electionarticlefromdrmink1.pdf

ilable at ScienceDirect

Electoral Studies 38 (2015) 1e9

Contents lists ava

Electoral Studies

journal homepage: www.elsevier.com/locate/electstud

Election administration and perceptions of fair elections*

Shaun Bowler a, Thomas Brunell b, Todd Donovan c, *, Paul Gronke d

a UC Riverside, USA b University of Texas, Dallas, USA c Western Washington University, USA d Reed College and Appalachia State Univeristy, USA

a r t i c l e i n f o

Article history: Received 23 September 2014 Received in revised form 8 December 2014 Accepted 13 January 2015 Available online 22 January 2015

Keywords: Election administration Electoral integrity Perceptions of elections Photo identification laws Public opinion Fair elections

* A earlier version of this paper was presented at t Project / MEDW Citizens, Parties, and Electoral Con shop, Montreal, Canada. 18 July 2014. * Corresponding author. Department of Politica

Bellingham, WA, USA. E-mail addresses: [email protected] (S.

utdallas.edu (T. Brunell), [email protected] [email protected] (P. Gronke).

http://dx.doi.org/10.1016/j.electstud.2015.01.004 0261-3794/© 2015 Published by Elsevier Ltd.

a b s t r a c t

Scholars of democracy proposes an important relationship between the quality of elections and democratic legitimacy, but there are few studies of how the conduct of elections af- fects perceptions of elections being fair. We examine how election administration and individual-level demographic traits affect public perceptions of fair elections in the US. Since administration of US elections is largely the responsibility of individual states we are able to exploit variation in the quality of how elections are conducted to assess effects of electoral administration on public perceptions. We find evidence that administrative performance is positively and significantly related to perceptions of elections being fair. Voter identification laws, in contrast, are not associated with greater confidence in elec- tions. We also find some evidence that speaks to the limits of these findings, as individual- level factors such as partisanship and minority status have larger effects than adminis- tration on perceptions of electoral fairness.

© 2015 Published by Elsevier Ltd.

1. Introduction

It is well established that, after an election, winners and losers differ in their attitudes about the winner's right to govern (Nadeu and Blais, 1993), their trust in government, their satisfaction with democracy, and their views that elections make officials respond to the public (Anderson and LoTempio, 2002; Banducci and Karp, 2003; Bowler and Donovan, 2007; Esaiasson, 2011; Singh et al., 2012). Yet as some basic level, democratic elections 'work' because (or if) losers and winners see the outcome as the result of a fair, legitimate process. One important theme from a recent

he Electoral Integrity text pre-IPSA work-

l Science, MS 9082,

Bowler), tbrunell@ (T. Donovan), paul.

body of research on electoral integrity is that the proce- dural quality of elections should contribute to democratic legitimacy (for an overview, see Norris, 2014; Birch, 2008). Part of the process by which supporters of losing parties and losing candidates see winners as having legitimate authority is that at some level, they view the electoral process as fair, and consent to the results of elections they lose (Anderson et al., 2005).

But how is it that people come to perceive outcomes of elections as legitimate and procedurally fair? In older, established democracies, it is likely that citizens have some base level of political socialization that causes them to view electoral procedures as fair in themselves. In these nations, the same social processes that transmit civic duty (Blais et al., 2004; Blais, 2006), patriotism, or even party loy- alties (Campbell et al., 1960; Niemi and Jennings, 1991) likely also build some reservoir of support for the outcomes of democratic institutions (Dalton, 2009). Regardless win- ning or losing, and regardless of procedural faults or glitches on election day, socialization processes may cause

S. Bowler et al. / Electoral Studies 38 (2015) 1e92

people in established democracies to see elections as routine events and to regularly accept election results as legitimate (Mozaffar and Schedler, 2002). Political social- ization is not, however, sufficient to explain how all people view electoral integrity at a particular point in time. Although socialization may well provide a reservoir or benchmark of support it is not plausible to suggest that the level of support remains unchanged through a life cycle of perhaps a dozen or more national elections. Several scholars note that younger generations are being socialized toward democracy differently (Denemark et al., 2012), with less deference to authority (Inglehart, 1990) and with civic duty acting as a weaker force in motivating political participation (Blais et al., 2004). The media environment that generates information about democratic institutions has also changed (Moy and Pfau, 2000) - a competitive, partisan media context can increase incentives news out- lets have to bring attention to procedural flaws in elections, and allegations of fraud.

Beyond any socialized acceptance of election results then, citizens' views of electoral legitimacy are conditioned by their perceptions of electoral and political performance (Norris, 2014, 2004; Elklit and Reynolds, 2005; Elklit and Reynolds, 2002). For example, Europeans who perceived that officials were bribed were less trusting of democratic institutions (Anderson and Tverdova, 2003). Russians who perceived elections as unfair were less supportive of po- litical parties, parliament, and their government (McAllister and White, 2011). Although we have evidence that satisfaction with democracy is related to broad mea- sures of procedural performance of government (Norris, 2004),1 and evidence that specific electoral rules (propor- tional representation and publicly financed elections) correspond with greater popular confidence in elections (Birch, 2008), we know less about how the quality of how election administration affects mass perceptions of elec- toral performance in established democracies.

2. The research question

In light of the preceding discussion our research ques- tion can be stated: To what extent are perceptions of electoral performance affected by the actual procedural quality of elections? By investigating this question, we can broaden our understanding of how citizens reason about political institutions in general. That is, are popular atti- tudes about democratic institutions, at least in part, struc- tured by the quality of institutional practice?

