Mobilized by Direct Democracy: Short- Term Versus Long-Term Effects ...

Mobilized by Direct Democracy: ShortTerm Versus Long-Term Effects and the Geography of Turnout in Ballot Measure Elections n Joshua J. Dyck, University at Buffalo, The State University of New York Nicholas R. Seabrook, University at Buffalo, The State University of New York Objectives. A number of recent studies find that direct democracy increases voter turnout. In this article, we ask: Who does direct democracy mobilize to vote and how are they mobilized? We distinguish between long-term and short-term effects on voter turnout, noting that much of the current literature has focused on participatory theory. Methods. Our research design harnesses the power of geographic information systems and examines turnout in special initiative-only elections using registered voter lists. Our model draws on individual and Census tract data, incorporated using a hierarchical generalized linear model. Results. The findings demonstrate how partisan context mitigates the potential for direct democracy to mobilize from the middle, and clarifies the dominance of short-term as opposed to long-term effects in increasing voter participation in ballot initiative elections. Conclusion. Mobilization via direct legislation occurs mostly because voters are actively mobilized by partisan campaigns, not because of an increase in participatory fervor.

The people are involved in public affairs by the conflict system . . . In a free political system it is difficult to avoid public involvement in conflict; the ordinary, regular operations of the government give rise to controversy, and controversy is catching. E. E. Schattschneider, The Semisovereign People (1960:135)

Recent empirical investigations have produced a single, widely replicated result: the presence of ballot initiatives in elections, particularly mid-term elections, appears to increase turnout rates by as much as 6 percentage points, depending on the number and nature of issues on the ballot (see n Direct correspondence to Joshua J. Dyck, Political Science Department, 520 Park Hall, University at Buffalo, SUNY, Buffalo, NY 14260 [email protected]. We are willing to share data and coding information with those wishing to replicate the study. This research was generously funded by the Baldy Center for Law and Social Policy at the University at Buffalo, SUNY. Special thanks to Fred Boehmke, James Gardner, Lynn Mather, and anonymous reviewers for helpful comments and suggestions. We alone are responsible for all the analyses, conclusions, and any errors present in the article. A previous version of this article was presented at the Annual Meeting of the American Political Science Association, Chicago, August 2007.

SOCIAL SCIENCE QUARTERLY, Volume 91, Number 1, March 2010 r 2010 by the Southwestern Social Science Association

Mobilized by Direct Democracy

189

Smith and Tolbert, 2004; Smith, 2001; Lacey, 2005). Where the literature is markedly more incomplete is in understanding the theoretic mechanism by which ballot measures turn a nonvoter into a voter. Participatory democratic theories have undergirded most of the research that has concluded that there are secondary effects to direct legislation like ballot initiatives and referendums. We present an alternative theoretic perspective to what motivates increases in turnout in ballot measure elections that is based on conflict, context, and the role of party in structuring and mobilizing voters in U.S. politics. We test our theoretic perspective using a unique data set that overcomes the problems of studying turnout effects when myriad candidate races and ballot questions appear on a single ballot in typical midterm and presidential year elections and includes valid voter responses for registered voters in California and Oregon. While our results demonstrate evidence of both participatory democratic theory (long-term) effects and partisan mobilization (short-term) effects, the magnitude of the partisan mobilization effect substantively swamps the participatory theory effect.

Institutions, Context, and Turnout

A great deal of recent scholarly attention has been paid to the way the presence of direct forms of adopting legislation, such as the ballot initiative and referendum, can create positive democratic citizenship externalities. Exposure to the initiative process has been demonstrated to increase external efficacy (Bowler and Donovan, 2002; Hero and Tolbert, 2004; Mendelsohn and Cutler, 2000), civic engagement (Smith, 2002; Smith and Tolbert, 2004), happiness (Frey and Stutzer, 2002), political awareness (Mendelsohn and Cutler, 2000; Smith, 2002; Smith and Tolbert, 2004), and voter turnout (Smith, 2001; Tolbert, McNeal, and Smith, 2003; Tolbert and Smith, 2005; Lacey, 2005).1 Despite the fact that evidence shows a turnout effect in direct democracy environments, the question remains: What are the characteristics of mobilized voters? In other words, which voters are turning out who might not have otherwise, and why? This article argues that direct legislation might stimulate voter turnout through two distinct mechanisms that have been largely muddied in the current literature. The first involves long-term socialization effects whereby, over time, citizens who are exposed to the initiative process regularly are in effect ‘‘educated by initiative,’’ and thus are likely to exhibit higher turnout rates in elections where initiatives are on the ballot than citizens in noninitiative states. This argument is drawn from Smith and Tolbert’s (2004) 1 Recent replications by Schlozman and Yohai (2008) and Dyck and Lascher (2009) have demonstrated that the links between ballot measures and efficacy and knowledge are not particularly robust. Additionally, Dyck (2009) demonstrates that ballot measures tend to decrease levels of trust in government.

190

Social Science Quarterly

comprehensive study of the effects of long-term exposure to ballot initiatives on citizens and political organizations in the U.S. states and is based on their empirical finding that voters in initiative states exhibit higher levels of a variety of positive democratic citizenship characteristics. The second involves short-term campaign and social context effects, whereby voters confronted with high stimulus ballot initiatives are either mobilized or demobilized by discussion and interpersonal relationships within their neighborhoods and social networks and exposure to political campaigns. These effects are contingent on the saliency of the initiatives on the ballot, the partisan homogeneity or heterogeneity of individual voters’ social networks, and the existence of strong partisan campaign messages and elite cues in the informational environment surrounding the election. This argument is rooted in the turnout literature, which emphasizes the partisan character of voter mobilization effects in U.S. politics (Rosenstone and Hansen, 1993; Campbell, 1987), and in the social context literature, which has consistently demonstrated the potential for political participation to be mitigated or enhanced by social interaction within individual voters’ communities and social networks (Huckfeldt, 1979; Mutz, 2002). Curiously, the partisan constraint of voting has often been talked about alongside educative effects (Smith and Tolbert, 2004:Ch. 6), but has not been directly connected as a competing theory in discussions of what motivates turnout. Direct Democracy and Turnout—Long-Term Effects Existing research has argued that direct democracy creates an educative environment that empowers citizens and that, over time, leads to greater participation. Voters are thereby socialized by exposure to the institution of direct democracy and as a result are motivated to turn out at higher rates in initiative states than in noninitiative states (Smith and Tolbert, 2004). This disparity has been growing over time as greater numbers of citizens become educated by the initiative process (Tolbert, Grummel, and Smith, 2001). This empirical result has its theoretical footing in the work of Barber (1984) and Pateman (1970), who argue that given more opportunities to participate, people will learn to become better democratic citizens. A counter to this argument, however, is that citizens favor greater opportunities to participate in principle, but do not want to put forth the time and effort to meet these goals (Hibbing and Theiss-Morse, 2002). Direct democracy imposes greater costs on voters, thus mobilizing some, while potentially demobilizing lower resource voters, who now face greater costs of time in voting and information gathering in preparing to vote. Nevertheless, the empirical evidence has consistently demonstrated that voters in initiative states have a propensity to turn out to vote at higher rates than voters in noninitiative states, an effect that survives the introduction of myriad controls for the many other factors that are known to be associated


