Electoral Reform, Party Mobilization and Voter Turnout. Robert Stein, Rice University

Electoral Reform, Party Mobilization and Voter Turnout Robert Stein, Rice University stein@rice.edu Chris Owens, Texas A&M University cowens@polisci.tamu.edu Jan Leighley, Texas A&M University leighley@polisci.tamu.edu

2 Abstract In this paper we explore the conditions under which candidates and party officials use electoral reforms (i.e., early voting) to mobilize voters. We draw upon a unique database that studies a sample of registered voters and matches their validated voting behavior with partisan campaign activities in each voter s respective county during the 2002 midterm election in Texas. We expect, first, that considerations of strategy and opportunity shape both parties use of early voting to mobilize supporters and second, that the impact of their efforts on voter turnout is mediated by these elite strategies. Our empirical evidence supports our central thesis: without the efforts of political parties and their candidates, electoral reforms are likely to have a marginal effect on voter turnout. An important modification to our original thesis, however, is that the effectiveness of electoral reforms is dependent not only on whether political elites choose to take advantage of such rules in developing their electoral strategies, but also on party members initial probability of voting absent any mobilization, a key feature distinguishing the effectiveness of Republican and Democratic mobilization efforts..

3 Low and possibly declining voter turnout over the past several decades has disturbed politicians and scholars alike. Some national and state officials have responded to this issue by proposing various institutional reforms such as election day registration, voting by mail, early voting and motor-voter registration aimed at reducing the costs of voting. In addition to increasing electoral participation in general, these proposals have also been justified as a means to increase voter turnout among historically underrepresented populations (i.e., the young, the poor and racial and ethnic minorities). Yet research has shown that the promise of these reforms has fallen short of expectations. Registration reform, for example, is not associated with greater turnout on election day (see, for example, Knack 1995). The disappointing effects of these policy adoptions pose an interesting puzzle with respect to electoral behavior. Why, when institutional costs are lowered, does the mass public not respond? We contend that the effectiveness of electoral reforms is contingent upon on the strategic behavior of elites. Without strategic decisions by elites to use electoral reforms to their advantage, electoral reforms will be unrelated to voter turnout. According to this logic, if vote maximizing candidates utilize these electoral reforms to mobilize voters in support of their candidacies, then voter turnout will increase. If candidates strategic calculations suggest that increased turnout is not to the campaign s advantage, then these electoral reforms would be irrelevant to candidates campaign strategies and, ceteris paribus, not affect turnout. Conceptually, this argument highlights the distinction between voters having the opportunity of voting by mail, registering on election day, or voting

4 early, for example, and candidates choosing to use these opportunities as part of their campaign strategy. Some researchers (e.g., Hansen and Rosenstone 1993; Aldrich and Simon 1986) have suggested that the efficacy of these reforms is dependent upon the campaign activities of candidates and their parties. That is, changing the rules of the game is not sufficient to increase turnout. Yet most research neglects the central role of candidates assessments of the utility of electoral reforms in winning elections. Another neglected issue in the literature is the potential impact of electoral reforms on the composition of the electorate. Electoral reforms might differentially impact voters by race, ethnicity, or class, for example. Although reformers have argued that reducing the costs of voting would increase representation of historically disadvantaged voters (i.e., the poor and less educated) in the electorate, there is limited research on this question. Moreover, we know of no research that has studied how elites strategic campaign decisions influence the relationship between electoral reforms and voter turnout among the poor and less educated. In an attempt to close this gap in the literature we offer a model of how party mobilization links the availability of early voting to voter turnout. Drawing upon a unique database that samples registered voters and matches their validated voting behavior with partisan campaign activities, we develop a research design for testing hypotheses derived from our explanation. We find that individuals who have greater opportunities to participate in early voting are no more likely to vote than individuals for whom such opportunities are more limited. However, when early voting opportunities are

5 exploited by partisan mobilization efforts, individuals are more likely to vote. Further, we find no evidence that electoral reforms enhance the likelihood that historically disadvantaged voters will cast a ballot. This relationship remains insignificant even in the presence of strong partisan mobilization. Together, these findings underscore the importance of elites in structuring mass participation and illustrate the importance of subjecting our conventional wisdom to empirical analysis. Electoral Reform and Party Mobilization Previous research has identified four major influences on individual decisions to participate social and demographic traits, psychological resources, electoral rules, and the mobilization efforts of parties and their candidates (Leighley 1995). Electoral reforms directed at increasing voter participation have centered on simplifying voter registration and increasing opportunities to vote (e.g., voting by mail and in-person early voting). The rationale underlying early voting and related electoral reforms (e.g., Motor Voter and voting by mail) has been the belief that providing more opportunities to vote (i.e., the number of days, hours or sites at which to vote) increases voter participation. Reducing the number of days prior to an election in which voters can register to vote and allowing individuals to register to vote when renewing their driver s licenses or at the polls on election day has increased voter registration (Wolfinger and Rosenstone 1980; Squire, Wolfinger and Glass, 1987). But studies of the direct effect of voter registration and balloting reforms on voter turnout suggest minimal responses on the part of the electorate. National legislation (i.e., National Voter Registration Act of 1993) to enable individuals to register to vote when they renew or obtain a driver s license has had only a modest impact on voter turnout (Knack 1995; Rhine 1995). Moreover, liberalized

