilemma is an important exemplar for a contribution to this type of dialogue. It took seriously the behavioral findings about limits to rationality. It distilled clear implications from theory and evaluated their ability to explain new data from simple, original experiments. It stands, therefore, not only as a major contribution to the optimistic literature on electoral accountability but also as a model for continuing the dialogue. Scholars in this field should heed its example to reengage more fully with one another.