191

with turnout. We hypothesize that direct democracy has a long-term effect on voter mobilization and turnout through its educative effects on the polity: as voters are exposed to the initiative process over numerous electoral cycles, they are socialized and educated to view the initiative as an increasingly legitimate and positive mechanism for political participation. Voting is a habit (Plutzer, 2002), and through repeated exposure to direct democratic environments, voters develop a habit of participating in ballot measure elections, a habit that over time should produce an aggregate increase in voter turnout, as more and more individual voters are exposed to the direct legislation. This effect should be evident both in initiative-only special elections and elections where initiatives are placed alongside candidates on the ballot. These long-term educative effects, we argue, are responsible for at least some of the increase in turnout that is attributed to direct democracy. H1: Voters socialized in states where they have been exposed to the initiative process should demonstrate a greater propensity to turn out in direct democratic elections than voters socialized in noninitiative states.

Direct Democracy and Turnout—Short-Term Effects Though these long-term effects, whereby voters are educated to participate by their exposure to direct democracy, clearly have some influence in mobilizing voters to turn out in response to the presence of ballot initiatives, we argue that they are not responsible for all or indeed most of the increase in turnout that has been demonstrated in the direct democracy literature. Another, perhaps more significant, part of the story is the potential for short-term factors within a specific electoral campaign to exert a similar influence. A crucial factor that conditions the potential for ballot initiatives to be used to mobilize voters is the saliency of the issues up for vote in a particular ballot measure election. Citizens in certain contexts exhibit a greater propensity for participating in politics than citizens in other contexts, even when controlling for other factors traditionally associated with political participation. We should therefore expect high-stimulus direct legislation elections to be associated with much more dramatic mobilization and turnout effects than low-stimulus elections, a finding that has been replicated with some frequency in the literature on direct democracy and turnout (Grummel, 2008; Lacey, 2005; Smith, 2001). H2: We should observe greater voter mobilization and turnout effects in highstimulus ballot initiative elections than in low-stimulus ballot initiative elections. Political scientists have been somewhat divided over the question of what types of voters are more likely to be mobilized by the presence of highly salient initiatives on the ballot. On the one hand, the progressive ideal of the disillusioned independent voter being brought to the polls by the

192


prospect of voting on issues rather than candidates has continued to gain traction in a number of recent studies. On the other hand, a great deal of research has continued to emphasize the highly partisan electoral environments that surround the most salient direct democratic contests. Though the idea of direct democracy attracting issue-oriented independent voters who would otherwise have stayed at home is appealing, it is very difficult to substantiate given what we already know about voter mobilization and turnout. Previous research has consistently demonstrated that independent voters are the least likely to turn out and the most difficult to mobilize across a multitude of electoral contexts. For example, Campbell (1987) demonstrates that the turnout surge in presidential elections is produced not by the mobilization of independents, as had previously been claimed, but by the greater participation levels of peripheral partisan voters. Rosenstone and Hansen (1993) dispute the traditional notion of the decision to participate in politics being driven by a simple weighing of the costs versus the benefits to each individual voter, an idea originally put forward by Downs (1957). They argue that the paradox of voting cannot be solved simply by examining the intangible benefits, since the costs of participation will always outweigh the benefits for the vast majority of individuals even when taking these into account. Instead, they demonstrate that the most important driving force behind political participation is mobilization: individuals are motivated to vote not based on the individual costs and benefits of the decision, but based on the extent to which they are mobilized by other external factors. The most important determinants of an individual’s decision whether to participate in a given election are whether or not he or she has been contacted by a political party, how close the election is, and the extent to which the individual can overcome rational ignorance by discussing politics with other individuals through interpersonal relationships. If we look at the mobilization efforts associated with ballot initiatives, we are led to the conclusion that partisan voters are more likely to be mobilized by the institution of direct democracy. First, the extent to which individuals are mobilized, either by being contacted by a political party or interest group or through discussions with other individuals, is very much dependent on saliency. Low-stimulus initiatives are likely to produce relatively minor mobilization effects, whereas high-stimulus initiatives are more likely to stimulate political discussion and to prompt more widespread voter contact on the part of organized interests. Second, given that the electoral environment surrounding highly salient ballot initiatives is often extremely partisan in nature,2 we should expect the majority of the mobilization effect to be among partisan voters. The fact that ballot initiatives have been shown to boost turnout in mid-term elections, but not in presidential years, lends 2 This is best understood by examining the literature on ballot initiatives and morality politics (Grummel, 2008; Haider-Markel and Meier, 1996).