6 voting by mail (Berinsky, Burns and Traugott 2001) and in-person early voting (Stein and Garcia-Monet 1997; Stein 1998) were also found to have an insignificant or marginal effect on increasing the likelihood an individual will vote. These findings conflict with the general belief that reducing the restrictiveness of electoral laws necessarily increases turnout, and especially so for lower-class individuals. Indeed, Lijphart (1997) argues that electoral reforms are justified for their ability to redress class inequalities in voter turnout, a particularly striking feature of the U.S. electorate (Leighley and Nagler 1992). Yet the empirical evidence that contemporary electoral reforms reduce class differences in turnout is mixed and weak. Nagler (1991: 1402) concludes that restrictive registration laws do not deter poorly-educated individuals from registering any more than the highly-educated. This implies that liberalizing these laws may increase registration overall, but will not equalize participation across classes. Conclusions regarding election reforms beyond registration are similar. Stein (1998) reports that resource-poor voters did not benefit from the adoption of in-person early voting, while Berinsky et al., (2001) find that voting by mail has little effect on the resource-poor (2001:178). Stein (1998) also reports that early voters appear to be more partisan, ideological, interested in politics, and disproportionately likely to have voted in the past. Simply put, electoral reforms have only been used by those who otherwise would have been most likely to vote without them. This minimal effects conclusion may be premature. When candidates and parties engage in voter mobilization activities, there is a high likelihood that their efforts will be fruitful (Gerber and Green 2000; Adams and Smith 1980; Huckfeldt and Sprague

7 1992; Wielhouwer 1995; Wielhouwer and Lockerbie 1994). And when party mobilization efforts rely on the use of electoral reforms such as absentee or early voting, they tend to increase voter turnout. Patterson and Caldeira (1982), for example, find absentee voting rates are higher in states with greater party mobilization efforts (1982:785), while Oliver (1996) finds that in states where absentee voting requirements are most liberal and where political parties invest time and resources to mobilize absentee voters, the levels of absentee voting rise and the characteristics of absentee voters change (1996:25). Party mobilization may thus be central to linking electoral reforms to voter behavior. Theory and hypotheses How might our assessment of the efficacy of electoral reforms differ were we to consider a model where reform effects are contingent upon elite strategic considerations? Our model of electoral reform effects posits an indirect relationship between electoral reforms and voter behavior. Electoral reforms intended to increase voter turnout are not self-actuating. That is, the implementation of these electoral reforms require agents political parties and their candidates to intervene between the opportunities created by state election laws to vote and eligible voters. The implementation of these reforms, however, relies on candidates strategic decisions to mobilize their supporters. Candidates and their parties know who their supporters are, the likelihood that they will ballot in an election, the costs of mobilizing these supporters and the probable impact voter mobilization will have on the outcome of an election. With this information, parties will make decisions about when and how to

8 mobilize their supporters. The efficacy of electoral reforms is thus dependent upon the decisions of candidates to use the electoral reform to mobilize supporters. This model identifies two central concepts, electoral reform and elite mobilization, and predicts that their interaction will yield significant and positive effects on voter turnout. We are agnostic about the direct independent effects of either electoral reforms or elite mobilization (as studied in previous research) on voter turnout. Instead our model predicts that the interaction of the availability of electoral reforms and the use of such reforms as part of elite mobilization efforts will influence voter turnout. The critical importance of elites' strategic decisions in this model demands elaboration (and caution) on at least two points (see Rosenstone and Hansen 1993). First, we assume (but do not further investigate) that candidates' and parties' assessments of the utility of incorporating electoral reforms as part of their campaign mobilization strategies are likely to vary with respect to electoral and political context, e.g., competitiveness, ballot length, and economic conditions. Verifying this assumption is likely an important step in better understanding the effects of electoral reforms in U.S. politics, though we do not seek to do that here. Second, we believe that party strategy requires leaders to focus on mobilizing their supporters, rather than turnout overall. Thus, aside from the contextual considerations noted above, party leaders also likely consider the personal characteristics and behaviors of their supporters in determining their most effective use of campaign resources. One of the key considerations following from this is the party leader's estimate of the probability of particular supporters turning out to vote in a particular election. While most electoral reforms are intended to benefit all voters, candidates are