Our primary question also has implications for the utility of efforts to improve the administration of elections. We know that, independent of the procedural quality of elections, a particular event or election rule can be viewed quite differently by different groups. In the US, for example, partisanship plays a major role in structuring whether or not people view key aspects of elections as unfair or corrupt. Party structures how people perceive the role of campaign finance in elections (Persily and Lammie, 2004),

1 Also see Putnam et al. (1994), who argue that government performs better where there is greater civic engagement.

how they view the relationship between campaign finance and the legitimacy of election results, how they viewed the legitimacy of the disputed 2000 presidential election (Craig et al., 2006), and how they viewed the utility and fairness of voter identification laws (Bowler and Donovan, 2013:30e31; Bentele and O'Brien, 2013). Indeed, at the mass and elite levels, Americans' attitudes about what does and what does not constitute electoral 'fraud' are defined sharply by their partisanship (Wilson and Brewer 2013; Ansolobehere and Persily 2008).

However, if we find that people view elections as more legitimate where objective measures show they are better administered, this would suggest that efforts that succeed at improving electoral performance can enhance the ability of democratic elections to impart legitimate political au- thority. We should note that there are some grounds for scepticism that election administration will have an inde- pendent effect on public opinion. Bowler and Donovan (2013) have demonstrated a wide range of electoral rules and reforms have little identifiable relationship with po- litical trust, efficacy, and citizen engagement with politics. These findings suggest a more limited role for “institutional effects” than one might expect given the argument that “institutions matter.” We might also see these results as suggesting a limited role for the effect of election admin- istration. One reason for such null results with respect to broad institutional changes is that although an electoral rule may exist, it need not necessarily be implemented in a fashion that citizens are able to detect. In this study, how- ever, we are not assessing how the presence or absence of an electoral institution affects attitudes, rather, we examine how the implementation of elections affects attitudes.

At this point we should note that the general hypothesis of interest is quite straightforward: better administration of elections should produce more positive views of the elec- toral process among mass publics. In order to test that general argument, however, we need to substantiate both that the US case is an appropriate case study and that appropriate measures of election administration exist.

3. The advantage of the American case

As Norris notes (2014), there are a number of problems with attempts at establishing causality when investigating the relationships between electoral performance and public attitudes about elections. For one, there are very few cases where we have survey data measuring attitudes about electoral performance collected before and after a jurisdiction transitioned to democratic elections. Even in established democracies, it is rare to find polls with suitable items conducted over a time span that is adequate enough to capture the potential effects on attitudes of problems with electoral performance. As such, cross-national studies of opinions taken as a snapshot in time have been our best chance for teasing out the effects of electoral performance on popular attitudes.

An additional research design problem is that of being able to measure electoral performance objectively, across a large number of jurisdictions (see Elklit and Reynolds, 2002). Up to this point, most studies have relied on sub- jective measures of electoral performance (e.g. corruptions

3 Pew worked with Charles Stewart III (MIT), Steven Ansolabehere (Harvard), Barry Burden (Wisconsin), Heather Gerken (Yale Law), Paul Gronke (Reed), Christopher Mann (Miami), Nathan Persily (Columbia

S. Bowler et al. / Electoral Studies 38 (2015) 1e9 3

perceptions indices, and/or experts' subjective perceptions of electoral performance). It is quite possible that popular perceptions of electoral performance are closely bound to objective measures of performance. Scholars, however, have not always been well-positioned to objectively mea- sure election performance with a standard that can be applied across a large number of jurisdictions. By exam- ining the US case, we can model individual citizens' per- ceptions as a function of an objective measure of electoral performance that is applied across a large number of cases. In sum, we can test if mass perceptions reflect the reality of how elections are conducted.

As noted earlier, the United States has one of the world's most decentralized systems of electoral administration. The level of decentralization makes for a range of variation in administration that, in turn, makes the US an especially interesting case for examining variation in the impact of administration. Non-US based scholars e and even many US students e may be surprised by the variation within the US: surely the Federal Election Commission plays a role in standardizing elections? Given both that it may be sur- prising to see how much variation exists and that the variation in administration is a rationale for choosing the US case it is worth spending a little time establishing just how varied is US experience.

There is no formal US equivalent of The Australian Election Commission, Elections Canada, or even the UK Electoral Commission. Although there are some federal statutes and Supreme Court rulings that create (weak) national guidelines for the conduct of elections, the pri- mary responsibility for administering elections remains in the hands of 50 different state governments.2 As examples of standardization we note that the Help America Vote Act (2002) set minimum standards for the maintenance of voter rolls, types of voting equipment used, and rules for provisional ballots. The Voting Rights Act (1965, and various amendments) regulates ballot information (non- English language) that certain county governments must use. The Federal Election Commission also helps to regulate and publicize campaign spending in a standard manner (for federal races) across the states.

Nevertheless, the individual states are left to fund and conduct federal elections. States have substantial discretion in how they may comply with national standards, and states have autonomy in setting rules for such things as voter registration processes and requirements, rules for showing identification, types of voting equipment used, rules for early voting, absentee voting, rules for recounts, and myriad other factors. As a concrete example consider voter registration. Although the Court has ruled that states cannot require a pre-election registration deadline greater than 30 days before an election there remains considerable variation. In practice state registration deadlines range from 0 (election day) to 30 days prior. States also have substantial discretion in determining the level of resources they invest in the conduct of federal elections, in managing voter registration (there is no national roll), in designing

2 Election administration is further decentralized in the US, with states delegating the conduct of elections to thousands of county governments.

ballots, setting rules for absentee voting and use of pro- vincial ballots, determining how (and whether) recounts and post-election audits are conducted, and in determining the number of polling places, their locations, and their hours of operation.

