Who Votes Early? A Comparative Study of Convenience Voting in Four Countries

Who Votes Early? A Comparative Study of Convenience Voting in Four Countries Peter Miller August 14, 2015 Abstract Comparative research has shed much light on the individual-level predictors of voting, notably the effects of age, education, partisan affiliation, political knowledge and a host of other variables that explain the likelihood of a given voter casting a ballot in an election. One assumption underlying this literature is that voting occurs at the end of the election campaign, after a process of information processing by the voter. There is an open question in the literature, however, on the degree to which voters self-select to cast a ballot before the end of the election campaign. Survey research shows the use of these convenience voting methods is increasing over time and across cases. This paper reports analyses from four cases: Finland, Switzerland, the United Kingdom, and the United States and assesses if early voters are systematically distinct from voters who cast a ballot on Election Day. The effects of convenience voting reforms are relatively unknown in a comparative perspective; this paper seeks to apply the collective knowledge of voting behavior to the increasingly common act of voting while the electoral campaign is still ongoing. 1

This paper is designed to assess the demographic and attitudinal factors that lead voters to cast their ballots while the election is still ongoing. Voting early is an increasingly common act, but we know remarkably little about it outside of the United States. This paper presents analyses of national election survey data from Finland, Switzerland, the United Kingdom, and the United States in order to identify the variables that lead to early voting. The paper begins with a discussion of the convenience voting literature based in the United States with regard to two topics: whether these reforms actually increase turnout in elections and if some voters are more likely to use the reforms than others. The paper then turns to the small literature related to comparative convenience voting, a description of the survey data used in the analyses, and then a series of logistic statistical models designed to identify which demographic and attitudinal factors are driving voters to decide for themselves when they are ready to cast their ballot. The paper concludes with a summary of key findings and prospects for future research in this area. Prior Literature Convenience voting reforms are a class of reforms designed to reduce the opportunity cost of voting and thereby increase turnout (Riker and Ordeshook, 1968, Aldrich, 1993). In the United States there are two general classes of convenience voting, distinguished by distance from the polling place: early inperson where a voter presents herself at a polling station prior to Election Day to complete the ballot and submit it in-person and mail voting where a voter completes her ballot away from the polls altogether and returns the ballot by post (Gronke et al., 2008). These reforms were first adopted on a large scale to accommodate Union soldiers away from their home precincts during the 1864 elections (Benton, 1915). The Civil War experience with soldier voting proved to be short-lived and these voting procedures remained dormant until the 1970s when states began to reduce ballot access requirements for all voters. These reforms have spread throughout the United States to the point that, as of 2015, 36 states and the District of Columbia provide for some form of convenience voting. Three of these states Colorado, Oregon, and Washington have eliminated polling places entirely and conduct elections only by mail ballots. 2

There are normative and practical implications to turnout in the United States and these voting reforms. Higher turnout is generally thought at least by political scientists and democratic theorists to be a social good. Do these reforms, then, lead to higher turnout? Research into the effects on turnout of these reforms is mixed. Early in-person voting increased turnout in Tennessee in 1994 (Richardson and Neeley, 1996), but subsequent research either finds no significant effect (Fitzgerald, 2005) or finds a negative effect of the reform on turnout (Burden et al., 2014, Richey, 2008). In contrast, Leighley & Nagler (2014) examine turnout in presidential elections between 1972 and 2008 using a cross-sectional time-series model and find no-excuse absentee voting increases turnout by about three percentage points, an effect they describe as one of if not the single most important of the changes made to election laws since the Civil Rights Act... (p. 115). All-mail voting was thought to increase turnout in Oregon elections (Richey, 2008, Southwell and Burchett, 2000), but that finding was not replicated in subsequent analyses and appears to be an artifact of the novel voting system in the initial all-mail voting elections (Gronke and Miller, 2012). Other states that have followed Oregon s lead to eliminate polling places exhibit a range of turnout effects. The reform has a small but significant effect on turnout in Washington State (Gerber, Huber and Hill, 2013, Southwell, 2011). In California, by contrast, the reform mandated in precincts with a small number of registered voters is associated with a lower turnout rate (Bergman and Yates, 2011, Kousser and Mullin, 2007). Furthermore, any positive effect on turnout dissipates quickly, usually within two to three electoral cycles (Giammo and Brox, 2010). There is mounting evidence, however, that all-mail balloting increases turnout in off-cycle, special elections (Gronke and Miller, 2012, Karp and Banducci, 2000, Kousser and Mullin, 2007). Voters with disabilities are also significantly more likely to vote using a mail ballot (Miller and Powell, Forthcoming, Schur and Kruse, 2014, Schur et al., 2002, Karp and Banducci, 2001, Schur and Kruse, 2000, Shields, Schriner and Schriner, 1998). It may also be the case that certain voters are more likely to use these reforms than other voters, which may bias any increase in turnout by mobilizing some voters instead of all voters equally. Early voters tend to be older (Barreto et al., 2006, Alvarez, Levin and Sinclair, 2012), better educated (Barreto et al., 2006, Dubin and Kalsow, 1996, Karp and Banducci, 2001, Monroe and Sylvester, 2011, Berinsky, 2005), and more interested in politics or the campaign (Berinsky, 2005, Karp and Banducci, 2001, Kropf 3