9 interested in mobilizing only their supporters. Moreover, the only voters who independently take advantage of these reforms are those who are likely to vote (i.e., informed and partisan voters). So how do candidates use electoral reforms? This is a calculation that candidates and parties make as a function of who they need to turn out, who will vote for them, and who is unlikely to vote in the first place. Candidates utilize electoral reforms as part of their mobilization strategies when they identify among their supporters voters in need of incentives to vote. This information might lead to the differential use and effects of reform-oriented party mobilization for parties with distinctive electorates. The core electoral constituency of each political party is well known and allows us to anticipate the mobilization strategies of each party. Republican candidates draw from a middle and upper class Anglo voter base located in suburban and rural communities. The voter base for Democrats is comprised of lower socio-economic status voters, minorities (African- American and Hispanic), and progressive/liberal Anglos located in urban areas of the state. Core Republican supporters are more likely to vote than their Democratic counterparts, largely because Republicans are disproportionately wealthier, older and better educated than Democrats. There is evidence in the literature to support our observation that partisans are more likely to mobilize only their core constituents. Huckfeldt and Sprague (1992) report party mobilization efforts are targeted to those voters who are consistent supporters of the party. Prior participation in a party s primaries is a significant and positive predictor that an individual will be contacted to vote by their respective party in a general election

10 Huckfeldt and Sprague also observe a strong and positive class bias in Republican party contacting: the probability of being contacted by the Republicans is dramatically affected by class and party (75). This suggests that core Republican constituents are significantly more likely than core Democrats to be wealthier and better educated. The lower status of core Democrats would suggest a greater need and likelihood for Democrats to mobilize their core supporters. Consequently we expect that the Democrats are more likely than Republicans to incorporate voter mobilization in their electoral strategies and utilize changes in the electoral laws (e.g., early voting) to effectively mobilize their core constituencies. We do not expect to observe higher voter turnout among Republican voters even when Republican party leaders pair their mobilization efforts with new electoral reforms. One logical consequence of a model that emphasizes elite strategy as mediating the effects of electoral reforms is that elite strategy will also determine whether such reforms are associated with changes in the class composition of the electorate. Most research on elite strategy suggests that party elites focus on known voters and their own party members (Herrnson 2002). This focus on existing voters and partisans might help to explain why previous research has found limited effects of electoral reforms on who votes. We argue that it is impossible to hypothesize what compositional effects any given electoral reform will have without accounting for the specific strategic decisions of elites. We examine whether in the presence of partisan mobilization activities electoral reforms affect the incidence of voting among the resource poor more than the resource rich. We thus consider the possibility that the null findings on changes in the

11 representativeness of the electorate result from past research failing to incorporate party mobilization strategies in analyzing the effect of electoral reforms. To test this elite-oriented model of the effects and effectiveness of electoral reforms, we focus specifically on early voting. We hypothesize: The likelihood an individual will vote is unrelated to the availability of early voting in the individual s county of residence. The likelihood an individual will vote is unrelated related to partisan efforts to mobilize voters in a county. The likelihood an individual will vote increases when the availability of early voting is used in Democratic partisan efforts to mobilize voters in a county. The effect of early voting electoral reforms on the incidence of voting in the presence of partisan mobilization activities is disproportionately greater for resource poor individuals than resource rich individuals. Research design, measures and data The 2002 Texas statewide election provides a rich setting in which to expect significant efforts on the part of one or both parties to mobilize their core supporters and target groups. In this election the Democratic candidates for governor and senator were a Hispanic and African-American male, respectively. Their Republican opponents were both Anglo males. Since 1988 voters in Texas have had the opportunity to cast their ballots in-person up to three weeks before a general election. 1 Texas voters do not have to establish any prior condition (e.g., illness, age, travel out of the state, etc.) in order to cast their vote

12 before the first Tuesday in November. Moreover, voting is conducted in-person at a number of satellite sites, including nontraditional balloting sites such as supermarkets, convenience stores and mobile voting sites. The incidence of early voting in Texas has averaged 30 percent of the total vote cast per election since 1991. The state sets a minimum number of early voting sites and hours of operation that each county must meet during the three-week period prior to each election. Counties are not prohibited from having more than the minimum number of early voting sites. To test our hypotheses we require a database that allows us to explain individual electoral participation as function of individual level factors (i.e., social-demographic traits and attitudes) and political characteristics including party campaign activities and election procedures. A sample of registered voters in Texas (N=1,019) 2 was interviewed in September 2002. Respondents were asked about their intention to vote in the November 2002 election, as well as standard demographic and attitudinal questions. After the November election the voting history of each respondent was annotated from his or her county board election. This information includes whether the respondent voted in the November 2002 election and how they voted: in-person on election day, in-person early or by mail. Shortly after the November 2002 election we surveyed the county party chairs in all 254 Texas counties. The survey instrument queried each party chair about their efforts to mobilize target voter groups for turnout in the November 2002 election. Party chairs were specifically asked about the party s efforts to register voters before early voting and election day and whether they encouraged mail-in and in-person early voting