The result of this is that the elections are conducted in one state at a very different level of quality than in another state. Moreover, unlike cross-national analysis, in the US case variation across jurisdictions in major cultural factors is much more limited. Although we cannot definitively establish that changes in electoral quality affected feelings about the legitimacy of US elections, the states provide an excellent opportunity for testing how variation in the quality of the conduct of elections affects popular percep- tions about whether or not elections were conducted fairly.

4. Measuring the performance of elections, and election laws

Despite the decentralization of election administration in the US, and the resulting variety in the conduct of elec- tions across the states, there are also sufficient features in common among the states such that election performance can be measured in a standardized manner. For example, federal elections are all conducted at the same time under the same electoral system, and each state maintains similar records about: 1) the number of absentee ballots unre- turned, 2) absentee ballots rejected, 3) the number of provisional ballots cast, 4) provisional ballots rejected, and 5) registrations rejected. Data are also available for each state on 6) wait times for voting, 7) registration rates, 8) registration problems, 9) whether or not the state allows registrations on-line, 10) whether it requires post-election audits, 11) the accuracy of voting technology (residual votes), 12) completeness of data records, 13) voting infor- mation lookup tools, 14) military and overseas ballots rejected, 15) military and overseas ballots not returned, 16) disability-related voting problems and 17) turnout.

Indeed, the Pew Center for the States has used these seventeen qualities of election administration to create an Elections Performance Index (EPI) that makes it possible to compare the quality of election performance across the states, circa 2012.3 Pew rates each state with an overall average score that represents a summary measure of how it performed on the seventeen items (with each item given equal weight).4 In 2012 the top scoring states on Pew's election performance index were North Dakota (86), Min- nesota (80), Wisconsin (79), Colorado (79) and Nevada (77). The lowest ranked were Mississippi (44), Oklahoma (54), California (54), Alabama (56) and New York (58).

In addition to estimating how state-level election per- formance may have affected attitudes about electoral

Law), Bob Stein (Rice), Daniel Tokaji (Ohio State Law), and a group of state and local election officials to create the index.

4 State-level measures of election performance are also available for 2008 and 2010. Most states' performance scores were improving from 2010 to 2012, and scores from those years are well correlated (.82).

S. Bowler et al. / Electoral Studies 38 (2015) 1e94

integrity in the US, we also account for how a specific election rule e photo identification e may have affected how people viewed the electoral process. A major reason for including this measure is that, as of 2012, the most contentious election rule in the United States was likely photo identification laws. In the mid 2000s a number of states began adopting strict laws that required voters to provide election officials a government issued photo identification when voting at a polling place. Republican- controlled state governments promoted the laws as a tool for preventing voter fraud (voter impersonation), while Democrats alleged incidences of voter impersonation were rare and that the laws were designed to supress turnout among potential Democratic voters (Bowler and Donovan, 2013; Bentele and O'Brien, 2013). It is a highly charged political issue that has potentially large implications for election results (Richman et al., 2014) and voter participa- tion but, more to the point, it is one that is talked about by both main political parties as a major issue in election administration.

The National Conference of State Legislatures (NCSL) maintains a database that records which states had photo identification laws in effect at the time of the 2012 US presidential election, and the type of identification requirement that the state had. The NCSL reports a five- point scale, ranging from 1) strict rules where photo identification is required, 2) strict identification rules that allow the use of non-photo documentation, 3) rules that allow photo identification to be requested (but a person can still vote if she lacks identification), 4) rules that allow non- photo identification to be requested, and 5) states with no identification requirements. We code states on a five-point scale according to these NCSL classifications, with a range from one (if the state had no identification requirement) to five (if the state had a strict photo requirement). At the 2012 election, 30 states had some form of identification requirement, with strict photo requirements in Georgia, Indiana, Kansas, and Tennessee.5

These two measures e the Pew EPI index and the NCSL photo id categorization e provide us with measures of the variation of state level election administration which we can then use to explore citizen opinion towards elections.

5. Measuring perceptions of electoral integrity in the US

Wave 6 of World Values Survey (2010e2014) included several items designed to measure respondent's percep- tions of electoral integrity, and some of these items were included on the 2012 American National Election Study. Respondents were prompted with, “in your view, how often do the following things occur in this country's elec- tions.” They were then asked (separately) if votes are counted fairly, and if election officials are fair. These two items reflect key principles of electoral integrity recognized

5 Mississippi, Pennsylvania, South Carolina, Texas, and Wisconsin had adopted strict photo id rules prior to the 2012 election, but these laws were not in effect at the time due to court or Department of Justice challenges.

by international institutions (Hall and Wang, 2008:43). With each item, response categories ranged from “very often”, “fairly often,” “not often,” to “not often at all.” The distribution of responses to these questions from the US and several other nations are reported in Figs. 1 and 2. Wave 6 of the WVS was conducted in many nations with limited experience with democratic elections. Data dis- played in Figs 1 and 2 reflect attitudes about elections in the four nations where the WVS reported data from estab- lished, affluent democracies, with six additional cases included for comparative perspective.