et al., 2008, Neeley and Richardson, 2001, Stein, 1998). It is often thought that strong partisan are more likely to vote early (Patterson and Caldeira, 1985, Berinsky, Burns and Traugott, 2001, Kropf et al., 2008, Stein, 1998) though that aspect of the typical early voting may be fading as early voting becomes more common in the United States (Alvarez, Levin and Sinclair, 2012). Parents with children are also more likely to vote early (Dubin and Kalsow, 1996). There appears to be less consensus on the relationship between relative income and early voting, with some studies finding higher income predicting voting prior to Election Day (Karp and Banducci, 2000, Monroe and Sylvester, 2011) and others finding lower income predicts early voting (Barreto et al., 2006, Stein, 1998). Convenience voting reforms in the United States tend to increase biases in the electorate (Berinsky, 2005, Rigby and Springer, 2011) and tend to mobilize voters already inclined to vote independent of the reforms (Neeley and Richardson, 2001, Karp and Banducci, 2001). While most of the academic literature focuses on the effect of convenience in the United States, it is by no means only an American phenomenon. Comparative studies have shown postal voting to increase turnout in advanced industrialized democracies (Massicotte, Blais and Yoshinaka, 2004, Franklin, 2002) and former British colonies (Qvortrup, 2005). Case studies of single countries have also identified a positive turnout effect in Switzerland (Luechinger, Rosinger and Stutzer, 2007), though less densely populated areas may have lower turnout due to a decrease in the civic ritual of voting (Funk, 2010). Turnout in the United Kingdom appears unaffected by postal voting, which was introduced shortly before the 2001 election (Curtice, 2005). Norris also finds beyond specific populations [the disabled and retired persons], the initiative failed to generate much greater turnout overall and nor did it generate greater social equality in the voting population (2005, 18). Rallings and Thrasher (2005), however, find turnout in local elections in 2004 increased in regions of the UK using postal voting. Pesonen (1992) found Finnish advance voters were more likely to be older, unmarried, welleducated, active party members, and very interested in politics in his study of the 1991 parliamentary election. What remains to be seen is if, as advance voting diffuses throughout the electorate, significant demographic and attitudinal differences can explain voting behavior. 4

Data and Methods This paper uses survey data from four countries Finland, Switzerland, the United Kingdom, and the United States to explore demographic factors that might be related to voting prior to Election Day. These four countries include a question about when a voter cast her ballot the vote mode in national surveys and repeat the question over time. 1 Finnish elections since the 1983 parliamentary elections (for a total of 8 electoral cycles) and the 1988 presidential elections (the 1994 and subsequent elections were two-round contests; for a total of 7 elections) are included in these analyses. These data are sourced from the Finnish Social Science Data Archive at the University of Tampere. 2 Pesonen (1992) explains the Finnish system of advance voting was adopted in 1969. Under this administrative system voters are able to complete a ballot in-person for a period of six days prior to the election at postal offices around the country and then the ballot is mailed to the tabulating authority. Swiss surveys include data from six parliamentary election cycles since 1991. 3 These data are available from FORS at the University of Lausanne. 4 Postal voting diffused among the Swiss cantons beginning in the 1960s (Luechinger, Rosinger and Stutzer, 2007, Funk, 2010). British national election surveys first asked about vote mode in 1997; in that election 2 respondents reported voting by post. Subsequent surveys in 2001, 2005, and 2010 also asked about vote mode and are included in these analyses. Survey data from these three elections are available from the University of Essex. 5 British voters were first granted unrestricted access to postal voting in the 2001 parliamentary elections (Curtice, 2005, Norris, 2005). The Current Population Survey provides data from the United States. 1 The 2008 national election survey in Canada, 2007 survey in Iceland, and the 2011 survey in New Zealand also include a question about vote mode, but do not repeat the question over time and are thus excluded from these analyses 2 See the Finnish Voter Barometer for various election years available from http://www.fsd.uta.fi/fi/. 3 The original cumulative dataset includes survey data from 1971-2011. Elections prior to 1991 do not include a question about vote mode and are thus dropped from the analyses. 4 See Selects: Swiss national election studies, cumulated file 1971-2011 [Dataset] Distributed by FORS, Laussane, 2013. www.selects.ch. 5 See http://bes2009-10.org/. The 2015 survey was conducted by a consortium of British universities and will be available from http://www.britishelectionstudy.com/. 5