13 among their supporters and selected groups of voters. Additional questions were asked about spending and time dedicated to different types of mobilization activities. We also collected information on the number of early voting sites operated by each county and their hours of operation Partisan early voting mobilization activities Party chairmen in each county were queried about their efforts to mobilize voters in their county for early voting. Specifically each chair was asked: Did your organization provide voters with transportation to the polls during early voting? (Yes=1, No=0) Did you organization encourage in-person early voting as part of its campaign strategy in this year s election? (Yes=1, No=0) How many weeks prior to election day on November 5 did your organization begin the encouragement of in-person early voting? (1=1 week, 2=2 weeks, 3=3 week or 4=4 weeks) Did the Democrat (Republican) party chair in this county encourage in-person early voting as part of their campaign strategy in this year s election (1=Yes, No=0) What activities did your organizations utilize to encourage by mail early voting: Distributed filled out registration forms (1=Yes, 0=No). Advertising in Spanish and English (1=Yes, 0=No). Social activities to promote early voting registration (1=Yes, 0=No). Promotions after religious activities (1=Yes, 0=No). Other, unspecified activity (1=Yes, 0=No).

14 A scale score of early voting mobilization activities was constructed from the above questions for each county party organization. The score is the summed responses to each question and ranges from zero (no activities performed in the 2002 election) to 12. The Cronbach Alpha for the early voting scale is.674 for the Democratic party and.700 for the Republican party. County level opportunities for early voting Opportunities for individuals to vote in-person before election day are function of the number of early voting sites in each county and each site s hours of operation during the three weeks before election day. Under state local there is a minimum number of early voting sites each county must operate before election day. Each site must be open a minimum of six hours per day, though there is no limit on the number of hours a site can be opened for in-person voting. Our measure of in-person early voting opportunities is the total number of hours in-person early voting sites were open in each county. 3 Control variables Democratic candidates are thought to be advantaged by greater voter turnout. This is because core Democratic supporters, including the African-Americans, Hispanics and resource-poor eligible voters are more likely to be infrequent voters and the object of Democratic mobilization activities. 4 The partisan competitiveness of the county should increase the likelihood that contending parties believe their mobilization efforts will be efficacious. Strong partisans 5, wealthier 6, older, and better educated voters 7 and voters with a strong interest in and knowledge about politics should be significantly selfmotivated to vote. Partisan competitiveness is scored as 1.0 minus the absolute difference between the normal Democratic and Republican vote in each county. 8

15 Knowledge and interest in politics was measured with six questions about national and Texas state politics. 9 Findings Our survey of party officials indicates that both parties took significant steps to mobilize their supporters through early voting opportunities in their respective counties (see descriptive appendix on county chair survey). Nearly two-thirds of all county chairs reported they encouraged their supporters to vote early. Democrats, however, were more likely to engage in efforts to mobilize their supporters through early voting than Republican party chairs. Approximately a third of the Republican party chairs reported their organization provided voters with rides to the polls for early voting. Nearly half (47%) of the Democratic party chairs provided their supporters with transportation to vote early. In every instance the Democratic party chairs reported a higher incidence of mobilizing their supporters for early voting. Democrats were more likely to report distributing registration forms for early voting, promoting early voting after religious services, at social gatherings and, advertising early voting in Spanish language media. This condition is clearly reflected in a higher scale score for Democratic early voting mobilization. Most importantly the findings from our survey of party chairs in Texas demonstrates that leaders of both parties, but particularly Democrats, took significant and non-trivial steps to use early voting as a means of mobilizing their core supporters. The fact that Democrats were more likely to use early voting to turnout their supporters might

16 be indicative of the lower socio-economic status of core Democrats and the party s need to assist their supporters to the polls. We estimate a model of validated voter turnout in the November 2002 election (1=voted in the November 2002 election, 0=did not vote in the November 2002 election) for voters by their partisan affiliation. Our assumption is that county party activists most likely to target members of their own party (i.e., self-identified Democrats and Republicans) and refrain from focusing mobilization activities on independents and other unaffiliated voters. Parties are only interested in maximizing voter support for their candidates, not voter turnout. Parties are risk aversive; they prefer to minimize turnout rather risk turning out a voter who might ballot for their opponent. Consequently parties focus their mobilizing activities on core supporters including self-identified partisan supports and demographic groups who share the party s policy preferences. Our model of turnout includes an index of partisan early voting mobilization activities, the number of hours of early voting in a respondent s county and the interaction of partisan mobilization activities and hours of early site operation in the county. Additional controls are included for education, age, family income, political knowledge strength of partisanship and race/ethnicity (i.e., Hispanic and African- American) Tables 1-3 report the logit regression coefficients for several models of voter turnout in the 2002 Texas election by partisan affiliation of the respondent. While the statistical significance of the coefficients can be used to test our hypotheses, logit coefficients are not easily interpreted. Consequently, the magnitude and significance of the effect of any independent variable on voting is determined by calculating the change