Overall, Americans appeared to have been moderately confident in the integrity of their elections in 2012 e more so than respondents from Columbia and Mexico, but, depending on the item, less confident than respondents from Australia, the Netherlands, and Germany. Over sev- enty per cent of US respondents stated, respectively, that election officials were fair and votes were counted fairly at least “fairly often.” By this measure, Americans' perceptions of elected officials appear similar to those of Uruguayans and Poles. But there is substantial variation in Americans' attitudes about elected officials, and evidence of pessi- mism. Barely one-fifth of US respondents were confident that election officials were fair “very often.” Less than one- third of Americans believed that votes were counted fairly “very often.” Americans, then, were far more sceptical about the integrity of vote counting than respondents from Australia, Germany, and the Netherlands. Our task in the analysis that follows is to understand how state-level electoral performance (as measured by the Pew EPI) and a key state election rule (identification laws) explains variation in Americans' attitudes about elections that are illustrated in Figs. 1 and 2.

6. Hypotheses and models

As noted above we expect that partisanship will struc- ture attitudes about the integrity of elections in substantial ways, particularly in terms of fairness of process. Sup- porters of the party in power (or depending on the timing of a survey, the winning party) are likely to be satisfied with the result and that satisfaction with may project on to their views on the legitimacy of the process. The opposite applies for electoral losers. In the US in 2012, Democrats were the incumbent party in the White House, they controlled a

Fig. 1. Perceptions of electoral integrity: Per cent of respondents saying election officials are fair “very often” and “fairly often.”

Fig. 2. Perceptions of electoral integrity: Per cent of respondents saying votes are counted fairly “very often” and “fairly often.” Sources: American National Election Study, 2012. World Values Survey 6: Australia 2012, Estonia 2011, Germany 2013, Netherlands 2012, Chile 2011, Columbia 2012, Uruguay 2011, Mexico 2012.

6 Response options included all (4%), most (25.6%), about half (31.4%), a few (36.4%), and none (1%). The variable is recoded such that 1 ¼ none, through 5 ¼ all.

7 Models are estimated with Stata xtmixed. When the models were also estimated with ologit, results are similar to what we report (in terms of which effects were significant).

8 Median state household income 2012, and EPI for 2012 were corre- lated at just .03. State income is represented in units such that 41.1 ¼ $41,100).

S. Bowler et al. / Electoral Studies 38 (2015) 1e9 5

majority in one chamber of Congress (the Senate), and won additional House and Senate seats when President Obama was re-elected in November. Given this and given the elections questions were asked after results of the election were known, we expect people who identified as Democrat to be more optimistic about electoral practices when asked in 2012, and we expect people who identified as Republican to be less so. Our models of perceptions of electoral integ- rity thus include respective dichotomous markers for Democrats and Republicans.

In addition, dichotomous measures identify white, Af- rican American, and Latino respondents (0/1), respectively. Given the historic institutionalization of African American and Latino voter suppression in the US (Kousser, 1999; Keyssar, 2000; Davidson and Fraga, 1988) we expect that members of these minority groups may be more likely to view the conduct of elections as unfair. There may also be effects here associated with the lack of descriptive repre- sentation. Members of groups who are historically under represented in elected offices (relative to their share of the population) may be more likely to view elections as unfair. Although African Americans have achieved levels of descriptive representation at some levels commensurate to their share of the population, this has been relatively recent. Latinos, in contrast, are represented at levels far below their share of the population. We control for gender as well. Although women are not a demographic minority, they do constitute a minority in terms of their descriptive representation. The enduring underrepresentation of women in American politics may cause women to view elections as unfair on multiple dimensions.

Furthermore, since higher levels of education are known to be associated with greater efficacy and trust (Niemi and Jennings, 1991; Craig et al., 1990), we expect respondents' with higher levels of education to be more likely to think that elections have a meaningful role, and, by extension to perceive that electoral processes are fair. Ed- ucation is measured here in five categories, ranging from less than high school (1) to graduate degree (5). Our models also include terms for age, media consumption (frequency of TV news viewing per week), and a measure of political trust. It is important that we control for age, given that younger cohorts may experience socialization processes

that leave them to be less deferential to elections and democratic institutions (Denemark et al., 2012). Media viewing is included to account for the possibility that people who frequently view TV news are more likely to be exposed to stories that feature suspicion about electoral malpractice and about political scandals, which may cause them to view elections as unfair.

Perceptions of the conduct of elections and of the people who conduct them could be part of an overarching set of attitudes about government and the integrity of public of- ficials in general. Those less trusting of government have been shown to be more likely to worry about problems with election procedures (Nunnally, 2011). The pre-election wave of the 2012 ANES included a standard battery of trust in government questions as well as an item asking “how many of the people running the government are corrupt?” Most Americans indicated they believed that “about half” or more of “people running the government” were corrupt in 2012.6 If we consider standard definitions of public corruption this ANES item demonstrates that public per- ceptions about the extent of corruption in America are grossly inaccurate. However, the item may capture pre- election cynicism about public officials that coloured how people responded to post-election questions about election officials. Given this item is similar to the post-election question about election officials, we report models that include and omit it. Including the item as a control provides a very conservative test for the effects of state-level per- formance on attitudes about elections.

Our state level elections variables include the NCSL measure of photo identification laws and the Pew EPI scores described above. Since we assume that attitudes about electoral practices are structured not only by parti- sanship, but also in response to how people experience the quality of elections, we expect that higher EPI scores will be correlated with more optimistic perceptions of the fairness of elections. Our expectations about photo identification laws are more conditional. Given the charged partisan context of these laws, we expect Republicans and Demo- crats viewed them differently, and thus viewed their po- tential effects differently. To test this, we estimate our models with interaction terms that test if Republicans were more confident in elections in states where stricter iden- tification rules were in place.