The CPS is jointly administered by the Census Bureau and the Bureau of Labor Statistics and best known as the official source of the unemployment rate in the United States. The November supplement to the CPS includes a set of questions related to voting behavior, including a question on vote mode. One shortcoming of this dataset, however, is that the survey does not ask about other predictors of voting behavior, like party identification, or political ideology. The provision of convenience voting is determined by state legislatures (the federal government plays a limited role in election administration). Unrestricted postal voting was first used in California in the 1970s; early in-person voting was first used in Texas in the 1980s. Both reforms have been adopted in other states over time (Richey, 2008, Gronke, Galanes-Rosenbaum and Miller, 2007, Leighley and Nagler, 2014). Insert Figure 1 about here Figure 1 shows the change in convenience voting in these four countries over time. The trend in each country indicates a growing proportion of the electorate is opting to vote prior to Election Day. The transition away from the polling place is most striking in Switzerland, where about 80% of the electorate voted early in the 2011 elections. About half of Finnish voters cast their ballots early, a marked change since the 1980s when less than a fifth of the electorate did so. However, the rate of early voting in Finnish parliamentary and presidential elections has remained more or less steady since the mid-1990s. The US and the UK, by contrast, exhibit lower total use of early voting, but also show an increase in early voting over time. 6 Logistic statistical models are used to assess the relationship between a variety of demographic factors and using convenience voting (i.e. voting prior to Election Day is coded as 1 and voting on Election Day is 0 ). Odds ratios are presented in place of log odds to ease the interpretation of coefficients. 7 6 The national rate of early voting in the United States may be deceptive, as state-level rates vary greatly. For example, Oregon and Washington are closer to Switzerland in terms of early voting than other states, while some states in the northeast and south may not provide for unrestricted convenience voting and thus have very low rates of early voting. 7 Recall that the typical logit coefficient is the log of the probability of success (i.e., 1 ) divided by the probability of failure (i.e., 0 ): logit(x)=log(s/f). The odds ratio is, then, the base of the natural logarithm e raised to the power of the log odds. 6

Findings Advance Voting in Finland The logistic models predicting advance voting in Finnish parliamentary elections are presented in Table 1. Consistent with much of the work based in the United States, an increase in age makes advance voting significantly more likely. For example, an increase in age of one year increases the chance of advance voting 2.7% in 1995. An increase in educational attainment by one level (e.g. from primary to secondary education) also increases the chance of advance voting, but only in the earlier elections. Presumably the difference in educational attainment between advance and polling place voters vanishes as a larger proportion of the electorate is voting prior to Election Day. Curiously, higher relative income reduces the likelihood of advance voting. Being married does not appear to significantly alter voting behavior, though an increase in the size of a household contrary to the results obtained in the US reduces the likelihood of advance voting. The Finnish surveys also include a number of attitudinal variables that might be related to advance voting, though these questions are occasionally dropped from a given survey. With those limitations, any relationship between interest in politics, self-placement on the right of the political spectrum, or agreeing that voting is a civic duty on the one hand and advance voting on the other appears tenuous. Insert Table 1 about here Table 2 reports the results of the surveys conducted for Finnish presidential elections. These results closely resemble the results obtained from the set of parliamentary election surveys. The gender of the voter, however, exhibits an interesting pattern. Women are not significantly different than men in terms of advance voting for most of these elections. Except in the 2006 presidential elections and the 2007 parliamentary elections. One possibility for this pattern is that the 2006 presidential election featured a woman, Tarja Halonen of the Social Democratic party, running for reelection. Perhaps her successful reelection contributed to women voting in advance in the parliamentary elections the next year. There are, however, some obvious complications to this possible link. For one, Halonen had already won an historic victory in the 2000 presidential election, but women were not more likely to advance vote in that election. Furthermore, Elisabeth Rehn of the 7

