Ranked Choice Voting in Maine from the Perspective of Local Election Officials

ELECTION LAW JOURNAL Volume 0, Number 0, 2021 # Mary Ann Liebert, Inc. DOI: 10.1089/elj.2020.0650 Ranked Choice Voting in Maine from the Perspective of Local Election Officials Joseph Anthony, Amy Fried, Robert Glover, and David C. Kimball Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. ABSTRACT In 2018, Maine became the first American state to use ranked choice voting in statewide and congressional elections. In a decentralized system of election administration local officials play an important role implementing new voting rules. We examine the implementation of ranked choice voting based on a survey of municipal clerks and interviews with several election officials in Maine. We test hypotheses derived from policy implementation theory in assessing election officials’ evaluations of ranked choice voting. We find that most municipal clerks in our sample are not enthusiastic about implementing ranked choice voting and do not want to continue its use in Maine. In addition, there are strong partisan divisions, with Democratic clerks offering much more positive assessments of ranked choice voting than Republican clerks. We also find differences in voter education efforts across jurisdictions. The findings raise concerns about the implementation of ranked choice voting in Maine. Keywords: ranked choice voting, election reform, election administration, local election officials, policy implementation INTRODUCTION ranked choice voting (RCV). Local election officials play a crucial role in election administration, and the municipal clerks in Maine were able to observe how RCV worked in their towns. RCV was later expanded in Maine to include presidential elections, however none of these contests went to RCV tabulation in 2020. We draw on policy implementation theory to test several hypotheses about how local election officials evaluate ranked choice voting. The clerks who answered our survey are not very fond of ranked choice voting. We also find substantial partisan differences in support for the goals of RCV. These partisan differences are strongly associated with overall assessments of RCV and support for continuing its use in Maine elections. By comparison, Maine clerks do not see significant new costs in implementing RCV and perceived administrative burdens are only modestly associated with their assessments of RCV. Finally, the views of Maine clerks do not seem to be influenced by voter support for RCV in their constituencies. In the case of ranked choice voting, partisanship seems to trump administrative burden and T his study examines the role of local election officials (LEO) as implementers of state election reforms. The primary data come from a survey of municipal clerks in Maine conducted after the 2018 general election, as well as interviews with many local officials, garnering their assessments of Joseph Anthony was a visiting assistant professor in the Department of Political Science at Oklahoma State University in Stillwater, Oklahoma, USA, and is currently Election Administration, Law, and Policy Fellow at Auburn University in Auburn, Alabama, USA. Amy Fried is a professor and chair of the Department of Political Science at the University of Maine in Orono, Maine, USA. Robert Glover is an associate professor of political science and honors at the University of Maine in Orono, Maine, USA. David C. Kimball is a professor and interim department chair in the Department of Political Science at the University of MissouriSt. Louis, in St. Louis, Missouri, USA. This research is supported by a grant from the Hewlett Foundation. Thanks to Shelly Crosby of the Maine Town and City Clerks’ Association for assisting us with the survey. 1 Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 2 constituent preferences in explaining how Maine municipal clerks evaluate ranked choice voting. This may raise concerns about how the reform is being implemented in Maine. More generally, implementation theory should consider the political views of local officials on policies that engender substantial partisan conflict. Ranked choice voting is an election reform that appears to be gaining attention and support in the United States. Until recently, this process of ranking candidates (also known as preferential voting or instant runoff voting) was more familiar in other corners of the world, such as Australia (Bowler and Grofman 2000; Reilly 2001; Farrell and McAllister 2006), Ireland (Farrell and Sinnott 2017; Endersby and Towle 2014) or parts of the UK (Clark 2013; Lundberg 2018). Rather than asking voters to indicate one preferred candidate, RCV allows voters to rank candidates in order of preference. While over a dozen municipalities across the United States employ ranked choice voting in some fashion, in 2018 Maine became the first state to use RCV for some state and federal elections. Maine also allows voters to rank as many preferences as there are candidates in ranked choice races. The ballot-counting process for ranked choice voting is more complex than plurality (winnertake-all) systems. In RCV contests with a single winner, voters’ first choices are counted, and if no candidate receives a majority, then the counting process continues. The candidate with the lowest share of the votes is eliminated in the next round of counting, and those voters’ second choices are allocated to the remaining candidates. This process continues until one candidate receives a majority of the available votes. Evaluations of ranked choice voting elections in the United States are mixed. Some find that ranked choice voting systems increase civility and decrease negative campaigning (Donovan, Tolbert, and Gracey 2016; Mauter 2014). Additionally, there is some evidence that RCV elections tend to attract more candidates to run for office, which can lead to a more representative set of candidates and elected officials (John, Smith, and Zack 2018). On the other hand, ranked choice voting makes the ballot more complicated, and voters may be confused by new rules that allow them to choose multiple candidates (Donovan, Tolbert, and Gracey 2019). Additionally, as is the case with Maine, some ballots may be split between RCV contests and races that ANTHONY ET AL. follow plurality rules. These new rules and mixed ballot designs could cause voter confusion. The adoption of ranked choice voting in Maine was a highly partisan affair, with clear expectations that the reform would help Democratic candidates in the short term. But it also came with some new administrative burdens. To what extent do these factors—increased burden and LEO partisanship—drive election officials’ views of reform, and the negative assessments of RCV in Maine? What might this say about other electoral reforms that introduce measurable burden? The answers to these questions are important because they point to different paths for election reform. If administrative burden is the main source of opposition to reform among local election officials, then the solution for reformers is to provide additional resources and support to local election officials. On the other hand, if partisanship is the main barrier to reform, then the solution is political—either build robust bipartisan coalitions (including election officials) to support reform or work to elect and appoint local officials from the political party that supports reform. WHY THE OPINIONS OF LOCAL ELECTION OFFICIALS MATTER In a decentralized system of election administration local officials play a crucial role in implementing election laws. The conditions facing local administrators vary dramatically, depending in part on the size and composition of the local electorate, and these conditions influence the attitudes of local officials (Moynihan and Silva 2008; Montjoy 2008). Local election officials typically have many responsibilities, which include hiring and training poll workers, purchasing and maintaining voting equipment, designing and printing ballots, and maintaining a list of registered voters. While local administrators are required to comply with state and federal election laws, those laws leave some room for interpretation and local officials may vary in how they implement those laws. There is considerable variation across local jurisdictions in measures of voting behavior like turnout, residual votes, and provisional ballots, and some of that variation is attributed to local election administration (e.g., Ansolabehere and Stewart 2005; Burden et al. 2013; Fullmer 2015). Local officials play an influential role in administering elections. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. RANKED CHOICE VOTING IN MAINE In addition, the views of local officials may shape how they implement election laws. A wide range of evidence indicates that the attitudes of local government officials influence how they enforce the law (e.g., Farris and Holman 2015). The ‘‘hyperdecentralization’’ of election administration is frequently noted as an obstacle to reform (e.g., Tokaji 2007). Local officials’ views toward election reforms help determine whether they embrace or resist those reforms (Moynihan and Silva 2008; Burden et al. 2012). Furthermore, the opinions of local officials toward election laws help predict local outputs, like the casting and counting of provisional ballots (Kropf, Vercellotti, and Kimball 2013). Similarly, the party affiliations of election officials are associated with their decisions to purge voter rolls (Stuart 2004), select early voting locations (McBrayer, Williams, and Eckelman 2020), choose titles for ballot measures (Lund 1998), and implement straightparty voting mechanisms (Hamilton and Ladd 1996). Under certain conditions the partisanship of local officials even seems to influence voter turnout (Burden et al. 2013).1 The opinions of local officials shape how they use their discretion in enforcing the law. Finally, local officials offer an important perspective on election law and administration. As ‘‘street-level bureaucrats’’ (Lipsky 1980), local administrators have more direct contact with voters and a closer perch to observe the voting experience than state or national officials. The public tends to evaluate local election officials more positively than state and national government institutions (Atkeson 2011; Adona and Gronke 2018). While legislatures may not always enact their preferred policies, local election officials are frequently asked to share their perspective on reform proposals with state and national policymakers (e.g., PCEA 2014). Local election authorities and the organizations that represent them can be very influential in the development and adoption of state election laws. For instance, the Help America Vote Act (HAVA) stipulated that in order to receive federal funds, states must engage with and receive input from state and local election authorities in developing state-based enabling legislation (Palazzolo and Ceaser 2005: 28). Organizations such as the National Association of State Election Directors (NASED) and the Election Center convened public hearings and meetings across the country to share information and ‘‘interpretations’’ of the newly passed HAVA (Palazzolo and Ceaser 2005: 21). 