Given the nature of our data, we estimate hierarchical linear models with random intercepts and five level-2 covariates: state EPI, state photo id laws, and controls for state median income, 2012 presidential vote margin in the state, and 2012 turnout.7 Income is included since state wealth may affect how much a state spends on the administration of elections,8 while margin accounts for

Table 1 Public attitudes about electoral integrity, United States.

Officials fair Vote count fair

I II III IV V VI

Level 1 TV news viewing .006 (.004) .002 (.004) .002 (.004) .001 (.004) �.001 (.004) �.001 (.004) Democrat .100** (.025) .037 (.025) .037 (.025) .133** (.026) .082** (.025) .083** (.026) Republican �.108** (.047) �.126** (.046) �.125** (.046) �.149** (.049) �.160** (.047) �.159** (.047) Black �.036 (.046) �.074 (.045) �.076 (.045) �.039 (.029) �.097* (.047) �.074 (.043) Latino �.058 (.032) �.086** (.032) �.082** (.032) �.070* (.033) �.092** (.032) �.091** (.032) Female �.125** (.023) �.115** (.021) �.114** (.021) �.118** (.022) �.108** (.021) �.108** (.021) Education .111** (.009) .085** (.009) .085** (.009) .111** (.009) .090** (.009) .090** (.009) Age (groups) .020** (.003) .014** (.003) .014** (.003) .016** (.004) .011** (.003) .011** (.003) Sample �.113** (.023) �.088** (.023) �.090** (.023) �.123** (.023) �.101** (.023) �.102** (.004) Officials corrupt? e �.243** (.011) �.242** (.012) e �.212** (.012) �.211** (.012) Level 2 State income .005** (.002) .003 (.002) .002 (.002) .004** (.002) .003 (.002) .002 (.002) Margin .002 (.002) .001 (.002) .002 (.002) .003 (.002) .002 (.002) .002 (.002) Turnout e e .0086** (.0032) e e .0049 (.0027) Election performance .0059** (.0023) .0056** (.0022) e .0039* (.0019) .0033 (.0019) e Photo ID law �.002 (.010) �.001 (.010) .006 (.011) �.002 (.002) �.010 (.009) �.005 (.009) Republican* ID law .026 (.015) .024 (.015) .024 (.015) .026 (.015) .023 (.015) .024 (.016) Constant 1.79** (.188) 2.79** (.189) 2.68** (.208) 2.21** (.170) 3.08** (.170) 3.03** (.185) Wald chi2 293.1** 745.0** 745.6** 300.0** 630.7** 630.9** Level 1 n 5307 5236 5236 5343 5261 5261 Level 2 n 50 50 50 50 50 50 Level 1 R2 .05 .12 .12 .05 .12 .12 Level 2 R2 .01 .01 .01 .01 .01 .01

Note: HLM with random intercept and level 2 covariates, estimated with Stata xtmixed; **p. < .01, *p. < .05 (two-tail). Source: 2012 ANES

S. Bowler et al. / Electoral Studies 38 (2015) 1e96

variation in electoral context. Turnout is included in some models as a separate measure of electoral quality e it is one of the 17 items in the EPI, and other items in the index (registration rates, registration problems, on-line registra- tion, etc.) and may reflect a better quality of election administration that is expressed in higher turnout. We estimate three models of both dependent variables; one model without the control that accounts for general atti- tudes about official corruption, one with that control, and one that includes turnout.

7. Results

Table 1 reports results of the models estimating Amer- icans' responses to these WVS/ANES questions about elec- toral integrity. The estimates of individual level factors are largely consistent with our expectations. Other things equal, partisans differed in how they viewed election offi- cials and vote counting after the 2012 contests. Republicans were consistently less confident that election officials and vote counts were fair, while Democrats were more confi- dent. The effect of Democratic partisanship on perceptions of election officials is muted when the control for attitudes about government officials is included. However, Demo- crats are consistently associated with greater confidence in vote counting. Women, Latino/as, African Americans,9

younger people, and the less educated, respectively, were less likely to respond that election officials were usually fair

9 The coefficients for African Americans in Model II, III and VI are sig- nificant at p. < .10 two-tail, or p. ¼ .05 one-tail; the coefficient for Latinos in Model I is significant at p ¼ .08 (two-tail).

and less likely to say that votes were usually counted fairly. That is, independent of partisanship, and independent of how well elections were conducted, women and minorities were less likely to think that US elections were fair.10 Table 2 (below) illustrates the substantive magnitude of these effects. Compared to a baseline respondent (a white male in a state with average EPI), women of colour were much less likely than others to see officials and vote counts as fair.

But what of state-level factors? We find that a state's 2012 Election Performance Index score was associated with how individuals responded to the question about the fair- ness of election officials and vote counts. Regardless of whether or not the models include controls for general cynicism about government officials, the coefficients for EPI are positive and statistically significant in estimates of perceptions that officials were fair and in estimates that vote counts were fair.11 In states where this objective measure indicates the conduct of elections to be of higher quality, respondents were significantly more likely to say elections were fair. As shown in Table 2, the substantive magnitude of this relationship is modest e at least when compared to the individual-level effect of partisanship and gender.

