Polling place hours and voter turnout

Transcription

1 Polling place hours and voter turnout Kyle A. Dropp 1 Polling place hours on Election Day vary considerably within and between states. Does this variation affect voter turnout? In the first study on the subject, I find that a 25% reduction in polling place hours exerts no influence on voter turnout. This study utilizes a regression discontinuity design and examines a Minnesota statute enabling jurisdictions with fewer than 500 residents to delay opening their polls by three hours on Election Day. Since areas with slightly more than 500 residents are indistinguishable from areas just below the threshold, the research design produces a credible estimate of the causal impact of reducing polling place hours of operation on voter turnout. This paper s main finding is that the hundreds of voting precincts with reduced hours do not exhibit lower levels of turnout in November Precinct level observational data from a Northeastern state lends support to this finding. 1 PhD candidate, Department of Political Science, Stanford University, dropp@stanford.edu 1

2 Introduction Election Day polling place hours for presidential and midterm contests vary considerably within and between states, affecting the convenience of the voting process. A robust research literature has shown that increasing the convenience associated with casting a ballot increases voter turnout. The current study employs a regression discontinuity design framework to test whether sizable reductions in polling place hours on Election Day reduce voter turnout. This research project examines a Minnesota statute authorizing areas with fewer than 500 residents to reduce polling place hours by up to three hours on Election Day. While approximately 600 jurisdictions limited their hours of operation to 10 hours in November 2010, polling places in all other jurisdictions remained open for 13 hours. Areas with slightly more than 500 residents are indistinguishable from areas just below the election law population threshold. This study s main finding is that voter turnout is no different in areas with expansive hours vis a vis those with limited operating hours. The null findings are consistent across a series of instrumental variables models and do not result from low statistical power the precinct level dataset contains 4,100 observations, variation in voter turnout rates is low and the hypothesized treatment effects are considerable. Further, time series, cross-sectional data from a Northeast state with wide variation in polling place hours confirms the null finding in the regression discontinuity design. These null findings stand in contrast to conventional wisdom and have important implications for party mobilization efforts, political debates over polling place accessibility, local budgets and public policy. This study is organized as follows. First, I discuss the literature and hypotheses. Next, I describe the Minnesota statute in detail. Then, I discuss data collection, research design and findings. The final section briefly concludes. Theory and Hypotheses Election Day polling place hours for presidential and midterm contests vary considerably within and between states. 2 A long line of research in political science has demonstrated that increasing costs associated with casting a ballot decreases voter turnout. Costs, defined broadly, may reference changes in election administration policies such as early voting, 2 Intrastate variation is concentrated in Northeast states such as Maine, New Hampshire and Vermont. In Vermont, for instance, polls across the state stay open between nine and 13 hours. Neighboring states also exhibit strikingly different patterns. Polls stay open 14 hours in Iowa but only 13 in adjacent Illinois, Missouri, Wisconsin, 12 in Nebraska, and 10 in parts of Minnesota. 2

3 registration deadlines, residency requirments, motor voter or same day registration; physical costs attributable to distance traveled, precinct consolidations or inclement weather; or effort required to obtain information about the candidates [Ansolabehere and Konisky, 2006, Brady and McNulty, 2011, Burden and Neiheisel, 2011, Gomez et al., 2007, Gronke et al., 2007, Haspel and Knotts, 2005, Knack, 1995, McNulty et al., 2009, Rosenstone and Wolfinger, 1978, Wolfinger et al., 1990]. Since the utility an individual receives from voting is a function of the overall costs and benefits associated with casting a ballot [Riker and Ordeshook, 1968], and voting costs rise in areas with limited Election Day polling place hours, I predict that limited polling place hours will reduce voter turnout. Many individuals have busy or constrained work and family schedules, and there is ample evidence that these busy schedules affect voting patterns. First, 26 percent of registered voters who failed to cast a ballot in November 2010 selected too busy, conflicting work or school schedule as their reason for not voting, according to the 2010 Current Population Survey s (CPS) November supplement. Non-voters cited conflicting schedules as their top reason for staying home on Election Day, and such work or school conflicts likely are exacerbated in locations with limited polling place hours in the morning or early evening. In areas with extended hours of operation, residents can vote before or after work; however, voters residing in jurisdictions with more limited hours must cast their vote during work hours. Second, and related, a sizable proportion of individuals can cast a ballot in person only in the morning. This subgroup is particularly susceptible to influence by minor shifts or reductions in precinct hours. Approximately one in six Americans can vote between 7:00 a.m. and 10:00 a.m. but at no other time throughout the day, according to a Spring 2012 survey. 3 Third, despite recent gains in popularity for alternative types of voting such as absentee-bymail or absentee-in-person voting, in-person Election Day voting remains the preeminent method. Three-in-four voters (75.1 percent) in 2010 cast their ballots on Election Day, according to the 2010 Current Population Survey. 4 In Minnesota, nearly all voting is done in person on Election Day. 5 In 2010, for example, absentee ballots represented only 6% of all ballots cast. 6 3 The survey asked respondents on Amazon Mechanical Turk to list all hourly time periods between 7 a.m. and 8 p.m. in which they could vote. 219 out of 1,355 (16%) respondents could vote prior to 10:00 a.m. but not at any time thereafter. 4 The 2010 CPS says 24.9 percent cast their ballots early, either in person or by mail pewhispanic.org/files/reports/141.pdf 5 Residents must have an excuse to cast an absentee by ballot by mail, which leads to low rates of absentee voting across the state