17 in the probability of voting in the 2002 election across the range of values for our main independent variables, controlling for all other independent variables at their mean values. We use CLARIFY (Tomz, Wittenberg, and King 1999) to determine the substantive significance of the changes in these probabilities. Opportunities for early voting Consistent with our own hypothesis and the findings of previous researchers we fail to confirm a strong positive relationship between the number of hours of early voting in a county and the likelihood a voter, Democrat, Republican or independent, voted in the 2002 Texas election. Among Democrats there is a positive but weak relationship between the likelihood a voter participated in the 2002 election and opportunities in their county to vote early (P <.08). For Republicans this relationship is positive but statistically insignificant and for independent voters the relationship is negative but statistically insignificant. [table 1 about here] Party early voting mobilization activities The efforts of each party to mobilize their supporters through early voting have an insignificant effect on the likelihood that supporters of each party balloted in the 2002 election. Furthermore neither each party s efforts to mobilize voters for early voting affected the likelihood an independent voted in the 2002 Texas election. [table 2 about here] Early voting opportunities and party mobilization activities Consistent with our thesis there is evidence that the interaction between only Democratic party efforts to mobilize voters through early voting and the opportunities to

18 vote early is positively related to the voter participation in the 2002 election. Republican efforts to mobilize their supporters in the presence of strong opportunities for early voting has a statistically insignificant effect the likelihood a Republican will ballot. The inclusion of the interaction between Democratic mobilization activities and opportunities to vote early significantly increases the predictive power of our model of validated voting from.109 to.233. 10 [table 3 about here] The magnitude of Democratic efforts to mobilize their supporters through early voting paired with greater hours of early voting in a county is modest. There is only a 4 percent change in the probability that a Democrat voted in the 2002 between one standard deviation below and above the mean value for this interaction. 11 Moreover, there is only.37 probability a Democrat participated the 2002 election in a county with both strong party mobilization activities and substantial opportunities to vote early compared to only a.27 probability that a Democrat voted in county with substantially less Democratic effort to mobilize partisan supporters and hours of early voting operations. In table 3 the coefficient for the hours of early voting represents the effect of early voting opportunities on individual voter participation in counties where partisan mobilization efforts were negligible (i.e., -1 standard deviation below the mean). For Democrats (model 1) this effect is significant and negative. This finding should not be interpreted to mean that greater hours of early voting decreases Democratic voter turnout. To the contrary, this finding is due to the relationship between partisan electoral support in a county and the incidence of hours of early voting.

19 Democrats in Texas and other states promoted early voting as a means of enhancing turnout among their core and resource-poor constituents (Stein and Garcia- Monet 1997). Not surprisingly it has been Democrats who have lobbied for more opportunities to vote early. The correlation between the normal Democratic vote in a county and the hours of early voting is r=.198 (p<.01). The correlation between the normal Republican vote in a county and the hours of early voting is r=.205 (p <.01). Because Democratic core voters are less likely to vote the Democratic Party has worked to create and expand early voting opportunities. Unfortunately without an accompanying effort to mobilize their supporters through early voting Democratic voters are not likely to benefit from opportunities to vote early. To the contrary, early voting is not selfactuating and Democratic voters remain victims of their resource-poor status when their party and candidates fail to defray the costs of voting. It is possible that the same persons and/or political parties in each county are responsible for determining the number of early voting sites and hours of operation as well as partisan efforts to mobilize voters for early voting. Texas state law requires that the county clerk determine the location and hours of operation for early voting. County clerks are elected in partisan elections. It is possible that county clerks confer with their party chairs when choosing early voting sites and hours of operation. If this is true our estimates of the impact of early voting and party mobilization on turnout may be biased. We have investigated this possibility and found it to be unlikely. First, if this condition is present we would expect a strong and positive relationship between mobilization efforts and opportunities for early voting in each county. This is not observed. We have regressed the number of hours of early voting in each county on Democratic and

20 Republican efforts in each county to mobilize voters. Both estimates are statistically insignificant. Second, we would expect that over time the opportunities for early voting would vary within counties as a function of changing political conditions. The correlation between the number of hours of early voting in each county in our sample in 1996 (Leighley 2001) and 2002 is.956, sufficiently large to indicate little variation in early voting opportunities over time. We speculate that state statutes (Stein and Garcia- Monet 1997) regulating early voting account for most of variation in early voting opportunities and that partisan influences are constrained by these requirements. 12 Electoral reform and class inequalities in turnout To test our final hypothesis that early voting reforms in the presence of partisan mobilization have stronger effects on resource poor individuals and thus reduce the class inequalities in turnout we estimate the same models reported above including interaction terms for education and income. Our analysis follows Nagler s (1991) strategy of using interaction terms consisting of class and various electoral reforms to identify whether the effect of an electoral reform varies across class. We thus add to the models above a three-way interaction term include education, hours of early voting and partisan mobilization effort (Democrat and Republican). We estimate a similar model for income. The direct effects for the three variables remain in the model as do the other control variables reported above. A positive and significant coefficient for our interaction terms would suggest that partisan implementation of early voting has a strong effect on the poor and less educated. Table 4 reports the coefficients for our several interaction terms. Consistent with our previous findings only among Democratic voters do we observe a significant increase