193

further credence to the notion that it is peripheral partisan voters who are being mobilized by direct democracy. These voters who are mobilized by the greater electoral stimulus of the presidential election are the same voters who are most likely to be mobilized by the stimulus of a highly salient ballot initiative in a midterm election (Campbell, 1987). H3: Low-stimulus ballot initiative elections should produce relatively minor mobilization effects that are approximately equal for both independent and partisan voters. High-stimulus ballot initiative elections should produce much greater mobilization effects that are significantly larger for partisan voters and significantly smaller for independents. Finally, while scholars have focused on institutional, educative, and policy motivations, little attention has been paid to the importance of social context in conditioning political participation in initiative elections.3 Within the social context literature, there exists an established finding that communities and social networks act as socializing agents to activate citizens’ involvement in politics (Berelson, Lazarsfeld, and McPhee, 1954; Huckfeldt, 1979). Context can be a force in shaping political attitudes: social context exerts a more significant influence than the media on individual vote choice in presidential elections (Beck et al., 2002). Social context also affects participation through interpersonal relationships (Straits, 1990) and social networking (Mutz, 2002). Additionally, there is some evidence that residential mobility decreases turnout by removing a person’s social network (Highton, 2000). Research has demonstrated that neighborhood effects condition the impact of the factors traditionally associated with turnout, specifically that partisans are less likely to vote in areas where opposite partisans constitute a large majority, and in areas with very high levels of in-migration (Gimpel, Dyck, and Shaw, 2004). Those voters who live in homogenous communities whose partisan disposition matches their own are more likely to turn out than those who are dispersed within neighborhoods where they are among fewer like-minded individuals (Mutz, 2002). While the former are likely to have their political views reinforced by interaction within their social networks, thus increasing their political knowledge and personal efficacy (Gimpel, Lay, and Schuknecht, 2003), the latter are more likely to withdraw from politics entirely to avoid conflict and disagreement with those around them (Mutz, 2002; Be´langer and Eagles, 2007). Thus, despite the fact that ballot initiatives allow for more complex preference expression, partisan political contexts may act to stunt the growth of this expression. Furthermore, the highly partisan electoral environment surrounding the most salient initiative elections may further reduce the 3 The work of Hero and Tolbert (2004) demonstrates how racial/ethnic contexts and direct democracy interactively affect efficacy, and Campbell, Wong, and Citrin (2006) show how political contexts affect vote choice on racially charged ballot initiatives.

194


potential for the institution of direct democracy to appeal to nonpartisan voters. Therefore, to the extent that independents are mobilized or demobilized in initiative elections, we should expect more partisan neighborhood contexts to act as a demobilizing force against independent voters, while at the same time stimulating peripheral partisan voters who are already more likely to be mobilized to show up at the polls. H4: We should expect mobilization in ballot initiative elections to be conditioned by social context. Individuals in more homogenous communities should exhibit higher turnout rates than individuals in more heterogeneous communities.

Research Design

To empirically evaluate who is mobilized by direct democracy, we use data from a source that is previously untapped in the literature on ballot initiatives and turnout: the complete lists of registered voters maintained by individual states. With the passage of the Help America Vote Act in 2002, states have started to centralize their voter registration procedures, making complete state files more readily available and easier to access. This study takes advantage of these developments to analyze the factors affecting voter mobilization and turnout at the individual level.4 In most electoral scenarios, initiatives are part of a bundle of electoral contests on the ballot. Americans are routinely asked to vote in a series of candidate-centered elections at both the national and state levels. Whether in primary or general elections, initiatives are an addendum to the ballot, which is headed by some other candidate contest(s). To be sure, there are some states (e.g., Washington) that routinely hold ballot initiative elections in odd years, along with other states that frequently call special elections (e.g., Oregon). However, almost every study examining turnout has focused on the comparison of turnout between states that have (or frequently use) initiatives and states that do not have or do not frequently use initiatives; some of these studies focus on self-reported individual turnout behavior, while others make use of aggregate data. Part of the reason single-state studies have not been feasible is because of the conflation of initiatives with other electoral contests. When a single state with ballot initiatives has exceedingly high turnout (by historical standards) in an election year, it is difficult to parse out who the ‘‘new voters’’ are, and how they are mobilized; a ‘‘mobilized’’ voter may be responding to a particular candidate, a ballot measure, or perhaps both simultaneously. This study attempts to avoid the problem of trying to parse out the mobilization effects of direct legislation in 4 On the advantages of using voter files in state politics research, see Cooper, Haspel, and Knotts (2009).


195

situations where both initiatives and candidates are on the ballot by examining these effects in a setting where they can be isolated from confounding electoral factors: the initiative-only special election. Studying a series of special elections of varying levels of saliency will allow us to conduct individual-level tests of the hypotheses outlined in the previous section, thus providing greater insight into the mobilization effects of direct democracy in a variety of electoral contexts. We were able to obtain data regarding four special elections since 2002—one in 2005 in California, and three in Oregon between 2003 and 2004. A great deal has already been written about the 2005 California special election, and the particularly fierce campaign environment that surrounded it (see Baldassare and Katz, 2007). Briefly, eight initiatives were proposed, four of which dealt with Governor Schwarzenegger’s reform package that he had been unable to pass through the legislature. The other four dealt with competing prescription drug proposals, parental consent on abortion, and electricity regulation. Interest and spending were relatively well dispersed across all the measures and all eight measures failed.5 The 2005 California special election, therefore, was undoubtedly a very high-stimulus contest: campaign spending by both political parties and interest groups was extremely high for an initiative-only election, and the high-profile informational environment surrounding the election makes it a likely environment to uncover significant voter mobilization efforts. Our theory predicts that in a high-stimulus ballot initiative election such as this we should see significantly higher turnout than in comparable lower-stimulus elections, and we should expect the largest voter mobilization effects to be found among partisan voters. In contrast to this high-stimulus California special election, in roughly the same time period the State of Oregon had three special elections that demonstrated a significantly lower level of campaign activity and interest, and thus serve as lower-stimulus contrasts to the mobilization effects in California.6 There is also considerable variation between the three Oregon 5

Almost $300 million were spent by the pharmaceutical industry alone in opposing Proposition 79’s prescription drug plan, and supporting the competing plan contained in Proposition 78, which the pharmaceutical lobby itself had pushed to place on the ballot in an attempt to try and confuse voters into voting no on both. Business and labor organizations combined to spend over $100 million on Proposition 75 regarding union member campaign contributions, whereas Proposition 73 on abortion notification attracted just $7 million in total spending, and just $2.5 million were spent by the energy lobby in opposition to Proposition 80’s proposed electricity regulations (California Secretary of State, Elections and Voter Information (totals include independent expenditures)). The more than $300 million directly spent on the election, including advertising and administrative costs, makes it the most expensive in California’s history (MSNBC, 2005). Polling data released before the election suggested that Californians were not only interested in the reform package. Thirtyeight percent of respondents named a Schwarzenegger initiative as the one of greatest interest, while 12 percent of respondents named one of the other initiatives. Additionally, 11 percent said none, 8 percent said all, and 31 percent said don’t know/refuse (Baldassare, 2005). 6 In January 2003, the Oregon electorate voted down Measure 28, which proposed to increase income taxes by 1 percentage point. In September 2003, Oregon voters approved