4 The Minnesota election statute (explained in detail below) shortened morning precinct hours, one of the busiest times to vote, according to exit polls and surveys. Between onequarter and one-third of voters say they cast their ballots between 7:00 a.m. and 10:00 a.m., according to surveys and exit polls [Fuchs and Becker, 1968, Busch and Lieske, 1985]. 7 Forty-five percent of respondents to the 1992 Current Population Survey, for example, said they cast their ballot before noon on Election Day. Finally, a series of recent studies demonstrate that similar shifts in election administration policy can decrease voter turnout. Decreases in voter turnout have resulted from consolidating precincts in Los Angeles County [Brady and McNulty, 2011], moving precinct polling places [Haspel and Knotts, 2005], increasing the distance from the individual s residence and polling place [Haspel and Knotts, 2005] and resulting from inclement weather on Election Day [Gomez et al., 2007]. In summary, five reasons suggest that areas with shortened polling place hours on Election Day will exhibit lower voter turnout: recent literature has shown that similar changes in election administration policies such as precinct consolidation reduce participation, a plurality of non-voters cite scheduling conflicts for staying home, the vast majority of ballots are still submitted on Election Day, many voters can only cast ballots in the morning and the morning period is one of the busiest on Election Day. The next sections take these predictions to the data. Polling place hours in Minnesota This section focuses on a Minnesota statute that authorizes townships under a specified population threshold to delay poll opening times on Election Day. The foregoing subsections introduce the statute, describe its provisions, compare eligible and ineligible municipalities, describe data collection efforts and present descriptive statistics such as compliance and balance tests. 7 There is little contemporary data available on the subject. In Alameda County, California in 1968, 8.2 percent voted before 8 a.m., 13.3 percent voted between 8 a.m. and 10 a.m., 14.0 percent voted between 10 a.m. and noon, 11.6 percent voted between noon and 2 p.m., 12.2 percent voted between 2 p.m. and 4 p.m., 11.3 percent voted between 4 p.m. and 5 p.m., 14.3 percent voted between 5 p.m. and 6 p.m., and 4.4 percent voted between 6 p.m. and 7 p.m.[fuchs and Becker, 1968] In 1972, 11.4 percent voted prior to 8 a.m., 31.9 percent voted between 8 a.m. and noon, 21.4 percent voted between noon and 4 p.m., 21.2 percent voted between 4 p.m. and 6 p.m., and 14.1 percent voted after 6 p.m, according to the Current Population Survey. In a 1981 Cleveland exit poll, Busch and Lieske find that 15.1 percent voted 6:30 a.m. to 9:00 a.m., 17.8 percent voted 9:00 a.m. to 11:30 a.m., 25.4 percent voted 11:30 a.m. to 2:00 p.m., 18.1 percent voted 2:00 p.m. to 4:30 p.m. and 23.6 percent voted 4:30 p.m. to 7:00 p.m.[busch and Lieske, 1985] Forty-five percent of respondents to the 1992 Current Population Survey said they cast their ballot before noon, compared with 21 percent who said they voted between noon and 4 p.m., 19 percent who voted between 4 p.m. and 6 p.m., and 15 percent who cast their ballot after 6 p.m. 4