3 In states that enacted ‘‘major reform’’ around HAVA, local election authorities were an integral part of the policy development and adoption process (Palazzolo and Ceaser 2005). In many other states, opposition from local election officials helped stymie or delay the adoption of new voting equipment and other reforms after the passage of HAVA (Lochner and Moncrief 2005; Maddox 2005; Kropf 2005; Krutz 2005). More recently, county clerks in Missouri helped defeat legislation offered by the GOP majority in the General Assembly to adopt closed primary elections (Crowley 2020). Given the power of LEOs on election policy development and implementation, we expect that they will be influential moving forward in implementing RCV in Maine, as well as in the passage of these kinds of reforms in other states. MAINE’S ADOPTION AND IMPLEMENTATION OF RANKED CHOICE VOTING Efforts to adopt RCV in Maine date back at least to 2001 (Santucci 2018; Armstrong 2019). After several failed attempts in the legislature, RCV supporters qualified a measure for the ballot in 2016 to adopt ranked choice voting for some state and federal elections. The reform effort was led by one of the major parties and some interest groups, but not by election officials. In fact, state and local election officials expressed some misgivings about the legality and cost of ranked choice voting, particularly in general elections (Armstrong 2019). Nevertheless, the ballot measure passed in the 2016 general election with 52 percent of the vote. Ranked choice voting was used for the first time for state and congressional primary elections in June 2018 and for congressional races in the November 2018 general election. Maine is a state of roughly 500 relatively small towns. The largest city (Portland) had a bit less than 40,000 voters in the 2016 presidential election. More than 10 percent of the state’s towns had fewer than 100 voters in the same election. The vast majority of municipalities have a single polling place for state and federal elections. Nevertheless, Maine has a relatively high degree of state centralization 1 For a more recent example of partisan conflict among election officials over voter purges see Smith and Wines (2020). ANTHONY ET AL. of election administration, particularly when it comes to ranked choice voting. Every town in Maine votes on paper ballots that can be scanned or counted by hand. In the 2018 elections the Secretary of State’s office designed and printed all ballots, including the RCV sections, and distributed them to each town. In RCV contests that required additional vote counting the Secretary of State’s office collected the ballots from each municipality and brought them to the state capital for further tabulation (Armstrong 2019). Nevertheless, municipal clerks in Maine have an important role in implementing ranked choice voting. Local clerks count the first-choice votes in each of the RCV contests. More importantly, municipal clerks need to train poll workers and educate voters on how ranked choice voting works and how to complete their ballots. The state played a supporting role here as well. The Secretary of State’s office produced ballot instructions, posters, videos, and other RCV training materials which were shared with the municipal clerks. Finally, municipal clerks are in a position to directly observe the voting experience with RCV in their localities. EXPLAINING CLERK EVALUATIONS OF RCV In this section we use policy implementation theory to develop hypotheses about how municipal clerks in Maine evaluate ranked choice voting. There is a well-developed framework in public administration for understanding the views of bureaucrats toward policy reforms that considers the attitudes of local officials, relationships between officials at different levels of government, and characteristics of the jurisdictions that local officials serve. In applying these ideas, we focus on partisanship and concern about administrative burdens as competing explanations for resistance to new voting rules among municipal clerks in Maine. American election administration is known for two unique qualities, localism and partisanship (Ewald 2009; Hasen 2012). Many decisions about how elections function are left up to local administrators, who are often elected on a partisan ticket or appointed by partisan officials. Partisan disagreements about election reforms are common in the United States, and partisan conflict is evident among election officials as well. As we note above, several studies find that party identification is associated with prefer- ences for reform among election officials, as well as their administrative choices (e.g., Stuart 2004; Hamilton and Ladd 1996). The adoption and implementation of ranked choice voting in Maine has been and remains a highly partisan issue. The adoption movement accelerated after the election of Republican Governor Paul LePage with a plurality of the popular vote in 2010 and again in 2014. There was a feeling in some quarters that Democratic and Independent candidates split the anti-LePage vote in both elections, allowing him to win under plurality rules (Santucci 2018; Armstrong 2019). The campaign to qualify an RCV measure for the ballot in 2016 was led by two experienced state legislators, one a Democrat and one an Independent. The RCV campaign was also supported by several liberal groups and the Independent candidate for governor in 2014. Governor LePage and the Maine Republican Party opposed the RCV ballot question, as did a conservative think tank closely aligned with the governor. Some GOP leaders saw ranked choice voting as an attack on Governor LePage that would weaken the ability of conservative candidates to win closely contested elections in Maine (Gratz, Mistler, and Leary 2018). Partisan divisions over ranked choice voting in Maine can be seen in the 2016 general election, when the measure to adopt ranked choice voting (Question 5) passed. Figure 1 plots voter support for the RCV ballot question by Donald Trump’s share of the presidential vote in Maine’s 502 municipalities (the data symbols are sized in proportion to the number of votes cast in each jurisdiction). There is a strong linear relationship between partisanship and the vote on Question 5 (the Pearson correlation 100 90 Percent Yes on RCV in 2016 Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 4 80 70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 Trump Vote Share in 2016 80 90 100 FIG. 1. Support for ranked choice voting measure by Trump support in Maine municipalities, 2016 general election. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. RANKED CHOICE VOTING IN MAINE between support for Trump and the RCV measure is .95). As expected, voter support for ranked choice voting declines as Trump’s vote share increases. Santucci (2018) estimates that roughly 80 percent of Democrats voted for the RCV measure while 80 percent of Republicans voted against the measure. The RCV measure also fared better in heavily populated municipalities and in jurisdictions along the southern coast of Maine (Gillespie, Levan, and Maisel 2019).2 The partisan conflict continued past 2016. In 2017, the Maine Supreme Court issued an advisory opinion stating that ranked choice voting was unconstitutional in the general election for governor and state legislative contests. The legislature then passed a law to delay the use of RCV until voters passed a constitutional amendment to allow RCV. Republican leaders amended the legislation to delay all RCV contests even though the court opinion was limited to elections for governor and state legislature (Armstrong 2019). In response, the same groups behind the 2016 Question 5 campaign gathered petition signatures to place a ‘‘people’s veto’’ referendum on the ballot to repeal the legislation delaying RCV in Maine. The people’s veto qualified for the ballot in 2018. In the meantime, GOP leaders in the state senate and the Maine Republican Party filed lawsuits to prevent the use of RCV in the June 2018 primary election. The lawsuits failed and ranked choice voting was used for the first time in the June 2018 primary elections for governor and congressional seats. Two contests, the Democratic nominations for governor and the 2nd congressional district, went to multiple rounds of ballot counting in order to determine the winner (Armstrong 2019). The RCV ‘‘people’s veto’’ was also passed by Maine voters in the June 2018 election, with 54 percent voting for the ballot measure. The 2018 statewide vote on ranked choice voting looks like a rerun of the 2016 RCV vote in important ways. At the municipal level, support for ranked choice voting in 2018 was strongly and positively correlated with support for RCV in 2016 (r = .90). Once again, voter support for RCV in 2018 was negatively correlated with support for Trump in 2016 (r = -.92). Partisan battles over RCV continued in late 2018. In the fall Governor LePage vetoed