Swedish People s Party competed in the second round of the 1994 presidential elections, but, again, women were not more likely to vote in advance of Election Day in that instance either. Further investigation of the salient issues in these elections is warranted to establish if there is any relationship between co-gender voting mobilization, akin to the co-ethnic mobilization patterns observed in elections in the United States. Insert Table 2 about here Postal Voting in Switzerland Table 3 presents some of the set of demographic and attitudinal predictors of postal voting from the Swiss election surveys. Older age is consistently related to an increase in the likelihood of postal voting, despite the rapid diffusion of postal voting throughout the Swiss electorate (see Figure 1). Education makes postal voting more likely, but not consistently across these six elections. Unlike the Finnish data, higher relative income in Switzerland makes postal voting significantly more likely. Women are not significantly more likely to vote early in Switzerland, except for a remarkably high effect in the 2011 elections (another case worthy of further investigation). Interest in politics strongly predicts voting early, while left-right placement of the voter or the strength of party affiliation has little effect on postal voting in Switzerland. Insert Table 3 about here Postal Voting in the United Kingdom The results of the analyses of survey data from the United Kingdom are presented in Table 4. Age, again, makes voting by post more likely. Relative income predicts postal voting only in the 2010 election. Women are markedly more likely to vote by post in 2005, but not in the other elections, a third intriguing case for further investigation of the relationship between gender and voting behavior. The remaining variables have no significant explanatory power, however. It remains to be seen if demographic factors predicted postal voting in the 2015 election in the United Kingdom. Insert Table 4 about here 8

Convenience Voting in the United States The United States is the last case examined in this paper. Results of these logistic models are presented in Table 5. Like the other cases discussed previously, these results indicate increased age and education make voting early whether in-person or by a mail ballot more likely. Relative income predicts voting before Election Day in the earlier elections in the dataset, but the effect largely disappears after 2000. Black voters are significantly less likely to vote early, except for the 2008 and 2012 elections which featured a credible co-ethnic candidate for President. Latino voters and other ethnic groups are consistently more likely to vote early. Women are also more likely to vote early but unlike the other cases addressed earlier, there is no spike in early voting for women in the United States. Lastly and consistent with other survey data addressed here but not with prior research, these data reveal that being married and an increase in household size are both related to a lower chance of voting early. Insert Table 5 about here Conclusions This paper used comparative election survey data to explore the demographic and attitudinal predictors of voting prior to Election Day. While it was once thought that mail ballots and convenience voting generally were a uniquely American method of election administration (Hamilton, 1988, 860), these data demonstrate convenience voting reforms are being adopted in other cases which can further inform theories of voting behavior. The national election survey data from four countries examined in this paper demonstrate that age is a consistent and significant predictor of voting early across different party and electoral systems. Educational attainment appears to identify voters likely to vote early, at least when rates of early voting are relatively low in a given country. Relative income, however, shows an intriguing relationship with voting behavior, as it is positively related to voting early in Switzerland and earlier elections in the United States, but negatively related with advance voting in Finland. The consideration of attitudinal factors is limited in this paper especially in the case of the United States though preliminary results suggest interest in politics among the Swiss drives the choice to vote prior to Election Day, suggesting the voters 9

who take advantage of these convenience voting reforms are most likely going to vote in any event. Finnish and British voters who choose to vote early, by contrast, are not significantly more interested in politics than polling place voters. It may be the case, then, that convenience voting reforms will not increase turnout in Switzerland as voters most likely to take advantage of the reforms are likely to vote in any event. Evidence from Finland and the United Kingdom is less conclusive as voters in those countries appear to be abandoning the polling place for reasons other than convenience. The results related to gender and voting behavior, in particular, suggest further study of the campaign issues and dynamics in Finland, Switzerland, and the United Kingdom may identify the cause behind the isolated cases where women were substantively and significantly more likely to vote early. Before considering how political campaigns alter strategies to persuade and mobilize voters in the face of increasing rates of convenience voting, it is helpful to identify the voters likely to take advantage of these administrative reforms. This paper provides some preliminary discussion of the factors that are driving the shift away from voting at the polls at the end of a campaign and toward a decentralized voting system where each voter decides when she is ready to cast her ballot. References Aldrich, John. 1993. Rational Choice and Turnout. American Journal of Political Science 37(1):246 278. Alvarez, R. Michael, Inés Levin and J. Andrew Sinclair. 2012. Making Voting Easier: Convenience Voting in the 2008 Presidential Election. Political Research Quarterly 65(2):248 262. Barreto, Matt, Matthew Streb, Mara Marks and Fernando Guerra. 2006. Do Absentee Voters Differ From Polling Place Voters? New Evidence From California. Public Opinion Quarterly 70(2):224 234. Benton, Josiah Henry. 1915. Voting in the Field: A Forgotten Chapter of the Civil War. Boston: Privately Printed. Bergman, Elizabeth and Philip Yates. 2011. Changing Election Methods: How Does Mandated Vote-By-Mail Affect Individual Registrants? Election Law Journal 10(2):115 127. 10