Models III and VI replace the EPI measure of election administration with a measure of state-level turnout. Turnout also has a significant, positive association with

10 Contrary to our expectation, we find no relationship between TV news consumption and attitudes about election officials and vote counts. 11 When the control for attitudes about corruption is included in Model V the estimated coefficient for EPI on perceptions of fair vote counting is slightly smaller, and significant at just p ¼ .09 (two-tail).

Table 2 Predicted probability of responding that election officials were fair and vote count was fair “very often.”

Officials fair Count fair

Baselinea .204 (.009) .333 (.013) Female .154 (.008) .267 (.012) Latino .161 (.016) .258 (.021) Latina .121 (.012) .203 (.019) Black, male .175 (.013) .279 (.017) Black, female .148 (.016) .215 (.015) Republican .182 (.009) .286 (.013) Democrat .244 (.012) .407 (.015) Lowest EPI state .154 (.020) .284 (.024) Highest EPI state .256 (.020) .378 (.022)

Note: Predicted probabilities generated from post-estimation Clarify simulations of ordered logit models (standard errors clustered by state).

a Baseline respondent is an independent, white male, with mean values on other variables (age, education, media viewing, perceptions of officials corrupt, state EPI, and state id rules). Source: ologit models replicating models I in Table 1, available from au- thors. Dependent variables range from 0 (“not at all often”) to 4 (“very often”).

S. Bowler et al. / Electoral Studies 38 (2015) 1e9 7

perceptions that election officials are fair (p. < .01, two-tail) and that votes were counted fairly (albeit at p. < .07, two- tail). It is possible that this reflects people who are more confident about the administration of elections being more likely to vote. However, that causal logic is not consistent with the fact that EPI, and components of EPI that remove turnout (see robustness tests below) predict greater con- fidence in elections. Turnout and EPI are highly corre- lated,12 and we expect this reflects a close relationship between the quality of election administration and turnout. This suggests the potential of a causal process where, over time, improvements in the quality of electoral adminis- tration may increase voter access and turnout while also improving voter confidence in elections.13

The potential effects of voter identification laws on perceptions of electoral legitimacy appear limited. None of our models yield any significant, direct relationship be- tween the strictness of a state's voter identification laws, and perceptions about the conduct of elections. We did anticipate that Republicans, as supporters of photo ID laws, might view the conduct of elections more positively where strict identification laws were in effect. There might be something to this, but the relationships, if any, are weak. The estimates show that although Republicans generally were less likely to view elections as fair, Republicans in states with strict identification rules were more likely to see elections as fair e but in each case the statistical sig- nificance (p ¼ . 09 two-tail) does not reach conventional levels.

Our mixed level models allow us an additional tool to assess the substantive magnitude of state-level versus individual-level factors. LR tests comparing the fit of mixed-level random effects ANOVA models to individual-

12 Being correlated at .73, both items cannot be included in the same model. When they are, neither term is significant. 13 Put differently, we have sound reason to expect that the quality of election administration, as measured by Pew, affects perceptions of electoral integrity. We have less reason to expect the opposite.

level OLS models reject the hypothesis that there is no cross-state variation in perceptions of officials being fair (p. < .0000) and in perceptions that vote counts were fair (p. < .000). However, we find that most (nearly all) of the variation in these attitudes is due to individual-level factors.14

Table 2 displays the substantive magnitude of the esti- mated relationships between key independent variables (state level EPI and individual-level demographic traits and perceptions of fair elections. For the sake of illustration, and given that individual-level factors are the major influence on these attitudes, the predicted probability of a respon- dent saying officials and vote counts were “very often” fair were estimated from ordered logit models (available from the authors).15 A respondent in a median EPI state had an estimated .204 probability of saying election officials were fair “very often,” and a .333 probability of saying vote counts were fair very often. The probability of respondent saying this in the state with the highest EPI was predicted to be about .05 higher for each item. In contrast, partisan- ship had a much larger estimated effect on perceptions of fair vote counts e with Democrats predicted to have a .12 greater probability than Republicans of saying counts were fair. Table 2 also illustrates the extent to which women of colour were less likely to view officials and vote counts as fair. Compared to other respondents in a median EPI state, a Latina had a .12 lower probability of saying counts were fair very often, and a .08 lower probability of saying officials were. These effects of party, race, ethnicity and gender on perceptions of fair elections thus rival or exceed the measured effects of electoral performance.

8. Robustness tests

We conducted additional analysis to assess the veracity of these results. First, we examined if the relationships we detected between EPI and attitudes were a quirk of the 2012 ANES, or if models estimated on a different survey platform produced similar relationships. Second, we decomposed the EPI measure to assess if there were di- mensions of the index that were associated with particular attitudes about elections. As for the first matter, the 2012 Cooperative Comparative Election Study (CCES) also included an item that asked respondents, “are election of- ficials fair?” As with results reported here in Table 1, esti- mates using CCES data yielded a significant positive relationship between EPI and perceptions that officials were fair; with the estimated size of the coefficients of EPI similar regardless of whether CCES or ANES data were used (CCES estimates available from the authors).

We conducted a factor analysis to assess which di- mensions of the EPI were related to perceptions of elec- tions. This produced five unique dimensions of electoral performance, allowing us to replicate the ANES models in

14 The inter-class correlation (ICC) for the percent variance explained by level-2 factors is .013 (1.3%) for attitudes about officials being fair, and .008 for attitudes about fair vote counts. 15 The relationships between EPI and both questions about elections are significant (p < .01) when estimated with ologit.