5 The statute In 2005, Minnesota revised an election law to grant specific jurisdictions the authority to reduce polling place hours on Election Day, and approximately 600 eligible areas shortened their hours of operation in November 2010, the most recent midterm contest. For historical details on the law dating back to 1981, see the footnote. 8 The statute addresses three main questions: 1) what are acceptable poll opening and closing times for statewide general elections, 2) which areas are eligible to reduce poll hours, and 3) what is the process for deciding whether to reduce hours? The statute sets standard opening and closing times for polls in statewide general elections across the state: Except as otherwise provided...hours for voting in every precinct in the state shall begin at 7:00 a.m. and shall extend continuously until 8:00 p.m. 9 The law s next subdivision exempts smaller jurisdictions outside of the Twin Cities metropolitan area 10 from the strict hours requirement. These precincts can open their polling locations as late as 10:00 a.m., up to three hours later than the 7:00 a.m. opening time prescribed in the aforementioned statute. All precincts must maintain the standard 8:00 p.m. closing time: 8 In 1981, 204C.05, Subdivision 1 was changed to say The governing body of a municipality of less than 1,000 inhabitants, located entirely outside the metropolitan area as defined in Minnesota statutes, Section , Subdivision 2, may fix a later time for voting to begin...it shall not be later than 9:00 a.m. for the state general election. Section refers to counties near the Twin Cities. In 1983, the clause was deleted and all polling places maintained full hours on Election Day https: // Two years later, in 1985, the statutes were amended to enable smaller jurisdictions to reduce hours: Subd. 1a....The governing body of a town with less than 500 inhabitants according to the most recent federal decennial census, which is located outside the metropolitan area as defined in section , subdivision 2, may fix a later time for voting to begin at state primary, special, or general elections, if approved by a vote of the town electors at the annual town meeting. The question of shorter voting hours must be included in the notice of the annual town meeting before the question may be submitted to the electors at the meeting. The later time may not be later than 10:00 a.m. for special, primary, or general elections. The town clerk shall either post or publish notice of the changed hours and notify the county auditor of the change 30 days before the election. Subd. 1b. [ELECTIONS; UNORGANIZED TERRITORY.] An unorganized territory or unorganized territories which constitute a voting district may have shorter voting hours if at least 20 percent of the registered voters residing in the voting district sign a petition for shorter hours and present it to the county auditor. The later time may not be later than 10:00 a.m. for special, primary, or general elections... Finally, in 2005, the definition of a metropolitan area switched from from to a more inclusive Compared with 1985, the following areas were considered metropolitan areas and forced to maintain full hours on Election Day: Chisago, Dakota (part), Hennepin (part), Isanti, Scott (part), Sherburne, and Wright. 9 Section 204C.05, Subdivision 1 10 The Twin Cities Metro Area is referenced as Section , subdivision 24 5

6 The governing body of a town with less than 500 inhabitants according to the most recent federal decennial census, which is located outside the metropolitan area as defined in section , subdivision 24, may fix a later time for voting to begin at state primary, special, or general elections, if approved by a vote of the town electors at the annual town meeting. The later time may not be later than 10:00 a.m. for special, primary, or general elections. 11 The later opening time must be approved at a town meeting, an annual low-key gathering where residents generally review year-end financial reports, discuss the budget and review current or future maintenance projects. Finally, while many eligible jurisdictions decide to reduce polling place hours of operation, more than 500 of the state s tiniest precincts conduct their elections solely by mail-in balloting. Residents in such areas are mailed ballots approximately 20 days before an election. The governing board of any municipality with fewer than 400 registered voters as of June 1 of the current election year may apply to the county auditor for permission to conduct balloting by mail. 12 Eligible areas To summarize the previous subsection, municipalities can reduce polling place hours if 1) they are located outside of the Twin Cities metropolitan area, 2) they have fewer than 500 inhabitants according to the 2000 decennial Census and 3) if town electors approve the change at an annual town meeting. Three thousand of the 4,100+ precincts in Minnesota are located outside of the Twin Cities, nearly 1,800 satisfy conditions 1 and 2, and about 600 reduced their polling place hours in There are hundreds of precincts slightly above or below the 500 person threshold, providing relatively high statistical power for a regression discontinuity design. Figure 1 on Page 14 displays a histogram of the population of precincts in the nonmetropolitan areas. There are 350 precincts either above or below the threshold within 100 persons, more than 500 precincts within 150 persons and 750 within 200 persons either way. [Figure 1 about here] 11 From Subd. 1a. Elections; organized town A municipality having fewer than 400 registered voters on June 1 of an election year and not located in a metropolitan county as defined by section may provide balloting by mail at any municipal, county, or state election with no polling place other than the office of the auditor or clerk or other locations designated by the auditor or clerk. The governing body may apply to the county auditor for permission to conduct balloting by mail. 6