196


contests, with two medium-stimulus elections conducted on the issue of whether the state should increase income taxes, and one low-stimulus election conducted on the issue of whether the state should incur general obligation debt for savings on pension liabilities. A cursory examination of turnout trends in these elections gives some weight to our characterization of their saliency. Oregon’s turnout in the 2006 midterm congressional election was 51.9 percent of the eligible electorate, while turnout rates on tax measures in January 2003 and February 2004 were at 49.3 percent and 45.2 percent, respectively (McDonald, 2002).7 In contrast, the September 2003 special election that asked voters to authorize the state to incur general obligation debt for savings on pension liabilities managed to generate a turnout rate of only 25.5 percent. Voter List Data and Design The data for this study come from the 2006 versions of the State of California’s voter list as well as the list for Multnomah County (Portland) in Oregon.8 Using voter lists for the study of turnout has some great virtues, but also limits. The single greatest advantage is that turnout is accurately recorded in the voter list—it is observed behavior and is therefore not plagued by the overreport problem inherent in surveys that ask individuals to self-report their vote history (Clausen, 1968; Anderson and Silver, 1986; Bernstein, Chadha, and Montjoy, 2001), or by the ecological inference problem inherent in designs that look at aggregate turnout to make inferences about individual-level behavior (King, 1997). Additionally, since we are computing statistical models on the population of registered voters, we have enough Level-1 observations to allow for meaningful clustering within communities. Given that research has shown that people are affected by neighborhood social networks and socialization, we are able to examine how social contexts condition the decision to vote. The greatest limitation of working with voter list data is that we cannot examine the same type of complex attitudinal questions available in most surveys. However, there is some useful information contained within the Measure 29, which authorized the state to incur general obligation debt for savings and pension liabilities. State Measure 30 was referred in February 2004 to a popular vote and was essentially a legislative ‘‘do-over’’ of Measure 30; once again, the measure failed. 7 The current rates are up to date on Michael McDonald’s website hhttp://elections.gmu.edu/index.htmli. 8 We obtained the entire Oregon voter list but, unfortunately, only Multnomah retained the special elections from 2003 and 2004 in its vote histories. Still, Multnomah is the largest county in the state and estimations were made with more than 500,000 Level-1 observations clustered in 170 Census tracts. While Multnomah is certainly not representative of the entire State of Oregon given the high degree of support for Democrats in the county, the electorate in this single county makes up approximately one-half of the entire electorate of the state. We have no a priori reason to expect that turnout effects would be markedly different in more Republican parts of the state, although we hasten to note that our data are not representative of Oregon on the whole.


197

voter file: both California and Oregon maintain party registration information, as well as past turnout history, age, and length registered to vote. In addition, the California voter list also contains data on each voter’s place of birth. With the use of geographic information systems, we are able to geographically identify the residence of each registered voter through geocoding, and use a spatial match to attach that resident to his or her Census tract, where a variety of socioeconomic and racial/ethnic information is available about the resident’s neighborhood. With individuals nested in Census tracts as the unit of analysis, we use hierarchical generalized linear modeling with a logit link function to model their participation (which is coded as a 0,1 dummy variable). A hierarchical model such as this allows us to estimate individual effects and context effects on turnout, as well as interactions between Level-1 and Level-2 predictors.9 At Level 1, we expect that vote history should explain the lion’s share of the variation in the dependent variable. In our analysis, we are most concerned with identifying who is mobilized by direct democracy and what contexts support or hinder voter participation by specific types of voters in ballot initiative elections. In particular, we are able to observe partisan and contextual mobilization factors, which we have termed short-term mobilization factors, versus those that would be affiliated with being socialized around direct democracy. In line with our short-term expectations, we expect party registration to be a larger determinant of voting behavior in highsalience than in low-salience elections. In addition, partisan context, measured as the percentage of registered voters of the different parties, by Census tract, ought to mitigate voting by nonpartisans/independents in high-salience elections. We estimate these effects by interacting the Level-2 variables for partisan context with the individual-level variable for party registration. We are able to account for the presence of long-term turnout effects by examining the place-of-birth data found in the California voter list.10 We code place of birth as a series of dummy variables—initiative state native, noninitiative state native, and immigrant (California native is the excluded category). Should there be a learning effect that occurs earlier in life at formative ages, or over time, we would expect noninitiative state natives and immigrants to be less likely to turn out to vote in a ballot initiative election. We also test for interactive context effects here—most notably among immigrants who live in areas that have the most Hispanic residents. This is of 9

Model 1 in Table 1 excludes the vote history controls; Model 2 includes them. Other work that uses methodology similar to ours controls for gender (Gimpel, Dyck, and Shaw, 2004; Be´langer and Eagles, 2007). Because gender has not always been asked on California registration cards, including the control would exclude a biased subsample of higher-likelihood voters (notably, those who are older and have been registered longer). Because of that, we exclude gender from the final model. However, the results with gender do not change any of the statistical or substantive interpretations presented here. A copy of these results is available from the authors upon request. 10 This is not available in the Oregon data.