Berinsky, Adam. 2005. The Perverse Consequences of Electoral Reform in the United States. American Politics Research 33(4):471 491. Berinsky, Adam, Nancy Burns and Michael Traugott. 2001. Who Votes by Mail? A Dynamic Model of Individual-Level Consequences of Voting-by- Mail Systems. Public Opinion Quarterly 65:178 197. Burden, Barry, David Canon, Kenneth Mayer and Donald Moynihan. 2014. Election Laws, Mobilization, and Turnout: The Unanticipated Consequences of Election Reform. American Journal of Political Science 58(1):95 109. Curtice, John. 2005. Turnout: Electors Stay Home - Again. Parliamentary Affairs 58(4):776 785. Dubin, Jeffrey and Gretchen Kalsow. 1996. Comparing Absentee and Precinct Voters: A View over Time. Political Behavior 18(4):369 392. Fitzgerald, Mary. 2005. Greater Convenience but not Greater Turnout: The Impact of Alternative Voting Methods on Electoral Participation in the United States. American Politics Research 33(6):842 867. Franklin, Mark. 2002. The Dynamics of Electoral Participation. In Comparing Democracies 2: New Challenges in the Study of Elections and Voting, ed. Lawrence LeDuc, Richard Niemi and Pippa Norris. Thousand Oaks: Sage pp. 148 168. Funk, Patricia. 2010. Social Incentives and Voter Turnout: Evidence from the Swiss Mail Ballot System. Journal of the European Economic Association 8(5):1077 1103. Gerber, Alan, Gregory Huber and Seth Hill. 2013. Identifying the Effect of All-Mail Elections on Turnout: Staggered Reform in the Evergreen State. Political Science Research and Methods 1(1):91 116. Giammo, Joseph and Brian Brox. 2010. Reducing the Costs of Participation: Are States Getting a Return on Early Voting? Political Research Quarterly 63(2):295 303. Gronke, Paul, Eva Galanes-Rosenbaum and Peter Miller. 2007. Early Voting and Turnout. PS: Political Science and Politics 40:639 645. 11

Gronke, Paul, Eva Galanes-Rosenbaum, Peter Miller and Daniel Toffey. 2008. Convenience Voting. Annual Review of Political Science 11:437 455. Gronke, Paul and Peter Miller. 2012. Voting by Mail and Turnout in Oregon: Revisiting Southwell and Burchett. American Politics Research 40(6):976 997. Hamilton, Randy. 1988. American All-Mail Balloting: A Decade s Experience. Public Administration Review 48(5):860 866. Karp, Jeffrey and Susan Banducci. 2000. Going Postal: How All-Mail Elections Influence Turnout. Political Behavior 22(3):223 239. Karp, Jeffrey and Susan Banducci. 2001. Absentee Voting, Mobilization and Participation. American Politics Research 29(2):183 195. Kousser, Thad and Megan Mullin. 2007. Does Voting by Mail Increase Participation? Using Matching to Analyze a Natural Experiment. Political Analysis 15:428 445. Kropf, Martha, Janine Parry, Jay Barth and E. Terrence Jones. 2008. Pursuing the Early Voter: Does the Early Bird get the Worm? Journal of Political Marketing 7(2):131 150. Leighley, Jan and Jonathan Nagler. 2014. Who Votes Now? Demographics, Issues, Inequality and Turnout in the United States. Princeton: Princeton University Press. Luechinger, Simon, Myra Rosinger and Alois Stutzer. 2007. The Impact of Postal Voting on Participation: Evidence for Switzerland. Swiss Political Science Review 13(2):167 202. Massicotte, Louis, André Blais and Antoine Yoshinaka. 2004. Establishing the Rules of the Game: Election Laws in Democracies. Toronto: University of Toronto Press. Miller, Peter and Sierra Powell. Forthcoming. Overcoming Voting Obstacles: The Use of Convenience Voting by Voters with Disabilities. American Politics Research. Monroe, Nathan and Dari Sylvester. 2011. Who Converts to Vote-by-Mail? Evidence from a Field Experiment. Election Law Journal 10(1):15 35. 12