7 Figure 2 displays the eligible counties, all counties outside of the Twin Cities Metropolitan area, in yellow. Figure 3 displays the actions taken by these jurisdictions. Areas shaded red maintained full polling place hours in 2010, yellow areas reduced hours, and white areas conducted mail-in elections. In many suburban and rural counties, small cities maintained full hours while the outlying areas reduced hours. This trend is most apparent in southwest Minnesota. [Figures 2 and 3 about here] Data collection Precinct level election returns for 2008 and 2010 were obtained from the Minnesota Secretary of State s office. The data includes total voter turnout, votes for each office and absentee balloting in each voting district. The Secretary of State s office does not maintain a statewide database of polling place hours of operation. Polling place opening and closing times for the 2010 midterm election were obtained from the state s 87 county auditors via phone correspondences and requests. 13 The voting returns were matched with Census data from the National Historical Geographic Information System site, supplementing the datasets with extensive voting district level demographic information. 14 Merged Census variables include racial composition, hispanic background, median age, housing unit tenure, rental status, urban vs. rural, marital status, employment status, work commute, educational attainment, college enrollment, citizenship, industry type, occupation type, median household income and poverty rate. The precinct s 2000 Census population was used to note whether it was eligible to reduce hours and serves as an instrument in the two stage least squares models (described in the next section). Descriptive statistics Table 1 on Page 16 displays the relationship between municipality size and polling place hours of operation. There is an inverse relationship between size and the decision to reduce hours on Election Day. The top half demonstrates that nearly all jurisdictions complied 13 Nearly all County auditors provided spreadsheets with this information. 14 First, the precinct level vote returns were matched with 2010 Census voting district files. Second, 2000 Block Group level demographic information was linked with the 2010 voting district level file. This paper uses 2000 data because there was no Census long form in the most recent decennial Census. A possible next step is to link five year American Community estimates to voting districts and to aggregate race, hispanic background, median age from the 2010 block level to the 2010 voting district level. 7

8 with the election statute: only 12 areas with populations above 500 reduced hours while more than 500 with populations under the threshold reduced their hours. The bottom half shows that there are hundreds of precincts on each side of the 500 person threshold. Precincts with fewer than 250 inhabitants were most likely to reduce their hours or conduct their elections by mail-in ballots. [Table 1 about here] Figure 4 on Page 17 displays the relationship between population size and the probability of reducing precinct hours. The plot demonstrates that the instrument satisfies the inclusion restriction, as it is correlated with the endogenous explanatory variable at conventional levels of significance. 15 Each precinct is represented by a black dot, and separate smoothed lines have been created for precincts above and below the 500 person population threshold. Precincts containing slightly less than 500 persons are much more likely to reduce hours. Approximately half of the smallest precincts limit polling place hours of operation and one in four precincts with between 400 and 500 persons limits hours. By contrast, only a handful of larger precincts reduce hours. No jurisdictions with 650 or more residents limit polling place hours. [Figure 4 about here] Table 2 on Page 18 demonstrates that precincts on both sides of the 500 person population threshold are nearly identical across demographic variables such as age, race, educational attainment, political views and median income. For instance, residents in voting districts above and below the cutoff by 150 persons typically are white, middle class, and married. Table 2 provides separate balance statistics for precincts within 150 persons of the threshold on either side and those within 250 persons of the threshold. [Table 2 about here] In summary, there are hundreds of voting districts on either side of the 500 person threshold identified in the election statute, areas on both sides are similar across a series of important demographic characteristics, and the threshold predicts precinct polling hours of operations. 15 The McFadden s Pseudo R-squared for a regression of an indicator for reducing hours on population is.18 and the coefficient is significant. 8