198


particular interest given the effects that ballot initiatives have had on the Hispanic community (Bowler, Nicholson, and Segura, 2006). Age group variables, coded as a series of dummies, are also included—we expect that older voters (age 65 and up is the excluded category) will be more likely to turn out than younger voters. Additionally, we include dummies for length of registration—we expect those who have been registered for a longer period of time to be more likely to vote. We also include a dummy variable for those registering in the last year to control for the potential that voters registered for the first time in California after the 2004 general election. At the context level (Census tract), we also expect social context to influence the probability of voting. We expect that higher socioeconomic status (measured as lower poverty rates, greater numbers of college graduates, and higher median income) to predict an increased likelihood of turnout. Additionally, we expect that African-American neighborhoods will exhibit lower turnout. Finally, we also control for living in a community where the costs of voting are likely higher because of longer commute times (Gimpel and Schuknecht, 2003; Dyck and Gimpel, 2005).11

Results

The results of our nonlinear HGLM model for the California data are presented in Table 1; the results from Oregon are presented in Table 2. The results presented are from the unit-specific HGLM model with robust standard errors. Model 1 for California does not include 2004 vote history controls, while Model 2 does. Without the vote history controls, very little in the model changes, so the interpretations will focus mostly on Model 2 unless otherwise noted. Because the model uses a logit link function for Level 1, the coefficients represent a beta unit change in the natural log of the odds-ratio of voting for a one-unit change in explanatory variable. This, of course, makes interpreting the results from Tables 1 and 2 difficult, especially the partisan context interactions. For ease of substantive interpretation, we present a series of figures highlighting the substantive importance of variables in the model.12 What can be surmised from Tables 1 and 2 is that in line with our expectations, vote history is a consistent and important predictor of turnout 11 This, of course, only makes sense in the context of California given that Oregon uses allmail elections, which presumably reduces the cost of voting imposed on commuters. For model consistency, however, we include the variable in all models. Our expectations are that this variable should only matter in the California models. 12 We were able to replicate the findings of this study using a postelection survey from the 2005 California special election conducted by the Public Policy Institute of California. The advantage of this analysis was that we were able to adequately control for attitudinal determinants of turnout, such as political interest. The replications corroborated our findings and do not add substantially to our argument; they are available from the authors upon request.

Intercept % Black % Commuting 301minutes % College degree (over 25) % Below poverty line Median HH income 1,000 Intercept Intercept Intercept % Democratic % Republican Intercept % Democratic % Republican Intercept Intercept Intercept Intercept Intercept Intercept % Hispanic Intercept % Hispanic Intercept

Overall means (b0)

Immigrant (b12)

Noninit. migrant (b11)

Aged 18–29 (b5) Aged 30–39 (b6) Aged 40–49 (b7) Aged 50–59 (b8) Aged 60–64 (b9) Init. state migrant (b10)

Independent (b4)

04 Primary voter (b1) 04 General voter (b2) Democrat (b3)

Level 2 (Census Tract)

Level 1 (Individual)

Fixed Effects (Unit-Specific Model)

0.2520 n n n 0.5600 n 3.5896 n n n 1.4009 n n n 6.3710 n n n 0.1100 n n n — — 0.0800 n n n 0.3660 n n n 0.6200 n n n 0.3553 n n n 1.1893 n n n 1.6203 n n n 1.7573 n n n 1.1641 n n n 0.7301 n n n 0.3519 n n n 0.0698 n n n 0.0117 n n n 0.0059 0.0414 n n n 0.1565 n n n 0.3204 n n n

b 0.0267 0.2981 0.2207 0.2221 0.4038 0.0000 — — 0.0027 0.0658 0.0587 0.0094 0.2195 0.2057 0.0082 0.0072 0.0061 0.0051 0.0046 0.0030 0.0135 0.0031 0.0137 0.0041

SE 1.2867 0.5712 0.0276 0.2464 0.0017 1.0000 — — 1.0833 1.4420 0.5379 0.7009 0.3044 0.1978 0.1725 0.3122 0.4819 0.7033 0.9326 1.0118 0.9941 0.9594 0.8551 0.7258

Odds Ratio

Model 1 (w/o Vote History Control) SE 0.0221 0.2167 0.1727 0.1655 0.3043 0.0000 0.0044 0.0067 0.0025 0.0610 0.0556 0.0065 0.1928 0.1909 0.0081 0.0069 0.0059 0.0050 0.0047 0.0031 0.0143 0.0032 0.0140 0.0040

b 2.0043 n n n 0.2956 2.4881 n n n 1.0555 n n n 4.4629 n n n 0.1100 n n n 1.5370 n n n 1.7974 n n n 0.0742 n n n 0.7437 n n n 0.0315 0.1495 n n n 1.0328 n n n 1.8033 n n n 1.2346 n n n 0.8012 n n n 0.5116 n n n 0.2370 n n n 0.0196 n n n 0.0079 n n 0.0306 n n 0.0328 n n n 0.1403 n n n 0.1929 n n n

0.1347 0.7441 0.0831 0.3480 0.0115 1.0000 4.6506 6.0338 1.0770 2.1036 1.0320 0.8612 0.3560 0.1648 0.2909 0.4488 0.5995 0.7890 0.9806 1.0080 0.9698 0.9677 0.8691 0.8246

Odds Ratio

Model 2 (with Vote History Control)

Hierarchical Generalized Linear Model of Registered Voter Turnout in the 2005 California Special Election

TABLE 1

Mobilized by Direct Democracy 199

% Hispanic Intercept Intercept Intercept Intercept nnn


NOTES: N for Level 1 5 11,148,738; N for Level 2 5 6,972;

New voter (b13) 1–5 Yr registrant (b14) 6–10 Yr registrant (b15) 11–20 Yr registrant (b16)



po0.01; nn

Odds Ratio 2.0761 0.6805 0.6039 0.7188 0.8705

SE 0.0164 0.0074 0.0058 0.0056 0.0045

po0.05; npo0.10.

0.7305 n n n 0.3850 n n n 0.5043 n n n 0.3301 n n n 0.1387 n n n

b

Model 1 (w/o Vote History Control)

TABLE 1—continued

0.5895 n n n 1.3511 n n n 0.1269 n n n 0.1421 n n n 0.0617 n n n

b

0.0157 0.0110 0.0049 0.0047 0.0042

SE

1.8031 3.8618 0.8808 0.8675 0.9402

Odds Ratio

Model 2 (with Vote History Control)

200 Social Science Quarterly

NOTES: N for Level 1 5 511,092; N for Level 2 5 170;

Aged 18–29 (b5) Aged 30–39 (b6) Aged 40–49 (b7) Aged 50–59 (b8) Aged 60–64 (b9) New voter (b13) 1–5 Yr registrant (b14) 6–10 Yr registrant (b15) 11–20 Yr registrant (b16)

Independent (b4)

nnn

po0.01;

Intercept % Black % Commuting 301minutes % Below poverty line Median HH income 1,000 % College degree (over 25) Intercept Intercept % Democratic % Republican Intercept % Democratic % Republican Intercept Intercept Intercept Intercept Intercept Intercept Intercept Intercept Intercept

Overall means (b0)

02 General voter (b2) Democrat (b3)




po0.05; npo0.10.