a bill to provide emergency funding for the additional costs of administering an election with ranked choice voting. The campaign for Maine’s 2nd congressional district was hard fought and ranked choice voting rules played a critical role in determining the winner. The contest featured Republican incumbent Bruce 5 Poliquin, Democratic challenger Jared Golden, and two Independent candidates. After the first-choice votes were counted, Poliquin led Golden by a little more than 2,000 votes. However, Poliquin only received a plurality of the vote (a bit over 46 percent), as the two Independent candidates garnered more than 23,000 votes between them. Under RCV rules, this triggered additional rounds of vote tabulation. Both Independent candidates had encouraged their supporters to select Golden as a second choice. Perhaps as expected, after those ballots were transferred to second and third choice candidates, Golden won a majority, defeating Poliquin by roughly 3,500 votes. Poliquin challenged the outcome in federal courts and requested a recount. The courts ruled against Poliquin and he eventually withdrew his request for a recount (Armstrong 2019). Governor LePage also criticized the use of ranked choice voting in that election. The governor wrote ‘‘stolen election’’ on the document certifying the official results of the 2nd congressional district contest (Thistle 2018). Finally, there were important differences in the way candidates campaigned under RCV rules in Maine. Democratic and Independent candidates generally embraced the new voting rules and often encouraged their supporters to give their second-choice votes to other candidates. In contrast, Republican candidates tended to criticize RCV and encouraged their supporters to only vote for their first choice (Armstrong 2019). In an exit poll of Maine voters in the November 2018 election, almost 81 percent of Democrats wanted to extend ranked choice voting to other elections while 72 percent of Republicans wanted to eliminate RCV (Shepherd 2018; Gillespie, Levan, and Maisel 2019).3 Given the intense partisan disagreements over RCV in Maine we expect that party affiliation will structure the assessments of town clerks toward ranked choice voting. H1: Democratic municipal clerks will evaluate ranked choice voting more positively than Republican or unenrolled clerks.4 2 Katherine Gillespie, Carrie Levan, and L. Sandy Maisel, ‘‘Ranked Choice Voting in Maine: Were the Critics’ Fears Justified?’’ Unpublished manuscript, Colby College (2019). 3 See Gillespie, Levan, and Maisel 2019 in note 2. 4 Maine does not offer the option of registering as Independent, but ‘‘unenrolled’’ is the most common choice. As of May 1, 2019, the distribution of registered voters in Maine was 35.4% unenrolled, 33.0% Democratic, 27.3% Republican, and 4.1% Green. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 6 ANTHONY ET AL. A related factor is support for the goals of election reform among local election officials. Implementation theory suggests that agreement with the objectives or main features of a new law reduces resistance to implementation (Sabatier and Mazmanian 1980; Montjoy and O’Toole 1979). Agreement on the goals of election reform is an important factor in local enforcement of HAVA after the law was passed by Congress in 2002 (Palazzolo and Ceaser 2005; Moynihan and Silva 2008). Similarly, agreement on the goals of election administration shapes whether poll workers faithfully execute the rules and procedures established by local election officials (Alvarez and Hall 2006). Much of the debate in Maine focused on the need for the main objectives of ranked choice voting (e.g., whether it was important to allow votes to transfer to second choice candidates or to produce a majority winner). Thus, agreement with the goals of reform should boost support for RCV among municipal clerks in Maine. H2: Municipal clerks who support the goals of ranked choice voting are more likely to see RCV as a positive change and are more likely to support its continued use in Maine. Apart from partisanship and support for reform objectives, the burden associated with implementing a new policy is another important factor for local officials. Resources are needed to implement significant policy changes but are not always provided. Local officials tend to resist policy mandates that are unfunded or when they have sunk costs in status quo policies or technologies (Montjoy and O’Toole 1979; Sabatier and Mazmanian 1980). Concerns about the local costs of election reform, particularly some HAVA provisions, are common among local election officials and these concerns tend to foster opposition to election reforms (Hale and Slaton 2008; Moynihan and Silva 2008; Creek and Karnes 2010; Montjoy 2010). Moynihan and colleagues develop the concept of ‘‘administrative burden’’ to describe whether a new policy is experienced as being onerous (Burden et al. 2012). Administrative burden can apply to officials implementing a new law and to citizens who must comply with a new law. Administrative burden includes the costs of learning a new policy and the psychological stress of implementing a new policy while still carrying out an existing workload (Moy- nihan, Herd, and Harvey 2014). Concerns about administrative burden are widespread among local election officials and are strongly associated with attitudes toward proposed election reforms (Burden et al. 2012). In addition, perceptions of burden can be more important than objective measures of burden when officials evaluate a policy (Burden et al. 2012). In Maine, the office of the Secretary of State (SOS) is responsible for printing all ballots statewide, and the ballots are of a uniform design. Relatedly, electoral contests that go to an RCV re-tabulation are also sent to the Secretary of State for centralized counting and verification. Therefore, LEOs in Maine may have faced limited administrative burden due to the implementation of RCV. However, the Maine Town and City Clerks’ Association testified to the Maine Legislature in 2013, before RCV passed, and afterwards, in 2017. It took no position on the bills before legislators but pointed to cost and logistical concerns associated with RCV.5 Thus, local clerks concerned about resource or administrative burdens may have a less favorable view of enforcing new policies like RCV. H3: Municipal clerks concerned about the increased costs and administrative burden associated with ranked choice voting are less likely to view RCV positively and are more likely to oppose its continued use in Maine. A related local factor is the size of the jurisdiction. Smaller rural jurisdictions tend to struggle to find the resources to comply with state and federal election reforms (Creek and Karnes 2010; Hall 2014). Larger jurisdictions tend to have the staff and budget needed to adapt to new policies. In addition, officials in large jurisdictions tend to see a greater need for innovation in election administration. As a result, local officials in larger urban jurisdictions tend to embrace election reforms more than 5 Although the Secretary of State’s office testifies to the Maine Legislature more often, in recent years the Maine Town and City Clerks’ Association has weighed in on a variety of election concerns beyond ranked choice voting, including early voting, vote by mail, photo identification requirements, and online voter registration, in some cases reporting the results of a survey of clerks. Majorities backed early voting (53% for, 36% against) and photo identification (52% for, 26% against) and opposed an expansion of vote by mail (19% for, 74% against). RANKED CHOICE VOTING IN MAINE officials in less populated jurisdictions (Burden et al. 2011; Adona et al. 2019). For example, Portland, the largest city in Maine, adopted RCV for local elections several years before the statewide adoption and their voters and election officials had more familiarity and experience with the system.6 Figure 1 also indicates that public support for RCV is stronger in larger Maine municipalities. Thus we expect clerks in larger municipalities to view ranked choice voting more positively. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. H4: Municipal clerks in large municipalities will evaluate RCV more positively than clerks in small towns. Similarly, we consider the congressional district where municipal clerks serve. While Maine has two congressional districts, the use of ranked choice voting has been more contentious in the 2nd congressional district (CD2).7 All municipal clerks in Maine needed to prepare poll workers and educate voters about RCV, and all clerks participated in the re-tabulation of votes for the 2018 Democratic primary for governor. Nevertheless, in the 1st congressional district (CD1) the primary and general election winners were determined by first choice votes alone, so no extra tabulation was needed. The 2018 Democratic primary and 2018 general election in the 2nd congressional district went to additional rounds of vote tabulation to determine the winner. Thus, municipal clerks in the 2nd congressional district needed to provide all of their ballots, memory cards, and other vote tabulation devices to the Secretary of State for the additional vote count. This means that LEOs in CD2 may have experienced more of an administrative burden than LEOs in CD1, which could lead to less positive evaluations of RCV from LEOs in CD2. Additionally, CD1 contains Portland and it is a more Democratic district than CD2. Congressional district 2 is also more rural and has many more small towns than CD1. In RCV contests some clerks in rural portions of the district served as ballot collection points for other nearby towns. This required additional paperwork and coordination with the state for these municipal clerks, a frequent talking point in our interviews. As noted above, the outcome of the 2018 general election in the 2nd district, with the leader in first choice votes ultimately losing the election after the ranked choice voting tabulation, was difficult for some to accept. The legal challenges 