21 in voting among low income voters when Democratic mobilization efforts are matched with significant opportunities to vote early. We do not, however, observe the same relationship for education. All of the coefficients for this measure of class are statistically insignificant. We accept, albeit in qualified manner, the hypothesis that Democrats efforts to mobilize their supporters in counties rich with opportunities to vote earlier, reduces class disparities in voter turnout. At first glance the results in Table 4 would appear to suggest that Republican efforts to mobilize low-income Republicans is not effective. However, when we compare the coefficient representing the effect of mobilization on low-income Republicans to that for Democrats we find that they are identical. The difference lies in the size of the standard errors in the Republicans column being twice that of the Democratic column. We believe this difference can be explained by two factors. First, as discussed earlier, lower socioeconomic status citizens tend to support the Democratic party. Therefore, when Democrats attempt to mobilize low-income identifiers, their pool of potential mobilization targets is much deeper than Republicans. For example, in our sample only 65 respondents with family incomes less than $35,000 a year identify with the Republican Party while 116 respondents in this income category identify with the Democratic Party. Because the Republicans targeted population is only about half the size of that for the Democrats, the standard error of these estimates is much larger. Second, Republicans targeting low-income voters are disadvantaged because most low- income Republicans live in rural or low population density areas while most lowincome Democratic identifiers live in urban or high population density areas (approximately 57% in both cases). This greater dispersion of low-income Republicans

22 coupled with a population almost half that of Democrats increases the cost and likelihood of error for Republican elites wanting to mobilize this population. These two observations regarding the larger standard errors associated with Republican mobilization suggest that Republican party officials are unlikely to gain a significant electoral advantage from mobilizing their low-income supporters through early voting. Conclusion Our empirical evidence supports our central thesis: without the efforts of political parties and their candidates, electoral reforms are likely to continue to have a marginal effect on voter turnout. Yet our thesis holds for only one political party and their partisan supporters. Only Democratic mobilization in the presence of strong opportunities to vote early (i.e., hours of early voting operations in a county) has a significant and positive effect on voter turnout. We find no evidence that Republican partisan mobilization activity in the presence of strong opportunities to vote early increases the likelihood that Republican partisans voted in the 2002 election. Thus, an important modification to our original thesis is that the effectiveness of electoral reforms is dependent not only on whether political elites choose to take advantage of such rules in developing their electoral strategies, but also on the individual s probability of voting absent any mobilization. Our data indicates a negative relationship between opportunities to vote early and the probability a Democratic voter participated in the 2002 election. The demographics of core Democratic supporters make them less likely to vote. Moreover, these voters are likely to reside in counties where efforts to increase voter turnout through early voting are more pronounced, producing the negative relationship we observe between voting and hours of early voting

23 poll operations for Democrats. Democrats may have worked hard to pass early voting legislation, but their continued efforts are necessary for the party to enjoy the benefits of this change in electoral rules. On the other hand, the demographics of Republican partisans make them likely voters. Consequently Republican mobilization efforts are not likely to significantly raise this already high intercept on voting. These analyses thus support the conventional wisdom, that Republican constituents have sufficient individual resources (e.g., income, time, efficacy) that partisan subsidies that result from mobilization efforts are not likely to significantly affect their likelihood of voting. More broadly, these findings help to explain the null findings of previous research on voting and registration reform. This earlier research used rather naïve models of mass political participation, where the assumption was made that a change in electoral rules would have a direct, significant effect on individual behavior. We believe the findings presented in this paper on early voting call into question this modeling assumption regardless of the type of reform one considers. If elites do not take advantage of new institutional rules, we should not expect to have them affect mass behavior. This research also underscores the importance of political elites in understanding who participates and the role of parties as linkage institutions. We thus echo Rosenstone and Hansen s (1993) claims that mass political participation reflects the strategic actions of political elites: elites target who they want to participate when, and the mass public responds.

24 Table 1 Logit estimates for voter turnout in the 2002 Texas Election (Robust Std. Error) Variable Model 1 (Dem) Model 2 (Rep) Model 3 (Ind) Constant -6.34*** -3.27 7.95*** (2.32) (2.28) (2.87) Education.292*** -.021.190 (.018) (.096) (.145) Competition 4.05 2.235 4.48 (2.99) (2.69) (.320 Family income.002 -.008**.017** (.005) (.002) (.005) Age.212**.264***.548*** (.107) (.076) (.140) Political Knowledge.118.190**.219* (.077) (.066) (.123) Black.014.322 -.027 (.319) (.931) (.199) Hispanic -.412-1.91 -.026 (.555) (.930) (.882) Strength of Party ID -.107.301 - (.292) (.212) - Early voting Hours.0002* (.0001).0001 (.0001) -.00007 (.0001) Pseudo R 2.120.094.210 N 199 231 134 Wald Chi 2 67.11 106.1 28.2 *P <.1 **P <.05 ***P<.01