0.1227 0.1457 0.3199 0.2843 0.0017 0.0903 0.0284 0.2844 0.4279 0.3545 0.2887 0.4372 0.3598 0.0310 0.0283 0.0262 0.0252 0.0352 0.1192 0.1202 0.1212 0.1234

0.9229 n n n 0.5060 n n n 0.9516 n n n 0.9230 n n n 0.0083 n n n 0.3813 n n n 3.4734 n n n 1.6912 n n n 2.9501 n n n 1.6296 n n n 1.7933 n n n 2.5650 n n n 1.7057 n n n 1.0815 n n n 0.6822 n n n 0.2399 n n n 0.0138 0.0545 0.5591 n n n 0.6725 n n n 0.5510 n n n 0.1719 nn

SE

b

Medium Salience

Jan ‘03 Special Election

1.8002 n n n 0.7770 n n n 0.2692 0.5410 n n n 0.0017 0.4351 n n n 2.7238 n n n 0.8882 n n n 1.3662 n n n 0.7072 n n n 1.1284 n n n 1.2923 n n n 1.0521 n n n 1.9153 n n n 1.6745 n n n 1.1961 n n n 0.7178 n n n 0.3489 n n n 0.8294 n n n 0.1799 n n 0.1994 n n n 0.0465

b SE 0.0761 0.1105 0.2464 0.1893 0.0010 0.0720 0.0257 0.2060 0.3179 0.2592 0.3085 0.4639 0.3817 0.0324 0.0278 0.0235 0.0213 0.0248 0.0755 0.0740 0.0752 0.0734

Low Salience

Sept ‘03 Special Election

0.9267 n n n 0.6552 n n n 1.2381 n n n 1.4456 n n n 0.0073 n n n 0.2757 n n n 3.0456 n n n 0.4521 n 0.9386 n n 0.0925 0.8002 n n n 0.9558 n n n 0.7572 n n n 1.1987 n n n 0.8055 n n n 0.4129 n n n 0.1037 n n n 0.0644 n n 0.7104 n n n 0.5185 n n n 0.4246 n n n 0.1732 n n n

b

0.0353 0.1116 0.2968 0.2511 0.0017 0.0864 0.0261 0.2602 0.3976 0.3178 0.2420 0.3697 0.2911 0.0312 0.0258 0.0236 0.0214 0.0258 0.0422 0.0309 0.0303 0.0293

SE

Medium/High Salience

Feb ‘04 Special Election

Hierarchical Generalized Linear Model of Registered Voter Turnout in Multnomah County in the 2003 and 2004 Oregon Special Elections

TABLE 2

Mobilized by Direct Democracy 201

202


behavior. Additionally, and not surprisingly, there is a clear effect of partisanship on voting, as independents were less likely than Democrats and Republicans to turn out to vote in all four elections. In addition, partisan context effects are inconsistent, with context universally demobilizing independent voters in the 2005 California special election, but actually mobilizing them in Oregon, particularly those in Democratic contexts. Finally, the effects for the place-of-birth variables suggest that Californians and those from ‘‘initiative’’ states were more likely to turn out to vote in 2005 than were noninitiative state migrants and immigrants. In Figure 1 we take up the substance of the long-term effects of direct democracy by considering the probability of voting in the special elections for previous nonvoters.13 While the effect of the difference between those from California and initiative states versus those from other states or foreign countries is statistically significant, we can see that the substance of these findings is relatively marginal. Immigrants, to be sure, are about 3 percentage points less likely than native Californians to vote, while noninitiative state migrants are about 0.75 of a percentage point less likely than Californians to vote. The effects are real, and perhaps surprising given the fairly blunt instrumentation of this concept. However, they also point to the fact that long-term effects on mobilization, while apparent, tell only a small part of the story. In Figure 2 we present the predicted probability of voting by party registration in the four separate elections. Recall that our expectation was that in higher-salience elections, like the 2005 California special elections, the mobilization of nonvoters would reflect a turnout bias, whereas in lower turnout, lower-salience elections, this bias would not be present. The differences in the probability of turning out between partisan and independent previous nonvoters in California in 2005 are substantial. Democrats are almost 40 percentage points more likely to vote than independents, while Republicans are about 30 percentage points more likely to vote than independents. Contrast that difference to the September 2003 Oregon special election, where all previous nonvoters are very unlikely to vote and the difference in turnout between Republicans and independent voters is a single percentage point. In the other two ‘‘medium-salience’’ elections, partisans are more likely to turn out than independents, but generally only by 3–5 percentage points. Finally, we consider the substantive impact of partisan context from California in 2005 in Figure 3. It is evident that increasingly Democratic and Republican partisan contexts further served to demobilize independent voters in the 2005 California special election. If we compare the proba13 We can consider the probabilities either in terms of those who have participated in the past or those who have not. For consistency, we present only the nonvoter probabilities, but note that we have computed and compared the substance of these effects if we move to a different place on the logit curve by considering voters. This type of induced interaction does not change our findings in any substantial way.


203 FIGURE 1

Predicted Probability of Voting by Place of Birth Predicted Probabilities of Voting in 2005 Special Election Among 2004 Nonvoting Independents

Probability of Voting

0.23 0.22 0.21 0.20 0.19 0.18 0.17 Place of Birth

bilities of voting from Oregon to California, independents in the January 2003 special election, the February 2004 special election, and the California 2005 special election all have approximately equal probabilities of voting. FIGURE 2 Predicted Probability of Voting by Party, Four Special Elections Predicted Probability of Voting Among 2002 Nonvoters in the January 2003 Oregon Special Election

0.7

0.7

0.6

0.6



Predicted Probability of Voting Among 2004 Nonvoters in the 2005 California Special Election

0.5 0.4 0.3 0.2 0.1 0.0

0.5 0.4 0.3 0.2 0.1 0

Independent

Democrat

Republican

Independent

0.7

0.7

0.6

0.6

0.5 0.4 0.3 0.2 0.1 0

Republican

Predicted Probability of Voting Among 2002 Nonvoters in the Februrary 2004 Oregon Special Election Probability of Voting