7 to the re-tabulation and result of the 2nd district contest in 2018 added to the controversy. H5: Municipal clerks in the 2nd congressional district will experience a greater administrative burden and will subsequently evaluate RCV more negatively than clerks in the 1st congressional district. Finally, we consider the views of voters being served by local election officials. Here we use principal-agent theory to examine the relationship between the principal—the voters of Maine (who indicated their preferences for RCV in two elections)— and the agent—the municipal clerks charged with implementing the new voting rules. Principal-agent theory is often used to understand challenges of public administration in a decentralized environment. Delegation problems occur when a principal who adopts a new policy relies on agents to implement the policy. In particular, if the principal is not able to closely monitor the agent’s behavior, then the agent may be able to avoid vigorous implementation of the new policy. Nevertheless, prior research offers some evidence that constituent sentiment may influence the attitudes and behavior of local election officials, particularly when local officials are elected to their posts (Burden et al. 2013). In Maine, most municipal clerks are selected from the community they serve (and some are directly elected), so their views about election policies may reflect local opinion. In addition, two statewide votes provide clear indicators of voter opinion on ranked choice voting in each municipality. If clerks are interested in operating as delegates and representing community sentiment, then their views about RCV may be associated with the voting results in their municipality. H6: Clerks’ evaluations of RCV will be correlated with the way their municipality voted on the RCV ballot measures. 6 In 2013, two years after first using ranked choice voting for city elections, the City Clerk of Portland, on behalf of the Maine Town and City Clerks’ Association, testified ‘‘neither for or against’’ a ranked choice bill before the Maine Legislature. Ms. Jones pointed to the city’s increased resources needed to tally the vote and educate voters, as well as their need to develop rules and regulations, and expressed concern about the capacities of smaller municipalities to implement such reforms. 7 We thank James Endersby for suggesting this hypothesis. 8 ANTHONY ET AL. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. DATA AND METHODS The primary data for this study is a survey of Maine municipal clerks conducted in the spring and summer of 2019. The municipal clerk oversees polling place operations and ballot counting in Maine. A separate municipal registrar administers voter registration, although in most towns the same person does both jobs. We designed the survey to measure the responses of local election officials on (1) the goals of RCV and poll worker training, (2) the costs and administrative burden associated with RCV, (3) relationships with state election officials and relevant advocacy organizations, (4) voter education efforts and voter competence with the new voting rules, and (5) overall assessments of ranked choice voting and how it worked in the 2018 elections. Appendix A1 provides descriptive statistics and wording for survey questions used in this study. Appendix Table A1 also provides a correlation matrix (Pearson’s r) for the study’s variables. We worked with the Maine Town and City Clerks’ Association (MTCCA) to distribute the survey to the state’s roughly 480 municipal clerks. We constructed the survey in Qualtrix and sent an e-mail link to the survey through the MTCCA e-mail listserv. We sent the first link to the survey on March 13, 2019, and we sent two subsequent requests with a link to the survey in the following weeks. In the process we discovered that a small number of municipal clerks serving small towns are not on the MTCCA listserv. We mailed printed surveys to a sample of municipal clerks that did not have access to the MTCCA listserv. In total, 110 clerks completed part of the survey and 99 Table 1. Comparing the Survey Sample to all Maine Municipalities Voting measure Support for Trump for president in 2016 election Support for GOP gubernatorial candidate in 2018 election Support for RCV ballot measures (averaged across both elections) Number of ballots cast in November 2018 election Observations RCV, ranked choice voting. Sample mean Statewide mean 49.0 53.4 47.6 50.8 49.9 46.8 2412 1281 N = 96 N = 502 completed the entire survey (roughly 21 percent of the clerks in the state).8 Our response rate is a bit lower than some other surveys of local election officials (e.g., Adona et al. 2019). Some descriptive data provide a snapshot of our sample of municipal clerks. The average clerk in our sample has roughly 14 years of experience in election administration and spends roughly 39 percent of work time on election-related activities. Almost half of the clerks have a college degree. These figures are quite similar to national surveys of local election officials (Adona et al. 2019). We also gathered aggregate data on voting behavior in each municipality, including support for the RCV ballot measures. The survey asked clerks the name of the municipality they served, which allowed us to merge municipal-level voting data with the survey data. The aggregate voting data allow us to compare the municipalities in our survey sample to all municipalities in Maine (in those instances where the clerk identified the municipality). The results, summarized in Table 1, indicate that the towns in our sample are more liberal (by about four points on each measure) than the average town in Maine.9 In addition, the average municipality in our sample is somewhat larger than the average town in the state. The mismatch may be due to the coverage error in the survey, described above. When we asked clerks in the survey to indicate their party affiliation, 41 percent are Republican, 38 percent are unenrolled, and 21 percent are Democrats. Municipal clerks are appointed or elected in nonpartisan elections and there is no publicly available data source on the party affiliation of all clerks in the state. Thus, we don’t know how representative our sample is on that indicator. For each comparison in Table 1 the sample mean is within half a standard deviation of the population mean, so we believe our sample is representative of the municipal clerks in Maine. 8 According to the directory provided by the state, as of 2018 there were 22 municipal clerks serving multiple towns, almost all very small towns. Thus, the sampling frame for our survey is approximately 480 clerks. 9 The statewide mean (across all municipalities) is several points to the right of the statewide vote percentages because the larger municipalities tend to be more Democratic and pro-ranked choice voting (RCV) than the average municipality. RANKED CHOICE VOTING IN MAINE 9 Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. Finally, we also gathered qualitative data through face-to-face and telephone interviews with 14 municipal clerks and two state election officials in Maine during the fall of 2019. These local election officials represent rural, mid-sized, and urban jurisdictions across all regions of the state. The officials interviewed represent both major political parties as well as the ‘‘unenrolled’’ affiliation, and these local officials also vary in their support of the change to RCV rules in Maine. The interviews allowed us to follow up on some of the issues covered in the survey and get a more in-depth understanding of the local experience with ranked choice voting. The names and jurisdictions of these local election officials have been kept anonymous throughout this article. 4.0 2.8* 2.1* 3.2 2.5* 2.0* segmented by the party affiliation of municipal clerks. [The overall mean score is roughly the same as the unenrolled mean score for each item.] Overall, Maine municipal clerks do not express strong support for the objectives of RCV. The mean score on each question is between 2 (minor disadvantage) and 3 (neither/neutral) on the fivepoint scale. Thus, clerks are more likely to rate the main features of RCV as disadvantages rather than advantages. Relatively speaking, allowing voters to indicate a preference for more than one candidate has the highest mean score and thus seems to be the most popular objective of RCV. The overall results, however, mask substantial partisan divisions. On each question, the mean for Democratic clerks is at least one point higher than the mean for GOP clerks. Democratic clerks tend to view each feature of RCV as more of an advantage than a disadvantage. The gap between the two parties is largest on two central features of RCV (allowing voters to indicate preferences for multiple candidates, and re-tabulating votes when there is no majority winner in the first round) that played a prominent role in the outcome of the 2nd congressional district race. Thus, Democratic clerks support the goals of RCV significantly more than Republican and unenrolled clerks. What do clerks see as the cost impact of ranked choice voting? We measured perceived administrative burden in two ways. In the first measure, the survey asked whether the implementation of ranked choice voting increased their election administration costs in three areas: poll worker training, Election Day staffing, and voter education. Clerks could choose from one of three responses (1 = no change, 2 = minor increase, 3 = significant increase). Table 3 shows the mean responses to each item for separate partisan groups. Municipal clerks generally don’t see major new costs associated with ranked choice voting. The 3.7 2.4* 1.8* Table 3. RCV Costs by Party Affiliation 3.3 2.7 2.2* 3.6 2.3* 2.0* RESULTS We start by examining support for the main goals of ranked choice voting. Following a strategy used by Moynihan and Silva (2008) the survey asked clerks to rate different features of ranked choice voting as advantages or disadvantages on a five-point scale, where 1 equals ‘‘strong disadvantage’’ and 5 equals ‘‘strong advantage.’’ These features include allowing voters to mark a preference for multiple candidates, re-tabulating ballots if no candidate has a majority, and reducing the ‘‘spoiler effect’’ of minor party candidates. Table 2 shows the mean responses to these questions, with the sample Table 2. Support for RCV Goals by Party Affiliation RCV feature Voters can show a preference for more than one candidate Candidates can urge voters to support other candidates Ballots are re-tallied if no candidate has a majority Ballots are sent to SOS if more rounds of counting are needed Reduces the ‘‘spoiler effect’’ if more candidates run Democratic Unenrolled Republican mean mean mean 1 = strong disadvantage, 5 = strong advantage. *Statistically different from Democratic mean (p < .05). RCV, ranked choice voting; SOS, Secretary of State. RCV cost impact Poll worker training Election Day staffing Voter outreach and education Democratic Unenrolled Republican mean mean mean 1.6 1.7 1.7 1.7 1.7 1.6 1 = no change, 2 = minor increase, 3 = significant increase. *statistically different from Democratic mean (p < .05). RCV, ranked choice voting. 1.7 1.9 1.6 Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 10 mean response for each item is between 1 (no change) and 2 (minor increase), so the clerks tend to report a minor impact on their local election costs due to RCV. We gave clerks the opportunity to report other budget impacts of RCV. Only seven availed themselves of that opportunity, mainly to mention additional equipment needs such as tabulators and thumb drives to count and transmit RCV results. Furthermore, there are no party differences in assessments of the specific cost impacts of RCV. Contrary to H5, clerks in the 2nd congressional district did not report higher RCV costs than clerks in the 1st district. We get a somewhat different picture from the survey when we ask whether ranked choice voting increases the ‘‘administrative burden’’ on local election officials. Almost two-thirds of the clerks in our sample agreed that RCV increased their administrative burden. This question also revealed partisan differences, as 83 percent of Republicans agreed that RCV increased their administrative burden as compared to 50 percent of Democratic clerks. We observe no difference in perceived administrative burden when comparing clerks from the two congressional districts. Municipal clerks in the 2nd district do not report higher costs or administrative burdens in implementing RCV. Interviews with election officials illuminated the nature and extent of the local administrative burden associated with RCV rules. Overall, the instances of increased burden imposed by RCV seemed manageable and were not onerous. Eight municipal clerks commented on how the tallying process for RCV elections delayed their ability to wrap up their local certification and election processes. As one clerk in a rural town stated, RCV elections ‘‘[make] the jobs of small-town clerks harder, because we are not able to ‘tidy up’ the elections until the [statewide] RCV count is done.’’ The clerk stated that this is important because small town clerks generally have many duties in addition to running elections, such as licensing, revenue collection, and other civic obligations. Therefore, the longer process of RCV vote tabulation cuts into the time local officials spend on other job responsibilities (personal communication: October 22, 2019). Most clerks interviewed stated that there were little to no changes in how elections were run on Election Day itself after RCV rules were implemented, though five clerks did note some changes. Of ANTHONY ET AL. those who noticed increased responsibility due to RCV rules, most noted extra time needed to program and test voting equipment before Election Day to ensure that the ranking and tabulation systems worked properly. For towns holding local elections on the same day as federal elections, Maine uses a mixed ballot design—where statewide and federal races are on an RCV ballot—while municipal and state legislative races are on a separate ballot. This difference in ballot design seemed to be the primary component increasing the administrative burden on some local officials (the design also may have created some voter confusion). The mixed ballot design meant that some clerks had to use two different scanning machines to count the different ballot styles. Other clerks noted that differences between plurality and RCV ballots meant that some contests had to be tallied by hand, while other contests were sent through electronic tallying machines (personal communications: October 21 and 22, 2019). These differing tabulation processes increased the time spent by local officials on counting votes. Only two of the local officials interviewed stated that their administrative burden in regard to costs increased due to RCV implementation. One clerk noted that her jurisdiction had to purchase additional flash drives to archive and send the ballot results to the Secretary of State’s office for counting, because the ranking results took up a great deal of space on these flash drives. Another clerk stated that the switch to RCV meant that the municipality had to pay for two additional poll workers and secure longer hours for the other poll workers, resulting in higher costs for the jurisdiction. While these instances of increased administrative burden are important to note, it appears that local election officials overall were largely not over-burdened in their new responsibilities after the shift to RCV rules. A common concern with ranked choice voting is whether voters properly understand and comply with the new voting rules (Neely and Cook 2008; Neely and McDaniel 2015; Donovan, Tolbert, and Gracey 2019). We also asked clerks to assess how voters performed with the new ranked choice voting rules. On a five-point scale (1 = strongly disagree, 5 = strongly agree) the clerks where asked whether voters in their municipality (1) clearly understood RCV rules, and (2) properly ranked candidates in RCV contests. The mean responses are reported in Figure 2. 11 Mean Rating of Voter Performance 2 3 4 5 The survey results indicate that Maine municipal clerks take a dim view of the voter’s ability to understand ranked choice voting rules. The mean score on each question is near 2 (disagree somewhat) on the five-point scale. More than two-thirds of the clerks in our survey do not believe that voters understood RCV rules. These results contrast sharply with surveys, including an exit poll of Maine voters in 2018, indicating that most voters find RCV rules easy to understand (Donovan, Tolbert, and Gracey 2019; Gillespie, Levan, and Maisel 2019).10 Moreover, the ballot image data released by the state show that in the 2nd congressional district race two-thirds of Poliquin supporters did not rank any other candidates (as Poliquin instructed) and two-thirds of the remaining voters did rank more than one candidate (as their preferred candidates instructed). The tendency of voters to follow instructions from party leaders or preferred candidates on how to complete a ranked ballot is consistent with other nations using ranked choice voting (Reilly 2019). Another measure of voter confusion is the number of ‘‘blank’’ ballots cast in a contest, due to overvotes or undervotes. In the 2018 general election, the controversial CD2 contest, which used RCV rules, had a lower rate of blank ballots (2.2%) than the competitive governor’s race (2.4%), which used plurality rules. The 2018 U.S. Senate race, which also used RCV rules, produced an even lower rate of blank ballots (1.8%). These rates are similar to the frequency of rejected ballots in other countries using ranked choice voting (Clark 2013: 61–62). The frequency of blank ballots in the 2018 Maine RCV contests is also lower than those observed in the same contests in prior years using 3.2 2.7 2.1 1.9 1.9 1.6 1 Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. RANKED CHOICE VOTING IN MAINE Democrat Understood RCV Unenrolled Republican Properly ranked candidates FIG. 2. Ratings of voter performance with RCV by party affiliation. plurality rules. Turning back to our survey, the frequency of blank ballots in the RCV contests at the municipal level was unrelated to the clerk’s assessments of voter understanding of RCV rules. Curiously, clerks were a bit more optimistic about voters’ ability to properly rank candidates than of voters’ understanding of the rules. Still, more than half of the clerks responding to the survey did not think their voters properly ranked RCV candidates. In addition, almost 85 percent of the clerks see the complicated nature of the ballot as a disadvantage of ranked choice voting. Once again, there are strong partisan differences as Democratic clerks offer substantially more positive assessments of their voters than Republican and unenrolled clerks. On each measure of voter performance the mean rating from Democratic clerks is significantly higher than the mean rating offered by Republican and unenrolled clerks (p < .05). In interviews, all of the local election officials stated that they received more questions from voters than in previous, non-RCV elections. Several election officials clarified that it was not necessarily the new RCV rules that most confused voters. Instead, voters were confused by the mixed design that included both plurality and RCV contests on the same ballot. One clerk noted that the mixed ballot design was particularly confusing to older voters in her jurisdiction (personal communication: October 22, 2019). A local election official in another rural jurisdiction remarked that it would be better for voters if the ‘‘entire ballot were comprised of RCV contests’’—or have them all be plurality contests (personal communication: October 22, 2019). Six of the local officials interviewed said that they believed most voters either did not properly rank their ballots, and/or voters only ranked one candidate in RCV contests. Two local officials thought that only around one-quarter of their voters properly and fully ranked