S. Bowler et al. / Electoral Studies 38 (2015) 1e98

Table 1 using the five factor (z) scores (rather than EPI) to estimate perceptions of elections. Four of the factors were associated with attitudes about fair elections, and one was not.16 One dimension of state-level performance charac- terized by higher registration rates, high turnout, and low rates of disability related voting problems was significantly and positively associated with perceptions that officials were fair, and that votes were counted fairly. A second dimension of administration characterised high rates of data completeness, and low rates of military ballots being rejected also had a significant, positive association with perceptions that officials were fair and votes were fairly counted. Two other factors were associated with specific attitudes: People in states with higher scores on a dimen- sion representing the presence of tools for looking up voting information were significantly more likely to say votes were counted fairly, while those in states scoring higher on a dimension that reflected online registration were significantly more likely to say officials were fair. This analysis suggest that rather than there being a single item in the EPI driving our results, there are multiple aspects of election administration captured by the EPI that are related to perceptions of elections.

9. Discussion

Even with some very conservative tests, we find that an objective measure of the quality of election administration explains some variation in perceptions of fair elections. In states that scored higher on a measure of administrative quality, people were more confident that election officials were fair, and that votes were counted fairly. Champions of reforms designed to improve the administration of elec- tions should find solace in these results. Our analysis demonstrates that people viewed elections as fairer where elections were conducted better. By extension, this implies that improvements in the governance of elections could also promote democratic legitimacy, as fair elections are a key feature of democratic processes.

We suggest it is one thing to find greater confidence where elections are ran reasonably well than where they are notoriously bad, but another thing to find greater confidence across places where, by international standards, all elections are conducted reasonably well. A sensible intuition might have us expect that elections in Robert Mugabe's Zimbabwe do much less in establishing legiti- mate authority there, than compared to well-administered elections in established democracies. Similarly, many might expect variation in election quality to shape perceptions of legitimacy across emerging democracies, where ‘quality’ ranges from elections with wide-spread fraud and intimi- dation to elections that are relatively free and fair. But our study demonstrates that even within an established de- mocracy, where systematic fraud is rare and where in- stitutions and socialization forces produce expectations

16 The first factor represents states that had higher rates of provisional ballots cast, provisional ballots rejected, and higher rates of absentee ballots rejected and unreturned. This dimension of electoral performance was not associated with attitudes about elections.

that elections will be reasonably fair, people do appear to notice when things are running very well and when they are running less well.

This said, there are several points of caution. Patterns of what, for a want of a better term, we might call a ‘demo- cratic divide’ in the US structured on demographic lines (race, gender) are both robust and worrying. One of the implications of these results is that large numbers of Americans remain sceptical of a key feature of their dem- ocratic process. It is unclear whether the kinds of factors identified in the Pew election performance index can be relied on to, eventually, bring all voters round to the idea that elections are conducted fairly. State election adminis- trators and reforms alike may therefore also find cause for concern e and rationale for more targeted action e in the opinions tied to demographic patterns. It may be that no matter how well things are administered, there are many people e particularly those from groups who are under represented -who might still see the electoral process as flawed, and unfair.

A second note of caution is that it is difficult to evaluate the substantive magnitude of the relationships we observe between election performance, and attitudes about fair elections. The relationships between the Pew measure of state-level election performance and attitudes are signifi- cant and robust across many alternative model specifica- tions, but little of the cross-state variation in these attitudes are explained by this measure of election performance. It seems unreasonable to expect large substantive effects from these features on attitudes about democratic pro- cesses; that is, we might not expect technical, administra- tive improvements by election officials to swamp the effects of individual partisanship, minority status, or the episodic effects of controversies such as bitterly contested recounts. It is nevertheless impressive that we find people to be modestly responsive to these rather routine features of elections.

Finally, it is worth underscoring some other limits of the results. It is notable, for example, that for all the sound and fury about voter photo ID there is very little evidence here or in additional results not-reported that photo ID had much effect on making people see the conduct of elections in a more positive light. It is also the case that the EPI measure had only a very modest effect on whether people thought votes were counted fairly, despite the large N of the sample. These points, in turn, raise a broader and possibly quite troubling question namely: in democracies where basic electoral practices are fairly well established, do marginal improvement in the quality of administration of elections 'matter' with respect to attitudes about de- mocracy that are most important? Our results suggest that e in broad terms e technical improvements to electoral administration can improve voter perceptions of elections being fair. These are worthy accomplishments. But another implication of these findings is that substantial numbers of people remain unpersuaded that American elections are fair. We return, then, to an earlier theme. There are limits to what we can expect electoral reform to accomplish, and that applies to reforming election administration as well. None of that should be taken as an argument against improving electoral processes, but, rather, as a suggestion

S. Bowler et al. / Electoral Studies 38 (2015) 1e9 9

to have modest expectations about what such reforms may accomplish.

References

Anderson, C., LoTempio, A., 2002. Winning, losing, and political trust in America. Br. J. Polit. Sci. 32 (2), 335e351.

Anderson, C., Tverdova, Y., 2003. Corruption, political allegiances, and attitudes toward government in contemporary democracies. Am. J. Polit. Sci. 47 (1), 91e109.

Anderson, C., Blais, A., Bowler, S., Donovan, T., Listhaug, O., 2005. Losers' Consent: Elections and Democratic Legitimacy. Oxford University Press.

Ansolobehere, S., Persily, N., 2008. Vote fraud in the eye of the beholder: the role of public opinion in the challenge to voter identification re- quirements. Harv. Law Rev. 121, 1737e1774.