25 Table 2 Logit estimates for voter turnout in the 2002 Texas Election (Robust Std. Error) Variable Model 1 (Dem) Model 2 (Rep) Model 3 (Ind) Model 4 (Ind) Constant -1.81* -.630-3.21*** -4.38*** (1.06) (.541) (1.22) (.917) Education.264*** -.056.126.305** (.074) (.119) (.171) (.126) Competition -6.08* -1.01-3.78-3.21* (3.27) (2.34) (3.26) (1.81) Family income.005 -.007**.016***.020*** (.004) (.002) (.005) (.005) Age.166*.172*.502***.550*** (.097) (.092) (.141) (.148) Political Knowledge.150* (.080).217*** (.001).252* (.151).190 (.139) Black -.023.048 -.031.149 (.323) (.941) (.273) (.233) Hispanic -.465-1.68* -.026 -.401 (.595) (.925) (.913) (.823) Strength of PID.065.274 - - (.373) (.244) - - Party Mobilization.068 -.006 -.033 -.011 (.091) (.044) (.128) (.032) Pseudo R 2.109.058.204.220 N 182 176 121 106 Wald Chi 2 54.63 114.2 26.74 50.62 *P <.1 **P <.05 ***P<.01

26 Table 3 Logit estimates for voter turnout in the 2002 Texas Election (Robust Std. Error) Variable Model 1 (Dem) Model 2 (Rep) Model 3 (Ind) Model 4 (Ind) Constant -7.55***.967 5.90*** 5.04*** (2.92) (3.19) (3.31) (2.80) Education.230*** -.015.137.286** (.081) (.098) (.163) (.138) Competition 5.29** -2.82 4.63.329 (2.76) (3.93) (3.54) (3.38) Family income.004 -.008***.017***.020*** (.005) (.001) (.005) (.005) Age.186**.217***.514***.567*** (.095) (.092) (.146) (.151) Political Knowledge.155** (.974).200*** (.052).260* (.146).210 (.150) Black.096.047 -.029 -.003 (.352) (.972) (.224) (.176) Hispanic.346-1.76 -.038 -.335 (.500) (1.07) (.929) (.750) Strength of PID.007.317 -.001.00006 (.316) (256) (.002) (.0003) Early voting Hours -.004** (.001).0002 (.0004).0001 (.0001) -.001 (.002) Party Mobilization.102 (.110) -.121 (.179) -.167 (.133) -.165 (.106) Mobilization X Hours.0003** (.0001).00002 (.00005).00003 (.00004).00008 (.0001) Pseudo R 2.233.094.232.214 N 182 175 105 121 Wald Chi 2 71.9 94.5 76.1 38.6 *P <.1 **P <.05 ***P<.01

27 Table 4 Logit estimates for income and education interactions (Robust Std. Error) All Respondents Income Interactions Income * Mobilization * Hours.00000003 (.00000004) Education Interactions Education*Mobilization*Hours -.00000002 (.0000001) Dem Respondents.00000009 (.00000005) Rep Respondents.000000009 (.000000009) -.00002.0000003 (.00002) (.0000004) Though not reported all estimates included controls for Education, Competition, Family Income, Age, Political Knowledge, Black, Hispanic, Strength of Party ID, Early Voting Hours, and Party Mobilization.

28 Appendix: Proportion of early voting mobilization activities performed by county party organizations Did your organization provide voters with rides to polls for early voting? Did your organization encourge in-person early voting? Weeks of early voting Did your opposing party organization encourage early voting? Did you distribute registration forms for early voting? Did you advertise in spanish and english for early voting? Did you use social activities to encourage early voting turnout? Did you promote early voting after religious activities? Other unspecified activities to promote early voting? Early voting mobilization score Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total Dem Rep Total N Mean Std. Deviation 127.4646.50072 140.3214.46870 267.3895.48856 127.6378.48254 140.6143.48851 267.6255.48491 127 2.7008 1.59300 140 2.9214 1.53645 267 2.8165 1.56321 127.6142.48872 140.5571.49851 267.5843.49377 127.2205.41621 140.1500.35835 267.1835.38782 127.3386.47510 140.1714.37824 267.2509.43437 127.2205.41621 140.1429.35118 267.1798.38472 127.1181.32402 140.0571.23295 267.0861.28110 127.2126.41077 140.2286.42142 267.2210.41568 127 2.9291 2.42730 140 2.3286 2.24552 267 2.6142 2.34862