their ballots. One election official, however, thought that the proportion of spoiled ballots with RCV was equivalent to the proportion in previous elections using only plurality rules (personal communication: October 21, 2019). Two election officials noted that RCV contests made for longer ballots, and as such several voters in each of their jurisdictions did not see or vote the back sides of their ballots. Local election officials 10 See Gillespie, Levan, and Maisel 2019 in note 2. Mean Rating of RCV 3 4 5 primary tool used by local election officials in informing the public about the new RCV rules and procedures. Our interviews revealed important differences in how some local election officials used outside materials as voter education guides, namely, the RCV instruction guide developed and distributed by the Maine League of Women Voters (LWV). Two local election officials reported that they were directed by the office of the SOS to not distribute the LWV guides to voters, as they arguably took a less-than-neutral stance on the question of RCV adoption. Several other local election officials, however, stated that they did provide these guides to voters; one local official even mailed the guide to voters in advance of Election Day. These interviews suggest that voters in different jurisdictions may have received different and varying amounts of information about RCV before and on Election Day. Our final set of survey questions examines overall evaluations of ranked choice voting in Maine. On a five-point scale (1 = strongly disagree, 5 = strongly agree) the clerks were asked to respond to two statements: (1) the process for counting votes in RCV contests worked effectively, and (2) ranked choice voting is improving the election process in Maine. The mean responses for these two items are reported by party affiliation in Figure 3. Previous measures indicated rather low marks for RCV and strong partisan differences. The same pattern holds here. Democratic clerks hold much more positive assessments of RCV than Republican and unenrolled clerks. Among survey respondents not a single GOP clerk, and only two unenrolled clerks, agree that RCV is improving 3.7 3.2 3.2 2 2 in larger jurisdictions also noted that the new RCV rules meant that it took some people longer to vote. One clerk also remarked on the length and the complexity of the ballot by saying, ‘‘[v]oters need[ed] a ruler to find the correct candidate on the RCV ballot’’ (personal communication: October 21, 2019). One clerk stated that while she did not receive a lot of questions about RCV on Election Day itself, she overheard and was part of several conversations afterward that indicated some voters may not have fully understood the ranking system. For instance, the clerk said that she talked with a few people in the town after Election Day who had ranked the same candidate more than once, had marked a ‘‘1’’ beside all candidates, or who had otherwise spoiled their second and third choice votes. A clerk in a different rural jurisdiction also noted that voters were confused about RCV on Election Day, asking her if they ‘‘had to vote for more than one’’ candidate (personal communication: October 22, 2019). The beliefs of municipal clerks about voter confusion had important consequences: two local officials opposed to RCV advised voters that they did not have to rank their ballots and that they should vote for only one candidate in RCV contests. Because the Maine Secretary of State’s office handled the tallying of RCV races, local election officials largely noted that poll workers (aka ‘‘election clerks’’) did not require a great deal more training for Election Day operations compared with previous plurality elections. Two local election officials did say that they had a slightly longer training session for poll workers on the morning of Election Day; these new training components largely focused on how poll workers should answer voters’ questions about RCV. One clerk also provided a one-page guide on RCV to voters on Election Day. In rural jurisdictions, where there was only one polling place, local election officials were almost always on hand as well to directly answer voters’ questions. In regard to voter outreach and education, the SOS also provided educational materials and webinars to inform the public about the new RCV rules. Local election officials, therefore, largely did not do ‘‘anything extra’’ to communicate with voters about RCV (personal communication: October 21, 2019). Most of the local officials interviewed who have an internet presence shared an animated video created by the Maine SOS on their Facebook pages and websites. This video seemed to be the ANTHONY ET AL. 1.7 1.3 1 Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 12 Democrat Unenrolled Counting worked effectively FIG. 3. Republican Improves election process Overall assessments of RCV by party affiliation. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. RANKED CHOICE VOTING IN MAINE 13 the election process in Maine. Even Democratic clerks were almost evenly split on that question, with almost equal numbers agreeing and disagreeing that RCV is improving elections. Responses are more positive to the more mundane question about counting votes in RCV contests. Strong partisan differences remain, however, as Democratic and unenrolled clerks assessed RCV ballot counting more favorably than Republican clerks. The survey also asked clerks whether Maine should keep RCV for future state and federal elections. Overall, 86 percent of respondents do not want to continue using RCV in Maine. None of the GOP clerks, and only four unenrolled clerks, want RCV to continue, while a majority of Democratic clerks want to keep RCV. Thus, while Republican and Democratic clerks offer similar assessments of the costs associated with implementing RCV, they differ substantially in overall evaluations and support of RCV. The strong party differences in assessments of ranked choice voting clearly support our first hypothesis and reflect the intense and longstanding partisan conflict over ranked choice voting in Maine. To test our remaining hypotheses we estimate regression functions with overall assessments of ranked choice, described in Figure 3 and the paragraphs above, as our dependent variables. The evidence above shows that partisanship is a predisposition that shapes opinions toward many aspects of ranked choice voting. In addition, GOP affiliation perfectly predicts opposition to ranked choice voting on two of our three outcome measures. To test our remaining hypotheses, we exclude partisanship from these analyses. We model overall assessments of RCV as a function of support for RCV goals, concerns about the impact of RCV on local costs, jurisdiction size, congressional district, and constituent support for RCV. We measure support for RCV goals with a scale constructed by averaging the five items reported in Table 2 (Cronbach’s a = .84, see question wording in the Appendix A1). To measure concerns about local costs we create a scale by averaging responses to the three questions summarized in Table 3 (a = .78). We measure jurisdiction size as the natural log of the number of ballots cast in the municipality in the November 2018 election. A dummy variable denotes clerks serving municipalities located in the 2nd congressional district. We measure constituent support as the average percent voting yes in the municipality served by the clerk on the two statewide RCV ballot measures. The results are presented in Table 4. Model 1 examines the process for counting RCV votes. Model 2 examines whether RCV is improving elections. Model 3 examines whether clerks want to keep RCV for future state and federal elections in Maine. The main predictor of attitudes toward ranked choice voting, in addition to partisanship, is support for the goals of RCV. Increased support for the goals of ranked choice voting is strongly associated with more positive assessments of the election reform. Moving from low support (10th percentile) to high support (90th percentile) of RCV goals increases the expected probability of agreeing that RCV is improving the election process in Maine by 31 percentage points. The same increase in support for the goals of RCV increases the predicted probability of wanting to keep RCV for future elections by 45 percentage points. These results provide clear support for our second hypothesis. Table 4. Predictors of Assessments of Ranked Choice Voting Independent variable Support for RCV goals RCV impact on local election costs Jurisdiction size (natural log of ballots cast in 2018 election) Voter support for RCV ballot measures (mean percent Yes) In 2nd congressional district N Pseudo-R2 Model 1: Process for counting RCV votes Model 2: RCV is improving elections Model 3: Keep RCV for future elections 1.30* (0.22) -0.66* (0.32) -0.01 (0.17) 2.27* (0.41) -0.08 (0.43) -0.40* (0.23) 2.50* (0.73) 0.49 (0.81) -0.91* (0.44) 0.03 (0.03) 0.03 (0.04) 0.08 (0.07) 0.19 (0.49) 90 .18 -1.10* (0.64) 91 .33 -2.62* (1.09) 89 .53 Models 1 and 2: Cell entries are ordinal logit coefficients (standard errors in parentheses). The dependent variable is a 5-point scale (1 = strongly disagree, 5 = strongly agree). Model 3: Cell entries are logit coefficients (standard errors in parentheses). The dependent variable is binary (1 = yes, 0 = no). *p < .05 (one-tailed). RCV, ranked choice voting. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 14 ANTHONY ET AL. In two of three tests we also find support for our fifth hypothesis. Municipal clerks in the 2nd congressional district offered more negative assessments of ranked choice voting. According to our estimates in Model 2, clerks serving in the 2nd congressional district were more likely than clerks in the 1st district to disagree with the statement that RCV was improving elections, by roughly 12 percentage points. According to Model 3, clerks in the 2nd district are less likely to support keeping RCV for future elections, by about 18 percentage points. We find very little support for the remaining hypotheses. Concerns about the impact of RCV on local election costs are modestly associated with opposition to ranked choice voting in Model 1, but not in the other two models.11 Contrary to our expectations, jurisdiction size is negatively associated with evaluations of ranked choice voting after controlling for other factors. Finally, constituent opinions on RCV are unrelated to the overall views of Maine municipal clerks toward the election reform. It is noteworthy that the party affiliation of municipal clerks is a strong predictor of support for ranked choice voting, but perceived costs and administrative burdens are at best modestly associated with overall clerk assessments of RCV. Furthermore, the evaluations of ranked choice voting offered by municipal clerks do not reflect the preferences of their voters. CONCLUSIONS This study examines the views of local election officials on the implementation of ranked choice voting in Maine. We find that Maine municipal clerks tend to dislike ranked choice voting overall. The clerks express low levels of support for the main goals of RCV and they tend to believe that voters do not understand the new voting rules. Even though they report only minor increases in local election costs, most Maine clerks do not believe that RCV has improved the election process and a large majority do not want to continue its use in Maine state and federal elections. We also find that partisanship does more than administrative headaches to shape the views of municipal clerks toward RCV. The main predictors of clerk attitudes toward this election reform are partisanship and support for the goals of RCV. Municipal clerks who identify as Democrats offer much more positive appraisals of ranked choice voting than Republican clerks. Unenrolled (Independent) municipal clerks typically fall in between Democratic and Republican clerks in their assessments of RCV in Maine. GOP clerks in our sample are unanimous in their opposition to the continued use of RCV in Maine. Democratic clerks are more varied in their evaluations, but they tend toward positive assessments of ranked choice voting. The partisan divisions among municipal clerks mimic similar divisions among the voters and the political class in Maine. However, in contrast to the strong partisan differences, the views of their constituents, as measured in two statewide ballot measures, are not reflected in the opinions of Maine municipal clerks toward ranked choice voting. The negative evaluations offered by municipal clerks differ substantially from the relatively positive views Maine voters expressed in a recent exit poll (Shepherd 2018; Gillespie, Levan, and Maisel 2019)12 and in voting twice to adopt the reform. In addition, clerk assessments are unrelated to the preferences of their constituents or measures of voter confusion based on observational data from recent Maine elections. Perhaps municipal clerks are observing problems with ranked choice voting that are not as evident to voters and reform supporters. On the other hand, clerk assessments of ranked choice voting seem to be driven by partisanship and opposition to the goals of RCV. Successful election reform depends, in part, on supportive implementation by local administrators. The partisan differences we observe are a concern because local election officials can use their discretion to influence voting behavior in ways that promote their own partisan preferences (e.g., Burden et al. 2013). An important part of implementing ranked choice voting is voter education—voters need to understand and comply with the new rules, and local election officials are a primary source of voter education. When it comes to voter education local election officials may discriminate against 11 When we replace the cost variable with the administrative burden measure, the estimated impact of administrative burden is modest, negative, and statistically significant in two of the three models. When we include controls for election administration experience and familiarity with RCV those two variables do not produce statistically significant coefficients and do not change the other results. For the sake of simplicity we leave those two variables out of the results reported in Table 4. 12 See Gillespie, Levan, and Maisel 2019 in note 2. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. RANKED CHOICE VOTING IN MAINE minority groups by withholding information or providing inaccurate information (White, Nathan, and Faller 2015). Furthermore, vigorous and accurate voter education efforts can increase voter turnout (Bright and Lynch 2017). RCV was backed by Maine voters in two referenda but municipal clerks are generally not enthusiastic about implementing RCV and seem to vary in terms of their voter education efforts. As noted above, in our interviews two clerks opposed to RCV reported that they instructed their constituents to vote for only one candidate in the RCV contests. The share of voters who ranked only one candidate in the 2nd congressional district contest in 2018 was higher in those two towns (57.8%) than in the rest of the district (49.9%). Rates of ballot exhaustion in the 2nd district contest were also higher in those two towns (6.2%) than in the rest of the district (4.8%). This evidence is suggestive, not causal, but more investigation is warranted to examine whether administrative decisions by municipal clerks influence voting behavior. Although the results are weaker, we also find that concerns about costs and administrative burdens are associated with opposition to RCV among Maine municipal clerks. The concerns regarding logistics and resources that the Maine Town and City Clerks’ Association raised in testimony to the Maine Legislature regarding RCV were reasonable and within their purview. The Maine Secretary of State’s office took on the major administrative tasks of gathering and tallying the ballots and prepared materials to educate voters. As more cities and states consider adopting RCV rules, it is important to enhance and expand voter education and understand how election reforms are implemented by state and local election officials and ensure they have adequate support. REFERENCES Adona, Natalie, and Paul Gronke. 2018. Understanding the Voter Experience. October. Washington, DC: Democracy Fund. Adona, Natalie, Paul Gronke, Paul Manson, and Sarah Cole. 2019. Stewards of Democracy: The Views of American Local Election Officials. June. Washington, DC: Democracy Fund. Alvarez, R. Michael, and Thad E. Hall. 2006. ‘‘Controlling Democracy: The Principal-Agent Problems in Election Administration.’’ Policy Studies Journal 34: 491–510. Ansolabehere, Stephen, and Charles Stewart III. 2005. ‘‘Residual Votes Attributable to Technology.’’ Journal of Politics 67: 365–389. 15 Armstrong, Katherine J. 2019. ‘‘Ranked Choice Voting in Maine.’’ Report commissioned by the William and Flora Hewlett Foundation. Available at <https://hewlett.org/wpcontent/uploads/2019/09/RCV-in-Maine-final-for-webposting.pdf>. Atkeson, Lonna Rae. 2011. ‘‘Voter Confidence in 2010: Local, State, and National Factors.’’ In Election Administration in the United States, eds. R. Michael Alvarez and Bernard Grofman. New York, NY: Cambridge University Press. Bowler, Shaun, and Bernard Grofman, eds. 2000. Elections in Australia, Ireland, and Malta under the Single Transferable Vote. Ann Arbor, MI: University of Michigan Press. Bright, Chelsie L. M., and Michael S. Lynch. 2017. ‘‘Kansas Voter ID Laws: Advertising and Its Effect on Turnout.’’ Political Research Quarterly 70: 340–347. Burden, Barry C., David T. Canon, Stephane Lavertu, Kenneth R. Mayer, and Donald P. Moynihan. 2013. ‘‘Selection Method, Partisanship, and the Administration of Elections.’’ American Politics Research 41: 903–936. Burden, Barry C., David T. Canon, Kenneth R. Mayer, and Donald P. Moynihan. 2011. ‘‘Early Voting and Election Day Registration in the Trenches: Local Officials’ Perceptions of Election Reform.’’ Election Law Journal 10: 89– 102. Burden, Barry C., David T. Canon, Kenneth R. Mayer, and Donald P. Moynihan. 2012. ‘‘The Effect of Administrative Burden on Bureaucratic Perception of Policies: Evidence from Election Administration.’’ Public Administration Review 72: 741–751. Clark, Alistair. 2013. ‘‘Second Time Lucky? The Continuing Adaptation of Voters and Parties to the Single Transferable Vote in Scotland.’’ Representation 49: 55–68. Creek, Heather M., and Kimberly A. Karnes. 2010. ‘‘Federalism and Election Law: Implementation Issues in Rural America.’’ Publius 40: 275–295. Crowley, Brendan. 2020. ‘‘Bill Seeks to ‘Close’ Missouri Primaries.’’ Jefferson City News Tribune, February 20. <https://www.newstribune.com/news/local/story/2020/feb/ 20/bill-seeks-to-close-missouri-primaries/817498/>. Donovan, Todd, Caroline Tolbert, and Kellen Gracey. 2016. ‘‘Campaign Civility Under Preferential and Plurality Voting.’’ Electoral Studies 42: 157–163. Donovan, Todd, Caroline Tolbert, and Kellen Gracey. 2019. ‘‘Self-Reported Understanding of Ranked-Choice Voting.’’ Social Science Quarterly 100: 1768–1776. Endersby, James W., and Michael J. Towle. 2014. ‘‘Making Wasted Votes Count: Turnout, Transfers, and Preferential Voting in Practice.’’ Electoral Studies 33:144–152. Ewald, Alec C. 2009. The Way We Vote: The Local Dimension of American Suffrage. Nashville: Vanderbilt University Press. Farrell, David M., and Ian McAllister. 2006. The Australian Electoral System: Origins, Variations and Consequences. Sydney, AU: UNSW Press. Farrell, David M., and Richard Sinnott. 2017. The Electoral System.’’ In Politics in the Republic of Ireland, eds. John Coakley and Michael Gallagher. London, UK: Routledge. Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. 16 Farris, Emily M., and Mirya R. Holman. 2015. ‘‘Public Officials and a ‘Private’ Matter: Attitudes and Policies in the County Sheriff Office Regarding Violence Against Women.’’ Social Science Quarterly 96: 1117–1135. Fullmer, Elliott B. 2015. ‘‘Early Voting: Do More Sites Lead to Higher Turnout?’’ Election Law Journal 14: 81–96. Gratz, Irwin, Steve Mistler, and Mal Leary. 