Predicted Probability of Voting Among 2002 Nonvoters in the September 2003 Oregon Special Election

Democrat

0.5 0.4 0.3 0.2 0.1 0

Independent

Democrat

Republican

Independent

Democrat

Republican

204

Social Science Quarterly FIGURE 3 Changes in Predicted Probabilities of Voting by Partisan Context Predicted Probability of Voting in the 2005 Special Election Among 2004 Nonvoting Independents 0.5


0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0

0.1

0.2

0.3

0.4 0.5 0.6 % Partisans in Tract Dem Context

0.7

0.8

0.9

1

Rep Context

When the salience and campaigns surrounding 2005 went into full swing, the entirety of the mobilization effect was among partisan voters. Figure 3 demonstrates that in the most partisan contexts, independent voters were least likely to turn out. This finding is not present in the Oregon models, but in Oregon there is also not nearly as large or noticeable a boost in partisan turnout as in the California model. In addition, most of the other control variables work as expected. At the individual level, we demonstrate that voting in elections is indeed a habit, with past participation increasing the odds of voting four and six times over for primary and general elections, respectively. Additionally, members of each age group are marginally less likely to vote than those 65 years and older, and the odds-ratios become progressively larger as we move from 18–29 year-olds (0.29) to 60–64 year-olds (0.97). Length of registration works in a similar manner, with the exception of new voters in Model 2. When we control for vote history, those who did not vote in 2004 because they were not registered are actually more likely than any other age cohort to be among those turning out to vote in 2005. However, length of registration does appear to matter, as those registered more than 20 years are more likely to vote. For Level-2 effects, we see the importance of various neighborhood contexts. Those who live in areas where more residents have long commutes are less likely to vote in California, but not in Oregon. Being a resident of a Californian community where commute times are one standard deviation above the mean reduces the odds-ratio to 0.742. As the percent of those above the poverty line increases to one standard deviation above its mean


205

(from 14 percent to 25 percent), the odds-ratio of an individual in that context voting is 0.61 when compared to an individual living in the mean context. The percentage of the population 25 or older with a college degree is also significant. Additionally, we see that income also seems to matter in all save the September 2003 model, a result that is due to collinearity with other Level-2 controls. Discussion

This article was motivated by a desire to clarify the seemingly robust effects discovered in the literature regarding direct democracy’s ability to increase voter turnout. Much of the literature has suggested that this finding has its theoretical footing in participatory democratic theory, which argues that given more opportunities for meaningful participation, voters will learn to become more active and engaged democratic citizens. To be sure, a great deal of other work has suggested that many initiatives, and particularly salient initiatives, are highly constrained by partisan politics. We attempt to reconcile this conflict. Employing a unique research design and data from special elections, where the meaningful mobilization effects of ballot measures can be isolated, we test hypotheses regarding both short-term factors (active mobilization) and long-term factors (learning) regarding mobilization in direct democratic elections. The findings suggest that while there is a real effect of having been socialized in a direct democracy context, which suggests an educative role for ballot measures, the effect is relatively small in comparison to the ability of campaigns to motivate peripheral voters to turn out. That mobilization, however, appears to occur most clearly among partisan voters. When elections are high-stimulus, high-spending affairs, scores of partisan voters make their way to the polls. However, when the elections are low-salience, lowturnout affairs, mobilized voters are proportionally allocated among partisans and independents. In addition, in salient contests like the 2005 California special election, partisan mobilization is mediated by partisan context. A skeptical reader might find this result unsurprising. However, the results highlight both an inconsistency in the literature regarding long-term mobilization effects in ballot measure elections, as well as a serious deficiency in the thinking of progressive reformers who saw ballot initiatives as a way to mobilize the disaffected middle. Institutions can be crafted to give voters more choices, but within the context of a two-party system, the presence of partisan social contexts and campaigns will continue to bias mobilization to partisan voters. Reforms throughout the 20th century sought to sap the power of political parties. Our evidence suggests that direct legislation, for better or worse, has had little effect in limiting the power of parties, and certainly has not mobilized an army of unaffiliated voters. If anything, initiatives allow parties an alternative route to mobilize supporters with an

206


inconsistent vote history. The turnout gains from ballot measures need to be understood in this context.

REFERENCES Anderson, Barbara A., and Brian D. Silver. 1986. ‘‘Measurement and Mismeasurement of the Validity of the Self-Reported Vote.’’ American Journal of Political Science 30:771–85. Baldassare, Mark. 2005. PPIC Statewide Survey, October 2005. Public Policy Institute of California. Available at hhttp://www.ppic.org/content/pubs/survey/S_1005MBS.pdfi. Baldassare, Mark, and Cheryl Katz. 2007. The Coming Age of Direct Democracy: California’s Recall and Beyond. New York: Rowman and Littlefield. Barber, Benjamin R. 1984. Strong Democracy: Participatory Politics for a New Age. Berkeley, CA: University of California Press. Beck, Paul Allen, Russell J. Dalton, Steven Greene, and Robert Huckfeldt. 2002. ‘‘The Social Calculus of Voting: Interpersonal, Media, and Organizational Influences on Presidential Choices.’’ American Political Science Review 96:57–73. Be´langer, Paul, and Munroe Eagles. 2007. ‘‘Partisan Cross-Pressure and Voter Turnout: The Influence on Micro and Macro Environments.’’ Social Science Quarterly 88:850–67. Berelson, Bernard R., Paul F. Lazarsfeld, and William N. McPhee. 1954. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago, IL: University of Chicago Press. Bernstein, Robert, Anita Chadha, and Robert Montjoy. 2001. ‘‘Overreporting Voting: Why it Happens and Why it Matters.’’ Public Opinion Quarterly 65:22–44. Bowler, Shaun, and Todd Donovan. 2002. ‘‘Democracy, Institutions and Attitudes About Citizen Influence on Government.’’ British Journal of Political Science 32:371–90. Bowler, Shaun, Stephen P. Nicholson, and Gary M. Segura. 2006. ‘‘Earthquakes and Aftershocks: Race, Direct Democracy, and Partisan Change.’’ American Journal of Political Science 50:146–59. Campbell, Andrea Louise, Cara Wong, and Jack Citrin. 2006. ‘‘‘Racial Threat’, Partisan Climate, and Direct Democracy: Contextual Effects in Three California Initiatives.’’ Political Behavior 28:129–50. Campbell, James E. 1987. ‘‘The Revised Theory of Surge and Decline.’’ American Political Science Review 31:965–79. Clausen, Aage R. 1968. ‘‘Response Validity: Vote Report.’’ Public Opinion Quarterly 32:588–606. Cooper, Christopher A., Moshe Haspel, and H. Gibbs Knotts. 2009. ‘‘The Value of Voterfiles for U.S. State Politics Research.’’ State Politics & Policy Quarterly 9:102–21. Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper. Dyck, Joshua J. 2009. ‘‘Initiated Distrust: Direct Democracy and Trust in Government.’’ American Politics Research 37:539–68. Dyck, Joshua J., and James G. Gimpel. 2005. ‘‘Distance, Turnout, and the Convenience of Voting.’’ Social Science Quarterly 86:531–48. Dyck, Joshua J., and Edward L. Lascher Jr. 2009. ‘‘Direct Democracy and Political Efficacy Reconsidered.’’ Political Behavior 31:401–27.