Banducci, S., Karp, J., 2003. How elections change the way citizens view the political system: campaigns, Media effects, and electoral out- comes in comparative perspective. Br. J. Polit. Sci. 33 (3), 443e467.

Bentele, K., O'Brien, E.E., 2013. Jim crow 2.0? why states consider and adopt restrictive voter access policies. Perspect. Polit. 11 (4), 1088e1116.

Birch, S., 2008. Electoral institutions and popular confidence in electoral processes: a cross-national analysis. Elect. Stud. 27 (2), 305e320.

Blais, A., Gidengil, E., Nevitte, N., 2004. Where does turnout decline come from? Eur. J. Polit. Res. 43 (2), 221e236.

Blais, A., 2006. What affects voter turnout? Annu. Rev. Polit. Sci. 9, 111e125.

Bowler, S., Donovan, T., 2013. The Limits of Electoral Reform. Oxford University Press, Oxford.

Bowler, S., Donovan, T., 2007. Reasoning about institutional change: winners, losers and support for electoral reforms. Br. J. Polit. Sci. 37 (3), 455e476.

Campbell, A., Converse, P.E., Miller, W.E., Stokes, D.E., 1960. The American Voter. Wiley, New York.

Craig, S., Martinez, M., Gainous, J., 2006. Winners, losers, and election context: voter response to the 2000 presidential election. Polit. Res. Q. 59 (4), 579e592.

Craig, S., Niemi, R., Silver, G., 1990. Political efficacy and trust: a report on the NES pilot study items. Polit. Behav. 12 (3), 289e314.

Dalton, R., 2009. The Good Citizen: How a Younger Generation Is Reshaping American Politics. CQ Press, Washington, DC.

Davidson, C., Fraga, L., 1988. Slating groups as parties in a nonpartisan setting. West. Polit. Q. 41 (2), 373e390.

Denemark, D., Donovan, T., Niemi, R., 2012. Generations and Democratic Attitudes in Advanced Democracies. Paper presented at the World

Congress of the International Political Science Association, Madrid, Spain. July 10.

Elklit, J., Reynolds, A., 2005. A framework for the systematic study of election quality. Democratization 12 (3), 147e162.

Elklit, J., Reynolds, A., 2002. The impact of election administration on the legitimacy of emerging democracies. J. Commonw. Comp. Polit. 40 (2), 86e119.

Esaiasson, P., 2011. Electoral losers revisited- how citizens react to defeat at the ballot box. Elect. Stud. 30 (1), 102e113.

Hall, T., Wang, T.A., 2008. International principles for electoral integrity. In: Michael Alvarez, R., Hall, T.E., Hyde, S.D. (Eds.), Election Fraud: Detecting and Deterring Electoral Manipulation. Brookings Institution Press, Washington, DC.

Inglehart, R., 1990. Culture Shift in Advanced Industrial Society. Princeton University Press, Princeton.

Keyssar, A., 2000. The Right to Vote. Basic Books, New York. Kousser, M., 1999. Color Blind Injustice: Minority Voting Rights and the

Undoing of the Second Reconstruction. University of North Carolina Press.

McAllister, I., White, S., 2011. Public perceptions of electoral fairness in Russia. Europe-Asia Stud. 63 (4), 663e683.

Moy, P., Pfau, M., 2000. With Malice toward All? the Media and Public Confidence in Democratic Institutions. Praeger, Westport, CT.

Mozaffar, S., Schedler, A., 2002. The comparative study of electoral governance. Int. Polit. Sci. Rev. 23 (1), 5e27.

Nadeu, R., Blais, A., 1993. Accepting the election outcome: the effect of participation on losers' consent. Br. J. Polit. Sci. 23 (4), 553e563.

Niemi, R.G., Jennings, M.K., 1991. Issues and inheritance in the formation of party identification. Am. J. Polit. Sci. 35, 969e988.

Norris, P., 2014. Why Electoral Integrity Matters. Cambridge University Press, New York.

Norris, P., 2004. Electoral Engineering: Voting Rules and Political Behavior. Cambridge University Press.

Nunnally, S., 2011. (Dis)counting on democracy to work: perceptions of electoral fairness in the 2008 presidential election. J. Black Stud. 42 (6), 923e942.

Persily, N., Lammie, K., 2004. Perceptions of corruption and campaign finance: when public opinion determines constitutional law. Univ. Pa. Law Rev. 153 (1), 119e180.

Putnam, R., Leonardi, R., Nanetti, R., 1994. Making Democracy Work: Civic Traditions in Modern Italy. Princeton University Press.

Richman, J.T., Gulshan, C., Earnest, D.C., 2014. Do non-citizens vote in U.S. Elections? Elect. Stud. 36, 149e157.

Singh, S., Karakoc, E., Blais, A., 2012. Differentiating winners: how elec- tions affect satisfaction with democracy. Elect. Stud. 31 (1), 201e211.

Wilson, D., Brewer, P., 2013. The foundations of public opinion on voter ID laws: political predispositions, racial resentment and information effects. Public Opin. Q. 78 (1).

  • Election administration and perceptions of fair elections
    • 1. Introduction
    • 2. The research question
    • 3. The advantage of the American case
    • 4. Measuring the performance of elections, and election laws
    • 5. Measuring perceptions of electoral integrity in the US
    • 6. Hypotheses and models
    • 7. Results
    • 8. Robustness tests
    • 9. Discussion
    • References