2018. ‘‘What Driving Poliquin’s Ranked-Choice Vote Challenge and Recount Request.’’ Maine Public Radio, December 3. <https:// www.mainepublic.org/post/nov-30-whats-driving-poliquinsranked-choice-vote-challenge-and-recount-request>. Hale, Kathleen, and Christa Daryl Slaton. 2008. ‘‘Building Capacity in Election Administration: Local Responses to Complexity and Interdependence.’’ Public Administration Review 68: 839–849. Hall, Thad E. 2014. ‘‘Voting from Abroad: Evaluating UOCAVA Voting.’’ In The Measure of American Elections, eds. Barry C. Burden and Charles Stewart III. New York, NY: Cambridge University Press. Hamilton, James T., and Helen F. Ladd. 1996. ‘‘Biased Ballots? The Impact of Ballot Structure on North Carolina Elections in 1992.’’ Public Choice 87: 259–280. Hasen, Richard L. 2012. The Voting Wars: From Florida 2000 to the Next Election Meltdown. New Haven, CT: Yale University Press. John, Sarah E., Haley Smith, and Elizabeth Zack. 2018. ‘‘The Alternative Vote: Do Changes in Single-Member Voting Systems Affect Descriptive Representation of Women and Minorities?’’ Electoral Studies 54: 90– 102. Kropf, Martha. 2005. ‘‘Dogs and Dead People: Incremental Election Reform in Missouri.’’ In Election Reform, eds. Daniel J. Palazzolo and James W. Ceaser. New York, NY: Lexington Books. Kropf, Martha, Tim Vercellotti, and David C. Kimball. 2013. ‘‘Representative Bureaucracy and Partisanship: The Implementation of Election Law.’’ Public Administration Review 73(2) (March/April): 242–252. Krutz, Glen S. 2005. ‘‘Arizona: Concerted Effort, Gridlock, and the Breakthrough.’’ In Election Reform, eds. Daniel J. Palazzolo and James W. Ceaser. New York, NY: Lexington Books. Lipsky, Michael. 1980. Street-Level Bureaucracy. New York, NY: Russell Sage Foundation. Lochner, Todd, and Gary Moncrief. 2005. ‘‘Idaho: Election Reform at the Margins.’’ In Election Reform, eds. Daniel J. Palazzolo and James W. Ceaser. New York, NY: Lexington Books. Lund, William A. 1998. ‘‘What’s in a Name? The Battle over Ballot Titles in Oregon.’’ Willamette Law Review 34: 143–167. Lundberg, Thomas C. 2018. ‘‘Electoral Systems in Context: United Kingdom.’’ In The Oxford Handbook of Electoral Systems, eds. Erik S. Herron, Robert J. Pekkanen, and Mathew S. Shugart. Oxford, UK: Oxford University Press. Maddox, H. W. Jerome. 2005. ‘‘Pennsylvania: New Policies, Old Politics.’’ In Election Reform, eds. Daniel J. Palazzolo and James W. Ceaser. New York, NY: Lexington Books. ANTHONY ET AL. Mauter, Erica L. 2014. ‘‘Ranked Choice Voting in Minneapolis 2013 Elections.’’ Master of Arts in Organizational Leadership Theses. Paper 21. St. Catherine University. McBrayer, Markie, R. Lucas Williams, and Andrea Eckelman. 2020. ‘‘Local Officials as Partisan Operatives: The Effect of County Officials on Early Voting Administration.’’ Social Science Quarterly 101: 1475–1488. Meier, Kenneth J. 1993. ‘‘Representative Bureaucracy: A Theoretical and Empirical Exposition.’’ In Research in Public Administration, vol. 2, ed. James L. Perry, 1–35. Greenwich, CT: JAI Press. Montjoy, Robert S. 2008. ‘‘The Public Administration of Elections.’’ Public Administration Review 68: 788–799. Montjoy, Robert S. 2010. ‘‘The Changing Nature . and Costs . of Election Administration.’’ Public Administration Review 70: 867–875. Montjoy, Robert S., and Laurence J. O’Toole. 1979. ‘‘Toward a Theory of Policy Implementation: An Organizational Perspective.’’ Public Administration Review 38: 465–476. Moynihan, Donald P., and Carol L. Silva. 2008. ‘‘The Administrators of Democracy: A Research Note on Local Election Officials.’’ Public Administration Review 68: 816–827. Moynihan, Donald, Pamela Herd, and Hope Harvey. 2014. ‘‘Administrative Burden: Learning, Psychological, and Compliance Costs in Citizen-State Interactions.’’ Journal of Public Administration Research and Theory 25: 43–69. Neely, Francis, and Corey Cook. 2008. ‘‘Whose Votes Count? Undervotes, Overvotes, and Ranking in San Francisco’s Instant-Runoff Election.’’ American Politics Research 36: 530–554. Neely, Francis, and Jason McDaniel. 2015. ‘‘Overvoting and the Equality of Voice Under Instant Runoff Voting in San Francisco.’’ California Journal of Politics and Policy 7: 1–32. Palazzolo, Daniel J., and James W. Ceaser. 2005. Election Reform: Politics and Policy. New York, NY: Lexington Books. Presidential Commission on Election Administration (PCEA). 2014. The Voting Experience: Report and Recommendations. Washington, DC. Reilly, Benjamin. 2001. Democracy in Divided Societies: Electoral Engineering for Conflict Management. Cambridge, UK: Cambridge University Press. Reilly, Benjamin. 2019. ‘‘A Tale of Two Continents: Ranked Choice Voting in Australia and America.’’ Paper presented at the annual meeting of the American Political Science Association, Washington, DC. Sabatier, Paul, and Daniel Mazmanian. 1980. ‘‘The Implementation of Public Policy: A Framework of Analysis.’’ Policy Studies Journal 8: 538–560. Santucci, Jack. 2018. ‘‘Maine Ranked-Choice Voting as a Case of Electoral-System Change.’’ Representation 54: 297–311. Shepherd, Michael. 2018. ‘‘What Exit Polling Reveals About Maine’s Experience with Ranked-Choice Voting.’’ Bangor Daily News, November 12. <https://bangordailynews.com/ 2018/11/12/politics/exit-polling-finds-narrow-majority-ofmainers-back-expansion-of-ranked-choice-voting%E2%80 %8B/>. Smith, Mitch, and Michael Wines. 2020. ‘‘In Closely Divided Wisconsin, the Battle for Votes Is Already Underway.’’ New York Times. January 14. <https://www.nytimes.com/ 2020/01/14/us/wisconsin-voter-purge.html>. RANKED CHOICE VOTING IN MAINE Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. Stuart, Guy. 2004. ‘‘Databases, Felons, and Voting: Bias and Partisanship of the Florida Felons List in the 2000 Elections.’’ Political Science Quarterly 119: 453– 475. Thistle, Scott. 2018. ‘‘LePage Takes Parting Shot at ‘Stolen Election’ That Cost Fellow Republican House Seat.’’ Portland Press Herald, December 28. <https://www.pressherald .com/2018/12/28/lepage-leaves-note-of-protest-on-goldenelection-certificate/>. Tokaji, Daniel P. 2007. ‘‘The Birth and Rebirth of Election Administration.’’ Election Law Journal 6: 118–131. White, Ariel R., Noah L. Nathan, and Julie K. Faller. 2015. ‘‘What Do I Need to Vote? Bureaucratic Discretion and Discrimination by Local Election Officials.’’ American Political Science Review 109: 129–142. 17 Address correspondence to: Joseph Anthony Department of Political Science 7080 Haley Center Auburn University Auburn, AL 36849 USA E-mail: jka0031@auburn.edu Received for publication May 25, 2020; received in revised form October 28, 2020; accepted November 1, 2020; published online February 8, 2021. Appendix A1. Survey Questions and Descriptive Statistics Dependent Variables: Assessment of RCV: How much do you agree or disagree with the following statements: (1 = strongly disagree; 5 = strongly agree): The process for counting votes in RCV contests worked effectively; Ranked choice voting is improving the election process in Maine. dicate a preference for more than one candidate for the same office; Candidates can urge supporters to make another candidate their second choice; Ballots are re-tallied if no candidate has a majority of votes; Ballots are sent to the Secretary of State if further rounds of counting are needed; Reduces the ‘‘spoiler effect’’ where new candidates may draw votes away from other candidates. Cronbach’s a = .84, Mean = 2.53, SD = 1.16 Item 1: Mean = 2.8, SD = 1.4 RCV Costs: Item 2: Mean = 1.8, SD = 1.3 Should Maine keep RCV for future state and federal elections? (1 = yes; 0 = no) Did the implementation of ranked choice voting increase your election administration costs in any of these areas? (1 = no change; 2 = minor increase; 3 = significant increase): Poll worker training; Election Day staffing; Voter education and outreach. Mean = .14 Cronbach’s a = .78, Mean = 1.71, SD = 0.65 Independent Variables: Administrative Burden: Support for RCV Goals: How much do you agree or disagree with the following statement: RCV increases the administrative burden on election officials like me: (1 = strongly disagree; 5 = strongly agree). Continue RCV: What do you regard as the advantages and disadvantages of ranked choice voting? Mark your most preferred response for each characteristic: (1 = strong disadvantage; 5 = strong advantage): Voters can in- Mean = 3.86, SD = 1.30 (Appendix continues/) 18 ANTHONY ET AL. Voter Performance: Mean = 14.2, SD = 10.9 How much do you agree or disagree with the following statement: (1 = strongly disagree; 5 = strongly agree): Education: Voters in my municipality clearly understood ranked choice voting rules Mean = 1.93, SD = 1.24 Voters in my municipality properly ranked candidates in RCV contests What is the highest level of education you have completed? (1 = completed some high school; 2 = high school graduate or equivalent; 3 = completed some college, but no degree; 4 = college graduate; 5 = completed some graduate school, but no degree; 6 = completed graduate school) Mean = 3.5, SD = 1.0 Party Identification: Downloaded by 18.234.180.153 from www.liebertpub.com at 06/17/21. For personal use only. Mean = 2.25, SD = 1.30 Experience: What is your party affiliation? (1 = Democratic, 2 = Unenrolled/NA, 3 = Republican) Approximately how many years have you worked in election administration? Unenrolled = 38%, cratic = 21% Republican = 41%, Demo- Appendix Table A1. Correlation Matrix of Study Variables Variable 1. Process for counting RCV votes 2. RCV is improving elections 3. Should Maine keep RCV 4. Party affiliation (GOP) 5. Support RCV goals 6. Costs of RCV 7. Administrative burden of RCV 8. Voter knowledge of RCV 9. Voter support for RCV 10. Jurisdiction size (natural log) 11. 2nd congressional district *p < .05 (N ranges from 92 to 112). 1 2 3 4 5 6 7 8 9 10 — .61* .43* -.46* .65* -.23* -.36* .47* .17 .06 -.03 — .79* -.52* .69* -.10 -.31* .57* .17 -.01 -.14 — -.50* .50* -.02 -.26* .50* .16 -.05 -.19 — -.44* .07 .36* -.32* -.18 -.17 .21* — -.11 -.24* .47* .16 .17 -.12 — .36* -.20 .21* .29* -.16 — -.46* -.03 .17 .08 — .18 .08 -.19 — .42* -.30* — -.47*