9 Research Design and Findings In general, observational approaches cannot provide a credible causal estimate of the impact of polling place hours on voter turnout. Extending hours of operation may amplify turnout, but, if officials in areas with lower turnout expectations extend polling place hours, then it will appear that such extensions reduce voter turnout. The challenge, therefore, is in finding polling places that are identical but for our primary independent variable polling place hours of operation. Jurisdictions in Minnesota with slightly more than 500 residents should be indistinguishable from those with slightly fewer than 500; however, the first subset maintains full hours on Election Day and the latter can delay poll opening times by up to three hours. Regression Discontinuity Design The defining characteristic of a regression discontinuity design is that the probability of receiving treatment varies discontinuously as a function of an underlying variable [Hahn et al., 2001]. In the primary specification, I treat polling place hours in 2010 as an endogenous explanatory variable and instrument for this variable using the 500-person population threshold. That is, the first stage equation regresses polling place hours on the population threshold, while the second stage regresses voter 2010 turnout on predictions from the first stage. Table 3 on Page 19 displays results for three instrumental variables models across two population thresholds. The first stage in these models was estimated by an OLS regression of polling place hours on the population threshold, along with demographic controls and previous turnout. Models 1-3 compare precincts with 500 to 750 persons to those with persons, while Models 4-6 compare precincts with 500 to 650 persons to those with residents. Overall, the main explanatory variable, Hours, does not significantly predict 2010 voter turnout in any of the models. Background control variables such as age, marital status, educational attainment and median income are all in the expected direction [Table 3 about here] Table 4 on Page 20 reports results for another set of instrumental variables models that used a first stage logistic regression (rather than OLS). 16 As in Table 3, there is no relationship between the decision to reduce polling place hours and voter participation. 16 Precincts that reduce hours are coded as 1 while those that maintain full hours are coded as 0. 9

10 [Table 4 about here] Finally, precincts were matched using genetic matching across the 500 resident threshold on age, race, educational attainment and median household income. There was no difference in turnout across the threshold. Given the relative balance displayed in Table 2 on Page 18, it is not surprising that these estimates are nearly identical to the instrumental variables estimates in Tables 3 and 4. Overall, differences in polling place hours across precincts exert little effect on voter turnout, despite the large dataset and relatively low variation in turnout across voter precincts. Variation in Polling Hours in Vermont Polling place hours of operation vary considerable within Vermont, staying open between nine and 13 hours on Election Day for general elections. 17 I test whether polling places with longer hours exhibit higher levels of turnout on Election Day using time-series, crosssectional data from the state. Overall, neither cross-sectional differences in hours, nor within precinct changes in polling place hours over time, affect political participation. I obtained data on polling place hours of operation in Vermont at the township level from 2004 through Next, I merged this data with voter turnout data from the Vermont Secretary of State s office and Census demographic data. Table 5 illustrates that there is no relationship between polling place hours on Election Day and voter turnout in Vermont. 18 Models 1-3 use the voting age population as the dependent variable, while Models 4-6 use the number of registered voters. Each of the models includes fixed effects for the four midterm and general election years. [Table 5 about here] Table 6 examines both cross-sectional differences in polling hours and within precinct changes over time. The variable Opening Time Difference is 0 if the polling place opened at 17 Vermont polling places are open between nine and 13 hours on Election Day. The modes are at nine and 12 hours. In the state, polling places must close at 7 p.m. but opening times range from 6 a.m. to 10 a.m. According to Kathleen Scheele, Vermont s Director or Elections and Campaign Finance, local BCAs [Boards of Civil Authority] have had the power to set the opening hours for over 30 years. There is always variations in Vermont primarily based upon the number of voters in the town. Nearly all townships larger than 4,000 residents have polling places that remain open 12 or 13 hours, while a majority of municipalities under 1,000 residents have a limited, nine hour voting period. 18 The main independent variable, Poll Opening Time, is coded 1 to 5, where precincts opening at 6 a.m. are 1 and precincts opening at 10 a.m. are 5. 10