Neeley, Grant and Lilliard Richardson. 2001. Who is Early Voting? An Individual Level Examination. The Social Science Journal 38:381 392. Norris, Pippa. 2005. Rearranging the Deckchairs on the Titanic? The Impact of Postal-Voting-on-Demand in Britain. American Political Science Association. Patterson, Samuel and Gregory Caldeira. 1985. Mailing in the Vote: Correlates and Consequences of Absentee Voting. American Journal of Political Science 29(4):766 788. Pesonen, Pertti. 1992. The Change in Finland from Two-Day Voting to One Day Voting. Behavioral Science Research 37:194 204. Qvortrup, Matt. 2005. First Past the Postman: Voting by Mail in Comparative Perspective. Political Quarterly 76(3):414 419. Rallings, Colin and Michael Thrasher. 2005. Not All Second Order Contests are the Same: Turnout and Party Choice at the Concurrent 2004 Local and European Parliament Elections in England. British Journal of Politics and International Relations 7(4):584 597. Richardson, Lilliard and Grant Neeley. 1996. The Impact of Early Voting on Turnout: The 1994 Elections in Tennessee. State and Local Government Review 28(3):173 179. Richey, Sean. 2008. Voting by Mail: Turnout and Institutional Reform in Oregon. Social Science Quarterly 89(4):902 915. Rigby, Elizabeth and Melanie Springer. 2011. Does Electoral Reform Increase (or Decrease) Political Inequality? Political Research Quarterly 64(2):420 434. Riker, William and Peter Ordeshook. 1968. A Theory of the Calculus of Voting. American Political Science Review 62(1):25 62. Schur, Lisa and Douglas Kruse. 2000. What Determines Voter Turnout? Lessons from Citizens with Disabilities. Social Science Quarterly 81(2):571 587. 13

Schur, Lisa and Douglas Kruse. 2014. Disability and Election Policies and Practices. In The Measure of American Elections, ed. Barry Burden and Charles III Stewart. New York: Cambridge University Press pp. 188 221. Schur, Lisa, Todd Shields, Douglas Kruse and Kay Schriner. 2002. Enabling Democracy: Disability and Voter Turnout. Political Research Quarterly 55(1):167 190. Shields, Todd, Kay Schriner and Ken Schriner. 1998. The Disability Voice in American Politics: Political Participation of People with Disabilities in the 1994 Election. Journal of Disability Policy Studies 9(2):33 52. Southwell, Priscilla. 2011. Letting the Counties Decide: Voter Turnout and the All-Mail Option in the State of Washington. Politics and Policy 39(6):979 996. Southwell, Priscilla and Justin Burchett. 2000. Does Changing the Rules Change the Players? The Effect of All-Mail Elections on the Composition of the Electorate. Social Science Quarterly 81(3):837 845. Stein, Robert. 1998. Introduction: Early Voting. Public Opinion Quarterly 62(1):57 69. 14

Figure 1: Convenience Voting Rates in Four Countries 15

Table 1: Advance Voting in Finnish Parliamentary Elections Predicting Advance Voting 1983 1987 1991 1995 1999 2003 2007 2011 Age 1.000 1.172*** 1.027*** 1.027*** 1.028*** 0.846 1.026*** 1.017*** (0.008) (0.063) (0.006) (0.007) (0.006) (0.086) (0.005) (0.005) Education 1.405** 1.589*** 1.107** 1.045 1.055 0.984 1.032 1.025 (0.187) (0.228) (0.047) (0.110) (0.102) (0.073) (0.068) (0.066) Relative Income 0.868** 0.898* 0.974 1.003 0.974 0.954 0.958** 0.941*** (0.059) (0.057) (0.022) (0.026) (0.021) (0.034) (0.021) (0.019) Female 1.248 0.981 1.159 0.914 1.140 1.169 1.397** 1.281 (0.294) (0.193) (0.176) (0.135) (0.157) (0.219) (0.225) (0.233) Married 0.759 0.630** 0.926 1.268 1.340 1.162 1.425* (0.157) (0.114) (0.202) (0.307) (0.306) (0.222) (0.276) Household Size 0.716 0.899 0.732*** 0.693*** 0.759*** 0.726*** (0.205) (0.067) (0.049) (0.064) (0.063) (0.066) Interest in Politics 1.175 1.175 1.173* 1.153 1.278** 1.118 (0.186) (0.151) (0.110) (0.138) (0.122) (0.110) Left-Right Self-Placement 1.006 1.050 1.106* 1.156*** 1.027 1.005 1.025 (0.049) (0.038) (0.065) (0.064) (0.046) (0.032) (0.032) Trust in Government 0.861 0.980 1.020 (0.112) (0.101) (0.092) Partisan Strength 1.119 1.077 1.210* 0.977 1.097 (0.101) (0.098) (0.127) (0.084) (0.087) Voting as a Civic Duty 1.099 1.233* 0.769** (0.147) (0.154) (0.079) Constant 0.102*** 0.043*** 0.082*** 0.220*** 0.227*** 2.608e+144 0.093*** 0.993 (0.073) (0.026) (0.042) (0.117) (0.115) (5.322e+146) (0.061) (0.613) Observations 923 854 841 798 989 578 991 959 Log Pseudolikelihood -278.2-371.9-543.6-530.8-620.5-369.7-629.5-619.7 Pseudo R 2 0.0271 0.0321 0.0432 0.0401 0.0772 0.0514 0.0829 0.0677 Odds ratios are reported in the table. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 16