29 References Adams, William C. and Dennis J. Smith 1980. Effects of Telephone Canvassing on Turnout and Preferences: A Field Experiment. Public Opinion Quarterly 44: 389-395. Aldrich, John H., and Dennis M. Simon. 1986. Turnout in American National Elections. in Research in Micropolitics, ed. Samuel Long. Greenwich, Conn. JAI Press. Berinsky, Adam J., Nancy Burns and Michael W Traugott. 2001. WHO VOTES BY MAIL? Public Opinion Quarterly: 65:178-198. DeNardo, James. 1980. Turnout and the Vote: The Joke s on the Democrats, American Political Science Review. 74:4-6-420. Gerber, Alan S. and Donald P. Green. 2000. The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout: A Field Experiment. American Political Science Review 94: 653-664. Huckfeldt, Robert and John D. Sprague. 1992. "Political Parties and Electoral Mobilization: Political Structure, Social Structure and the Party Canvass." American Political Science Review 86: 70-86. Knack, Stephen. 1995. Does Motor Voter Work? Evidence from State-Level Data. The Journal of Politics 57: 796-811. Lijphart, Arend. 1997. Unequal Participation: Democracy s Unresolved Dilemma. American Political Science Review 91: 1-14. Leighley, Jan E. 1995. Attitudes, Opportunities, and Incentives: A Field Essay on Political Participation. Political Research Quarterly 48:181-209. Leighley, Jan E. 2001. Strength in Numbers? The Political Mobilization of Racial and Ethnic Minorities. Princeton, NJ: Princeton University Press. Leighley, Jan E. and Jonathan Nagler. 1992. " Individual and Systemic Influences on Turnout: Who Votes? 1984." Journal of Politics 54: 718-740. Nagler, Jonathan 1991. The Effect of Registration Laws and Education on U.S. Voter Turnout. American Political Science Review 85: 1393-1405.

30 Oliver, J. Eric. 1996. Who Votes at Home?: The Influence of State Law and Party Activity on Absenteee Voting and Overall Turnout. American Journal of Political Science 40:498-513. Patterson, Samuel C. and Gregory Caldeira. 1982. Contextual Influences on Participation in U.S. Legislative Elections. Legislative Studies Quarterly 7: 359-381. Rhine, Staci L. 1995. Registration reform and turnout change in the American states. American Politics Quarterly 23: 409-437 Rosenstone, Stephen J., and John Mark Hansen. 1993. Mobilization, Participation, and Democracy in America. New York: Macmillan Publishing Company.. Squire, Peverill, Raymond E. Wolfinger and David P. Glass. 1987. Residential Mobility and Voter Turnout. American Political Science Review, 81: 45-66. Stein, Robert M. and Patricia A. Garcia-Monet. 1997. Voting Early, but Not Often. Social Science Quarterly 78:657-671. Stein, Robert M. 1998. Early Voting. Public Opinion Quarterly 67:205-217. Tomz, Michael, Jason Wittenberg, and Gary King,. 2001. CLARIFY: Software for Interpreting and Presenting Statistical Results. Version 2.0. Cambridge, MA: Harvard University, June 1. <http://gking.harvard.edu/>. Wielhouwer, Peter W. 1995. Strategic Canvassing by the Political Parties, 1952-1990. American Review of Politics 16: 213-238. Wielhouwer, Peter W. and Brad Lockerbie. 1994. "Party Contacting and Political Participation, 1952-1990." American Journal of Political Science 38: 211-29. Wolfinger, Raymond E. and Steven Rosenstone. 1980. Who Votes? New Haven, CT: Yale University Press.

31 Endnotes 1 Six other states including Colorado, Oregon, Tennessee, Virginia, etc. allow their voters to vote in-person before election day. 2 Telephone interviewers were conducted with 1,019 registered voters in 110 Texas counties. Interviews were conducted by the University of Houston Center for Public Policy between September 3-10 and September 12-15. The response rate for the survey is 38% and the error rate is +/- 3% 3 Several alternative measures were tested including the per capita hours of in-person early voting in each county and the hours in-person voting sites were open per square mile in each county. Both measures account for the possibility that opportunities to vote early are constrained by congestion or distance. Our findings do not vary significantly with these alternative measures of early voting opportunities. 4 See DeNardo (1980) for a contrary view. 5 Strong partisans, both Democrats and Republicans are code 1, respondents who are leaning Democrats, Republicans, Independents and other partisan affiliations are coded 0. 6 Total family income was recorded. 7 1-less than high school, 2=high school graduate, 3=some college/technical school, 4=college graduate, 5=post-graduate education. 8 The normal vote is the mean percent of the two party vote each party received in elections for governor, senator and Lt. Governor in 1994, 1998 and 2002. 9 The questions (answer) included: How may years is the Texas Governor s term of office (four); How often does the Texas Legislature in Austin meet (every two years); Which party has the most members in the Texas State House of Representatives (Republican). Correct answers were score 1 and incorrect answers were score 0. A political knowledge scale is constructed from the sum of correct answers; How is the majority Leader of the U.S. Senate (Trent Lott); How much of a majority is required for the U.S. Senate and House of Representatives to override a presidential veto (two-thirds); Which political party has the most members in the U.S. House of Representatives (Republican). Cronbach Alpha=.781. 10 The incremental F-test for the difference between these two pseudo-r 2 s is 4.5, p <.001. 11 The standard deviation is calculated for the interaction term i.e., the product of early voting hours in each county and report party mobilization activities.

12 There is another important reason why the number of early voting sites and hours of operation do not vary over time. Voters expect to ballot at the same site (and during the same days and hours) across elections. County clerks avoid changing voting sites over time in an effort to maintain turnout and satisfy the county electorates. 32