11 the same time across two election cycles, positive if the polling place opened later (shorter hours) in the second comparison year, and negative if the polling place opened earlier in the second year. Models 1-4 regress the change in turnout for elections two years apart (e.g., 2004 to 2006; 2006 to 2008; 2008 to 2010) on the opening time difference and poll opening time, while Models 5-8 examine change only across midterm contests (e.g., 2006 to 2010) or presidential elections (e.g., 2004 to 2008). The full models with demographic control variables suggest that poll opening times exert little to no influence on voter turnout across the state. [Table 6 about here] Conclusion This study s main finding is that voter turnout is no different in areas with expansive hours vis a vis those with limited operating hours. The null findings are consistent across a series of instrumental variables models and do not result from low statistical power the precinct level dataset contains 4,100 observations, variation in voter turnout rates is low and the hypothesized treatment effects are considerable. Further, time series, cross-sectional data from a Northeast state with wide variation in polling place hours confirms the null finding in the regression discontinuity design. The finding is unintuitive, since nearly all Minnesota voters cast their ballots in person on Election Day and exit polls and other studies show that a significant proportion of voters typically cast their ballots in the affected morning hours. It suggests that municipalities may be able to save money on Election Day without concomitantly reducing voter participation rates. This section addresses the generalizability and potential limitations of this study. The statute under examination applies to suburban and rural areas in the state of Minnesota, along with the state of Vermont, thus excluding diverse, urban areas. Changes in polling place hours of operation affect urban residents differently than suburban or rural voters. The treatment may not have been strong enough. While hundreds of precincts opened at 10 a.m. rather than 7 a.m., this disruption may not have been a strong enough treatment. Morning voters could return a few hours later. Future studies can test when closing a poll exerts a larger impact on turnout, compared with opening a poll late. Voters may be well informed of these polling place hours changes. Residents in Minnesota must approve election law changes at an annual town meeting and the county auditor and 11

12 local officials must provide notice of polling place hours and other election information prior to each election. Finally, voting is habitual, and if these hours have been in place for multiple election cycles, voters may adjust their schedules accordingly. References S. Ansolabehere and D. M. Konisky. The introduction of voter registration and its effect on turnout. Political Analysis, 14(1):83âĂŞ100, H. E. Brady and J. E. McNulty. Turning out to vote: The costs of finding and getting to the polling place. American Political Science Review, 105(1): , B. C. Burden and J. R. Neiheisel. Election administration and the pure effect of voter registration on turnout. Political Research Quarterly, R. J. Busch and J. A. Lieske. Does time of voting affect exit poll results? Public Opinion Quarterly, 49(1):94 104, D. A. Fuchs and J. Becker. A brief report on the time of day when people vote. The Public Opinion Quarterly, 32(3): , B. T. Gomez, T. G. Hansford, and G. A. Krause. The republicans should pray for rain: Weather, turnout, and voting in US presidential elections. Journal of Politics, 69(3): , P. Gronke, E. Galanes-Rosenbaum, and P. A. Miller. Early voting and turnout. PS: Political Science & Politics, 40(04): , Jinyong Hahn, Petra Todd, and Wilbert Van der Klaauw. Identification and estimation of treatment effects with a regression-discontinuity design. Econometrica, 69(1): , Moshe Haspel and H. Gibbs Knotts. Location, location, location: Precinct placement and the costs of voting. Journal of Politics, 67(2): , Stephen Knack. Does Motor voter work? evidence from state-level data. The Journal of Politics, 57(03): , J. E. McNulty, C. M. Dowling, and M. H. Ariotti. Driving saints to sin: How increasing the difficulty of voting dissuades even the most motivated voters. Political Analysis, 17 (4): , W. H. Riker and P. C. Ordeshook. A theory of the calculus of voting. The American Political Science Review, 62(1):25 42,

13 S. J. Rosenstone and R. E. Wolfinger. The effect of registration laws on voter turnout. The American Political Science Review, pages 22 45, W. Van der Klaauw. Estimating the effect of financial aid offers on college enrollment: A regression-discontinuity approach. International Economic Review, 43(4): , R. E. Wolfinger, D. P. Glass, and P. Squire. Predictors of electoral turnout: An international comparison. Review of Policy Research, 9(3): ,

14 Figure 1: Voting precinct population in Minnesota Frequency Population, 2000 Census 14

15 Figure 2: Areas eligible to reduce hours Figure 3: Polling place hours. Red is full, yellow is reduced, white is mail-in 15