Table 2: Advance Voting in Finnish Presidential Elections Predicting Advance Voting 1988 1994-1 1994-2 2000-1 2000-2 2006-1 2006-2 Age 1.074 1.030*** 1.025*** 1.025*** 1.037*** 1.035*** 1.039*** (0.053) (0.005) (0.005) (0.006) (0.007) (0.006) (0.007) Education 1.189 1.269** 1.030 1.032 1.035 0.914 0.956 (0.152) (0.122) (0.097) (0.101) (0.103) (0.098) (0.103) Relative Income 0.831*** 0.944*** 0.951** 0.980 0.949* (0.043) (0.020) (0.020) (0.027) (0.027) Female 1.397* 1.089 1.193 0.931 1.025 1.431** 1.521*** (0.253) (0.150) (0.162) (0.126) (0.140) (0.217) (0.231) Married 0.826 1.066 0.944 1.050 0.962 (0.166) (0.250) (0.217) (0.218) (0.206) Household Size 0.865*** 0.864*** 0.869** 0.874* (0.048) (0.046) (0.060) (0.060) Left-Right Self-Placement 1.016 1.070 0.945 0.978 1.043 (0.049) (0.062) (0.053) (0.060) (0.065) Voting as a Civic Duty 1.133* (0.083) Interest in Politics 1.129* 1.038 (0.082) (0.080) Constant 0.168*** 0.101*** 0.231*** 0.279*** 0.264*** 0.327** 0.252*** (0.085) (0.037) (0.081) (0.137) (0.127) (0.165) (0.129) Observations 799 933 928 1,013 1,063 828 845 Log Pseudolikelihood -409.1-604.3-620.7-661.8-676.4-538.7-540.5 Pseudo R 2 0.0318 0.0308 0.0287 0.0404 0.0600 0.0611 0.0770 Odds ratios are reported in the table. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 17

Table 3: Postal Voting in Switzerland Predicting Postal Voting 1991 1995 1999 2003 2007 2011 Age 1.000 1.008*** 1.010*** 1.008*** 1.013*** 1.021*** (0.007) (0.002) (0.004) (0.003) (0.004) (0.004) Education 1.085 1.021 1.063** 1.049** 1.029 0.979 (0.064) (0.018) (0.030) (0.022) (0.032) (0.028) Relative Income 1.078*** 1.058** 1.077*** 1.078*** 1.098*** (0.015) (0.024) (0.018) (0.026) (0.025) Female 1.216 1.075 1.116 1.158* 1.209 1.504*** (0.275) (0.079) (0.131) (0.101) (0.162) (0.187) Married 0.617** 0.595*** 0.777** 0.836** 0.811 0.597*** (0.149) (0.045) (0.095) (0.076) (0.114) (0.085) Left-Right Self-Placement 0.932 0.952*** 0.963 0.999 0.993 0.975 (0.056) (0.016) (0.022) (0.016) (0.023) (0.022) Interest in Politics 1.401** 1.184*** 1.656*** 1.335*** 1.397*** 1.094 (0.203) (0.064) (0.147) (0.091) (0.146) (0.111) Partisan Strength 0.402*** 0.925 0.666 0.838 (0.080) (0.136) (0.167) (0.173) Constant 0.100*** 0.189*** 0.145*** 0.293*** 0.289*** 1.074 (0.067) (0.046) (0.058) (0.087) (0.138) (0.436) Observations 564 3,784 1,338 2,705 1,275 2,202 Log Pseudolikelihood -253.3-2,357-884.8-1,646-741.2-941.1 Pseudo R 2 0.0333 0.0201 0.0441 0.0197 0.0232 0.0257 Odds ratios are reported. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 18

