The Future of Election Administration

and other journals. His research focuses on several aspects of elections, cam-paigns, and voting. Mann conducts research on election administration, par-ticularly the e ff ects of changes to permit (and occasionally restrict) pre-election day voting alternatives. Th is research includes partnerships with public election agencies to conduct fi eld experiments and analyses of observation data.


Introduction
. is assessment is based on studies of voting place locations in single jurisdictions. To date, there have been no multi-jurisdictional studies of polling place practices, although some data about polling places are collected in the Survey of the Performance of American Elections (Stewart 2017).
is chapter reports the ndings from a national study of polling places and polling place practices in 26 election jurisdictions and 17 states across the United States during the 2016 presidential election. We evaluate polling places on three dimensions including their accessibility to voters, the quality of the facility/location, and barriers to voting. We measure the variation on these characteristics between and within jurisdictions in order to determine the origin of variance in polling place attributes and practices. We nd that polling place operations, facilities, and practices in 2016 exhibit an overall high quality. Contrary to prior research on these characteristics in a case study of Los Angeles County in 2004 (Barreto et al. 2009), we do not nd that polling place quality varies by race, ethnicity, or the socioeconomic composition of voters at each polling location. Variation in polling place operations, facilities, and practices in 2016 appears to be a function of county and state level factors. 85 and Markovits 2010) and penultimate to voter participation. e Barreto et al. (2009) study of Los Angeles (L.A.) County, CA polling places during the 2004 primary election serves as the touchstone for our own national study. ey report signi cant variation in polling place access, quality, and operations within Los Angeles County. eir principal nding was that ["L]ow-income and minority communities tend to have 'lower quality' precincts, which tended to depress voter turnout" (2009,445). We ask what variation, if any, is observed in polling places across, rather than just within voting jurisdictions (here, counties) and whether this variation is attributable to conditions within or between voting jurisdictions and their respective states.
Reliance on polling place studies in single jurisdictions risks confounding the e ects of polling attributes with di erences in other factors across counties and states. Studying polling places across states and jurisdictions allows us to apportion the variance in polling place attributes and performance to state and county e ects that might be omitted and unobserved in studies of just one voting jurisdiction. Accurately attributing the source of variation in polling place attributes and operations is important to identify where policy interventions might be adopted to enhance the quality and performance of polling places.

Overview of the Project
Research teams recruited from local colleges and universities and located in over 26 election jurisdictions and 17 states across the US observed polling place operations and voters as they entered the queue at their respective polling places on November 8, 2016. 1 A common set of protocols was used across all jurisdictions participating in the data collection (Mann et al. 2018). e jurisdictions that comprise our data set constitute a sample of convenience, because they depend on who responded to the call to participate in the study. e obvious bias induced by this sampling method, compared to drawing a representative sample of voters or polling places, is that jurisdictions without a college or university are unlikely to be included in the study. However, as the list of jurisdictions in Table 6.1 makes clear, the jurisdictions that were in the study were distributed geographically across the country and across urban, suburban, and rural locations. us, while not representative, the collection of precincts is varied enough that important empirical insights can perhaps be gleaned from the data. e jurisdictions studied closely approximate the demographic makeup of the 2016 electorate, as illustrated in Table 6.2. Within jurisdictions, polling places were selected randomly by participating faculty. e unit of random sampling was the polling place rather than physical location, since multiple polling places may be physically located in a single facility (e.g., library, school, community center). In some cases, multiple polling places were selected at a single location. ere were instances where local conditions necessitated deviating from random selection; this was most often due to di culty traveling to voting location or wanting to observe campus voting locations. When such circumstances occurred, the teams were instructed not to select locations expected to have problems or lines (to ensure that selection did not constitute sampling on the dependent variable). e protocol for observing polling place attributes and operations was based on previous research (Barreto et al. 2009;Herron and Smith 2016;Spencer and Markovits 2010;Stewart 2015). Pairs of student-researchers were assigned to observe election day polling places for twohour periods. Researchers were tasked with collecting several pieces of information about voters' experience including length of lines, time waiting to vote, and time to cast a ballot. Each research team was also responsible for lling out a form that described the physical characteristics of the polling place they visited. is form is based on the Barreto et al. (2009)  . e observer's form recorded information about the approach to the polling place (visibility from street, ease of parking, etc.), exterior polling place characteristics (outdoor light, access to parking, accessibility of entrance, etc.), interior polling place characteristics (lighting conditions, waiting area signage, etc.), polling place operations (informational instructions, working machines and scanners, etc.), and a sketch of the polling place layout.
Taking our lead from Barreto et al. (2009), we were interested in knowing how easy it was for the voter to nd and access their election day polling place, whether the location was easy to use, and whether there were any barriers and/or enhancements to voting in the polling places. We measure a voter's accessibility to their polling place with ve 'check list' items including: 1. e polling place address in clear sight (1 = yes, 0 = no) 2. e polling place was readily visible from the street (1 = yes, 0 = no) 3. Flags, banners, or signs made the polling place visible (1 = yes, 0 = no) 4. e polling location was (very easy = 4, somewhat easy = 3, somewhat di cult = 2, very di cult = 1) 5. e outside lighting was adequate (1 = yes, 0 = no) e quality of the polling place location was de ned in terms of seven characteristics including: 1. Adequate parking nearby (1 = yes, 0 = no) 2. Polling place entrance was handicapped accessible (1 = yes, 0 = no) 3. Restrooms were clearly marked (1 = yes, 0 = no) 4. Interior well lit for reading (1 = yes, 0 = no) 5. How small or large was the inside of the polling place (1 = very small, 2 = somewhat small, 3 = medium, 4 = somewhat large, 5 = very large) 6. What kind of waiting area was present (1 = none, 2 = small standing area, 3 = large standing area, 4 = sofas and chairs) 7. Additional amenities for voters (1 = yes, 0 = no) 2 Barriers to voting included whether instructions were posted in the polling location to assist voters to check into vote (0 = yes, 1 = no), to operate voting machines or ballot scanners (0 = yes, 1 = no), and how to complete a ballot (0 = yes, 1 = no) and whether all voting machines and scanners were working (0 = yes, 1 = no). 3