16 Table 1: Precincts by Polling Place Hours and 2000 Population Full Reduced Mail-in Total Metropolitan Non-Metropolitan Under 250 people

17 Figure 4: 2000 Population and probability of reducing hours in 2010 Probability of reducing hours Population 17

18 Table 2: Balance summary across Census demographic variables Precinct Population % Dayton - Gov % White % Married % Never Married % Urban % Renter % Poverty % Age % Age % Citizens Median HH Income 41,836 42,085 41,866 41,159 Cases

19 Table 3: IV estimates polling place hours and 2010 voter turnout, Minnesota - OLS first stage +/- 250 residents +/- 150 residents Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Hours (range is 10-13) (0.0058) (0.0038) (0.0039) (0.0128) (0.0082) (0.0079) % White * * * (0.0513) (0.0343) (0.0688) (0.0428) Median HH * * Income (1,000s) (0.0006) (0.0004) (0.0008) (0.0005) % College * * * * (0.0589) (0.0392) (0.0853) (0.0527) % Married * (0.0548) (0.0364) (0.0785) (0.0486) % 20 yrs Housing Tenure (0.0320) (0.0213) (0.0482) (0.0298) % Age * * * * (0.0405) (0.0276) (0.0599) (0.0384) Turnout * * * * (0.0210) (0.0214) (0.0273) (0.0279) Constant * * (0.0817) (0.0522) (0.0565) (0.1514) (0.1056) (0.0950) N adj. R Resid. sd Standard errors in parentheses indicates significance at p <

20 Table 4: IV estimates of polling place hours and 2010 voter turnout, Minnesota - logit first stage +/- 250 residents +/- 150 residents Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Reduced Hours (0.0058) (0.0112) (0.0039) (0.0353) (0.0239) (0.0218) % White * * * (0.0513) (0.0343) (0.0728) (0.0452) Median HH * * Income (1,000s) (0.0006) (0.0004) (0.0008) (0.0005) % College * * * * (0.0589) (0.0392) (0.0855) (0.0528) % Married * (0.0548) (0.0364) (0.0813) (0.0502) % 20 yrs Housing Tenure (0.0320) (0.0213) (0.0482) (0.0298) % Age * * * * (0.0405) (0.0276) (0.0604) (0.0386) Turnout * * * * (0.0209) (0.0214) (0.0273) (0.0279) Constant * * * * (0.0817) (0.0153) (0.0565) (0.0704) (0.0204) (0.0442) N adj. R Resid. sd Standard errors in parentheses indicates significance at p <

21 Table 5: OLS models that regress turnout on poll opening hours, Vermont (1) (2) (3) (4) (5) (6) %VAP %VAP %VAP %RVs %RVs %RVs Poll Opening Time * (6 to 10 a.m.) (0.003) (0.002) (0.002) (0.002) (0.002) (0.002) Log of Median HH Income * * * * (0.017) (0.019) (0.016) (0.018) % College * * (0.027) (0.029) (0.026) (0.028) Voting Age Population * * * * (square root) (0.000) (0.000) (0.000) (0.000) Constant * * * * * * (0.022) (0.177) (0.199) (0.016) (0.171) (0.192) N adj. R Resid. sd Year Fixed Effects Yes Yes Yes Yes Yes Yes County Fixed Effects No No Yes No No Yes Standard errors in parentheses indicates significance at p <

22 Table 6: Difference in Turnout Models, Vermont Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 % VAP Turnout Diff - Election t vs. t-1 % VAP Turnout Diff - Election t vs. t-2 Opening Time Difference (0.002) (0.002) (0.012) (0.012) (0.002) (0.002) (0.012) (0.012) Log of Median HH Income * * (0.011) (0.011) (0.012) (0.012) % College (0.018) (0.018) (0.020) (0.020) Polling Opening Time * * (0.001) (0.001) (0.001) (0.001) Polling Opening Time * Opening Time Difference (0.002) (0.002) (0.002) (0.001) Constant * * * * * * * * (0.002) (0.111) (0.010) (0.116) (0.002) (0.123) (0.011) (0.130) N adj. R Resid. sd Year Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Standard errors in parentheses indicates significance at p <