Table 4: Postal Voting in the United Kingdom Predicting Postal Voting 2001 2005 2010 Age 1.042* 1.035*** 1.032*** (0.022) (0.012) (0.007) Relative Income 1.127 1.029 1.077** (0.120) (0.062) (0.032) Female 1.029 2.034** 1.179 (0.544) (0.702) (0.236) Married 0.430 0.743 0.810 (0.250) (0.261) (0.171) Interest in Politics 0.840 0.894 0.879 (0.176) (0.183) (0.090) Partisan Strength 1.445 0.733 1.239 (0.603) (0.196) (0.176) Voting as a Civic Duty 1.003 1.019 1.067 (0.302) (0.200) (0.129) Trust in Government 1.061 0.993 1.033 (0.086) (0.060) (0.046) Constant 0.003*** 0.031*** 0.015*** (0.005) (0.030) (0.011) Observations 278 358 758 Log Pseudolikelihood -55.27-137.8-345.8 Pseudo R 2 0.0971 0.0714 0.0465 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 19

Table 5: Early Voting in the United States Predicting Early Voting 1996 1998 2000 2002 2004 2006 2008 2010 2012 Age 1.016*** 1.018*** 1.016*** 1.020*** 1.015*** 1.019*** 1.013*** 1.017*** 1.016*** (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Education 1.075*** 1.028** 1.074*** 1.075*** 1.062*** 1.069*** 1.077*** 1.060*** 1.078*** (0.012) (0.013) (0.011) (0.011) (0.008) (0.010) (0.008) (0.008) (0.007) Relative Income 1.030*** 1.021*** 1.021*** 0.995 1.007* 1.002 1.015*** 1.004 1.002 (0.006) (0.007) (0.006) (0.005) (0.004) (0.005) (0.003) (0.004) (0.003) Black 0.533*** 0.549*** 0.482*** 0.551*** 0.580*** 0.478*** 1.120*** 0.664*** 1.067** (0.040) (0.045) (0.032) (0.036) (0.028) (0.030) (0.040) (0.029) (0.033) Latino 1.179* 1.383*** 1.300*** 1.389*** 1.338*** 1.541*** 1.411*** 1.468*** 1.469*** (0.099) (0.121) (0.086) (0.099) (0.067) (0.089) (0.059) (0.073) (0.054) Other 1.192* 1.410*** 1.223** 1.264*** 1.485*** 1.444*** 1.280*** 1.554*** 1.435*** (0.111) (0.138) (0.101) (0.104) (0.075) (0.086) (0.060) (0.078) (0.058) Female 1.064* 1.063* 1.074** 1.010 1.015 1.077*** 1.107*** 1.048** 1.080*** (0.035) (0.039) (0.031) (0.030) (0.022) (0.028) (0.022) (0.023) (0.020) Married 0.863*** 0.837*** 0.860*** 0.834*** 0.833*** 0.898*** 0.953** 0.906*** 0.933*** (0.035) (0.039) (0.031) (0.031) (0.022) (0.029) (0.023) (0.024) (0.020) Household Size 0.881*** 0.893*** 0.912*** 0.941*** 0.924*** 0.905*** 0.876*** 0.910*** 0.946*** (0.014) (0.016) (0.012) (0.013) (0.009) (0.011) (0.008) (0.010) (0.008) Constant 0.044*** 0.054*** 0.066*** 0.066*** 0.129*** 0.101*** 0.211*** 0.151*** 0.162*** (0.005) (0.008) (0.007) (0.007) (0.010) (0.010) (0.014) (0.012) (0.009) Observations 45,294 35,291 43,030 41,123 54,064 40,171 50,983 44,568 59,294 Log Pseudolikelihood -14,066-11,148-16,354-15,989-26,186-19,004-30,042-24,450-35,897 Pseudo R 2 0.0226 0.0228 0.0227 0.0235 0.0206 0.0275 0.0196 0.0233 0.0173 Odds ratios are reported. Robust standard errors in parentheses. White is the reference racial category. *** p<0.01, ** p<0.05, * p<0.1 20