Findings
We obtained information on the attributes of 528 polling places used on election day in 2016 in 26 jurisdictions and 17 states. e distribution of polling traits that Barreto et al. (2009) identi ed as measures of accessibility, quality, and barriers to voting are reported in Tables 6.3, 6.4, and 6.5, respectively. We also report the distribution of the same traits for Los Angeles County polling places for the 2004 primary election and for our sample (N = 38) of 2016 Los Angeles County polling locations (Table 6.6). 4

Polling Place Accessibility
Our ndings point to a greater degree of accessibility than Barreto et al. (2009)  3 In addition, Barreto et al. (2009) identi ed whether poll workers lived nearby the polling place (1 = yes, 0 = no) and whether a photo ID was asked for when checking into vote. We have excluded these two items from our composite score of polling place barriers to voting. We are uncertain how poll workers who live in the neighborhood are either a barrier or enhancement to voting. No discussion of this measure is included Barreto et al. (2009). A portion of our sample of voting jurisdictions is in states that require photographic identi cation in order to vote. Consequently, this is not a discretionary action on the part of either poll workers or county election o cials. Barreto et al. (2009) also report whether a "Voter Bill of Rights" was visibly posted. 4 e dimensionality of the three sets of indices varies considerably. e Cronbach Alpha scores for accessibility are 0.52, 0.45 for polling place quality and 0.7 for barriers to voting. County polling places. e results of our ndings about accessibility are reported in Table 6.3. Nearly two-thirds (65.5%) of our national sample of polling locations were rated 'very easy' to nd and less than 8% were rated as either 'somewhat di cult' or 'di cult to nd.' Only 47% of 2004 Los Angeles polling places were rated very easy to nd. Twice as many Los Angeles County polling places in 2004 (16%) than observed in our 2016 national sample were rated 'very di cult' or 'somewhat di cult' to nd.
Between sixty and nearly ninety percent of our 2016 national sample of polling places were rated clearly visible by their signage, unobstructed street addresses and their proximity to major roadways. ese ratings, with the exception of whether the site clearly displayed its street address, closely match those reported in the 2004 Los Angeles primary. Outside lighting was highly rated in 2016, especially in parking lots adjacent to polling places. Nearly 90% of our national sample of polling places were rated as having adequate outside lighting, an important feature for early morning and late evening voters. Only 77% of the 2004 Los Angeles primary polling places were rated as having adequate outside lighting. A composite accessibility score is constructed from the summation of ratings for our ve indices of polling place access. e composite measure ranges between 2 and 8, with a mean of 6.7 and a standard deviation of 13. e proportion of polling places scoring at the higher end of the accessibility score is skewed.

Quality of Polling Places
e quality of polling places in 2016 was on par with the accessibility of these voting locations. is is also true for four of six measures of polling place quality when we compare 2004 and 2016 polling locations in L.A. County. In 2016, interior lighting and restroom signage were rated signi cantly lower than in 2004. Our ndings about quality are reported in Table 6.4.
In excess of 90% of 2016 polling places were rated as well lit for reading (95%) and handicapped access (94%). Parking at three-fourths of our national sample of polling places was rated 'very easy' to nd. Nearly half (42%) of the interior spaces of polling places in 2016 were rated as 'somewhat large' or 'very large.' Waiting areas in 46% of the sample of 2016 polling places had large standing areas and/or chairs and sofas for voters waiting to vote. A quarter of polling places in 2016 provided amenities to voters while they waited to vote, including water, co ee, cookies, and popcorn. Access to clearly marked restrooms was reported in only 54% of 2016 election day polling places, a lower proportion than reported in the 2004 Los Angeles primary election. e 2016 national sample of polling places exhibited higher scores on all other indices of polling place quality than reported for the 2004 Los Angeles primary.
A composite measure of polling place quality in 2016 has a range of 5-17, with an average polling place quality score of 12 and a standard deviation of 2.2. Missing data for several indices of polling place quality (e.g., restrooms and amenities) signi cantly reduce the proportion of polling places for which we can construct a composite measure of polling place quality (N = 350). 5 Unlike accessibility, our composite score for polling place quality is normally distributed with little evidence of any skewness toward either end of the measure.