207

Frey, Bruno S., and Alois Stutzer. 2002. ‘‘What Can Economists Learn from Happiness Research?’’ Journal of Economic Literature 40:402–35. Gimpel, James G., Joshua J. Dyck, and Daron R. Shaw. 2004. ‘‘Registrants, Voters, and Turnout Variability Across Neighborhoods.’’ Political Behavior 26:343–75. Gimpel, James G., J. Celeste Lay, and Jason E. Schuknecht. 2003. Cultivating Democracy: Civic Environments and Political Socialization in America. Washington, DC: Brookings Institution Press. Gimpel, James G., and Jason E. Schuknecht. 2003. ‘‘Political Participation and the Accessibility of the Ballot Box.’’ Political Geography 22:471–88. Grummel, John A. 2008. ‘‘Morality Policy, Direct Democracy, and Turnout.’’ State Politics and Policy Quarterly 8:282–92. Haider-Markel, Donald P., and Kenneth J. Meier. 1996. ‘‘The Politics of Gay and Lesbian Rights: Expanding the Scope of the Conflict.’’ Journal of Politics 58:332–49. Hero, Rodney E., and Caroline J. Tolbert. 2004. ‘‘Minority Voices and Citizen Attitudes About Government Responsiveness in the American States: Do Social and Institutional Context Matter?’’ British Journal of Political Science 34:109–21. Hibbing, John R., and Elizabeth Theiss-Morse. 2002. Stealth Democracy: Americans Beliefs About How Government Should Work. New York: Cambridge University Press. Highton, Benjamin. 2000. ‘‘Residential Mobility, Community Mobility, and Electoral Participation.’’ Political Behavior 22:109–20. Huckfeldt, Robert R. 1979. ‘‘Political Participation and the Neighborhood Social Context.’’ American Journal of Political Science 23:579–92. King, Gary. 1997. A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data. Princeton, NJ: Princeton University Press. Lacey, Robert J. 2005. ‘‘The Electoral Allure of Direct Democracy: The Effect of Initiative Salience on Voting, 1990–96.’’ State Politics and Policy Quarterly 5:168–81. McDonald, Michael P. 2002. ‘‘State Turnout Rates Among Those Eligible to Vote.’’ State Politics and Policy Quarterly 2:199–212. Mendelsohn, Matthew, and Fred Cutler. 2000. ‘‘The Effect of Referendums on Democratic Citizens: Information, Politicization, Efficacy and Tolerance.’’ British Journal of Political Science 30:685–98. MSNBC. 2005. Schwarzenegger Weakened in Election: ‘No Silver Lining’ as California Voters Reject All Four Measures. MCNBS News Services November. Available at hwww.msnbc.msn.com/id/9975371/i. Mutz, Diana C. 2002. ‘‘The Consequences of Cross-Cutting Networks for Political Participation.’’ American Journal of Political Science 46:838–55. Pateman, Carole. 1970. Participation and Democratic Theory. New York: Cambridge University Press. Plutzer, Eric. 2002. ‘‘Becoming a Habitual Voter: Inertia, Resources, and Growth in Young Adulthood.’’ American Political Science Review 96(1):41–56. Rosenstone, Steven J., and John Mark Hansen. 1993. Mobilization, Participation, and Democracy in America. New York: Longman. Schattschneider, E. E. 1960. The Semisovereign People: A Realist’s View of Democracy in America. New York: Holt, Rinehart and Winston.

208


Schlozman, Daniel, and Ian Yohai. 2008. ‘‘‘How Initiatives Don’t Always Make Citizens: Ballot Initiatives in the American States, 1978–2004.’’ Political Behavior 30:469–89. Smith, Daniel A., and Caroline J. Tolbert. 2004. ‘‘The Initiative to Party: Partisanship and Ballot Initiatives in California.’’ Party Politics 7:739–57. ———. 2004. Educated by Initiative: The Effects of Direct Democracy on Citizens and Political Organizations in the American States. Ann Arbor, MI: University of Michigan Press. Smith, Mark A. 2001. ‘‘The Contingent Effects of Ballot Initiatives and Candidate Races on Turnout.’’ American Journal of Political Science 45:700–06. ———. 2002. ‘‘Ballot Initiatives and the Democratic Citizen.’’ Journal of Politics 64:892– 903. Straits, Bruce C. 1990. ‘‘The Social Context of Voter Turnout.’’ Public Opinion Quarterly 54:64–73. Tolbert, Caroline J., John A. Grummel, and Daniel A. Smith. 2001. ‘‘The Effects of Ballot Initiatives on Voter Turnout in the American States.’’ American Politics Research 29:625–48. Tolbert, Caroline J., Ramona S. McNeal, and Daniel A. Smith. 2003. ‘‘Enhancing Civic Engagement: The Effect of Direct Democracy on Political Participation and Knowledge.’’ State Politics and Policy Quarterly 3:23–41. Tolbert, Caroline J., and Daniel A. Smith. 2005. ‘‘The Educative Effects of Ballot Initiatives on Voter Turnout.’’ American Politics Research 33:283–309.