Barriers to Voting
In excess of 75% of all polling places surveyed in 2016 had posted instructions for voting, checking into vote, and using voting machines or optical scanners; this gure is comparable to what was reported for Los Angeles polling locations in 2004. Only a scant 4% of 2016 polling places were observed to have problems with either polling machines or optical scanners for paper ballots. Polling place barriers to voting in L.A. County were comparably rated in 2004 and 2016. Our ndings about barriers to voting are reported in Table 6.5.
Overall, few if any signi cant barriers to voting were identi ed in our national sample of voting places. e composite barrier score ranges between zero and ve with a mean 0.6 and a standard deviation of 1, indicating voters experienced few if any barriers to voting in our 2016 sample of voting locations.
Accessible, high quality facilities and a lack of barriers to voting characterize polling places in the 2016 presidential election. Is this nding consistent across and within jurisdictions? Barreto, Cohen-Marks, and Woods report that "data …. reveal variation in polling place quality across precincts (2009,5)." Moreover, the authors go on to conclude that the "quality of polling places varies across the diverse neighborhoods of Los Angeles," where diversity is de ned in terms of the racial/ethnic and socioeconomic composition of polling place voters.

The Source of Variation in Polling Place Attributes
Limited to only Los Angeles County, the Barreto et al. (2009) nding cannot re ect di erences between jurisdictions. To test whether the variation in polling place attributes in our study is a function of within-county polling place attributes or related to di erences among jurisdictions (i.e., counties), we regressed each composite score of polling place attributes on the racial and socioeconomic composition of the electorate in each voting precinct and a dummy variable for the polling place's jurisdiction (i.e., county). 6 If Barreto et al. (2009) are correct, variation in polling place attributes will be signi cantly related to the racial and socioeconomic makeup of the polling place, independent of jurisdictional e ects. Table 6.7 reports regression models for the composite scores of polling place accessibility, quality, and barriers to voting. e models were estimated with xed e ects for counties. e key variables of interest are the proportion of voters in each polling precinct by race and ethnicity (i.e., Black, Hispanic, and Other). e excluded category for race/ethnicity is White. e coe cients for Black, Hispanic, and Other shares of precinct voters represent the e ect of a larger racial/ethnic share of the electorate on the polling place attribute relative to same White share of the precinct's electorate.
ere is little support for the Barreto et al. (2009) nding that polling place attributes were related to the racial and/or socioeconomic composition of a polling places voters. Polling place accessibility is unrelated to any racial, ethnic, or socioeconomic measure. Similarly, polling place quality is unrelated to any racial, ethnic, or socioeconomic measure. Only the proportion of Hispanic persons is signi cantly and positively related to the barriers to voting at polling places relative to the same share of the electorate that is White. Polling places with a higher proportion of voters who are of Hispanic origin than White face more barriers to voting relative to polling places with the same share of White voters. is e ect is statistically insigni cant in the xed e ects model.
To assess the proportion of variance in polling place attributes explained by unspeci ed jurisdictional factors, we have estimated the regression models reported in Table 6.7 with and without xed e ects for jurisdiction. e di erence in R-squares between the models with and without county xed e ects provides us with a relative measure of how much variance in polling place attributes is explained by unobserved jurisdictional factors. Over 90% of explained variation in accessibility, quality, and barriers to voting is attributable to jurisdictional level factors and not racial or socioeconomic conditions unique to within-county polling places. Discussion e ndings from our national sample of polling place practices show the quality and performance of election polling practices to be on balance strong. Our national sample indicated that polling places were accessible, well managed, and with few barriers to voters. Our descriptive ndings match what others have reported for single jurisdiction studies (e.g., Barreto et al. 2009) in other elections. We found no evidence that accessibility to polling places, their quality and practices varied by race, ethnicity or the socioeconomic makeup of the persons who voted at our sample of polling places. Variation in the composition of voters according to polling location does not a ect polling place quality or practices. We identi ed the variation in polling place quality and practices to reside at the level of the county and state. is is not unexpected. State governments are largely responsible for legislating how, when, and where elections are conducted. Counties and other sub-state jurisdictions are responsible for implementing these laws. Counties and their election o cials have some statutory and administrative discretion in conducting elections as prescribed by state law. Identifying the source of variation in polling place practices might begin with a comparative analysis of state election laws and procedures and their implementation at the county level.
We have not examined the consequences of polling places quality and practices on voting. ere are a host of dependent conditions that should be the subject of future research including waiting in line to vote, checking into vote, time to cast a ballot and voter turnout. Barreto et al. (2009) report that poor polling place access, quality, and operations depress polling place turnout. If this is true, is the e ect of polling place attributes on turnout direct or mediated? For example, does the inadequacy of waiting room space at a polling place increase time to check in and vote, increasing the number of voters who leave the polling place without voting? Or does the e ect of overall polling place quality deter voters from voting in the future? Answers to these questions are consequential for identifying remedies to poor voter participation.